INTERIM FINAL TECHNICAL REPORT
November 1, 1999, through October 31, 2000
Project Title: FINE COAL CLEANING WITH AN AUTOMATED ENHANCED DENSE
MEDIUM GRAVITY CIRCUIT
ICCI Project Number: 99-1/3.1A-1
Principal Investigator: Dr. Y. P. Chugh, Southern Illinois University
Other Investigators: Dr. D. P. Patil, Southern Illinois University; Mr. Amit Patwardhan, Southern Illinois University
ICCI Project Manager: Dr. Ken Ho, Illinois Clean Coal Institute
ABSTRACT
The main objective of this project is to automate the operation of the dense medium Falcon concentrator to provide optimal yield and grade under varying feed characteristics.
Based on the analysis of washability data of different Illinois coal samples,
the Jader Fuels (Equality, IL, Illinois No. 2 and No. 3 seam) and American
Coal (Galatia, IL, Illinois No. 5 seam) samples were selected for the present
work. Water-injection cyclone experiments were carried out to remove the
-325 mesh size fraction from the feed coal. The most important parameters
of the water-injection cyclone are vortex finder diameter, water-injection
rate and spigot diameter for both Jader and American coals. The truncated
cone is a significant parameter for Jader and American coals at 90% confidence
level. The feed pressure significantly affects the classification of Jader
coal, where as for American coal, it does not have much effect. The hydrocyclone
experiments (without water-injection) show that, for cut point
(D95) the most important parameters are vortex finder diameter
and percent solids. The percent solids also affect the cut point, which increases
with an increase in percent solids. The size distribution of the underflow
of the cyclone without water injection for American coal show that about
40% of -500 mesh fraction present in the feed reports to underflow. The results
of water-injection cyclone show that only 14% of the -500 mesh fraction present
in the feed reports to underflow.
The dense medium Falcon experiments show that the feed ash and underflow
rate have the greatest effect on the product ash content. The product ash
is a strong quadratic function of underflow rate. The increase in feed ash
increases the product ash, reducing the product yield. An increase in feed
ash content from 12.77% to 18.44% increased the product ash from about 8%
to 10.5% with a decrease in yield from about 92 % to 82%.
A Levenberg-Marquardt gradient search algorithm has been implemented to perform
a constrained optimization routine which identifies the operating conditions
(Bowl speed, and underflow rate) that will provide the maximum combustible
recovery for any given product quality constraints.
All the planned tasks have been carried out on schedule with the exception
of the automation and control task. The difficulties encountered have been
presented to ICCI and a four-month no-cost extension has been obtained.
EXECUTIVE SUMMARY
The research conducted during the past two decades on processing of coal fines has changed the description of fine coal as an environmental liability to a potential economic resource. The development of "Enhanced Gravity Techniques" is a product of research conducted during this period aimed at improving the efficiency of fine particle processing. Research funded in a large part by the state of Illinois has shown that enhanced gravity concentration has the ability to improve the rejection of both ash forming minerals and coal pyrite while maximizing the recovery of fine coal. The process efficiently treats particles of sizes of 1 mm, down to as small as 44 mm.
Based on the previous studies conducted at SIU using an enhanced gravity
separator called the Falcon Concentrator, it has been observed that though
the high centrifugal acceleration helps to achieve efficient separation of
finer fraction (100x325 mesh), it also leads to faster settling of the coarser
fraction (28x100 mesh), thus reducing the recovery of the coarser coal fraction.
Hence, an innovative method was developed at SIUC with the funding provided
by ICCI, to improve the separation of coarser coal particles in the Falcon
Concentrator. The method uses dense medium (DM) in the Falcon concentrator
thus reducing the effect of size on separation of coal particles. Previous
studies with the dense medium Falcon concentrator (DMFC) showed that the
separation efficiency depends on the amount of minus 325-mesh present in
the feed coal. Also, the efficiency of dense medium operation depends on
the feed and media characteristics. This variation in the feed and media
characteristics is widely observed in coal preparation plants. Hence, to
address these problems, studies were conducted using a water-injection cyclone
to reduce the fines content in the feed coal. Automation potential for the
DM Falcon concentrator is being evaluated to ensure an optimal product yield
and grade under varying feed characteristics.
The coal samples for the present study were selected after analyzing the
washability data of various Illinois coals. As the DMFC is a more efficient
separator than the water-only Falcon concentrator, it is necessary to use
coal samples having a "difficult to clean" coal characteristics. The -16
mesh washability data for coals from Galatia (American Coal), Marissa, Wabash
and Jader Fuel were analyzed to select two coal samples for the current work.
The run-of-mine (ROM) coal of Galatia mine contains about 14% (by wt.) of
-16-mesh coal, having an ash content of 26%. The near gravity material (NGM)
in the region 1.4 to 1.8 densities (gm/cm3) is about 3%. The presence
of lower amount of NGM indicates that a moderately efficient process may
be used to clean the coal. However, the presence of high amount of fines
indicates that the coal requires an efficient gravity separator.
The ROM coal from Marissa mine contains about 42% (by wt.) of -16-mesh size
fraction having an ash content of 25%. The NGM in the density range 1.4 to
1.8 is about 4 to 6%, requiring an efficient separator for the cleaning this
coal. For the Wabash coal, the ROM contains about 10% (by wt.) -16-mesh fraction
having an ash content of about 7%. The NGM in the density range 1.4 to 1.8
is around 1%. Therefore, a moderately efficient separator may be utilized
for the cleaning this coal. The washability data on composite Jader Fuel
classifying cyclone feed (seams 2+3) indicates that the -16 mesh sample contains
about 21% ash. The NGM in the density range 1.4 to 1.8 is about 4 to 17%.
Hence, this coal requires an efficient separator for cleaning.
Samples from the classifying cyclone feed were collected in separate barrels
from Jader and Galatia coal preparation plants. These samples were homogenized
and a representative sample was extracted for characterization of each sample.
The standard deviation values for ash and total sulfur for Jader samples
in different barrels were 1.10 and 0.18 respectively.
The representative samples were subjected to particle size analysis and
washability tests and the resulting size and density fractions were analyzed
for ash, total sulfur and heating values. The Jader coal contains about 19%
ash, having 4% total sulfur. The sample contains around 15% of -400 mesh
fraction having an ash and total sulfur content of about 43% and 6% respectively.
The ash and total sulfur content increases with decrease in size, indicating
the necessity of using an enhanced gravity separator. The American coal contains
about 29% ash having 2.26% total sulfur. The amount of -400 mesh particle
size fraction is about 28% with an ash content of about 69%.
The washability analysis data for the 16x325 mesh particle size fraction
shows that the NGM within the specific gravity rage of 1.5 to 1.8 is in between
1 to 3% for Jader coal and 0.3 to1.5% for American coal. This finding suggests
that the particles are liberated and a moderately efficient gravity separator
is sufficient to ensure high separation efficiency. However, since about
35% of the material is between 48x325 mesh size fraction, it is necessary
to use an enhanced gravity technique for achieving the required separation
performance. The theoretical product ash and yield that can be obtained from
Jader and American coals respectively are 7% and 85%, and 5% and 90%.
An AMDEL online coal slurry analyzer is being used at Southern Illinois
University (SIU) to measure the feed ash content and feed solids ratio. This
information is used in an online optimization and control routine for the
purposes of adjusting operating parameter values, especially underflow mass
flow rate or specific gravity of cut, when the assay and/or mass flow rate
in the feed stream to the DMFC varies. Thus, on-stream analysis of the product
and/or tailing streams by the AMDEL will provide efficient control of the
product grade through manipulation of medium density or underflow rate. An
Amdel ash analyzer was installed at SIU's high-bay testing facility. About
40 coal samples having varying ash contents (10%-60%) and percent solids
(8%-40%) were used to calibrate the ash analyzer. Calibration equations were
developed using the following parameters.
Scatter - X-ray
Neutron count rate
Fe correction
Aeration
It was observed that calibration curves developed using samples having a
narrow ash ranges (10%-30%, 30%-60%) provided better ash accuracy
(± 1.25%) than using a single range (10%-60%).
However, the accuracy of the measurement of percent solids did not significantly
improve using a closer range of percent solids. The percent solids accuracy
was ±0.75%.
Statistically designed experiments were carried using the Plackett-Burman
experimental design on the Jader and American coal samples to identify the
important design and operating variables that affect the performance of a
water-injection cyclone. The following parameters were used for the experimental
program.
Percent solids (5% - 20 % by weight)
Feed pressure (10 - 30 psi)
Vortex finder diameter (0.75 - 1.25 in)
Spigot diameter (0.6 - 0.9 in)
Truncated cone diameter (0.75 - 1.0 in)
Water injection rate (4 - 11 gpm)
The most important parameters affecting the water-injection cyclone performance
are identified as the vortex finder diameter, water-injection rate and spigot
diameter for both Jader and American coals. The truncated cone is a significant
parameter for Jader coal at 95% confidence level. For American coal, the
cone is a significant parameter at 90% confidence level. The percent solids
do not have a significant effect on the partition number. In normal operation
of the cyclone, the spigot diameter does not have much effect on partition
number. Usually, it is expected that a larger spigot diameter will result
in a higher percentage of feed material reporting to the underflow. The feed
pressure significantly affects the partition number for the Jader coal, where
as for the American coal, it does not have much effect. The increase in feed
pressure generally increases the centrifugal force, thus providing smaller
D50 values. Alias structures for the 8 run Plackett-Burman
design show that main effect F (water-injection) is partially confounded
with B (feed pressure) and C (vortex finder diameter) two-factor interaction.
Based on the analysis of the Plackett-Burman design, further experiments
according to the Box-Behnken design were conducted with the American coal
varying vortex finder diameter, water-injection rate and feed pressure.
The hydrocyclone experiments without water injection show that, for the size
separation cut point, the most important parameters are vortex finder diameter
and percent solids. An increase in vortex finder diameter increases the cut
point because of higher volumetric flow rate of solids to the overflow, which
removes higher fraction of stratified material from the center of the
hydrocyclone, thus increasing the cut point. The percent solids also affect
the cut point, which increases with increase in percent solids. A feed containing
higher percent solids, does not allow the conditions favorable to proper
settling of particles, thus increasing the misplacement of coarse fraction
to overflow. A lower percent solids is required for a sharper cut point.
As the feed pressure increases, the centrifugal force on the particles increases,
thus forcing the particles to move along the wall of the cyclone, and eventually
be discharged as the underflow. Hence, more particles will report to underflow
with increase in pressure, resulting in a decreased size cut point.
The results of near optimum size distribution of the overflow and underflow
of the cyclone without water injection show that about 40% (corrected 19%)
of -500 mesh (average size 12.5 micron) fraction present in the feed reports
to underflow. Where as about 90% of 41 micron fraction reports to underflow.
This data clearly shows the amount of very fine particles short-circuiting
to the underflow fraction. The preliminary results of water-injection cyclone
show that only about 14% (uncorrected) of -500 mesh fraction present in the
feed reports to underflow.
Statistically designed experiments were carried using the Box-Behnken design
on the American coal to quantify the effects of the operating parameters
associated with the dense medium-based Falcon using the following operating
parameters.
Feed ash (7% - 20%)
Bowl speed, (20 - 40 Hz)
Tailings underflow rate (5 lpm - 15 lpm)
The results of the dense medium Falcon tests show that feed ash and underflow
rate have the greatest effect on product ash content. The curvilinear nature
of the product ash curve suggests that the product ash is a strong quadratic
function of underflow rate, which is supported by a relatively low t-statistic
obtained for C2 term. The increase in feed ash increases the product
ash with decrease in yield. An increase in feed ash content from 12.77% to
18.44% increased the product ash from about 8% to 10.5% with a decrease in
yield from about 92% to 82%. Hence, the change in feed properties significantly
affects the process performance. Therefore, it is necessary to automate the
densese medium Falcon to maximize the product yield for a given product ash
content for varying feed properties. The higher bowl speed increases the
centrifugal force, thus providing accumulation of magnetite particle near
the wall of bowl. This process increases the specific gravity of cut increasing
the product ash and with a marginal increase in yield.
A Levenberg-Marquardt gradient search algorithm has been implemented to perform
a constrained optimization routine which identifies the operating conditions
(Bowl speed and underflow rate) that will provide the maximum product yield
for any given product quality constraints. The hardware and software setup
has been completed to adjust the Falcon variables based on the developed
empirical model.
INTERIM FINAL TECHNICAL REPORT
November 1, 1999, through October 31, 2000
Project Title: FINE COAL CLEANING WITH AN AUTOMATED ENHANCED DENSE MEDIUM GRAVITY CIRCUIT
ICCI Project Number: 99-1/3.1A-1
Principal Investigator: Dr. Y. P. Chugh, Southern Illinois University
Other Investigators: Dr. D. P. Patil, Southern Illinois University
Mr. Amit Patwardhan, Southern Illinois University
ICCI Project Manager: Dr. Ken Ho, Illinois Clean Coal Institute
ABSTRACT
The main objective of this project is to automate the operation of the dense medium Falcon concentrator to provide optimal yield and grade under varying feed characteristics.
Based on the analysis of washability data of different Illinois coal samples, the Jader Fuels (Equality, IL, Illinois No. 2 and No. 3 seam) and American Coal (Galatia, IL, Illinois No. 5 seam) samples were selected for the present work. Water-injection cyclone experiments were carried out to remove the -325 mesh size fraction from the feed coal. The most important parameters of the water-injection cyclone are vortex finder diameter, water-injection rate and spigot diameter for both Jader and American coals. The truncated cone is a significant parameter for Jader and American coals at 90% confidence level. The feed pressure significantly affects the classification of Jader coal, where as for American coal, it does not have much effect. The hydrocyclone experiments (without water-injection) show that, for cut point (D95) the most important parameters are vortex finder diameter and percent solids. The percent solids also affect the cut point, which increases with an increase in percent solids. The size distribution of the underflow of the cyclone without water injection for American coal show that about 40% of -500 mesh fraction present in the feed reports to underflow. The results of water-injection cyclone show that only 14% of the -500 mesh fraction present in the feed reports to underflow.
The dense medium Falcon experiments show that the feed ash and underflow rate have the greatest effect on the product ash content. The product ash is a strong quadratic function of underflow rate. The increase in feed ash increases the product ash, reducing the product yield. An increase in feed ash content from 12.77% to 18.44% increased the product ash from about 8% to 10.5% with a decrease in yield from about 92 % to 82%.
A Levenberg-Marquardt gradient search algorithm has been implemented to perform a constrained optimization routine which identifies the operating conditions (Bowl speed, and underflow rate) that will provide the maximum combustible recovery for any given product quality constraints.
All the planned tasks have been carried out on schedule with the exception of the automation and control task. The difficulties encountered have been presented to ICCI and a four-month no-cost extension has been obtained.
EXECUTIVE SUMMARY
The research conducted during the past two decades on processing of coal fines has changed the description of fine coal as an environmental liability to a potential economic resource. The development of “Enhanced Gravity Techniques” is a product of research conducted during this period aimed at improving the efficiency of fine particle processing. Research funded in a large part by the state of Illinois has shown that enhanced gravity concentration has the ability to improve the rejection of both ash forming minerals and coal pyrite while maximizing the recovery of fine coal. The process efficiently treats particles of sizes of 1 mm, down to as small as 44 mm.
Based on the previous studies conducted at SIU using an enhanced gravity separator called the Falcon Concentrator, it has been observed that though the high centrifugal acceleration helps to achieve efficient separation of finer fraction (100x325 mesh), it also leads to faster settling of the coarser fraction (28x100 mesh), thus reducing the recovery of the coarser coal fraction. Hence, an innovative method was developed at SIUC with the funding provided by ICCI, to improve the separation of coarser coal particles in the Falcon Concentrator. The method uses dense medium (DM) in the Falcon concentrator thus reducing the effect of size on separation of coal particles. Previous studies with the dense medium Falcon concentrator (DMFC) showed that the separation efficiency depends on the amount of minus 325-mesh present in the feed coal. Also, the efficiency of dense medium operation depends on the feed and media characteristics. This variation in the feed and media characteristics is widely observed in coal preparation plants. Hence, to address these problems, studies were conducted using a water-injection cyclone to reduce the fines content in the feed coal. Automation potential for the DM Falcon concentrator is being evaluated to ensure an optimal product yield and grade under varying feed characteristics.
The coal samples for the present study were selected after analyzing the washability data of various Illinois coals. As the DMFC is a more efficient separator than the water-only Falcon concentrator, it is necessary to use coal samples having a “difficult to clean” coal characteristics. The -16 mesh washability data for coals from Galatia (American Coal), Marissa, Wabash and Jader Fuel were analyzed to select two coal samples for the current work. The run-of-mine (ROM) coal of Galatia mine contains about 14% (by wt.) of -16-mesh coal, having an ash content of 26%. The near gravity material (NGM) in the region 1.4 to 1.8 densities (gm/cm3) is about 3%. The presence of lower amount of NGM indicates that a moderately efficient process may be used to clean the coal. However, the presence of high amount of fines indicates that the coal requires an efficient gravity separator.
The ROM coal from Marissa mine contains about 42% (by wt.) of -16-mesh size fraction having an ash content of 25%. The NGM in the density range 1.4 to 1.8 is about 4 to 6%, requiring an efficient separator for the cleaning this coal. For the Wabash coal, the ROM contains about 10% (by wt.) -16-mesh fraction having an ash content of about 7%. The NGM in the density range 1.4 to 1.8 is around 1%. Therefore, a moderately efficient separator may be utilized for the cleaning this coal. The washability data on composite Jader Fuel classifying cyclone feed (seams 2+3) indicates that the -16 mesh sample contains about 21% ash. The NGM in the density range 1.4 to 1.8 is about 4 to 17%. Hence, this coal requires an efficient separator for cleaning.
Samples from the classifying cyclone feed were collected in separate barrels from Jader and Galatia coal preparation plants. These samples were homogenized and a representative sample was extracted for characterization of each sample. The standard deviation values for ash and total sulfur for Jader samples in different barrels were 1.10 and 0.18 respectively.
The representative samples were subjected to particle size analysis and washability tests and the resulting size and density fractions were analyzed for ash, total sulfur and heating values. The Jader coal contains about 19% ash, having 4% total sulfur. The sample contains around 15% of -400 mesh fraction having an ash and total sulfur content of about 43% and 6% respectively. The ash and total sulfur content increases with decrease in size, indicating the necessity of using an enhanced gravity separator. The American coal contains about 29% ash having 2.26% total sulfur. The amount of -400 mesh particle size fraction is about 28% with an ash content of about 69%.
The washability analysis data for the 16x325 mesh particle size fraction shows that the NGM within the specific gravity rage of 1.5 to 1.8 is in between 1 to 3% for Jader coal and 0.3 to1.5% for American coal. This finding suggests that the particles are liberated and a moderately efficient gravity separator is sufficient to ensure high separation efficiency. However, since about 35% of the material is between 48x325 mesh size fraction, it is necessary to use an enhanced gravity technique for achieving the required separation performance. The theoretical product ash and yield that can be obtained from Jader and American coals respectively are 7% and 85%, and 5% and 90%.
An AMDEL online coal slurry analyzer is being used at Southern Illinois University (SIU) to measure the feed ash content and feed solids ratio. This information is used in an online optimization and control routine for the purposes of adjusting operating parameter values, especially underflow mass flow rate or specific gravity of cut, when the assay and/or mass flow rate in the feed stream to the DMFC varies. Thus, on-stream analysis of the product and/or tailing streams by the AMDEL will provide efficient control of the product grade through manipulation of medium density or underflow rate. An Amdel ash analyzer was installed at SIU’s high-bay testing facility. About 40 coal samples having varying ash contents (10%-60%) and percent solids (8%-40%) were used to calibrate the ash analyzer. Calibration equations were developed using the following parameters.
Scatter - X-ray
Neutron count rate
Fe correction
Aeration
It was observed that calibration curves developed using samples having a narrow ash ranges (10%-30%, 30%-60%) provided better ash accuracy (± 1.25%) than using a single range (10%-60%). However, the accuracy of the measurement of percent solids did not significantly improve using a closer range of percent solids. The percent solids accuracy was ±0.75%.
Statistically designed experiments were carried using the Plackett-Burman experimental design on the Jader and American coal samples to identify the important design and operating variables that affect the performance of a water-injection cyclone. The following parameters were used for the experimental program.
Percent solids (5% - 20 % by weight)
Feed pressure (10 - 30 psi)
Vortex finder diameter (0.75 - 1.25 in)
Spigot diameter (0.6 - 0.9 in)
Truncated cone diameter (0.75 - 1.0 in)
Water injection rate (4 - 11 gpm)
The most important parameters affecting the water-injection cyclone performance are identified as the vortex finder diameter, water-injection rate and spigot diameter for both Jader and American coals. The truncated cone is a significant parameter for Jader coal at 95% confidence level. For American coal, the cone is a significant parameter at 90% confidence level. The percent solids do not have a significant effect on the partition number. In normal operation of the cyclone, the spigot diameter does not have much effect on partition number. Usually, it is expected that a larger spigot diameter will result in a higher percentage of feed material reporting to the underflow. The feed pressure significantly affects the partition number for the Jader coal, where as for the American coal, it does not have much effect. The increase in feed pressure generally increases the centrifugal force, thus providing smaller D50 values. Alias structures for the 8 run Plackett-Burman design show that main effect F (water-injection) is partially confounded with B (feed pressure) and C (vortex finder diameter) two-factor interaction. Based on the analysis of the Plackett-Burman design, further experiments according to the Box-Behnken design were conducted with the American coal varying vortex finder diameter, water-injection rate and feed pressure.
The hydrocyclone experiments without water injection show that, for the size separation cut point, the most important parameters are vortex finder diameter and percent solids. An increase in vortex finder diameter increases the cut point because of higher volumetric flow rate of solids to the overflow, which removes higher fraction of stratified material from the center of the hydrocyclone, thus increasing the cut point. The percent solids also affect the cut point, which increases with increase in percent solids. A feed containing higher percent solids, does not allow the conditions favorable to proper settling of particles, thus increasing the misplacement of coarse fraction to overflow. A lower percent solids is required for a sharper cut point. As the feed pressure increases, the centrifugal force on the particles increases, thus forcing the particles to move along the wall of the cyclone, and eventually be discharged as the underflow. Hence, more particles will report to underflow with increase in pressure, resulting in a decreased size cut point.
The results of near optimum size distribution of the overflow and underflow of the cyclone without water injection show that about 40% (corrected 19%) of -500 mesh (average size 12.5 micron) fraction present in the feed reports to underflow. Where as about 90% of 41 micron fraction reports to underflow. This data clearly shows the amount of very fine particles short-circuiting to the underflow fraction. The preliminary results of water-injection cyclone show that only about 14% (uncorrected) of -500 mesh fraction present in the feed reports to underflow.
Statistically designed experiments were carried using the Box-Behnken design on the American coal to quantify the effects of the operating parameters associated with the dense medium-based Falcon using the following operating parameters.
Feed ash (7% - 20%)
Bowl speed, (20 - 40 Hz)
Tailings underflow rate (5 lpm - 15 lpm)
The results of the dense medium Falcon tests show that feed ash and underflow rate have the greatest effect on product ash content. The curvilinear nature of the product ash curve suggests that the product ash is a strong quadratic function of underflow rate, which is supported by a relatively low t-statistic obtained for C2 term. The increase in feed ash increases the product ash with decrease in yield. An increase in feed ash content from 12.77% to 18.44% increased the product ash from about 8% to 10.5% with a decrease in yield from about 92% to 82%. Hence, the change in feed properties significantly affects the process performance. Therefore, it is necessary to automate the densese medium Falcon to maximize the product yield for a given product ash content for varying feed properties. The higher bowl speed increases the centrifugal force, thus providing accumulation of magnetite particle near the wall of bowl. This process increases the specific gravity of cut increasing the product ash and with a marginal increase in yield.
A Levenberg-Marquardt gradient search algorithm has been implemented to perform a constrained optimization routine which identifies the operating conditions (Bowl speed and underflow rate) that will provide the maximum product yield for any given product quality constraints. The hardware and software setup has been completed to adjust the Falcon variables based on the developed empirical model.