University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | CHAPTER 4. DETAILED TESTING OF FLOTATION ROTOR DESIGN
at high tip speeds while drawing low power. Furthermore, the drastic phenomenologi-
cal difference in power reduction and gas dispersion (Figure 5.9) may provide insight
on the operational advantages present in this rotor.
4. Intermsoffloodinglimit, solidsuspension, andoverallareaofstableoperation, theVT
- B2 substantially outperforms its counterparts. This rotor is capable of suspending
solids at a relatively low power draw, and this suspension capacity is not as sensitive to
airadditionastheotherrotors. ThetwoadditionalbladesoftheVT-B2provideaddi-
tional area for gas dispersion. This mechanical feature necessitates an increased power
number, but it also provided additional capacity to overcome the flooding condition.
5. Bulk comparisons of the air holdup versus power data and the rate constant versus
power data show strikingly similar trends. While a one-to-one correlation is not neces-
sarily suggested, the same results are determined from test. Theoretically, this result
is justified since air holdup is related to the available surface area which is related to
flotation performance.
6. FactorssuchasS andotherdimensionlessparameters(Reynoldsnumber, PowerNum-
b
ber, and Aeration Number) are capable of producing consistent relationships to perfor-
mance criteria for each rotor. The magnitude of the laboratory values generally match
those reported for industrial machines. The continued use of these values in scaling
studies is recommended.
4.5 Bibliography
Banisi, S., & Finch, J. (1994). Technical note reconciliation of bubble size estimation
methods using drift flux analysis. Minerals Engineering, 7(12), 1555–1559.
Fitzgibbon, A., Pilu, M., & Fisher, R. (1999). Direct least square fitting of ellipses. Pattern
Analysis and Machine Intelligence, IEEE Transactions on, 21(5), 476–480.
Hernandez-Aguilar, J., Gomez, C., & Finch, J. (2002). A technique for the direct mea-
surement of bubble size distributions in industrial flotation cells. In Proceedings of the 34th
annual meeting of the canadian mineral processors (pp. 389–402).
Silva, R., Echeverri, L., Olson, T., Foreman, D., Yang, Y., & Caldwell, K. (2012). The
effect of laboratory cell design on flotation machine hydrodynamics, solid suspension and
particle recovery. In 2012 sme annual meeting and exhibit preprints (pp. 452–457). SME.
122 |
Virginia Tech | Chapter 5
Conclusions and Recommendations
The primary objective of this research was to characterize and compare small-scale
flotation rotors while developing scaling relationships and future design criteria. In order to
complete this task, several experimental setups were initially constructed to measure various
flotation performance parameters. In choosing the appropriate parameters to measure, the-
oretical factors, practical limitations, and industrial experience were equally acknowledged.
As a result, most testing focused on four distinct evaluation criteria: power draw, gas disper-
sion, operational robustness, and flotation performance. The measurement and analytical
approach to these criteria needed to provide reliable data which retains significance as the
rotors are scaled to the industrial size. Furthermore, the data must provide a basis for
fair comparison among the alternative rotor designs, and the experimental techniques must
coincide with the expected paradigms and costs of a typical flotation laboratory.
This report extensively described the experimental techniques used to evaluate the four
performance criteria. The theoretical background was first established, and the required
equipment and analytical means were explained. When applicable, measurements were nor-
malized by or compared against power input. In the laboratory setting, power input is the
most easily determined cost associated with flotation equipment, and this measurement may
be scaled to the full-size by dimensionless relationships. Furthermore, when comparing alter-
native rotors, differences in power draw measured at the laboratory scale will retain relative
relationships as the scale is increased (although the absolute magnitude of the difference
may radically change).
Testing for this project task was primarily conducted in two campaigns. The original,
exploratory campaign focused on the analysis of 14 rotors in a select range of operating
conditions. The second, detailed campaign included on a more rigorous analysis of three
123 |
Virginia Tech | CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS
selectrotorsinseveraloperatingconditions. Alltestswereperformedwith2.75inchdiameter
rotors in tanks 10 and 14 inch diameter tanks (9 - 35 liters). The goals of the exploratory
campaign were to establish and refine the experimental procedures, verify the existence of
performance difference between rotors, and narrow the field of potentially viable prototypes
by eliminating underperforming or “fatally-flawed” alternatives. From this campaign, the
SP - A1, VT - A4, and VT - B2 were selected for further experimentation. The goals of the
detailed campaign were to implement standard modes of evaluation and analysis, examine
uncertainty and repeatability in the measurements, thoroughly characterize the remaining
rotor prototypes, and establish insight on the causes of performance differences.
Following the successful completion of both laboratory campaigns, the following obser-
vations and design criteria are derived:
1. Given the strengths of the VT - A4 in the areas of power-normalized gas dispersion
andflotationperformance,flotationsystemsdesiringthesecharacteristicsshoulddesign
rotors to exhibit low inherent power numbers and operate near the boiling condition
(seeFigure4.3, Figure4.12, Table4.6, andFigure4.22. Unfortunately, theseattributes
general realize a reduction in operational robustness, especially with regard to the
sanding condition (see Figure 4.15, and Figure 4.16).
2. Alternatively, the strengths of the VT - B2 rotor in the areas of operational robustness
should entice high power numbers and operational away from the boiling condition
with solid suspension is problematic (see Figure 4.3, Figure 4.14, Figure 4.15, and
Figure 4.16).
3. Given the measured linear relationship between k and S , rotors desiring high flotation
b
rates should seek to maximize S (see Figure 3.13, Figure 3.14, and Figure 4.18).
b
4. Plots showing global air holdup plotted against power show striking general similar-
ity to plots showing flotation rate constant plotted against power. Consequently, air
holdup can be used as an acceptably surrogate to flotation testing when comparing
machine designs (see Figure 4.20 and Figure 4.22).
5. Care should be taken when comparing batch and continuous data. S ratios should
b
be used to scale the measured flotation rates, and appropriate plug-flow or perfectly-
mixed models should be selected. Note that in order to achieve the same recovery up
to a seven-fold increase may be necessary in residence time when changing from batch
to a continuous reactor (see Figure 4.23, compare residence time at 65% recovery).
124 |
Virginia Tech | CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS
Finally, the author of this thesis recommends the following three topics for further and
continued study:
1. Additional quantification of scale factors and verification of lab-scale to full-scale ap-
plicability. While the tested rotors consistently show variations in performance criteria
on the lab scale, the absolute magnitude of these differences is not expected to remain
proportional as the scale is increased. Testing and characterization at the larger scales
may assist in identifying scale-factor which can reliably predict the magnitude of these
differences as the scale is increased.
2. Larger scale implementation of the prototype rotors. Given the strengths each rotor
has consistently manifested at the laboratory scale, further development and imple-
mentation at a larger scale is recommended.
3. Additional refinement of laboratory testing and analytical techniques. While the meth-
ods described in this report have fulfilled the goals of obtaining a fair and consistent
basis of comparison between the rotor alternatives, further development and revision
may provide additional insight to the fundamental difference between each rotor. Also,
byincorporatingautomation(levelcontrols, airholdupmeasurements, solidsuspension
determination), the degree of testing reliability and reproducibility will be enhanced.
125 |
Virginia Tech | Elutriation Technology in Heavy Mineral Separations
Matthew Donnel Eisenmann
ABSTRACT
Hindered-bed separators have been used in several different mineral processing fields
for many years. Recent improvements in designs have led to the development of the
CrossFlow separator. This new design employs a tangential feed system that has shown
promise in several applications. This paper investigates the use of this relatively new
technology to upgrade heavy mineral concentrates using Florida type ores. The intended use
of this separatory device in this particular application is the removal of gangue quartz from
other valuable heavy minerals such as ilmenite, leucoxene, rutile, zircon, and staurolite. The
results of two different pilot-scale in-plant testing investigations are discussed. In general,
quartz rejections in excess of 80% were achieved while maintaining TiO and heavy mineral
2
recoveries above 98% and 99%, respectively.
In addition to field test work, two separate unit models have been developed. The
first model is an empirical investigation into understanding unit operation and functionality.
The second model is a statistical prediction of unit operation based on specific field test
work. These models can be used to effectively scale-up a CrossFlow unit for full-scale
installation at any Florida heavy mineral sands operation. Emphasis is placed on unit
capacity and other operational parameters such as elutriation flowrate and bed level.
i |
Virginia Tech | ACKNOWLEDGEMENTS
The author wishes to express his deepest appreciation to Dr. Gerald Luttrell. His
support, guidance, and knowledge throughout this thesis work was invaluable. The author
would also like to thank Dr. Greg Adel and Dr. Roe-Hoan Yoon for their advice throughout
undergraduate and graduate studies.
The author would like to thank Dupont White Pigments and Mineral Products for
their continued support. Special thanks to Vincente Stutts, Process Engineer, and Francisco
Parisi, R&D Supervisor, for their understanding, guidance, and assistance.
Sincere appreciation is extended to all those graduate students who spent many late
nights at the Plantation Road Laboratory including: Jaisen Kohmuench, Tim McKeon, Ian
Sherrell, and even Brian Halford. Special thanks to the Department of Mining and Minerals
Engineering Administrative and Secretarial Staff.
The author would also like to express his thanks to his parents, siblings, and God; for
their constant influence on his continued education. They all provided love and support
throughout this entire process, as well.
Lastly, the author would like to express his love and appreciation to his wife, Colleen,
for her love, support, understanding, and caring attitude that made this entire thesis possible.
ii |
Virginia Tech | CHAPTER 1
1.1 Introduction
Titanium Dioxide (TiO ) is a material used in a variety of commercial products.
2
Relatively pure titanium dioxide is used extensively as a pigment in the food, ceramics,
paper, and paint industries. Titanium is also used in metal production where its low weight
and high strength are beneficial in many different applications. The primary source of TiO
2
is a mineral called ilmenite. In 1998 the world production of ilmenite was over 4 million
metric tons. Along with the United States, other major producers include South Africa,
Australia, and Canada.
In the United States, Florida and Virginia are the two key production areas for
titanium bearing minerals. These sites occur primarily as placer type deposits and span
several miles. The TiO in these deposits is primarily contained within the heavy minerals
2
ilmenite (FeTiO ), rutile (TiO ), and leucoxene (FeTiO ). Also contained within these
3 2 3
deposits are significant amounts of zircon (ZrSiO ) and staurolite (Fe+2Al [SiO ] O [OH] ).
4 4 4 2 2 2
Zircon is used in zirconium metal and various abrasive and foundry applications; and
staurolite is used primarily as an abrasive in the sand blasting industry. Due to their high
specific gravities, the above listed specific minerals are termed heavy minerals throughout
the industry.
The initial processing of these heavy mineral sands is done primarily by the use of
spiral separators. This type of separatory equipment uses the differences in densities of the
material to concentrate the heavy minerals through low gravity means. Characteristics such
as low unit efficiency and low capacity have plagued this type of wet separation in the past.
1 |
Virginia Tech | Until recent years, heavy mineral concentration was performed solely by spiral separators.
However, several different gravity based separation techniques have been developed and
implemented in various mineral processing applications.
Of these techniques, hindered-bed classifiers have been used quite frequently in
various sections of the heavy mineral industry. Their low capital cost, ease of operation, and
high capacities has led to promising results in several different applications. These devices
have also been used quite successfully in other industries such as coal, mica, sand and gravel,
and phosphate. This type of density based separator relies on two key physical properties of
the mineral being separated, size and density. In a mono-density application this device is
quite efficient in performing a size separation of the feed material. In a mono-size
application a hindered-bed classifier is extremely efficient in performing a density separation.
However, when the feed material contains varying mixtures of both size and density, unit
performance is significantly reduced.
Several investigations have been performed on the implementation of hindered-bed
separators and their use within the heavy minerals industry. They have proven to be
successful when used in combination with the spiral circuits. Recent work has shown
promise in specifically using this device as a final stage in order to upgrade the concentrate
generated by the spiral circuit. A novel device known as the CrossFlow has been examined
for this particular application. The CrossFlow uses a new feed introduction system which
has been proven effective in several heavy mineral applications. Both laboratory and pilot-
scale test work in this application are discussed within this report.
2 |
Virginia Tech | 1.2 Literature Review
Hindered-bed classifiers have been used extensively throughout the minerals industry
for several years. These classifiers are typically used to categorize mineral particles
according to both size and density. They have been proven useful in recent years for density
based separations in fine coal applications (Mankosa et al., 1995; Reed et al., 1995; Honaker,
1996). Also, these devices have been of particular interest for carbon recovery from both
active fine coal streams and refuse ponds. With low capital cost and high capacities,
hindered-bed classifiers have found their way into many different sectors of the mining
industry.
Most hindered-bed classifiers operate on the same basic principles. First, material is
introduced into a tank in slurry form. Water, termed teeter water, is then injected into the
tank bottom. This process generates a zone of fluidized particles above the teeter water
injection point. This zone is where separation of the particles takes place based on both
differences in gravity and particle size. Within the fluidized zone, particles with a low
settling velocity report to the unit overflow. Under normal operation, both fine particles and
lower density particles have low enough settling velocities to report to this location. During
the same process, coarser particles and heavy material with much greater settling velocities
work their way through the unit and are eventually ejected from the bottom outlet of the
device. The quantity of teeter water added to the unit and height location of the fluidized bed
are the two main factors that determine the separation process.
These separators typically work quite efficient in applications where the feed material
is either mono-density or mono-size. However, when the feed presented to the unit is varied
in both size and density, efficient separation becomes difficult. Under this application the
3 |
Virginia Tech | shortcoming of this particular unit is that coarse low-density particles tend to accumulate in
the teeter bed region. As other particles begin to congregate in this same zone, these coarse
low-density particles are ultimately forced through the bottom unit discharge by mass action.
To compensate for this unit phenomenon, the teeter water rate is typically increased, but this
causes fine, dense material to become misplaced to the overflow. Both of these processes
adversely affect the unit’s overall separation efficiency.
Recent test work with Australian based mineral sands concentration has shown
several successful circuit combinations using both spiral and hindered-bed separation
technology (Elder et al., 2001). This test work employed the use of a typical hindered-bed
separator as a final upgrading step after spiral concentration of heavy minerals. Although the
results were promising, high heavy mineral recovery was only successful for low levels of
quartz rejection. Table I shows the data derived during this investigation. For higher
rejection levels of 81.7%, heavy mineral recovery drops to 96.6%. It may be possible that
Australian mineralogy adversely affects the performance of this particular setup. The
specific unit design employed for this testing could also negatively impact performance.
Table I: Investigative Hindered-bed Separation Test Work on Heavy Mineral Concentrate
(Elder et al.,2001)
4 |
Virginia Tech | Pilot-scale testing of a conventional hindered-bed separator for heavy mineral
separation, which later developed into full-scale installation, was performed by McKnight et
al., 1995. This particular application was the removal of quartz gangue from zircon rich
tailings. These tailings contained approximately 3-8% Zircon, 18-22% Al O , and 1% TiO .
2 3 2
Quartz rejections above 80% were achieved with zircon recoveries higher than 90%. In this
particular application the density difference between zircon, 4.7, and quartz, 2.7, was
sufficient to allow proper separation despite the tremendous size discrepancies between the
two minerals. However, in near-density applications this conventional hindered-bed
approach would prove insufficient (Dunn et al, 2000).
To overcome the shortcomings with typical hindered-bed separators, a novel device
known as the CrossFlow separator was developed. This unit relies on an alternate feed
presentation system that reduces the amount of fine, dense particles that become misplaced to
the overflow launder. The CrossFlow is comprised primarily of a rectangular tank, as seen in
Figure 1, that is divided into two regions; an upper separation chamber and a lower
dewatering cone. Within this novel device, feed is presented tangentially across the upper
portion of the unit. This tangential feed entry system allows for a lower velocity introduction
of feed across the top of the classifier. This feed presentation design allows the unit to
operate more efficiently than other conventional apparatuses.
5 |
Virginia Tech | Given the advantages of the CrossFlow elutriator, testing began on various samples of
heavy mineral concentrate from a typical Florida mineral sands operation (Dunn et al, 2000).
The samples were comprised of several different minerals with differences in both size and
density. The majority of the concentrate samples were comprised mostly of ilmenite,
leucoxene, zircon, staurolite, and quartz. In this particular application, quartz represents the
gangue/trash material that must be removed in order to produce pure heavy mineral products.
Individual mineral characteristics are discussed later within this report.
1.3 Typical Heavy Mineral Operations
1.3.1 Processing Overview
Several heavy mineral operations are located in northern Florida. The deposits are
relatively shallow and span many hundreds of acres stretching from south to north. The
majority of heavy minerals contained within these deposits are present within the first 50 feet
of ground depth. This thickness reduces to a mere 15 feet toward the outer edges of the
deposit on both the west and east sides. The in-situ feed matrix is comprised mostly of sand,
clay, and “hard pan”. Hard pan is a composite material containing individual sand grains
cemented together by humate, clay, and fine organics. This cemented material is located
throughout the deposits in thin veinlets ranging from 1 to 4 feet thick.
The in-situ sand matrix is first excavated by means of a dredge. This dredging
technique involves the use of a rotary cutter head to excavate and loosen the feed matrix.
The dredge pivots around a spud system through use of several different swing lines. The
dredging techniques developed by the mining industry are basically retrofitted canal dredging
systems. The excavated material is then pumped directly to a rotating trommel, which
7 |
Virginia Tech | removes plus 1/4” material. The oversize material is primarily made up of roots, other
organic material, and hard pan. Although hard pan contains economic levels of TiO bearing
2
minerals, it is extremely abrasive and difficult to process, and is thus rejected from the
process immediately.
After the oversize has been removed, the feed material is sent through several stages
of spiral classifiers. This spiral circuit is responsible for upgrading the feed from
approximately 3-5% heavy mineral to a concentrate grade of 75-85%. The spiral circuit
consists of four stages with numerous re-circulating loads. The four circuits used in this
process are rougher, cleaner, finisher, and scavenger. Each of these stages consists of spirals
specifically designed for certain separations. This circuit configuration is extremely effective
in generating low grade, “throw away” tailings. These tailings are immediately pumped for
re-contouring and subsequent reclamation activities at the mine site. However, this four
stage configuration has proven inefficient in upgrading the concentrate above 85% heavy
mineral without suffering high losses in both TiO and heavy mineral recovery. Figure 2
2
shows the general flowsheet for this concentrate after gravity separation.
8 |
Virginia Tech | important, as it also adversely affects dry mill separation. The “clean” concentrate is then
transported to several dry mill facilities, where electrostatic and magnetic separation
techniques are used to segregate the various titanium bearing minerals and staurolite
products.
Tailings generated from the dry mill consist mainly of quartz, zircon, and alumino-
silicates. However, small amounts of TiO bearing minerals and staurolite are still present
2
from the previous imperfect separation. These dry mill tailings are further processed using
spirals and additional dry milling equipment to produce a saleable zircon product. This
zircon product must be extremely low in trash material such as alumina, iron, and titanium.
The extremely high product quality required makes recovery of zircon a difficult process.
Many stages of spirals and dry milling equipment are required to meet these specifications.
1.3.2 Florida TiO Mineralogy
2
In the formation of these placer type deposits, the heavy mineral grains were
selectively concentrated during various geologic transportational mechanisms. Evidence of
these processes can be readily seen in “black sand” concentrations commonly found in most
ocean beach environments. However, millions of tons of heavy mineral concentrate are
required for deposits to be economically justifiable. Such deposits must have undergone
lengthy geologic conditioning in order to reach concentrations of this magnitude. One
example of this process is the Old Hickory Deposit owned and operated by Iluka Resources
Limited, located in southeastern Virginia. Other deposits are located throughout northern
Florida which are owned and operated by both Iluka Resources Limited and DuPont White
Pigment and Mineral Products.
10 |
Virginia Tech | The heavy mineral assemblages contained within these deposits are comprised mostly
of ilmenite, leucoxene, rutile, zircon, staurolite, and kyanite. Other minerals such as
monazite, spinel, and garnet are also present, but in relatively low quantities. Table II gives
general information on the density and composition of these minerals. The TiO minerals
2
which comprise the majority of production from these deposits are ilmenite, leucoxene, and
rutile. Zircon and staurolite products are also produced in significant quantities from these
deposits.
Table II: Summary of Mineral Characteristics
Chemical Specific % of HM
Mineral Type
Composition Gravity Suite
Ilmenite FeTiO 4.2 - 4.5 30
3
Leucoxene TiO 3.6 - 4.3 10
2
Rutile TiO 4.2 - 4.3 3
2
Zircon ZrSiO 4.7 16
4
Staurolite Fe,AL(SiO )4(OH) 3.7 18
4
Kyanite Al O (SiO ) 3.2 - 3.7 15
2 3 4
Of these minerals, the four that are of most economic interest are ilmenite, rutile,
leucoxene, and zircon. Many other heavy minerals are present in this deposit, but are not of
sufficient economic interest. The heavy mineral concentrate generated from the wet mill
process also contains small amounts of clay and quartz, which were not properly separated.
A set of four samples from this deposit where used for investigation using reflected light
microscopy techniques. The textures and assemblages found within these samples are
discussed.
11 |
Virginia Tech | For both primary products, ilmenite and zircon, chemical make-up as determined by
x-ray fluorescence (XRF) is of much importance. The stoichiometric amount of iron (Fe),
aluminum oxide (Al O ), and zirconium oxide (ZrO ) are also of importance. It is thus very
2 3 2
important to determine what characteristics of the heavy mineral feed affect its chemical
composition and separation efficiency.
Stoichiometric ilmenite contains 53% TiO , while, ilmenite in placer deposits varies
2
widely in its actual titanium concentrations. This titanium concentration is dependent upon
both the initial factors that occurred during deposition and those from subsequent weathering.
Ilmenite is particularly resistant to weathering and displays a wide variety of textures in
placer deposits. These textures can range from single crystals to polycrystalline aggregates.
At very high temperatures ilmenite and hematite are in solid solution with one another. At
lower temperatures the hematite will exsolve out of the solution and form lamellae structures
within grains and at grain boundaries. These exsolution lamellae of hematite are often
formed in crystallographic orientations.
During weathering and erosion, selective dissolution of the more soluble hematite
proceeds from the margins of the grains and continues inward. This process leaves behind
empty pits within the ilmenite grains. During this process of leaching, the ilmenite grain
becomes increasingly enriched in its titanium concentration. The degree to which this
leaching has occurred is important because iron is not a desired constituent of the ilmenite
product. Figure 3 shows reflected light microscopy slides of an ilmenite grain that has not
undergone complete leaching. The leaching and removal of iron occurs primarily at the grain
boundaries and along internal discontinuities. Hematite is dissolved first from the ilmenite
grains due to its greater solubility.
12 |
Virginia Tech | pale cream color under reflected light microscopy. Typical zircon grains are both
nonmagnetic and nonconductive, although some grains exhibit para-magnetic properties due
to small internal impurities in their grain structure. Zircon grains are often elongated and
more angular than ilmenite due to breakage along cleavage planes during weathering. These
grains show strong white internal reflection under crossed polars, as seen in Figure 6.
Reflected light microscopy is a very important diagnostic tool in identifying various levels of
internal impurities that may affect both separation and recovery.
FIGURE 6: ZIRCON GRAIN SHOWING CHARACTERISTIC WHITE INTERNAL REFLECTIONS
Rutile is also present in the mineral samples, but in much lower concentrations.
Rutile has almost a pure TiO assemblage. It is a non-magnetic, conductor which often tends
2
to be finer in nature than the other heavy minerals present. Under reflected light, rutile is
most noticeable by its internal reflections that are reddish brown. The existence of twinning
is also an important diagnostic tool when locating rutile under the microscope. Figure 7
15 |
Virginia Tech | total grain structure. In this circumstance, the specific gravity of the grain becomes 2.8,
instead of 3.7 for a uniform ilmenite grain. When submersed under water, these void spaces
also generate buoyancy forces. The existence of buoyancy forces within dense minerals
present in an elutriation device, cause significant reduction in unit separation efficiency.
These grains are inherently more difficult to separate in a density based separatory device.
1.4 Problem Statement
Typical concentrate generated from the wet mill facilities is comprised of only 70-
85% heavy mineral. This particular composition is the maximum concentration possible
using the existing spiral circuits without suffering excessive recovery penalties. The excess
quartz fraction present must be scrubbed, hauled, and dried along with the valuable heavy
minerals prior to dry mill separation. This quartz fraction also reduces dry mill capacity and
lowers the magnetic and electrostatic separation efficiencies. The potential benefits of
installing an effective upgrading technology at the wet mill facilities are numerous and
include:
• Reduced haulage costs
• Reduced chemical consumption during attrition scrubbing
• Reduced drying costs
• Increased dry mill TiO recovery as a result of higher grade feed
2
Due to space restrictions at both wet mills, the addition of a fifth spiral circuit to
achieve the required concentrate upgrade is not seen as a feasible solution. Another negative
for spiral separation is the need for many secondary devices such as distributors, launders,
piping, and additional pumps if gravity flow is not possible. The addition of another stage of
17 |
Virginia Tech | spirals would also complicate the existing circuit quite drastically making the selection of
overall circuit optimum settings extremely difficult. The particular properties of the
concentrate produced through the current circuit are such that an alternate separation device
would be required to achieve the desired results.
Based on available data, an elutriator is one such device that employs different
separation mechanisms to achieve mineral separation than the current spiral technology
employed. Installation of an elutriation device would require less space and less ancillary
equipment. The device would also be considerably less difficult to operate and maintain.
The goal of this final elutriation stage would be to generate a “throw away” tailings stream
with maximum recovery of the valuable TiO bearing minerals present in the feed material.
2
1.5 Heavy Mineral Upgrading – Initial Test Work
Initial testing of the CrossFlow separator as an effective separation device for Florida
heavy minerals began by using lab-scale equipment (Dunn et al, 2000). The goal of these
initial tests was to evaluate the potential for effective heavy mineral upgrade by means of
hydraulic classification. This test work used a 5 x 15 cm laboratory scale CrossFlow unit to
simulate the process. The design of the unit was typical of the CrossFlow design principles,
an upper tangential feed presentation system and a lower dewatering zone. Figure 8 shows a
picture of the lab-scale CrossFlow used for testing. The disadvantage of this setup was the
use of a manual control valve to regulate the underflow rate and fluidized bed level. The unit
was initially fed at a target nominal feed rate of 1 tph/ft2. Efforts were made by the operators
to ensure the teeter water rate was minimized for the desired amount of quartz rejection
required.
18 |
Virginia Tech | CHAPTER 2
2.1 CrossFlow Testing
2.1.1 Overview
The first stage in scale-up elutriation test work began with continuous open circuit
testing on a small lab/pilot-scale device. The CrossFlow unit used during this period of
testing was installed for continuous field operation. It was installed directly in line with the
final wet mill concentrate. This particular location was placed directly after the attrition
scrub system and prior to dewatering and stacking. This positioning allowed introduction of
clean “scrubbed” concentrate into the test unit. Clean well water was used for the elutriation
network in order to additionally scrub and rinse the individual concentrate grains, thus
enhancing downstream dry processing.
2.1.2 Equipment Setup
The test unit was constructed by Eriez Magnetics and designed for continuous
operation for approximately 0.5 to 1.0 tph of feed material. To accomplish this task, the unit
was constructed with a cross-sectional area measuring 4” x 16”, allowing for feed
introduction of approximately 1.0 to 2.0 tph per ft2 of cross-sectional area. The unit was
constructed from stainless steel in order to prevent erosion and rust from any residual caustic
(NaOH) still present in the feed stream due to attrition scrubbing. An engineering diagram of
the unit constructed can be seen in Figure 11. To facilitate continuous operation, the unit was
outfitted with a pneumatic underflow valve controlled through use of a microprocessor based
controller system.
23 |
Virginia Tech | The unit was fed directly from the final wet mill concentrate piping system. This
particular point in the process circuit was located directly after the attrition scrub system and
prior to dewatering and stacking. To accomplish installation at this particular location, a 2”
bypass line was installed on the pipe feeding the stacker cyclone. This bypass feed contained
significant amounts of pressure and was far lower in percent solids than desired. Because of
these reasons, it was necessary to install an intermittent sump that could store material from
the bypass and allow for greater control of the unit feed. This setup allowed the user to
adjust feed rate with ease and also allowed the feed material to dewater to the necessary
process percent solids.
Elutriation water was delivered to the process from a fresh water well located onsite
by over 600 feet of high-pressure hose. Although the pressure and flowrate in this particular
line fluctuated quite drastically, a 500-gallon intermittent storage tank was installed which
could pump water directly to the elutriation network. This storage and pumping system
allowed for a relatively constant volume and pressure to be delivered to the unit elutriation
network. The water fed to the unit through this system was extremely clean, showing no
signs of added viscosity affects from humate and clay loading. All other process water
located onsite contained varying degrees of humate and clay, making use of well water most
desirable. The well water used typically had a pH of 5.5 and a viscosity of 1.0 cp for the
majority of test work. The acidic nature of the well water used for testing was expected from
prior experience with typical Florida ground water. Elutriation flowrate was controlled by
use of a Cole Parmer flow meter. This particular unit is a fixed orifice flow sensor, which
allows greater ease of operation compared to mechanical or ultrasonic sensors.
25 |
Virginia Tech | This particular lab-scale CrossFlow was identical in design to the previous unit used,
however, the underflow was controlled by a pneumatic actuator, which was regulated by a
pressure sensor. The use of this particular valve control system allowed for continuous
adjustment of the underflow rate in order to maintain a specified bed level and pressure
within the unit. The pressure sensor used in this process correlates the pressure at a
particular point within the tank to the underflow rate necessary to maintain this setting. This
measurement is an indirect reading of the teeter bed interface location/height within the unit.
Being able to control elutriation water rate and bed level/pressure independently is extremely
important to achieve proper testing conditions.
The unit was fitted with an industrial pressure transmitter capable of monitoring the
ranges of pressures expected during testing. As stated previously, these pressure readings are
an indirect measure of the actual teeter bed level within the unit. The transmitter selected for
this application was manufactured by Omega, model #PX726-300WCGI. This particular
transmitter was fitted with a weatherproof enclosure and all wetted parts were compatible
with corrosive media. The most important factor for selection of this model transmitter was
its rated 0.15% accuracy while used in difficult and hazardous conditions. Installation of this
instrument was made possible by use of the ½” npt adapter located on the test unit’s exterior.
More detailed information on this particular model can be seen in Table III.
Table III: Detailed Specifications on Pressure Transmitter
Output Accuracy Repeatability Stability Respons Time
4 - 20 mA 0.15% 0.05% 0.25% of URL 50 ms
26 |
Virginia Tech | Required adjustments to the process are based on the information available from the
transmitter output. To accomplish this task, a 2600 series microprocessor based controller
manufactured by Love Controls was selected. This particular controller was capable of both
manual program functions and fuzzy logic control. It allowed for direct input of process rate
functions, allowing for more user defined process control. The controller output was
connected directly to the pneumatic pinch valve, allowing for teeter bed level control.
Extensive work was required in order to calibrate the valve functions for use with the
controller output.
Feed material provided by the intermittent sump was sent directly to a 2” Mosley
Hydrocyclone. This device allowed significant increase in feed percent solids, which is
known to be beneficial to unit operation. The underflow from this hydrocyclone was fed
directly into the feed box on the CrossFlow unit. Overflow from this hydrocyclone was sent
to a tank to be discarded later. Although very small portions of the feed material did report
to this overflow stream, it was decided that this loss was inconsequential to the overall
performance of this equipment setup. Pilot-scale and full-scale installations would require a
properly sized hydrocyclone to ensure minimal loss of material during the dewatering
process.
2.1.3 Testing Program
Initial unit application began with preliminary testing aimed at identifying relative
high and low operational conditions. The conditions tested included feed rate, teeter water
rate, and bed level. Although other variables existed, they were not directly controllable with
27 |
Virginia Tech | the current testing arrangement. Given this specific testing arrangement, examples of such
uncontrolled variables are feed grade and feed percent solids.
The feed rate used during this test program varied from 1.0 to 1.5 tph/ft2. This
tonnage range appeared to offer the most quiescent flow in the separation tank. The target
percent solids for this tonnage, given the cyclone feed system used, was approximately 50%.
The elutriation flowrate necessary for separation given these conditions was 2.0 to 4.0
gpm/ft2. Flowrates below 2.0 gpm/ft2 caused the separation tank to clog and those above 4.0
gpm/ft2 led to significant unit overflow of heavy mineral.
A total of 72 individual tests were conducted. Sampling was done approximately
every four hours of unit operation. This practice allowed for the unit to experience a wide
variety of feed conditions and operational fluctuations. However, this timed procedure
generated occasional data points during both wet mill startup and shutdown. This led to
drastic changes, beyond expectations, in unit feed heavy mineral grade. The feed rate, teeter
water rate, and bed level previously used for normal process operations were incorrect for
these particular periods. Inability to control these variations in unit feed grade led test
operators to make small adjustments to teeter water rate and bed level. The majority of these
changes were done by visual observation of the heavy mineral present in the unit overflow.
Future test work will be focused around development of models capable of determining
optimum operational settings for use during periods of such drastic changes in concentrate
grade.
Samples were taken at three different locations around the CrossFlow separator. The
first sample taken was located at the CrossFlow overflow stream. A bypass valve was
installed in order to facilitate sample collection. The second sample, CrossFlow underflow,
28 |
Virginia Tech | was also taken at this time. The last sample point was unit feed located at the underflow of
the hydrocyclone. This process stream was simply diverted for sample collection. This
sample was taken last due to its drastic influence on the unit overflow. Collection of this
stream prior to the overflow would have resulted in erroneous readings of the overflow
stream if steady state had not resumed within the unit. The diamond symbols shown in
Figure 12 represent the sample points used during this phase of test work.
FIGURE 12: LAB SCALE CIRCUIT DESIGN AT THE MAXVILLE WET MILL STACKER
The samples taken from these locations were decanted and promptly sent for drying.
Temperature during the drying process was not an important factor since no degradation of
the material is experienced at high temperatures. Frequent stirring of the samples during the
initial part of the drying process was necessary to prevent “popping” of the material. The
samples were next weighed in order to obtain dry weight values, and then split into smaller
29 |
Virginia Tech | representative samples by use of a Jones Riffler. Approximately 100 grams of material was
needed for the heavy mineral analysis for low grade samples, 0-15% HM, and about 50
grams for higher grades, 15-100% HM. Another 50 grams of sample was required for TiO
2
analysis by x-ray fluorescence techniques. Retains were also split out from the original
sample and stored for various analysis which might become necessary at a later date.
Heavy Mineral analysis is accomplished by using a material of known specific
gravity to separate the various minerals. Current industry standards are to use LMT, Lithium
Metatungstate, as the separation media. This particular chemical can be heated in order to
alter the specific gravity to the required value. Typically a value of 2.7 is used for this
particular analysis. At this value, light material such as quartz and other gangue minerals are
removed from the heavy mineral assemblage. Two separate fractions are retained from this
process, a light mineral fraction and a heavy mineral fraction. These samples must be rinsed
with warm water in order to remove the excess LMT left on individual mineral grains. The
samples are then dried and weighed, in order to compare with each other, so that the percent
heavy mineral of the original sample may be determined. This analysis procedure is used
quite frequently for exploratory drilling programs as well as wet process analysis. This
procedure is not typically used for material once it enters the dry mill processing facilities.
During dry mill separation, other characteristics such as conductivity and magnetic
susceptibly are more useful in understanding the individual sample characteristics.
TiO analysis is accomplished by using x-ray fluorescence technology. Wavelength-
2
dispersive x-ray fluorescence spectrometry, “XRF”, is a non-destructive analytical technique
used to identify and determine the concentrations of the elements present in solids, powders
and liquids. XRF is capable of measuring all elements from beryllium (atomic number 4) to
30 |
Virginia Tech | uranium (atomic number 92) and beyond (at trace levels, often below one part per million,
and up to 100%). Figure 13 depicts the equipment setup and process used for this
measurement process.
FIGURE 13: DEPICTION OF THE XRF PROCESS
2.2 Mass Balancing
2.2.1 Introduction
The primary focus of the sampling and laboratory work done during testing is to
generate viable data in order to allow for interpretations and conclusions of the process under
investigation. In order to achieve this, the data must fit some predetermined level of quality
and accuracy. Test work designed to compare extremely small process variations must be
handled in a more controlled fashion than those of a large process simulation. To this end,
the level of accuracy needed to draw the desired conclusions from an experiment will dictate
the procedures used during sampling and laboratory processing.
The usefulness of sampling data as a process-modeling tool is drastically dependent
on the quality of the input data. The data used for modeling can come from a number of
31 |
Virginia Tech | different sources, such as human sampling, automated sampling, process outputs, and process
inputs. Data collected from these various sampling and testing procedures is quite frequently
susceptible to many different forms of errors. While some of these can be attributed solely to
human error, others are more innate within the sampling and laboratory processes
themselves. These errors have many different underlying causes, but the majority can be
characterized as follows:
• Statistical Effects
• Sampling Procedures
• Assay Procedures
• Sizing Procedures
• Plant Process Fluctuations
Due to the innate nature of these errors, various mass balancing programs/models are
used within the industry. The purpose of these mass balancing techniques is to generate
statistically viable data which remains as close as possible to the actual measured data. The
overall procedure for mass balancing is as follows:
• Data Collection
• Data Analysis
• Observation of Data Fit
• Statistically Refine Experimental Values
The basis for mass balancing algorithms used within industry are compositional
differences of the various streams generated by the process equipment. A typical mass
balancing routine takes all selected streams and calculates the smallest set of data
adjustments required, which in turn make the total data set consistent. Consistency of the
32 |
Virginia Tech | sample data is based on a simple principle, input must equal output. An example of this
principle would be that the TiO content entering the CrossFlow must equal the unit’s output
2
of TiO . Information such as this is fundamental throughout the entire mass balancing
2
routine.
Consider a process which contains three streams having assays of a, b, and c (where
a, b and c may be assays of size, TiO , or any other conserved property specific to the
2
process). A simple schematic of this situation is seen in Figure14. If we assume the flowrate
in stream of assay a is 100 tph, then:
100a = x*b + (100 - x)*c
where x is the flowrate in stream of assay b, then:
x = 100(a-c)/(b-c)
The most important aspect of this equation is that there must be some differences between
the assay values. If the process where simply a splitter device and the assays remained
constant:
a = b = c
and therefore x = 0/0 which is undefined.
Expressed in a different way, values can only be estimated for mass balancing if a
process imposes a difference on its product streams. If no difference is imposed, the
information obtained by sampling cannot be used to calculate erroneous or missing data. It
follows that the most useful properties used for mass balancing are those that generate the
largest difference in the process streams sampled. This means that heavy mineral assays
work well around a spiral circuit while size analysis works well around a hydrocyclone
33 |
Virginia Tech | circuit. The true ability of a mass balance program lies in its capacity to use a wide range of
assays and flowrates over an entire process flowsheet.
FIGURE 14: SIMPLE PROCESS DEFINITION FOR MASS BALANCING
2.2.2 Mass Balance Program Design
In an effort to better understand and draw statistically viable data from the samples
taken during testing, a mass balance spreadsheet was constructed. The spreadsheet designed
was written in Microsoft Excel format. It was designed to take user defined sample data and
convert it to statistically viable data.
The developed mass balance spreadsheet first takes user entered experimental data
into the Experimental Values section. This data is then duplicated into the Adjusted Values
section. From these values, the relative change and weighted sum of squares are calculated.
Several constraints have been entered into the spreadsheet. A total of four constraints
dealing with mass, HM tonnage, TiO tonnage, and ZrO tonnage are used in this program.
2 2
These constraints force the output of the system into equaling the input for both mass and
assay values. Next, a “solver” routine is used which alters the data in the Adjusted Values
section. The goal of this solver routine is to calculate the experimental values needed to meet
34 |
Virginia Tech | Complete mineral composition analysis using XRF analysis was run on each of the
individual test runs. Figure 15 shows the various data for these runs. It can be seen that
ZrO , zircon, generally maintained higher recoveries than TiO . This phenomenon was
2 2
expected and can be attributed to the higher specific gravity of the zircon grains, 4.7.
Rejection of the alumino-silicates, AL O , was typically less than 10%. Alumino-silicates
2 3
are significantly less dense than other heavy minerals, 3.2 – 3.7 specific gravity, and would
normally be expected to report to overflow in an elutriation system such as the one tested.
These particular heavy minerals, specifically kyanite and sillmanite, have no significant
value and must typically be removed during subsequent processing. Removal of this material
is another beneficial feature of the proposed elutriation system.
The average TiO recovery and ZrO recovery experienced during the test work were
2 2
97.68% and 98.05%, respectively. At these levels of recovery, TiO in the unit feed was
2
upgraded from an average 26.01% to an underflow grade of 31.18%. The average overflow
experienced during the testing program was only 2.95% TiO . Although these results are
2
typically better than those achieved by Dunn et al, the same degree of upgrading needs to be
achieved while generating an overflow stream less than 2% TiO in order for this project to
2
become industrially feasible.
36 |
Virginia Tech | also scattered in this testing range. It is believed that the +30 mesh quartz present in the feed
and underflow causes this scatter of the data. This coarse quartz fraction, known as “rice
rock”, is extremely sporadic in quantity and thus difficult to accurately evaluate. Future
testing will remove this material prior to laboratory analysis in order to achieve more
accurate quartz rejection information.
Using the SiO data obtained through XRF analysis, a less sporadic trend can be seen
2
in Figure 16 (B). The only mineral present in the heavy mineral concentrate which contains
specific levels of SiO is quartz. However, the SiO data and heavy mineral separation data
2 2
show limited correlation to one another. It is believed that the particular SiO analysis
2
program used during the XRF analysis incorporates silicon readings from other minerals such
as staurolite, kyanite, sillminite, and zircon. Therefore, the SiO data obtained from XRF for
2
this test program cannot be used as an accurate indication of quartz content. Although some
correlation of SiO to the content of quartz, staurolite, kyanite, sillminite, and zircon can be
2
achieved, it is neither recommended nor needed for this test program.
38 |
Virginia Tech | 100
90
80
70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Quartz Rejection (%)
39
tnecreP
100
90
80
70
TiO2 Recovery
60 OverFlow Mass Yield
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
SiO Rejection (%)
2
tnecreP TiO2 Recovery
OverFlow Mass Yield
(A) (B)
FIGURE 16: TIO 2 RECOVERY AND OVERFLOW YIELD VERSUS QUARTZ AND SIO 2 REJECTION
Further examination of TiO recovery shows a relatively strong trend when compared
2
to overflow mass yield. Although test work was carried out at several different operational
settings and feed grades, the data suggests that TiO recovery is directly related to the
2
quantity, or percentage, of material being forced into the unit overflow stream. In order to
accomplish the different levels of mass rejection, an operator can manipulate both elutriation
flowrate and bed level. Feed grade in terms of heavy mineral content, as seen in Figure 17,
also exhibits drastic influence over the percentage of material in the unit overflow. It is
plausible for high levels of quartz flowing upward in the test unit toward the overflow to trap |
Virginia Tech | or entrain large amounts of fine heavy mineral. The majority of these lost heavy minerals,
when introduced into an empty elutriation system under the same operational settings,
typically would report to the underflow stream. Feed grade, however, is relatively
uncontrollable using the testing setup developed for this stage of test work. In order to affect
feed grade, drastic changes to the multi-stage wet mill spiral circuit supplying this test setup
would be required.
Quartz rejection as a function of overflow mass yield gives insight into the size of
quartz present in the concentrate. The largest portion of quartz rejection occurs in the initial,
lower levels of mass rejection. This phenomenon would tend to describe the easy removal of
fine/medium quartz grains. However, as mass rejection increases, the rate of quartz rejection
decreases significantly. The medium/coarse quartz remaining in the concentrate becomes
increasingly difficult to separate above mass rejections of 20%, resulting in the associated
loss of fine heavy minerals such as leucoxene and rutile. It can be stated that the most
efficient rejection of quartz occurs were less than 20% mass rejection is desired. Operation
above these settings results in significant recovery penalties for both heavy mineral and TiO .
2
40 |
Virginia Tech | a function of these dramatic fluctuations would need to be reduced. A constant recovery
above 97% would be a successful achievement under these circumstances.
FIGURE 18: CROSSFLOW TIO
2
RECOVERY TIME SERIES PLOT
The average heavy mineral feed grade encountered during this stage of test work was
72.57%. This value was much lower than expected, when compared to the plant
performance goal of 85%. While some of the lower grades experienced may be due to plant
operational conditions, others may be associated with the location of the bypass line feeding
the CrossFlow unit. One possibility is that heavy minerals in slurry tend to settle towards the
bottom of the flow in a horizontal pipe, making collection of representative unit feed
difficult. Another possibility for this lower than expected feed grade might be due to the
presence of large amounts of coarse quartz. This coarse material is not considered part of the
heavy mineral concentrate assemblage and is typically dry screened before any dry mill
42 |
Virginia Tech | separation occurs. For this reason, future test work will remove this material from heavy
mineral determination.
Although the feed encountered during test work was lower than expected, the
CrossFlow unit was successful in upgrading this material to an average grade of 86.84%
heavy mineral. Figure 19 shows trend fit data for the CrossFlow feed and product streams
encountered during testing. The average heavy mineral concentrate upgrade produced during
the entire testing was 14.27%. The upgrade produced during periods of low feed grade was
much more pronounced than during periods of high feed grade. However, during these
periods of high upgrading, heavy mineral recovery decreased significantly. This can be
attributed to an increase in retention time of the heavy mineral matrix within the unit.
Operation under these particular conditions would require decreased unit bed pressure in
order to reduce the amount of heavy mineral present in the upper separation zone. These
lower grade periods operating at low pressure levels allow a thicker bed of light material to
develop within the separation zone. It is also feasible that during periods of high overflow
rates, fine heavy mineral particles tend to become trapped in the upward movement of light
material being rejected to the overflow stream. It may prove beneficial to unit performance
to lower teeter water rates and/or bed levels during these situations, reducing particles caught
in carryover and those lost due to increased retention time.
43 |
Virginia Tech | remained in a relatively narrow range. A normal probability plot of the experienced TiO
2
recoveries is shown in Figure 27. The standard deviation of 2.07% for this process was much
higher than expected. The probability plot for TiO recovery does not exhibit a typically
2
normal pattern. Few points fall within the 95% confidence intervals predicted by a normal
distribution function. There are too many points in the left tail (points above the limits) as
well as too many points in the right tail (points above the limits) compared to that of the
expected normal distribution function. The Anderson-Darling statistic for this situation is
3.304. This measurement is an interpretation of how far the plot points fall from the fitted
line for the probability plot. The statistic is a weighted squared distance from the plot points
to the fitted line with larger weights in the tails of the distribution. A smaller Anderson-
Darling statistic indicates that the distribution fits the data better. For some limited purposes,
assuming a normal distribution for this data would be acceptable. However, an attempt was
made to find a better result using six other common distributions including: lognormal base
10, lognormal base e, Weibull, extreme value, exponential, logistic, and loglogistic
distributions. Of these distributions, extreme value exhibits the best Anderson-Darling
statistic, 1.16. Thus for determining maximum likelihood estimates of the population
parameters and percentiles of the distribution for this data, an extreme value distribution
should be used.
52 |
Virginia Tech | 99
95
90
80
70
60
50
40
30
20
10
5
1
90 95 100 105
Data
53
tnecreP
Normal Probability Plot for TiO2 Rec.
ML Estimates - 95% CI
ML Estimates
Mean 97.6848
StDev 2.07446
Goodness of Fit
AD* 3.304
FIGURE 27: NORMAL PROBABILITY PLOT FOR TIO 2 RECOVERY
A normal probability plot of the experienced HM recoveries can be seen in Figure 28.
Standard deviation for this process is slightly higher than the TiO data, 2.72%. The
2
probability plot for HM recovery also does not exhibit a typical normal distribution pattern.
The majority of the points fall outside the 95% confidence intervals predicted by a normal
distribution. There are too many points in the left tail (points above the limits) as well as too
many points in the center (points below the limits) compared to what is expected in a normal
distribution function. Assuming a normal distribution for these data would not be acceptable,
given the AD value of 4.97. After examining all other typical distributions, it was found that
an extreme value distribution gives a slightly better AD value of 3.54. Neither the normal or
extreme value distributions appear to accurately predict HM recovery for this test work. |
Virginia Tech | CHAPTER 3
3.1 Model Development
3.1.1 Overview and Objectives
For this particular set of test work, two individual unit models will be developed. The
first model is based on numerous empirical equations that together represent the intricacies
involved in a CrossFlow elutriation device. This model will allow the reader general
understanding of unit operational functions, however, will not attempt to answer or predict
exact unit operational outputs. The second model developed will be based entirely from field
testing data obtained from a pilot-scale unit. This field data will be input into statistical
software, MiniTAB, and used to model the particular pilot-scale unit given the same
operational conditions. Data obtained from this model should be able to accurately predict
unit operation and allow the reader to properly scale-up a production size unit. The testing
program and developments will be discussed later.
3.2 Empirical Modeling
3.2.1 Model Introduction
The primary object of the proposed empirical model is to develop a mathematical
description of the processes involved within the CrossFlow unit. The goal of the model is to
represent the particle size and density within the cell as a function of height, z. This
information can then be used to estimate the particle concentrations anywhere within the cell
and also determine the concentrations within the overflow and underflow. The proposed
model will assume that the unit has been operating at steady state, therefore, the teeter-bed
has already been formed. In this case, the initial fluid flow conditions while the bed is
56 |
Virginia Tech | forming will not be required. The proposed model will encompass the entire unit above the
elutriation network. Two sections in the model will separate this zone, the area above the
teeter-bed and the area below the teeter-bed. A second model for the dewatering cone is not
necessary since no settling occurs, velocity is constant, and the feed material entering equals
the underflow material exiting.
The proposed setup of the model is quite similar to that of the actual CrossFlow unit,
as seen in Figure 30. Feed enters through the center of the cell and discharges at the teeter-
bed level. The feed material then flows in the positive and negative x directions as
separation occurs. Both water and air addition to the cell occurs at the elutriation network
level, z = 0. Material that rises, due to low density or attachment to bubbles, reports to the
overflow product stream. Material that does not rise, eventually settles into the teeter-bed
reports to the underflow via mass-action.
57 |
Virginia Tech | 3.2.2 Microscopic Population Balance Model of the CrossFlow Separator
Given a general microscopic population balance model as follows:
dy d ( ) d ( ) d ( ) (cid:229) J d ( ) • •
+ v y + v y + vy + vy + D- A = 0
d
(
it
)
dx
(
ii)x dy
(
iii)y dz
(
iv)z
j=1
dz
( vj )
j ( vi) ( vii)
where:
dy
= Accumulation
dt
d ( ) d ( ) d ( )
v y , v y , vy = Continuous changes due to particle motion
dx x dy y dz z
(cid:229) J d ( )
v y = Continuous changes in property space
dz j
j=1 j
•
D = Disappearance
•
A = Appearance
The assumptions which must be used in order to analyze the population balance are:
1. Two properties of interest; size, d , and density, r .
2. The bank is completely mixed in the y direction.
3. The bank is completely mixed in the x direction. This assumption, although
somewhat incorrect, simplifies the preliminary model development.
4. Particle size does not change continuously.
5. No disappearance of particles from one density class or size class to another exists
6. No appearance of particles from one density class or size class to another exists.
7. The cell is operating under steady state conditions.
59 |
Virginia Tech | • Q volume flow rate of water addition at z = 0
w :
• Q volume flow rate of underflow at z = 0, same as flow rate through dewatering cone
u :
d ( )
• f ; volume fraction solids as a function of height, z, within the cell; requires sampling
dz
at various heights throughout the cell at various flow and feed rates
d
• Re; Reynolds Number as a function of height, z, within the cell; determination of
dz
d ( )
f and flow and feed rates will allow calculation of this term
dz
3.2.3 Model Conclusions
The developed CrossFlow model contains many unknown quantities and functions.
While some are relatively simple to calculate, although time consuming, others are much
more involved. While the determination of the volume fraction solids and Reynolds numbers
as a function of height z within the cell are beyond the scope of this project, it is hoped that
an analytical solution to these terms can be achieved in the near future.
Use of this model should aid the reader in the general understanding of the CrossFlow
unit. However, exact predictions of unit recovery performance and operational settings are
outside the range of the particular model developed. Additionally, a unit model that
determines conditions without the assumption of complete mixing in the x direction. This
assumption makes computation much simpler, however, recent studies by Komuench et al.
have shown this assumption to be incorrect. This model enhancement would allow for more
realistic model calculations, but requires significantly more computations.
65 |
Virginia Tech | 3.3 Statistical Unit Modeling
3.3.1 Response Surface Modeling
In order to accurately predict unit performance and determine scale-up information, a
more elaborate model for unit operation was required. Statistical modeling of unit operation
would allow the precise predictions of unit operational conditions needed for engineering
analysis. Response surface modeling techniques were employed through use of statistical
software, primarily MiniTAB. These techniques were used throughout the testing and
analysis programs.
Response surface methods are used to examine the relationship between one or more
response variables and a set of quantitative experimental variables or factors. These methods
are often used after identification of specific controllable factors has taken place. They are
then used to find the particular settings that optimize the response (recovery, rejection, etc.).
Response surface designs are usually chosen when curvature, or nonlinear interaction, is
suspected within the response surface. Response surface methods are typically employed in
order to:
• find operating conditions that produce the desired response
• find operating conditions that satisfy operating or process specifications
• identify new operating conditions that produce demonstrated improvement of
the response
• develop a model of the relationships between defined operational conditions
and the response
66 |
Virginia Tech | Many response surface applications are sequential in nature in that they require more
than one stage of experimentation and analysis. The steps shown below are typical of a
response surface experiment.
• Choose a response surface design for the experiment.
• Create a response surface design, central composite or Box-Behnken.
• Modify the design by renaming the factors, changing the factor levels,
replicating the design, and randomizing the design.
• Perform the experiment and collect the response data.
• Enter the data into MiniTAB.
• Analyze the response surface design in order to fit a model to the
experimental data.
• Generate wireframe plots in order to visualize response surface patterns.
• Optimize the desired responses by using a response optimizer to obtain a
numerical and graphical analysis.
Depending on the experiment, steps can be performed in different orders, a given step can be
performed more than once, or a step can be entirely eliminated.
3.3.2 Response Surface Design Selection
Before the use of statistical software, determination of the most appropriate design for
the particular experiment is required. Choosing the correct design for a given experiment
ensures that the response surface is fit in the most efficient manner. MiniTAB’s software
package provides both central composites and Box-Behnken designs. When choosing a
design it is most important to:
67 |
Virginia Tech | • Identify the number of factors that are of interest.
• Determine the number of test runs physically possible, given time and sampling
requirements.
• Ensure adequate coverage of the region of interest on the response surface.
• Determine the impact that other considerations (such as cost, time, or the
availability of facilities) have on your choice of a design.
Depending on the particular experiment, there are other considerations that make a particular
design most desirable. It is important to use designs that show consistent performance in the
criteria considered important, such as the ability to:
• Increase the order of the design sequentially.
• Perform the experiment in orthogonal blocks. Orthogonally blocked designs allow
for model terms and block effects to be estimated independently and minimize the
variation in the estimated coefficients.
• Rotate the design. Rotatable designs provide the desirable property of constant
prediction variance at all points that are equidistant from the design center, thus
improving the quality of the prediction.
• Detect model lack of fit.
Given the requirements of the experiments, a Box-Behnken design was chosen. This
particular design can be created as a blocked or unblocked design. The illustration below,
Figure 31, shows the general layout of a three-factor Box-Behnken design. The points on the
diagram represent the experimental runs that are performed and used for response analysis.
Box-Behnken designs are often used when performing non-sequential experiments. That is,
planning to perform the experiment only once. These designs allow efficient estimation of
68 |
Virginia Tech | CHAPTER 4
4.1 Pilot-Scale CrossFlow Test Work
4.1.1 Overview
The next stage in elutriation testing as an effective heavy mineral upgrade tool began
with continuous open circuit testing on a larger pilot-scale CrossFlow. The unit used during
this period of testing was installed in the field and arranged for continuous field operation.
The unit constructed for this test work had a cross-sectional area of 2’ x 2’. This pilot unit
was installed directly in line with the final wet mill concentrate at a Florida mineral sands
operation. This particular unit was positioned in the exact feed location as the 4”x16” lab
unit previously tested, immediately following the attrition scrub system and prior to
dewatering and stacking. It was surmised that scale-up testing using a unit of this size and
capacity would generate much more realistic data for any future full-scale installation plans.
Although significant modifications were required to functionally install the pilot scale
unit, the basic flowsheet remained the same. Figure 32 shows the basic overview of the
flowsheet used for pilot-scale operation. Feed material was taken directly from the wet mill
final concentrate line and fed to the CrossFlow’s tangential feed box. Two 4” hydrocyclones
were used to dewater the feed slurry to the necessary 50-60% solids recommended by the
manufacturer. Two additional 4” hydrocyclones and a 7.5 HP pump with a variable speed
drive were required in order to transport and stack the underflow process stream. In addition
to these ancillary equipment needs, additional steel work was required in order to sustain the
weight of the larger unit.
72 |
Virginia Tech | the pilot-scale unit. The hydrocyclones would be fed directly from the bypass feed line
installed prior to the full-scale stacker cyclone. As previously stated, pressure and percent
solids of this bypass line were incorrect for unit feed conditions for laboratory scale testing.
However, installation of the larger feed hydrocyclone would allow for effective operation
under these high pressures and low percent solids present. The installed hydrocyclone had
been previously tested in-line to determine its operational ability. Minimal loss of fines in
the overflow stream was experienced with underflow percent solids reporting in the range of
40–70%.
The larger volume of elutriation water required for the pilot unit posed another
installation problem. Required flowrates were initially estimated at 8–16 gpm, 2–4 gpm/ft2.
The 110 volt chemical pump previously used during lab-scale testing would prove to be
inadequate. Instead, a ¾” by 1”, 220 volt pump was installed. This particular pump was
fixed speed and capable of supplying 35 gpm under the testing conditions.
Initial shakedown testing of the pilot-scale facilities included the use of a RCM
Industries fixed aperture flowmeter. However, the fluctuations in flow readings experienced
with this model flowmeter were much greater than desired and made precise flowrate
changes difficult. Instead, a Cole-Parmer 5000 series variable area flowmeter was installed.
This particular meter was a direct reading unit constructed from solid acrylic and was capable
of reading flowrates from 0.1 to 20.0 gpm. Product specifications for this model showed an
accuracy of 2% full scale, well within the required field. This type of meter makes use of a
single calibrated stainless steel float, whose size and weight determines the relative flowrate
present within the chamber. The inner chamber containing the float is bored at precise
specifications from acrylic to allow for accurate flowrate readings. This particular unit was
75 |
Virginia Tech | well suited for field use due to its high pressure capacity, 100 psi, as well as its high
temperature capacity, 150(cid:176) F.
Underflow stacking of the high tonnages was also a new issue for the pilot-scale
activities to address. Test stand facilities would need to be at least 20 ft in height in order to
allow sufficient height for stacking. Instead, unit underflow was directed to a sump pump
where it could be sent to stacking equipment. The underflow material, combined with added
makeup water, would then be pumped directly to two stacker hydrocyclones. The
hydrocyclones chosen for this task were 4” Lynatex models. The small apex diameter of this
model would allow for higher underflow percent solids rates. In order to facilitate stacking, a
percent solids greater than 65% would need to be created by the new hydrocyclones.
Overflow from these units would be sent back to the underflow sump, and thus re-circulate
the water. This system would also allow for re-circulation of any fine material that might
become misplaced in the overflow streams. Re-circulating this material would increase its
probability of reporting to the stacking underflow stream.
The sump pump chosen for transport of the unit underflow was a Warman 1.5” by 1”
centrifugal pump. The impeller chosen for installation under these particular conditions was
constructed of natural rubber in order to reduce capital cost. A 7.5 HP motor capable of
reaching speeds of 2000 rpms was also chosen. This equipment setup would allow for ample
capacity of the various underflow rates predicted. A variable speed drive unit was installed
such that pump rpms could be changed quickly. This arrangement would allow important
changes in pump discharge flowrate and pressure, vital to hydrocyclone performance under
varying conditions.
76 |
Virginia Tech | The previous microprocessor based fuzzy logic controller operated during lab-scale
testing was successfully adapted for use in the new system. However, some minor changes
to the program logic were necessary to incorporate this equipment into the new setup. The
most difficult aspect of this configuration was the setting of the response rate for the valve
operation. As compared to the lab-scale unit, pressure changes in the pilot-scale unit were
much slower and required careful programming of the process logic. If the valve response
rate were too fast, steady state conditions would never predominate within the unit and
successful separation would be quite difficult.
In addition to the tangential feed system used by the CrossFlow, a deflection plate
was added. This plate forced solids downward into the separation bed. This addition appears
to reduce the amount of fine heavy mineral lost to the unit overflow. These particles, having
been introduced farther down into the cell, must be forced through significant amounts of
material prior to overflow rejection. This development is believed to offer higher levels of
TiO recovery than the previous setup. Figure 34 shows the location of the deflection plate
2
within the cell.
77 |
Virginia Tech | 4.2 Pilot-Scale Model Test Work Results
Test work using the pilot-scale setup was carried out based on the testing
program previously described in Chapter 3. A total of 30 test runs were performed in order
to extract sufficient information necessary to generate a statistical model for unit operation.
Those aspects of the program which were controllable were changed between each test run,
those variables not controllable were tracked and recorded. As previously stated, the key
operational conditions under examination are feed rate, elutriation flow rate, and bed level.
Although not specifically controllable, feed percent solids and feed grade were also tracked
for model development. Table VIII shows the actual results experienced during this phase of
testing. Sample 22 was removed due to error or mistakes in the sample laboratory analysis.
Mass balancing of the test data was performed using the program developed in
Chapter 3. Unfortunately, there were excessive tonnage fluctuations experienced for both the
overflow and underflow unit streams. Several test runs required relative changes of 50% or
more to the overflow and underflow stream tonnage, much too high for the balancing
software to make accurate predictions. Reaction rate of the pneumatic underflow valve was
attributed as the primary source of these fluctuations. The reaction rate of the valve system
forced periodic “purges” in order to maintain bed level at the specified level. Instead of mass
balancing, the following common recovery and rejection equations were used for this stage
of test work:
Concentrate(Feed - Tailings)
Recovery =
Feed(Concentrate- Tailings)
Tailings(Concentrate- Feed)
Re jection =
Feed(Concentrate- Tailings)
80 |
Virginia Tech | performance. The larger unit capacity and feed arrangement for the pilot unit are believed to
have reduced these feed fluctuations quite drastically. Dramatic fluctuations in the feed
material would have made test work extremely difficult since the pilot unit had a retention
time of approximately 20 minutes, depending on feed rate.
Table IX: Average Pilot-Scale Model Testing Results
Feed Underflow Overflow
Test Runs Ave. TiO Ave. HM Ave. TiO Ave. HM Ave. TiO Ave. HM
2 2 2
30 26.64 76.47 32.63 91.28 2.55 5.87
Ave. Recovery Ave. Rejection
TiO HM Quartz
2
97.87 98.42 67.27
Quartz rejection was slightly lower than desired for the testing program. An average
rejection of 67.27% was experienced for the entire test program. This phenomenon is
explained by the high and low operational conditions used for testing. Certain test designs
either had extremely low elutriation flowrates or low bed pressure levels, neither setting
being optimal for quartz rejection. Model development from the test work should allow for
significant improvement in rejection at similar TiO and Heavy Mineral recoveries. Figures
2
36 and 37 depict the various recoveries as a function of both quartz rejection levels and feed
rate. Although limited reduction in TiO recovery occurs for higher feed rates, Heavy
2
Mineral recovery is left completely unaffected. Feed tonnages above 2 tph/ft2 are needed in
order to assess the impacts on recovery and quartz rejection.
84 |
Virginia Tech | Although the feed entering the system fluctuated quite drastically, unit underflow remained
relatively constant for the entire testing program. Figure 38 shows descriptive statistics for
the feed TiO grade. Typical wet mill performance during this phase of testing gave a
2
standard deviation for concentrate grade of 3.0%. Fluctuations in this plant performance are
attributed to such things as tonnage shifts in dredging, mechanical sisues and operational
errors. Figure 39 shows that the CrossFlow pilot unit was successful in reducing these
fluctuations by approximately 44%. Unit TiO output grade averaged 32.63% with standard
2
deviation of 1.68%. This drastic reduction in fluctuation would significantly improve
subsequent dry mill processing. Constant output grades would allow dry mill operators to
configure plant performance around relatively tight ranges of feed grade, improving overall
dry mill recovery and performance.
Descriptive Statistics
Variable: Feed TiO2
Anderson-Darling Normality Test
A-Squared: 0.923
P-Value: 0.016
Mean 26.6432
StDev 3.0041
Variance 9.02447
Skewness -6.3E-01
Kurtosis -6.7E-01
N 28
20 22 24 26 28 30
Minimum 19.7600
1st Quartile 23.9700
Median 27.4000
3rd Quartile 29.4450
95% Confidence Interval for Mu Maximum 30.2900
95% Confidence Interval for Mu
25.4784 27.8081
24.5 25.5 26.5 27.5 28.5 29.5 95% Confidence Interval for Sigma
2.3751 4.0890
95% Confidence Interval for Median
95% Confidence Interval for Median
24.7507 29.1824
FIGURE 38: BASIC STATISTICS FOR UNIT FEED (%TIO 2)
86 |
Virginia Tech | Data from the test program can be analyzed through several techniques. However,
for the investigation multiple linear regression is used. This technique allows for model
development using the following terms:
• linear terms
• linear terms and all squared terms
• linear terms and all two-way interactions
• linear terms, all squared terms, and all two-way interactions (the default)
• a subset of linear terms, squared terms, and two-way interactions
The model fit chosen will determine the nature of the effects, linear or curvilinear, that can be
detected from the experimental data.
4.3.2 Linear Regression Analysis for TiO and HM
2
Based from data in Chapter 2, unit feed grade is known to contribute the vast majority
of prediction for unit recovery. Simple linear and polynomial regression are used to
determine the extent of feed grade on recovery. This procedure performs regression with
linear and polynomial (second or third order) terms of a single predictor variable and plots a
regression line through the data. Polynomial regression is one method for modeling curvature
in the relationship between a response variable (Y) and a predictor variable (X) by extending
the simple linear regression model to include X2 and X3 as predictors. The estimation
methods used for this analysis is the least squares approach.
The quadratic model developed in Figure 40 appears to provide a good fit to the data
for TiO recovery. The R2 indicates that feed grade accounts for 68.9% of the variation in
2
TiO recovery. A visual inspection of the plot reveals that the data are randomly spread
2
88 |
Virginia Tech | interactions present between X variables. MiniTAB provides a matrix plot option that
quickly allows the interpretations of these interactions. Any interactions between terms can
be seen by the plots through visible trends, either linear, logarithmic, or polynomial. Figure
42 shows the matrix plot developed from % Solids, TPH, GPM, Bed Level, and Feed TiO .
2
In general, there appears no strong interactions involving GPM, Bed Level, and Feed TiO .
2
However, TPH and % Solids does show a relatively strong interaction. This phenomenon is
typical for a hydrocyclone feeding arrangement where an increase in tonnage corresponds to
an associated increase in discharge percent solids. Using both TPH and % Solids in
regression analysis is not advised given this occurrence. Further analysis is needed in order
to assess which variable plays a more important role in model development.
69.6325
% Solids
53.8775
6.8875
TPH
4.1025
15
GPM
13
95
Bed Level
93
27.9875
Feed TiO2
23.3825
5
3.8 7 7 5
6
9.6 3 2 5 4.1 0 2 5 6.8 8 7 5 1 3 1 5 9 3 9 5
2
3.3 8 2 25 7.9 8 7 5
FIGURE 42: MATRIX PLOT ANALYSIS OF X INTERACTIONS FOR TIO 2
91 |
Virginia Tech | In order to assess the use of either % Solids or TPH, regression analysis using the best
subset tool was used. The best subsets output from MiniTAB can be seen in Table X. Each
line of the output represents a different model comprised of different variables or predictors.
The predictors used in each model are indicated by use of an X in the table. From this data, it
can be clearly seen that Feed TiO accounts for as much as 62.4% of the predictive model.
2
GPM also plays an important role with a R-sq(adj) value of 25.0. Since % Solids and TPH
cannot both be used for model development, this table allows for understanding as to which
factor plays a larger role in model development. If the model where to be developed using %
Solids, GPM, Bed Level, and Feed TiO , the model fit would be 83.2% R-sq(adj). If TPH is
2
added to the model and % Solids dropped, the fit remains 83.9%. For further modeling
development, TPH will be used instead of % Solids. In the design of a full-scale unit, TPH
plays a much larger role by determining the exact unit cross-sectional area needed for the
intended feed tonnages. It is important to note that the TPH term used for analysis is not
entirely independent and will consist of a TPH portion and an associated % Solids portion.
92 |
Virginia Tech | Table X: Best Subsets Regression for TiO Recovery
2
93
%
Solids TPH GPM
Bed
Level
2
Feed
TiO
Vars R-Sq R-Sq(adj) C-p S
1 64.0 62.4 30.1 0.74843 X
1 28.1 25.0 81.0 1.05740 X
2 84.7 83.3 2.8 0.49960 X X
2 66.3 63.3 28.8 0.73996 X X
3 85.8 83.8 3.2 0.49213 X X X
3 85.5 83.5 3.5 0.49656 X X X
4 86.6 83.9 4.0 0.48956 X X X X
4 86.0 83.2 4.8 0.49985 X X X X
5 86.6 83.1 6.0 0.50212 X X X X X
Multiple regression analysis of TiO recovery began by using a quadratic model that
2
uses linear, square, and interaction terms. Each square or interaction term that displays a p-
value greater than 0.05 is removed from the model. Although the linear term Feed TiO
2
exhibits a p-value greater than 0.05, it was left in the model based on its influence in the
interaction term TPH x Feed TiO . It can be seen from Table XI that the R-sq(adj) value for
2
the regression model developed is 83.3%. Based on the coefficients and terms used, equation
[3] represents the best-fit model for the given pilot-scale test work. This equation will most
likely display curvature for some settings due to the inclusion of the square term TPH2 and
the interaction term TPH x Feed TiO .
2 |
Virginia Tech | Table XI: Estimated Regression Coefficients for TiO Recovery
2
Term Coef SE Coef T P
Constant 98.883 11.578 8.541 0.000
TPH -0.079 0.0684 -1.161 0.260
GPM -0.362 0.0679 -5.336 0.000
Bed Level -0.104 0.0796 -1.310 0.206
Feed TiO 0.783 1.0196 0.768 0.452
2
Feed TiO 2 -0.009 0.0195 -0.461 0.650
2
S = 0.4995 R-Sq = 86.8% R-Sq(adj) = 83.3%
TiO Recovery = 98.883 – 0.079(TPH) – 0.362(GPM) – 0.104(Bed Level)
2
+ 0.783(Feed TiO ) - 0.009(Feed TiO )2 [3]
2 2
Regression analysis for the unit Output TiO began through the same process. Full
2
quadratic terms were first used for the model, then select terms were removed which
possessed p-values less than 0.05. However, the use of several terms containing high p-
values was necessary in order to create a model that showed an acceptable relative fit.
Removal of these terms forced the overall fit, R-sq(adj), below the 84.2% value and also
generated higher residual values. Leaving these terms in allows for a better fit of the model
to the actual test data, however, will not accurately predict a model for a larger population of
test work. This predictive equation displayed in Table XII can be used to predict the Output
TiO for the specific testing data observed, however, will not accurately predict a continued
2
testing population. Equation [4] represents the predicted Output TiO based on TPH, GPM,
2
Bed Level, and Feed TiO . Running a 3rd replicate of the testing series would allow the
2
model to base its fit off more substantial information and improve the overall population
prediction.
94 |
Virginia Tech | these outputs given various inputs. Displayed are two overlaid contour plots on a single
graph. The two factors, Feed TiO and GPM, are used as the two axes in the plots and the
2
third and fourth factors, TPH and Bed Level, have been held at levels 6.0 and 94
respectively. The white area inside each plot shows the range of Feed TiO and GPM where
2
the criteria for both response variables are satisfied. Responses used for this specific graph
are a TiO recovery above 98% and an Output TiO above 33%. This plot, in combination
2 2
with the predictive models, can be used to find the best operating conditions for maximizing
recovery and output grade. To achieve these process outputs, feed grade entering the unit
can drop to as low as 27%. The ideal GPM for this situation is 14 GPM, as displayed by the
graph. This elutriation flowrate allows for the upgrading of the lower feed TiO grades while
2
outputting the proper responses previously described.
97 |
Virginia Tech | 4.3.4 Multiple Regression Analysis for HM
The first step in multiple regression analysis for heavy mineral recovery and output
grade is to determine the various interactions present between X variables. The same process
as for TiO is used, generation of a matrix plot through use of MiniTAB. This graph allows
2
quick interpretations of these interactions for the variables selected. Figure 45 shows the
matrix plot developed from % Solids, TPH, GPM, Bed Level, and Feed HM. In general,
there appears no strong interactions involving GPM, Bed Level, and Feed HM. However,
TPH and % Solids again show a relatively strong interaction. Thus, using both TPH and %
Solids in the regression analysis for heavy mineral is not advised given this occurrence.
Further analysis is needed in order to assess which variable plays a more important role in
model development, TPH or % Solids.
69.6325
% Solids
53.8775
6.8875
TPH
4.1025
15
GPM
13
95
Bed Level
93
80.575
Feed HM
65.965
99.09
HM Rec
97.35
53.8 7 75 6 9.63 2 5 4.1 0 25 6.8 87 5 1 3 1 5 9 3 9 5 6 5.9 65 8 0.57 5 9 7.3 5 9 9.0 9
FIGURE 45: MATRIX PLOT ANALYSIS OF X INTERACTIONS FOR HM
99 |
Virginia Tech | In order to assess the use of either % Solids or TPH, regression analysis using the best
subset tool was used. The best subsets output from MiniTAB can be seen in Table XIII.
Each line of the output represents a different model comprised of different variables or
predictors. The predictors used in each model are indicated by use of an X in the table.
From this data, it can be clearly seen that Feed HM accounts for as much as 68.3% of the
predictive model. GPM also plays an important role with a R-sq(adj) value of 15.6. Since %
Solids and TPH cannot both be used for model development, this table allows for
understanding as to which factor plays a larger role in model development. If the model
where to be developed using TPH, GPM, Bed Level, and Feed HM, the model fit would be
87.3% R-sq(adj). If % Solids is added to the model and TPH dropped, the fit drops slightly
to 86.6%. For further modeling development using heavy mineral analysis, TPH will be used
instead of % Solids. However, initial observation using the best subsets regression
information shows that GPM, Bed Level, and Feed HM account for a R-Sq(adj) value of
87.2. Based from this information, TPH does not appear to make a significant influence on
the model, although further analysis is needed.
100 |
Virginia Tech | Table XIII: Best Subsets Regression for HM Recovery
101
%
Solids TPH GPM
Bed
Level
Feed
HM
Vars R-Sq R-Sq(adj) C-p S
1 69.8 68.3 29.4 0.62575 X
1 19.6 15.6 108.3 1.02090 X
2 88.3 87.1 2.4 0.39982 X X
2 70.6 67.5 30.2 0.63376 X X
3 89.1 87.2 3.2 0.39720 X X X
3 88.9 87.0 3.5 0.40029 X X X
4 89.7 87.3 4.1 0.39574 X X X X
4 89.2 86.6 5.0 0.40684 X X X X
5 89.8 86.6 6.0 0.40630 X X X X X
Multiple regression analysis of HM recovery began by using the same quadratic
model as for TiO . This model uses linear, square, and interaction terms to develop a
2
predictive model of unit operation. Each square or interaction term that displays a p-value
greater than 0.05 is removed from the model. Through this process, the TPH linear, square,
and interaction terms were determined to play no significant role in the overall model. Also,
during this analysis, test runs #7, 18, 23, and 28 were removed for their excessive residuals
based from the developed model. The error in these points is either the development of
laboratory analysis error or field test work error. Removal of these data points both increase
overall model fit and reduces the average residual value for the test work. It can be seen
from Table XIV that the R-sq(adj) value for the regression model developed is 95.1%. Based
on the coefficients and terms used, equation [5] represents the best-fit model for the given
pilot-scale test work. This equation will most likely display curvature for some settings due |
Virginia Tech | to the inclusion of the square term Feed HM and the interaction terms GPM x Feed HM and
Bed Level x Feed HM.
Table XIV: Estimated Regression Coefficients for HM Recovery
Term Coef SE Coef T P
Constant 249.717 53.4195 4.675 0.000
GPM -2.783 0.436 -6.384 0.000
Bed Level -1.464 0.5906 -2.479 0.024
Feed HM -1.436 0.6039 -2.377 0.029
Feed HM2 -0.003 0.0012 -2.228 0.040
GPM x Feed HM 0.031 0.0055 5.634 0.000
Bed Level x Feed HM 0.016 0.0073 2.247 0.038
S = 0.2367 R-Sq = 96.4% R-Sq(adj) = 95.1%
HM Recovery = 249.717 – 2.783(GPM) – 1.464(Bed Level) – 1.436(Feed HM)
– 0.003(Feed HM)2 + 0.031(GPM x Feed HM) + 0.016(Bed Level x Feed HM) [5]
Regression analysis for the unit Output HM began through the same process. Full
quadratic terms were first used for the model, then select terms were removed which
possessed p-values less than 0.05. All terms remaining contain extremely low p-values
which represents a relatively high influence from each factor on the developed model.
Besides linear terms, the square term Feed HM and the interaction term TPH x Feed HM
were left in the model analysis. Unlike the HM recovery regression analysis, TPH and its
square and interaction terms play significant roles in unit output HM grade. In addition to
those test runs removed in the HM recovery analysis, test runs #9 and 15 were also removed
for their high residual values relative to the developed model. Table XV shows the
regression analysis information given the specific test runs and terms used in its
102 |
Virginia Tech | development. An R-sq(adj) value of 92.1% was achievable under the given circumstances.
Equation [6] represents the predicted Output HM based on TPH, GPM, Bed Level, and Feed
HM. Running a 3rd replicate of the testing series would allow the model to base its fit off
more substantial information and improve the overall population prediction.
Table XV: Estimated Regression Coefficients for Output HM Grade
Term Coef SE Coef T P
Constant -18.000 33.96 -0.530 0.604
TPH 24.060 5.1928 4.634 0.000
GPM 2.150 0.1922 11.165 0.000
Bed Level -0.580 0.1993 -2.922 0.011
Feed HM 1.360 0.3003 4.528 0.000
TPH2 -0.980 0.1562 -6.296 0.000
TPH x Feed HM -0.170 0.0482 -3.525 0.003
S = 1.283 R-Sq = 94.4% R-Sq(adj) = 92.1%
HM Output = -18.000 + 24.060(TPH) + 2.150(GPM) – 0.580(Bed Level)
+ 1.360(Feed HM) – 0.980(TPH)2 – 0.170(TPH x Feed HM) [6]
In order to understand the HM recovery equation developed, graphical analysis is
needed. To examine more than two interactions a 3-D surface plot is used. Figure 46 shows
one such contour plot where Feed HM and GPM are used as the X and Y variables, while
HM recovery is used as the Z value. For this particular instance, Bed Level is held at 94 and
TPH is not used. It can be seen that GPM displays a relatively linear relationship with HM
recovery. If significantly lower and higher water rates are used, this model would most
likely display curvature for HM recovery as a function of GPM. However, the flowrates
used in the model test work are relatively close which makes linear regression the only
103 |
Virginia Tech | Figure 47 is one such contour plot for these outputs given various inputs. Displayed are two
overlaid contour plots on a single graph. The two factors, Feed HM and GPM, are used as
the two axes in the plots and the third and fourth factors, TPH and Bed Level, has been held
at levels 6.0 and 94 respectively. The white area inside each plot shows the range of Feed
HM and GPM where the criteria for both response variables are satisfied. Responses used
for this specific graph are a HM recovery above 98% and an Output HM above 90%. This
plot, in combination with the predictive models, can be used to find the best operating
conditions for maximizing recovery and HM output grade. To achieve these process outputs,
feed grade entering the unit cannot drop below approximately 72% HM. The ideal GPM for
this situation is 14 GPM, as displayed by the graph. This elutriation flowrate allows for the
upgrading of the lower Feed HM grades while outputting the proper responses previously
described.
105 |
Virginia Tech | 4.5 Economic Analysis
As previously stated, installation of a CrossFlow unit has several economic
advantages. The largest financial incentive is the reduction of haulage requirements to the
dry mill facilities. This material also carries an additional tailings haulage necessary for
zircon retrieval. The cost of haulage to the dry mill facility and zircon wet mill facilities are
estimated at $1.75 and $2.50 per ton of concentrate, respectfully. Given a yearly plant
production of 355,000 tons of concentrate, a reduction in haulage of 10% would result in
annual variable cost savings of $150,875.
This system also saves the cost of drying unnecessary material in the dry mill
facilities. These costs are estimated at $1.50 per ton of feed material. A 10% reduction of
material needing drying would thus result in annual variable cost savings of $53,250. This
calculation does not take into account any maintenance costs associated with drying wet mill
concentrate, only the cost of fuel consumption.
Installation costs of the proposed system is estimated at $200,000. This cost includes
construction of the CrossFlow unit, necessary support structures, electrical services, water
services, and engineering fees. Assuming no added variable cost for the proposed system,
capital cost of $200,000 and yearly savings of $204,125 the Internal Rate of Return (IRR) of
this project is estimated at 75% with a Net Present Value (NPV) of $372,897. Total
economic calculations, including taxes, depreciation, and depletion, can be seen in Table
XVI. This economic exercise used a tax rate of 38% with the typical graduated depreciation
rates over 5 years.
109 |
Virginia Tech | CHAPTER 5
Conclusions
Lab/pilot-scale testing was accomplished using a 4” x 16” CrossFlow unit in continuous
field operation. The primary goal of this test work was to evaluate the feasibility of using
elutriation technology for heavy mineral concentration. These tests were carried out for
numerous feed grades and tonnages. Various operational parameters were also examined
during these tests. As previously stated, the feed material used was from a mineral sands
operation in northern Florida. The results achieved from this testing program are as follows:
• Quartz rejections greater than 80% were possible under certain operational
conditions. TiO recoveries at these rejection levels were above 95%.
2
• Elutriation flowrates were reduced to as low as 2 gpm/ft2 under certain conditions.
This reduction allows for significant savings of fresh water required for unit
operation.
• Use of a deflection plate showed significant recovery/rejection improvement over
the typical CrossFlow feed arrangement.
• Feed tonnages as high as 1.5 tph were successful in achieving the testing goals.
This allows for reduced unit size and cost.
Pilot scale testing was accomplished using a 2’ x 2’ CrossFlow unit in continuous field
operation. These tests were carried out for numerous feed grades, tonnages, elutriation flow
rates and bed levels. The results achieved from this testing program are as follows:
111 |
Virginia Tech | • Quartz rejections greater than 80% were possible under many different
operational conditions. TiO recoveries at these rejection levels were
2
approximately 98%.
• Elutriation flow rate was optimized to 14 gpm, 3.5 gpm/ft2, for several operational
conditions.
• Use of a deflection plate showed significant recovery/rejection improvement over
the typical CrossFlow feed arrangement.
• Feed tonnages as high as 1.5 tph were successful in achieving the testing goals.
This allows for reduced unit size and cost.
• Standard deviation for the wet mill concentrate TiO grade was reduced from
2
3.0% to 1.68%
Along with experimental testing, both empirical and statistical models for unit operation
were developed. Given a feed grade of 28% TiO and a feed rate of 70 tph for full-scale
2
operation, the statistical model was successful in determining the following:
• A minimum elutriation flow rate of 245 gpm (3.5 gpm/ft2) would allow for
successful quartz rejection.
• A feed rate of 1.5 tph/ft2 could be used to achieve successful unit results. This
tonnage requires the use of a 47 ft2 CrossFlow unit. A typical 7’ x 7’ square
configuration would achieve the proper cross-sectional feed rate.
• An overall quartz rejection of 80% with a TiO recovery of 98% is possible for
2
these particular unit operational conditions.
112 |
Virginia Tech | CHAPTER 6
Recommendations for Future Work
Improve Performance of Hindered-Bed Classifiers for Heavy Mineral Separations.
1. Laboratory testing focusing on optimal settings for the added baffle plate are needed
in order to better understand this item’s impact on unit performance. It is
recommended that this baffle plate be set at various heights and distances from the
feed presentation system to accomplish this task.
2. Full-scale installation of a CrossFlow unit in order to achieve the benefits previously
mentioned within this report. Elaborate testing would be needed for this application
to achieve the highest possible success. This testing should be used in order to
generate model information that could be used by operators for precise field
operation.
Improved Rejection of Fine TiO from Zircon Wet Mill Concentrate.
2
1. Conducting testing to investigate the possible benefits of using elutriation technology
in order to clean zircon wet mill concentrate. This particular concentration is
comprised mainly of zircon, ilmenite, leucoxene, rutile, kyanite, sillmanite, and
staurolite. This separation would be quite difficult given the near density of minerals
such as ilmenite, 3.7, to the zircon product, 4.7.
113 |
Virginia Tech | 10/4/2015 VT Fair Use Analysis Results
Draft 09/01/2009
(Questions? Concerns? Contact Gail McMillan, Director of the Digital Library and Archives at Virginia Tech's
University Libraries: [email protected])
(Please ensure that Javascript is enabled on your browser before using this tool.)
Virginia Tech ETD Fair Use Analysis Results
This is not a replacement for professional legal advice but an effort to assist you in making
a sound decision.
Name: Scott Koermer
Description of item under review for fair use: T.P Meloy, D.A Whaley, M.C Williams, Flotation tree analysis —
reexamined, International Journal of Mineral Processing, Volume 55, Issue 1, October 1998, Pages 21-39, ISSN 0301-
7516, http://dx.doi.org/10.1016/S0301-7516(98)00023-4.
Report generated on: 10-04-2015 at : 14:37:31
Based on the information you provided:
Factor 1
Your consideration of the purpose and character of your use of the copyright
work weighs: in favor of fair use
Factor 2
Your consideration of the nature of the copyrighted work you used weighs: in
favor of fair use
Factor 3
Your consideration of the amount and substantiality of your use of the
copyrighted work weighs: in favor of fair use
Factor 4
Your consideration of the effect or potential effect on the market after your use of
the copyrighted work weighs: in favor of fair use
in
Based on the information you provided, your use of the copyrighted work weighs:
favor of fair use
http://etd.vt.edu/fairuse/analyzer/results.php 1/2 |
Virginia Tech | 10/4/2015 VT Fair Use Analysis Results
Draft 09/01/2009
(Questions? Concerns? Contact Gail McMillan, Director of the Digital Library and Archives at Virginia Tech's
University Libraries: [email protected])
(Please ensure that Javascript is enabled on your browser before using this tool.)
Virginia Tech ETD Fair Use Analysis Results
This is not a replacement for professional legal advice but an effort to assist you in making
a sound decision.
Name: Scott Koermer
Description of item under review for fair use: Zhang, S., Forssberg, B.A., Moss, W., 1999. “Separation Mechanisms
and Criteria of a Rotating Eddy-Current Separator Operation” Resources, Conservation and Recycling, Vol. 25, No. 3-
4, pp. 215-232.0301-7516, http://dx.doi.org/10.1016/S0301-7516(98)00023-4.
Report generated on: 10-04-2015 at : 14:38:30
Based on the information you provided:
Factor 1
Your consideration of the purpose and character of your use of the copyright
work weighs: in favor of fair use
Factor 2
Your consideration of the nature of the copyrighted work you used weighs: in
favor of fair use
Factor 3
Your consideration of the amount and substantiality of your use of the
copyrighted work weighs: in favor of fair use
Factor 4
Your consideration of the effect or potential effect on the market after your use of
the copyrighted work weighs: in favor of fair use
in
Based on the information you provided, your use of the copyrighted work weighs:
favor of fair use
http://etd.vt.edu/fairuse/analyzer/results.php 1/2 |
Virginia Tech | 10/4/2015 VT Fair Use Analysis Results
Draft 09/01/2009
(Questions? Concerns? Contact Gail McMillan, Director of the Digital Library and Archives at Virginia Tech's
University Libraries: [email protected])
(Please ensure that Javascript is enabled on your browser before using this tool.)
Virginia Tech ETD Fair Use Analysis Results
This is not a replacement for professional legal advice but an effort to assist you in making
a sound decision.
Name: Scott Koermer
Description of item under review for fair use: Wilson, R.J., T.J. Veasey, D.M. Squires, The application of mineral
processing techniques for the recovery of metal from post-consumer wastes, Minerals Engineering, Volume 7, Issue 8,
August 1994, Pages 975-984, ISSN 0892-6875, http://dx.doi.org/10.1016/0892-6875(94)90027-2.
Report generated on: 10-04-2015 at : 14:28:39
Based on the information you provided:
Factor 1
Your consideration of the purpose and character of your use of the copyright
work weighs: in favor of fair use
Factor 2
Your consideration of the nature of the copyrighted work you used weighs: in
favor of fair use
Factor 3
Your consideration of the amount and substantiality of your use of the
copyrighted work weighs: in favor of fair use
Factor 4
Your consideration of the effect or potential effect on the market after your use of
the copyrighted work weighs: in favor of fair use
in
Based on the information you provided, your use of the copyrighted work weighs:
favor of fair use
http://etd.vt.edu/fairuse/analyzer/results.php 1/2 |
Virginia Tech | PUBLISHERS OF:
Recycling Today magazine
Recycling Today Global Edition magazine
Construction & Demolition Recycling magazine
Plastics Recycling magazine
Renewable Energy from Waste magazine
Equipment & Services Buyers’ Guides
Recycling Markets Directories
Mr. Scott Koermer
Virginia Polytechnic Institute and State University
Blacksburg, VA
Scott,
Thank you again for your contributed article “Gauging yield and recovery,” which was published in the
September/October issue of Recycling Today Global Edition magazine.
Please be assured that you have our company’s permission to use your drafted article, or the article in any
other form, as part of a submitted thesis.
Should you need to access the published version of the article, in can be found here:
http://www.recyclingtodayglobal.com/article/rtge0915-shredded-scrap-separation-performance
and starting on page 34 of the digital editon of the magazine, which can be found here:
http://www.recyclingtodayglobal.com/fileuploads/digital-editions/rtge/digital/20150910/index.html .
Please let me know if you need any additional proof of permission from Recycling Today Media Group or
its parent company GIE Media, and best of luck in organizing a successful thesis!
Respectfully Yours,
Brian Taylor,
Editor,
Recycling Today Media Group
5811 Canal Road / Valley View, Ohio 44125
Phone: 800-456-0707 / Fax: 216-525-0515
www.RecyclingToday.com
www.CDRecycler.com / www.SDBmagazine.com
PROVIDING INDUSTRY RESOURCES SINCE 1963. |
Virginia Tech | IDENTIFICATION OF IMPROVED STRATIGIES
FOR PROCESSING FINE COAL
by
Zulfiqar Ali
ABSTRACT
In modern coal preparation plants, solid-solid and solid-liquid separation processes used
to treat fine coal are least efficient and most costly operations. For example, field studies indicate
that the froth flotation process, which is normally used to treat minus (-0.2 mm) fine coal, often
recovers less than 65 to 70% of the organic matter in this size range. Fine coal separation
processes are also inherently less effective in removing pyrite than that of coarse coal
separations. Moreover, while fines may represent 10% or less of the total run-of-mine feed, this
size fraction often contains one-third or more of the total moisture in the delivered product. In
order to address these issues, several multistage coal processing circuits were set up and
experimentally tested to demonstrate the potential improvements in fine coal upgrading that may
be realistically achievable using an “optimized” fine coal processing flowsheet. On the basis of
results obtained from this research, engineering criteria was also developed that may be used to
identify optimum circuit configurations for the processing different fine coal streams.
In the current study, several fine coal cleaning alternatives were evaluated in laboratory,
bench-scale and pilot-scale test programs. Fine coal processes compared in the first phase of this
work included spirals, water-only cyclones, teeter-bed separators and froth flotation. The
performance of each technology was compared based on separation efficiencies derived from
combustible rejection versus ash rejection plots. The resulting data was used to identify size
ranges most appropriate for the various alternative processes. As a follow-up to this effort, a |
Virginia Tech | second phase of pilot-scale and in-plant testing was conducted to identify new types of spiral
circuit configurations that improve fine coal separations. The experimental data from this effort
indicates that a four-stage spiral with second- and fourth-stage middlings recycle offered the best
option for improved separation efficiency, clean coal yield and combustible recovery. The newly
developed spiral circuitry was capable of increasing cumulative clean coal yield by 1.9 % at the
same clean coal ash as compared to that of achieved using existing conventional compound
spiral technology. Moreover, the experimental results also proved that slurry repluping after two
turns is not effective in improving separation performance of spiral circuits.
The third phase of work conducted in this study focused on the development of methods
for improving the partitioning of pyrite within fine coal circuits. The investigation, which
included both laboratory and pilot-scale test programs, indicated that density-based separations
are generally effective in reducing sulfur due to the large density difference between pyrite and
coal. On the other hand, the data also showed that sulfur rejections obtained in froth flotation are
often poor due to the natural floatability of pyrite. Unfortunately, engineering analyses showed
that pyrite removal from the flotation feed using density separators would be impractical due to
the large volumetric flow of slurry that would need to be treated. On the other hand, further
analyses indicated that the preferential partitioning of pyrite to the underflow streams of
classifying cyclones and fine wire sieves could be exploited to concentrate pyrite into low-
volume secondary streams that could be treated in a cost effective manner to remove pyrite prior
to flotation. Therefore, on the basis of results obtained from this experimental study, a combined
flotation-spiral circuitry was developed for enhanced ash and sulfur rejections from fine coal
circuits.
iii |
Virginia Tech | ACKNOWLEDGEMENTS
It is my pleasure to acknowledge everyone who contributed directly or indirectly to
complete this research work. First of all, my special thanks to my research advisor, Dr. Jerry
Luttrell, who was always there for guidance, suggestions, help, support and encouragement
during my research work at Virginia Tech. Jerry, it was an honor for me to work with you and
there is no doubt in saying that this work would not have been possible without you. I am also
truly indebted and thankful to my respected committee members, Dr. Greg Adel, Dr. Emily
Sarver and Dr. Jaisen Kohmuench for their valuable comments, suggestions and advices.
During my experimental work, Bob Bratton was there to help whenever I was stuck.
Thank you Bob, for your invaluable help and suggestions to improve my research work. My
thanks and appreciation was due to Jim Waddell and John Matherly for their invaluable help
during experimental set up. Special thanks are due to Kathryn Dew, Carol Trutt and to Gwen
Davis for their administrative assistance throughout my graduate studies.
I would like to thank Alpha Natural Resources, Arch Coal, Inc. and their employees for
assisting in my in-plant experimental testing programs. I am also obliged to the management and
employees of Cardinal, Knight Hawk, Prairie Eagle and Creek palm preparation plants for
providing much needed fine coal samples for my experiments. Appreciation is also extended to
Eriez Manufacturing for their assistance during teeter-bed and HydroFloat™ testing.
The financial support provided by the University of Engineering and Technology, Lahore,
Pakistan and Taggarat Global LLC and Nano Drying Technologies LLC is greatly
acknowledged.
I would also like to thank many friends for their nice company throughout my stay in
Blacksburg namely Jawad Raza, Imran Akhtar, Shamim Javid, Arshad Mehmood, Karim Akhtar,
vi |
Virginia Tech | CHAPTER 1 - INTRODUCTION
1.1 Motivation
Coal is one of the most abundantly available energy sources in the world and is found
almost in every country. The top five largest proven reserves of coal are in the United States,
Russia, China, Australia and India, respectively. Table 1.1 shows the distribution of proven coal
reserves. Currently, there are over 860 billion tonnes of proven coal reserves. In 2010, world coal
consumption grew by 7.6% and now coal accounts for 29.6% of world total energy, 40% of
world electricity and 66% of world steel production (BP, 2011).
The United States leads the world with a little over 237 billion short tons of recoverable
coal reserves, i.e., approximately 27.6% of the total world coal reserves. It is estimated that at the
current production rate of 984.6 million tonnes per year (USEIA, 2010), the U.S. reserves would
last for 240 years. Coal is the largest single fuel used for electricity generation in the U.S. and
accounts for 42% of electric power generation. Since 2000, about 90% of all the coal consumed
in U.S. has been used for electric power generation as shown by Figure 1.1.
Run-of-mine coal often contains inorganic impurities in the form of mineral matter. The
mineral matter consists of noncombustible materials such as shale, slate and clay. These
unwanted contaminants reduce the coal heating value, leave behind an undesirable ash residue,
and increase the transporting cost of coal. These impurities can also alter the suitability of coal
for the manufacture of metallurgical coke or generation of petrochemicals and synthetic fuels.
1 |
Virginia Tech | 100
90
80
70
60
50
40
30
20
10
0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Excess surface moisture also reduces the heating value of coal, leads to severe handling
and freezing problems and increases the overall transportation costs of coal to consumer sites
(Kaytaz et al., 1994). Therefore, strict limitations on the heating value, ash and moisture contents
of purchased coal are imposed in coal purchasing agreements.
Coal processing is an important step to satisfy the run-of-mine coal quality specifications
as per end-user demand. Coal processing also removes impurities such as sulfur and trace
elements like mercury, thus cleaned coal is more environmentally friendly. In short, coal
processing increases the heating value, lowers the transport cost, reduces the particulate
emissions and improves the marketability of the run-of-mine coal. There are 286 coal processing
plants in the United States, which cleaned approximately 67% (660 million short tons per year)
of the total coal consumed in the United States (Coal Age, 2010).
)%(
noitpmusnoC
laoC
Year
Electric Power Coke Plants Other Industrial Residental & Commercial
Figure 1.1 Coal consumption by sectors from 2000 to 2010 (US EIA, 2011).
2 |
Virginia Tech | Although not shown in Figure 1.2, coal processing usually starts with crushing, i.e., run-of-mine
coal lumps that are too large to pass through the processing plant are crushed down to an
appropriate size. The crushed coal is then separated into appropriate size fractions as coarse,
small, fine and ultrafine. Screens are employed for sizing coarser particles, while combinations
of sieves and classifying cyclones are used for sizing finer particles. Figure 1.3 shows coal
particles sizing equipment for various size ranges.
The objective of coal processing is to separate impurities from valuable carbonaceous
material. In order to remove these impurities, modern coal processing plants incorporate a
number of solid-solid separation methods such as dense medium vessels, water-only cyclones,
teeter bed separators, coal spirals and froth flotation. Figure 1.3 shows the effectiveness of
different types of conventional coal cleaning separators relative to the coal particles sizes.
The final step of coal preparation is solid-liquid separation, or dewatering, which
removes unwanted surface moisture and produces a relatively dry concentrate. Dewatering
methods are broadly classified in to three main groups: sedimentation, filtration and thermal
drying (Wills & Napier-Munn, 2006). Primarily screens are used to remove excess moisture
from coarser (+5 mm) coal particles. Finer particles, which used to have higher moisture contents
than that of coarser ones due to their greater surface area, are dewatered using centrifugal
methods or filtration systems (Luttrell et al., 2007). Figure 1.3 shows several different types of
mechanical dewatering methods for different ranges of particle sizes.
5 |
Virginia Tech | Figure 1.3 Range of coal particle sizes that can be effectively treated by conventional coal
processing methods (Luttrell, 2012).
1.2 Problem Statement
The solid-solid and solid-liquid separation processes used to treat fine (-1 mm) coal are
the least efficient and most costly operations used in modern coal processing facilities (Figure
1.4). Field studies indicate that the froth flotation process, which is normally used to recover coal
finer than 0.1-0.2 mm, often recover less than 65 to 70% of the organic matter in this size range.
Moreover, this surface based separation process is inherently less effective in removing pyrite
than density based separation processes used to treat coarse coal. The lower particle size that can
be effectively treated by water based density separators is severely limited by the low mass of
small particles. Moreover, fines often represent 10 percent or less of the total run-of-mine feed.
However, this size fraction may contain one third or more of the total moisture of the delivered
6 |
Virginia Tech | well as various types of surface-based separators such as conventional or column flotation
machines. In many cases, the separation processes are used in multistage circuits and are
integrated with various types of classification operations either before or after the cleaning step.
In addition to technical and economic reasons, operator preferences and vendor biases also
appear to contribute to the large variations that are observed in how fine coal is cleaned and
dewatered. As a result, a standard “optimum” flowsheet for fine coal processing, which can be
flexible and adaptable to accommodate changes in feed coal characteristics, does not exist at this
time.
1.3 Research Objectives
The objective of this research is to develop an engineering criterion that can be used to
identify optimum circuit configurations for the processing of fine coal streams. This dissertation
describes several potential improvements in fine coal processing circuitry. Several multistage
circuits, including laboratory, bench-scale, pilot scale-and in-plant test circuits, were set up and
tested to demonstrate the potential improvements in fine coal recovery that may be realistically
achievable using an “optimized” fine coal processing flowsheet. This research work also focused
on the development, testing and evaluation of a modified compound spiral to treat fine coal
feedstocks. A new combined flotation-spiral circuitry for desulfurization of high sulfur fine coal
was also developed and experimentally tested. Finally, this research work also evaluated an
innovative fine coal dewatering technique called Nano Drying Technology (NDT™). This
innovative process is capable of physically removing moisture from fine coal at ambient
temperature using molecular sieve technology.
8 |
Virginia Tech | 1.4 Contributions
The key contributions of this research and development work are as follows:
• Developed an improved fine coal processing circuitry based on feed size characteristics.
• Developed an expanded stage compound spiral circuit (4-3-4-3 turn spiral circuit with
second and fourth stage middlings recycle).
• Developed a new and innovative enhanced fine coal sulfur rejection circuit using
combined spiral-flotation circuitry.
• Evaluated the effectiveness of a new low-temperature fine coal drying process.
1.5 Dissertation Organization
This dissertation is composed of seven chapters. The first chapter “Chapter 1 -
Introduction” provides an introduction to the research topic, a detailed problem statement and a
listing of research objectives. The second chapter “Chapter 2 - Literature Review” is a brief
report on current status of research on fine coal processing. Specifically, this chapter describes in
detail the research and developments in coal spiral technology (i.e., history, construction, design
variables and operating variables). A brief introduction of several other fine coal processing
separators used in this research work is also included in the literature review.
The next four chapters focus on the experimental testing and engineering evaluation of
various new approaches for fine coal cleaning and dewatering. The first of these chapters
“Chapter 3 - Performance Comparison of Fine Coal Cleaning Alternatives” discusses all the
laboratory-scale and pilot-scale test results obtained from the detailed testing of common
technologies used to clean fine coal (i.e., spirals, teeter bed separators, water-only cyclones and
froth flotation). The resultant data suggest, for the particular coal investigated in this study, that
the most effective processes for each size range were generally (i) froth flotation for feeds finer
9 |
Virginia Tech | than about 0.3 mm, (ii) spirals for feeds sized to 1 x 0.3 mm, and (iii) teeter-bed systems
(particularly the HydroFloat™ technology) for feeds larger than 1 mm. Water-only cyclones are
not recommended as stand-alone units due to the potential for high coal losses when secondary
back-up units are not available within the plant circuitry.
The next chapter “Chapter 4 - Engineering Development of Expanded Stage Compound
Spiral” provides results of field tests conducted with a prototype compound spiral that was
modified to improve coal recovery and enhance the selectivity of fine coal processing. Detailed
in-plant experimental tests results, along with separation performance data comparisons, are
presented in this chapter for five different spiral circuit configurations. The performance
comparison indicates that, amongst all the spiral circuits tested, a modified compound spiral with
more cleaning stages and partial middlings recycle is the best option for improved separation
efficiency, clean coal yield and combustible recovery. Preliminary calculations indicate that this
new modified spiral circuit is capable of increasing the clean coal yield by 1.9%, while
maintaining the same ash contents as achieved by existing compound spiral circuits.
This spiral technology research was further expanded in the next chapter “Chapter 5 -
Enhanced Sulfur Rejection Using Combined Spiral Flotation Circuits” to investigate new
methods for improving fine coal desulfurization. Specifically, this chapter discusses the
experimental set up, test results, and technical evaluation of an innovative combined spiral-
flotation circuitry. This chapter also describes in detail the study conducted to evaluate the
partitioning of pyrite within fine coal circuits. The sulfur and ash separation performances of
different fine coal cleaning alternatives were also presented and compared. On the basis of this
study, a spiral followed by a froth flotation cleaning process is recommended for the cleaning of
high sulfur ultrafine (minus 0.25 mm) coal feeds. This chapter describes the rationale for the
10 |
Virginia Tech | design of this new fine coal cleaning circuit and finally a generic flowsheet is also proposed for
any coal preparation plant treating high sulfur ultrafine coal feeds.
The next chapter “Chapter 6 - Engineering Development of Micro Sieve Drying Process”
discusses the theoretical basis for an innovative fine coal drying process. Experimental results
obtained from bench- and pilot-scale testing of this novel approach to fine coal dewatering are
also presented and discussed in this chapter. The results obtained from the experimental work
indicates that the NDT™ system can effectively dewater fine (1 mm x 0) coal from slightly more
than 30% surface moisture to single-digit values. Test data obtained using a pilot-scale NDT™
plant further validated this capability using a continuous prototype facility. The data presented in
this chapter also showed that the performance of the NDT™ system is not dictated or constrained
by particle size, i.e., it works equally well on 1 mm x 0 coal as it does on 325 mesh x 0 coal.
The seventh and final chapter of the dissertation “Chapter 7 – Summary and
Conclusions” provides an overall summary of the findings, conclusions and recommendations
resulting from this research and development study.
11 |
Virginia Tech | CHAPTER 2 - REVIEW OF LITERATURE
2.1 Fine Coal Processing
Coal loss during fine (-1 mm) coal processing is perhaps the greatest among all other size
fractions. For example, froth flotation processes typically recover only 60-80% of the organic
matter contained in fine coal feeds (Bethell, 1998). A study conducted by Cavallaro et al. (1991)
indicates that the reserves of low-ash coal in the central Appalachian region could be nearly
doubled by efficiently cleaning at a particle topsize of 1 mm (Figure 2.1). Moreover, surface-
based separation processes, such as froth flotation, which are generally used to treat fine (0.15 x
0.044 mm) coal, are less effective in removing pyritic sulfur than density-based processes used to
treat the coarser sizes of coal (Adel and Wang, 2005). Therefore, fine coal desulfurization is
often poor in many coal processing facilities treating high sulfur run-of-mine coal seams.
Field studies indicate that the yield
and quality of clean coal products from fine
coal circuitry may be significantly increased
by improving the efficiency obtained for
particle size separations at 150- and 45-μm.
Unfortunately, screens used for sizing fine
particles, and particularly those finer than
0.5 mm, tend to blind easily, wear quickly
and suffer from low throughput and poor
efficiency (Mohanty, 2003). Another
problem associated with fine coal screening
Figure 2.1 Effect of decreasing top size on
coal availability (Cavallero et al., 1991).
Used under fair use, 2012.
15 |
Virginia Tech | is the inefficient removal of ultrafine mineral sediments, such as high-ash clay particles, from
fine coal feeds (Mohanty et al., 2002). Moreover, fine (-1 mm) coal particles may represent as
little as 10% of the total run-of-mine coal, but often contain one-third or more of the total
moisture in the final coal product. Existing fine coal dewatering processes, such as filtration,
centrifuges and thermal drying, are expensive and consume large amounts of energy (Osborne,
1988; Le Roux et al., 2005). The lack of an efficient, inexpensive and safe drying process is one
of the primary reasons that about 2 billion tons of fine coal was discarded by the United States
coal preparation plants into waste impoundments (Orr, 2002).
2.2 Spirals
2.2.1 Spiral Technology
A spiral is composed of a helical channel of modified semicircular cross-section wound
around central column. These are flowing film concentrators and have been found varied
applications in coal and mineral processing industry. Generally, a feed pulp containing between
15 to 45% solids by weight and in the size range of 3 mm to 75 µm is introduced at the top end
of spiral trough. The pulp then gradually flows spirally downwards and, during this motion, the
particles tend to stratify into different streams depending upon their particle size and specific
gravity. Separation is achieved by the combined action of stratification, film sizing and
centrifugal/gravitational forces. Finally, adjustable splitters and/or cutters are used to divert the
separated particles into clean, middlings and refuse streams (Davies et al., 1991; Wills and
Napier-Munn, 2006). Figure 2.2 is a schematic cross-section of a coal spiral trough. In addition,
coal spirals have been successfully used for the treatment of iron ore, chromite, heavy mineral
sand deposits.
16 |
Virginia Tech | Figure 2.2 Cross-section of flow through a spiral trough.
Spirals are considered as one of the simplest, most effective and lowest cost fine (1 x 0.15
mm) coal processing technology (Kohmuench, 2000). Coal cleaning using spiral concentrators
was started back in 1947 when Hudson Coal Company installed 48 spirals to clean fine
anthracite coal in eastern Pennsylvania. During the 1950s, a number of researchers tried to use
spirals to clean fine bituminous coal, but their attempts were not successful (Denin & Wilson,
1948). Finally, since first introducing design changes in the 1980s that made spirals larger
(which improve throughput capacity) and lighter (due to fiberglass and urethane construction),
spiral separators have become one of the most popular fine coal cleaning separators. Apart from
their simplicity, coal spirals have many other advantages such as low capital and operating cost,
simple to operate, no moving parts, no regent requirements, stable cut point with size, good clean
coal recovery and high reject ash levels. On the flip side, spirals have low throughput capacity
17 |
Virginia Tech | compared to other water-based separators such as water-only cyclones. Spirals also operate with
a relatively high specific gravity cut point and can treat only a very limited feed size range
(MacHunter et al., 2003; Luttrell et al., 2003).
2.2.2 Historical Development
The concept of separation by spirals dates back to 1943 when Humphreys Mineral
Industries introduced their first spiral for concentrating mineral ores. These early units were used
to upgrade a wide variety of ores including gold, silver, tin, ilmenite, rutile, zircon, monazite,
iron, barite, fluorspar, mica and phosphate (Davies et al., 1991; Leonard, 1991). For many years
following the introduction of the first units, spirals tended to be used only for comparatively easy
separations. However, subsequent research and development by different spiral manufacturers
around the world led to much wider use of spirals in mineral processing applications and
eventually to their adoption for the coal cleaning (Richard et al., 1985).
The early spirals utilized relatively simple profiles and were designed with a single start
and 4-6 turns. These units were constructed from semicircular sections that were bolted together.
One of the disadvantages of early spiral separators is that they were constructed from cast iron
and weighed about one tonne each. In Australia, sand miners used truck tire sections that were
cut and reattached together to form a spiral instead of using excessively heavy spirals. In 1947,
spirals made from asbestos reinforced concrete were introduced for rutile and zircon extraction.
The decade of 1950 marked a major advance in spiral technology when Ernst Reichert used
fiberglass as a construction material for spirals. The use of fiberglass made it possible to made
light weight, non -corrosive continuous helices and allowed two or three helices to be mounted
on one central column (Davies et al., 1991, Hunter et al., 1985). Another development in the
subsequent spirals is the replacement of rotating disc cutters for concentrate removal with the
18 |
Virginia Tech | concentrate cut from the pulp stream by finger-type splitters. All early designs used wash-water
channels to overcome the sanding/beaching problem of inner trough section. In late 1970s and
early 1980s, Australian researchers came up with modified spiral geometry that resulted in
complete wash-waterless light-weight spirals (Richards & Palmer, 1997). The most recent breed
of spiral is made of fiberglass spray-coated with polyurethane (Das et al., 2007).
2.2.3 Particle Separation Mechanism
While spirals are conceptually very simple in terms of design, the particle separation
mechanism that occurs along the flow path is relatively complex. Feed slurry introduced at the
top of the spiral gradually flows downward under gravity through the spiral trough. Within the
rotating flowing film, coal particles are subjected to gravitational, centrifugal and Bagnold forces
(Bagnold, 1954; Kapur and Meloy, 1999). The combined action of these forces causes lighter
particles to move towards the outer wall and the denser particles move to the central column.
The particle separation mechanism on spirals has been a continuous source of confusion
in the literature. Some researchers believe that there are two basic types of fluid flow along a
spiral trough, which are (i) a primary axially downward flow and (ii) a secondary cross channel
flow (Figure 2.3). The hydraulic phenomenon responsible for both of these flows has been
explained by a number of researches (Holland-Batt, 1990, 1992, 1994, 1998; Richards and
Palmer, 1997; Kapur and Meloy, 1998). According to Holland-Batt (1990), the interaction
between the fluid flow and the particles results in separation of particles of different densities.
On one hand, light particles are carried in the cross flow from the inner region of the trough
towards the outer region and settle at the bottom of the channel. In this case, light particles are
picked up and carried down by the primary flow, which eventually transport the particles out of
the separator. On the other hand, dense particles in the outer region quickly fall to the bottom of
19 |
Virginia Tech | the channel and are carried towards the inner region by the cross flow. These dense particles are
too heavy to be picked up and carried back into the outer region. Hence, dense remain in the
inner region and are carried by the primary flow down the separator (Holland-Batt, 1989).
Richards and Palmer (1997) divided the cross section of the spiral trough into three zones
as shown in Figure 2.3. The inner zone is occupied by a bed of slow moving heavy particles. In
the outer region, or recovery zone, heavy particles must settle into lower layers in order to be
transported towards the center of the spirals. The intermediate transition zone contains composite
“middlings” particles and is located between the inner and outer zones (Richards and Palmer,
1997).
Figure 2.3 Separation mechanism and primary and secondary flow pattern on a spiral
trough (Richards & Palmer, 1997). Used under fair use, 2012.
According to Luttrell et al. (2007), the above mentioned description of particle separation
fails to recognize two counter-rotating flows that actually present across the spiral profile. These
20 |
Virginia Tech | two rotating flows converge along a line of separation as shown by Figure 2.4. The clockwise
flow in the lower rotation zone is responsible for moving lighter particles towards the outer wall
of the spiral. Heavier particles contained in this flowing stream settle down and are carried
inward towards to the inner side of the spiral trough. The clockwise rotation is responsible for
providing a dense concentrate that is relatively free of light particles. In contrast, the counter
clockwise flow in the upper rotating section stratifies particles along the outer wall according to
density. Unfortunately, some denser particles in the upper flow zone settle against the wall and
are trapped there by rising current of the counter clockwise flow. Studies indicate that these
entrapped high density particles rarely cross from the upper to the lower flow zones and
eventually report with the low density product regardless of density. To eliminate these
entrapped particles, Luttrell et al. (2000) recommends recleaning of primary low-density product
using a secondary stage of spirals.
Figure 2.4 Separation regions across a spiral profile (Luttrell et al., 2007).
Used under fair use, 2012.
21 |
Virginia Tech | 2.2.4 Particle Separation Forces
The forces responsible for particle separation within a spiral trough have been studied by
a number of researchers (Atasoy and Spottiswood, 1995; Kapur and Meloy, 1999; Atasoy, 1987;
Holland-Batt and Holtham, 1992; Luttrell et al., 2000). Forces involved in the particle separation
within the spiral trough include hindered settling, interstitial trickling, centrifugal, frictional,
gravitational, drag and Bagnold forces (Bagnold, 1954; Kapur and Meloy, 1999). Among these,
the Bagnold force is distinctive. Data that shows its existence during particle separation within a
spiral trough was first shown by Holtham in 1992. He concluded that Bagnold forces arises due
to increase in inter particle interaction at high pulp densities and at high shear rates. It was also
found that Bagnold forces weaken at a solid percentage below about 50% (Holtham, 1992). The
Bagnold force is a dispersive force that is directly proportional to the shear rate and square of the
particle diameter. The Bagnold force varies along the depth of the flowing film and, depending
on its magnitude, causes particles to move upward or downward in the flowing film. Studies
show that Bagnold forces are in effect within the inner region of spiral, where particles are in
bed-load motion and the percentage solids are more than 50% (Atasoy and Spottiswood, 1995).
Kapur and Meloy (1999) concluded that amongst all the forces acting on a particle, no single
force dominates the others and, hence, separations are based on differences in the rate of change
of all these individual forces with respect to particle size, shape, density and radial position.
2.2.5 Spiral Design Parameters
Spiral design is critical to effective separation performance and has been the subject of
both experimental and computational studies over the last three decades (Kapor and Meloy,
1998; Holland-Batt, 1989; Holtham, 1990; Stokes, 2000). Primary spiral design parameters
include spiral pitch, diameter, trough slope, length and profile. Secondary design parameters
22 |
Virginia Tech | include wash-water configurations, feed box arrangements, splitters locations, repulper locations
and construction materials. The design process consists of a number of interactive stages and
usually starts with the scale-up, followed by volute shape determination, and finally with a
sanding analysis (Davies et al., 1991; Holland-Batt,1990). Some of the key design variables are
described in greater detail in the following sections.
Pitch and Diameter: The ideal pitch of a spiral trough for a particular feed type is one
that ensures particle fluidity. In general, the pitch is steeper for heavy high-density feed material
and is shallower for light particles (Davies et al., 1991). The diameter of a spiral depends upon
the capacity and the separation size. For a given separation size, the capacity of a spiral is a
function of trough area and it decreases with the particle size (Hollan-Batt, 1985). Due to the low
density and low unit value of coal, spirals used in coal applications have a lower pitch and a
larger diameter than mineral spirals (Luttrell et al., 2007).
Profile: Profile is a very important design parameter for spirals. In particular, the shape
of the inner and flatter section of a spiral controls the migration of high density particles towards
central column. The inner profile terminates and curves steeply upward at the outer side as a
vertical water-retaining wall. The profile is generally designed to give a targeted relative density
of separation, but in practice the profile often represents a compromise between fluidity and
selectivity. Davies (1991) also considered the number and method of product stream divisions
and overall pulp capacity as important parameters that influence the profile design. Optimizing
studies of profile shape, with a focus of material handling aspects, were conducted by Holland-
Batt (1995). This work discussed various profile shapes and concluded that trough shape has a
profound effect on the nature of fluid flow phenomena and, hence, on the separation efficiency
as well. This study concluded that, for a given pitch, one shape will produce excellent metallurgy
23 |
Virginia Tech | but poor material handling behavior, while other shapes may excel at transporting solids but
perform poorly in terms of separation (Holland-Batt, 1995).
Flow length: In addition to pitch and profile, spiral length is a critical parameter that has
been thoroughly investigated by a number of researchers (Davies, 1991; Kohmuench, 2000;
Wildon and MacHunter; 1997; Atasoy and Spottiswood, 1995). Length, which reflects the total
distance over which slurry travels as it passed down a spiral, is normally referenced by the
number of complete 360o turns utilized by the spiral design. This parameter should not be
confused with spiral height, which varies depending on the pitch employed. The aim of these
studies was to optimize, and perhaps standardize, the required number of turns on a spiral for an
efficient separation process for different ores. In early 1960’s, Australian coal spirals employed
as few as two full turns, while modern spirals employ as many as seven turns or more to achieve
the required separation. In general, a minimum of five to six turns are recommended to achieve
maximum separation efficiency (Holland-Batt, 1995). Other research (Weldon and MacHunter,
1997) has suggested that four turns are optimum for most spiral applications and that acceptable
separations can be achieved using even shorter two or three turn spirals, depending on the
density distribution of the feed particles (Figure 2.5). According to Atasoy and Spottiswood
(1995), the optimum length of a coal spiral is a function of feed size. Shorter spirals are more
effective for separation of coarse particles, while longer spirals are better for fine particles.
24 |
Virginia Tech | (DGR = Clean Ash/Feed Ash)
Figure 2.5 Plot for optimum number of spiral turns (showing 4, 5 and 6 turn coal spirals)
(Wildon & MacHunter, 1997). Used under fair use, 2012.
Repulping: The installation of repulpers along the spiral length generally improves the
separation efficiency of mineral spirals. The repulpers reinitiate the separation process by
capturing, mixing and reintroducing the high velocity slurry stream with the relatively sluggish
middling stream. The idea of repulping was introduced based on work originally done by
Holland-Batt in 1995. According to this study, spiral fluid flow reaches steady-state after only
two turns, while mineral recovery slowly continues for up to four or more turns. As a result,
repulpers are generally installed after three or four turns when high density products are removed
through central column. Several spiral manufacturers have introduced designs that have
successfully incorporated repulping (MacHunter et al., 2003). The effectiveness of repulping in
coal applications has drawn mixed opinions. In his PhD dissertation, Kohmuench (2000) argued
that although repulping in coal spirals reduces gravity cut point but is found to be less effective
in improving separation efficiency because the relatively low specific gravity of coal particles
25 |
Virginia Tech | requires more turns to be effectively separated. Holland-Batt (1995) also stated that repulping
can destroy a partial separation occurring with finer material and thus can decrease the separation
efficiency. Similarly, Atasoy and Spottiswood (1995) noted that if a mineral spiral treating 4.0
SG solids requires only one turn for an effective separation, then it can be expected that a coal
spiral will need approximately five or six turns to achieve a good separation.
Construction Materials: The earliest spiral was manufactured from cast iron. This practice
remained unchanged until the 1950’s when fiberglass was introduced as a construction material
for the spiral structure by Ernst Reichert. Until the late 1970’s, rubber was used as a lining
material for spiral trough and for feed and product boxes. Rubber lining was generally effective,
but expensive and difficult to apply to complex shapes. In 1980’s, sprayed polyurethane was
introduced as a lining material. Finally, in 1988, the first mono-polymer spiral was commercially
introduced, which is available in either ceramic or polyurethane. The major advantages of the
mono polymer spiral construction include improved wettability and fluidity and superior
resistance to reagents and to acid and spark attacks. Today, reverse casting is now well
established for the construction of spirals, which made it possible to accurately fabricate feed and
product boxes, splitters, repulpers and other components of the distribution and laundering
system in heavy duty sections (MacHunter et al., 2003).
Feed Box: The spiral feed box is used to introduce the slurry to the spiral trough in a
direction parallel to the walls of the trough. Ideally, the feed box should be hydraulically
designed to provide equilibrium of the flow pattern as early as possible without splashing or
surging. It is also generally accepted that the design of the feed box should facilitate the
distribution of solids evenly throughout the slurry (Holland-Batt, 1995). More recently, however,
some researchers have proposed that selective segregation and distribution of larger particles
26 |
Virginia Tech | towards the inside of the trough may be desirable to reduce unwanted entrapment of dense
particles in the high velocity flow region (Luttrell, 2012).
Product Splitters: In order to make appropriate low-density, middling and high-density
products, spirals are fitted with two adjustable splitters at the discharge end of the spiral. There
are several different types of splitters such as small finger splitters, banana splitters and slide
splitters. Splitters located at the end of the last turn are normally of the pivoting-blade type.
These splitters are placed either on the trough surface or, in some cases, may be embedded in or
positioned exterior to the trough surface. The splitters may be positioned in parallel or offset
slightly to permit total elimination of either the middlings or concentrated products. Multiple
start spirals are linked through a common shaft to control the same splitter positioning. Some
mineral spirals are also equipped with one or more auxiliary splitters to remove separated solids
(Holland-Batt, 1995).
For spiral applications in the coal industry, two splitters are used to separate coal,
middlings and refuse products. Generally, the outer splitter nearer the wall is capable of making
a density cut between 1.55 and 2.0 SG, while the inner splitter nearer the support pole can make
a density cut 1.8 and 2.4 SG (Mikhail et al., 1988). In order to maintain efficiency, a constant
density cut amongst the entire spiral bank should be targeted, which requires the position of the
splitters to be the same for all the spiral units (Luttrell, 2007). Typically, the outer splitter should
be placed to provide an acceptable clean coal product quality, which typically requires a position
of approximately 3 inches from outside wall. Likewise, the inner splitter should be placed to
provide a reject product that is acceptable for discard, which normally requires a splitter position
of approximately 10.5 inches from outside wall. The middlings product resulting from these
positions can be diverted to the clean product, refuse stream or recycled back to feed. Some
27 |
Virginia Tech | spirals have an additional primary refuse splitter, called a cutter, located after three or four turns.
The purpose of the cutter is to remove high density refuse as soon as possible so as to improve
separation efficiency and increase refuse loading capacity.
Ancillary Components: In order to collect products from main and auxiliary splitters,
some form of product receivers are also associated with the spiral assembly. Their designs vary
depending upon the manufacturer and model of spiral. The main characteristics of product
receivers include satisfactory performance in terms of wear resistance, avoidance of splashing of
products and suitable material handling characteristics (Holland-Batt, 1995). Spirals are usually
installed in multiple banks and the necessary ancillaries required for these banks include a frame
to support the spiral bank, main distributors to split feed pulp equally amongst each start and
launders to transport the concentrate, middling and tailing flows. Spiral banks are fed via an
overhead feed distributor which ideally distributes the feed slurry equally and homogenously to
every spiral unit in the bank.
2.2.6 Conventional and Compound Spirals
There are a number of spiral manufacturer around the world. Some manufacturers
incorporate proprietary features into their conventional single-stage spirals and claim that design
improves the separation performance. Studies conducted by Honaker and Wang (1991) evaluated
the separation performance of four conventional single-stage spirals made by different
manufacturers. This investigation concluded that there is a little difference in the separation
performance among all these spiral designs (Figure 2.6). Therefore, it is not surprising that most
manufacturers have instead focused much of their R&D efforts on the development of compound
spiral designs. Compound spirals incorporate two stages of spiral processing in a single spiral
assembly. Typically, compound spirals consist of three or four turns of primary spirals
28 |
Virginia Tech | 2.2.7 Spiral Operating Variables
The separating performance of spirals is greatly influenced by a number of operating
variables that are under the control of the plant operators. In coal preparation plants, spiral
circuits are not often run at their maximum separation efficiency because of poor feed sizing,
incorrect splitter settings, inappropriate solid and volumetric flow rates and uneven feed
distribution (Luttrell et al., 2000). Therefore, to avoid these issues, a brief discussion of how
these operating variables influence spiral performance is provided in the following sections.
Particle Size: There have been conflicting opinions about the optimum particle size
range for coal spiral circuitry. For coal applications, some early researchers were of the view that
the appropriate coal feed size for spirals is 3 x 0.1 mm (Kapur and Meloy, 1999; Davis et al.,
1991), while others suggest that a feed size as broad as 3 x 0.05 mm can be effectively treated by
this technology (Holland-Batt, 1992). Another point of view (Luttrell et al., 2007) is that coal
particles coarser than 1 mm or finer than 0.2 mm are not cleaned as effectively in spirals and
should instead be upgraded by dense medium processes (for plus 1 mm) and froth flotation (for
minus 0.2 mm).
Slurry Flow Rate: For an efficient separation, spirals should be provided with an adequate
and stable slurry flow rate. The optimum slurry flow rate varies according to the spiral diameter.
Typically, for most of the industrial units, the optimum flow rate is between 30-40 gallons per
minute (GPM) per start for a particle size range of 1 x 0.15 mm (Luttrell et al., 2007). A
volumetric flow rate on higher side is maintained for a coarser feeds, while a lower slurry flow
rate is required for a finer feeds (Honaker et al., 2006). At a constant tonnage of dry solids, a
lower flow rate may result in sanding or beaching problems along the spiral trough. A higher
flow can also cause high density particles to report with the water, which reduces the quality of
30 |
Virginia Tech | the low density product by lowering the separation efficiency and increasing the density cut point
(Kohmuench, 2000).
Atasoy and Spottiswood (1995) studied the effect of residence time on the separation
performance of spirals. They concluded that the residence time has a mixed effect on the
separation efficiency of the particles of different densities and size classes. For example,
residence time does not play a significant role in the separation of low-density coarse (3.35 x 1.7
mm) coal particles. In contrast, increased residence time has an unfavorable effect on higher
density (SG > 1.45) particles of the same size class because they tend to move towards the clean
coal stream with time.
Solids Feed Rate: Recent research (Luttrell et al., 2003; 2007) indicates that the density
cut point (SG ) and Ecart Probable (Ep) increases sharply with an increase in the dry solids feed
50
rate to a spiral. As shown in Figure 2.7(a), a decrease in the solids feed rate improves the product
quality (lowers the clean coal ash content), but decreases the recovery of product solids (reduces
the recovery of combustible organic matter). An increase in dry solids feed rate to more than 3
tonnes per hour per spiral start seriously impacts the separation efficiency (Holland-Batt, 1994;
Li et al., 1993). Contrary to this, some spiral manufacturers claim that their spiral designs can
handle a feed rate as high as 4.5 tonne per hour per start without impacting the separation
efficiency (Luttrell, 2012).
31 |
Virginia Tech | Figure 2.7 Effect of dry feed rate on spiral performance (a) on gravity cut-point, (b)
on separation efficiency (Luttrell et al., 2003). Used under fair use, 2012.
Feed Solids Content: At constant feed percent solids, an increase in the feed rate
increases the volumetric flow of slurry down the spiral. This increases the centrifugal force
exerted on the particles, forcing more material to report to the low density product that, in turn,
results in a higher density cut point (Mikhail et al., 1988). As shown in Figure 2.6(b), a similar
relationship between dry feed rate and cutpoint was also reported by Luttrell (2003). If the dry
tonnage is fixed, then an increase in the feed solids content decreases the slurry flow rate and
lowers the specific gravity cut point. For coal applications, this action decreases combustible
recovery and improves clean coal ash (Luttrell et al., 2003). Mikhail (1988) is of the point of
view that feed rate may actually have a greater effect on separation cut point than even splitter
position.
32 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.