University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | 3.3.5 Differently Sized Synthetic Plumes
The first set of synthetic plumes is designed to test the appearance of a plume within the highest
elevation portion of the Desert Creek reservoir. Figure 3.15 shows the DWS of the Aneth Unit
data at 1,740 meters overlaying the Desert Creek reservoir elevation map shown in Figure 3.2.
There is no ray path coverage for the majority of the high elevation portion of the Desert Creek
reservoir. Through synthetic tests, it will be determined if plumes of varying radii are large
enough to be detected with the given event-receiver arrangement if a CO plume is residing in
2
this region.
Figure 3.15: DWS Overlay of the Aneth Unit at 1,740 meter depth (Adapted from [48])
The low velocity zone varies in size for three separate scenarios: one with a 250 meter radius
plume, another with a 500 meter radius plume, and a last test with a 1,000 meter radius plume.
All of these different radii extend 100 meters vertically to surpass the average thickness of the
Desert Creek reservoir. Each of these plume radii correspond to 500x500x100 meter,
1000x1000x100 meter, and 2000x2000x100 meter rectangular prism low velocity zones, and are
all centered at 1,800 meters easting, -400 meters northing, and 1,770 meters in depth. All the
plumes have a velocity reduction of 2 km/s from the background velocity. Plots of each different
plume size in relation to the event and station locations are given in Figure 3.16.
45 |
Virginia Tech | Figure 3.16: Locations and Extents of Differing Diameter Synthetic Plumes
To continue further investigating the ability of tomoDD to detect a low velocity zone with the
given event-receiver arrangement, a plume with an infinite size in the northing and easting
direction is created to simulate the complete saturation of the Desert Creek reservoir. A low
velocity zone is created that reduces the background velocity by 2 km/s, extends infinitely in the
northing and easting direction, and extends vertically 100 meters, from 1,720 meters to 1,820
meters in depth. These extents fully encompass the Desert Creek reservoir as well as up to 90
meters below the reservoir. The travel time calculator is used to generate travel times between
the events and receivers, all of which involve travel paths traversing the low velocity layer.
3.3.6 Different Location Synthetic Plumes
The next three synthetic conditions vary the plume location rather than plume size. Using the
Aneth Unit receiver and event locations, three different synthetic plume locations are analyzed.
The locations chosen are not necessarily where CO is expected to accumulate, but rather three
2
locations at varying positions between the events and receivers where ray path coverage is high.
Plumes in these locations are most likely to generate a velocity perturbation, however, the
differences, or lack thereof, between the different plume location simulations should demonstrate
the potential accuracy of tomography with the given event-receiver arrangement.
46 |
Virginia Tech | All the synthetic plumes are 500x500x100 meter rectangular prisms, and have a velocity
reduction of 2 km/s from the background velocity. Plume location A is centered at 600 meters
easting, -200 meters northing, and 1,600 meters in depth. Plume location B is centered at 800
meters easting, -300 meters northing, and 1,700 meters in depth. Plume location C is centered at
1,000 meters easting, -400 meters northing, and 1,800 meters in depth. The plume locations are
shown in Figure 3.17.
Figure 3.17: Plume Locations and Extents for Locations A, B, and C
3.3.7 Methods for Aneth Results
The processing of data for the Aneth Unit begins with refining the events provided. The
northern cluster of events does not contain enough events to provide reliable ray path coverage
of the desired area. The northern cluster and the event located near the monitor well are removed
to allow for an increased resolution in the MOD file. The remaining 1,166 events were separated
into 4 distinct time periods. The number of events in each time period and the dates
encompassed are given in Table 3.1.
47 |
Virginia Tech | Table 3.1: Events Contained in Each Time Period
Time Start Number of
End Date
Period Date Events
1 4/25/2008 8/31/2008 126
2 9/1/2008 10/31/2008 323
3 11/1/2008 12/31/2008 510
4 1/1/2009 3/16/2009 207
These time periods were selected to ensure that a sufficient number of events existed in each
period for tomographic analysis, considering the removal of events during the relocation process.
The Aneth Unit data are treated in the same manner as the synthetic data for double-difference
tomography. When choosing closely linked events, the maximum epicentral separation was
given an initial value of 40 meters and later narrowed in tomoDD to 20 meters. The maximum
number of neighbors was set to the number of events in that time period. The minimum number
of links was set to 1, and the maximum number of links was chosen as the number of stations,
22. The background velocity model chosen was shown in Figure 3.7, and is the same velocity
model used for the synthetic plume simulations.
A baseline velocity model for the Aneth Unit data will be assumed, performing a similar function
as the control test for the synthetic plume simulations. There are no data provided for any period
of time prior to CO injection that could be used as a baseline, however, the first time period
2
should most closely simulate the absence of a plume. Subtracting the initial time period from the
remaining time periods should eliminate some program irregularities and highlight areas of
increasing CO concentration.
2
Nodes with DWS values of 0 are removed from the velocity model during post-processing. As
part of a separate analysis, nodes with the lowest confidence values are also removed. Three
confidence thresholds are set: the bottom 25%, 50%, and 75% DWS values. Tomograms will be
shown that include only the most confident voxels: the top 25%, 50%, and 75% confidence
levels. The limited data remaining after a confidence limit is established has the potential to
result a low resolution tomogram, creating a net loss of accuracy in the resultant tomograms.
48 |
Virginia Tech | 3.4 Results and Discussion
3.4.1 No Low Velocity Zone (Control)
To better understand the results from tomoDD, an analysis of the synthetic data, designed to
simulate Aneth conditions, is presented first. The first synthetic test is one in which no low
velocity zone was placed in the synthetic travel time calculation velocity model. Due to the
layered nature of the velocity model, two tomograms of this simulation are shown. The first
tomogram narrows the velocity range to isolate the 1,789 meter to 2,000 meter layer, which will
be referred to as the 1,789 layer. Isolating this layer allows for visual representation of velocity
changes that would otherwise be too small to see with the velocity scale encompassing the entire
range of velocities present in the velocity model. The second tomogram narrows the velocity
range to isolate the 1,400 meter to 1,789 meter layer, which will be referred to as the 1,400 layer.
Tomograms for both of these layers are shown in
Figure 3.18. Isolations of the remaining layers are not performed, as artificial low velocity zones
were not placed near these regions.
Figure 3.18: Synthetic control tomograms for 1,789 layer (left) and 1,400 layer (right)
Figure 3.19 shows the difference between this control test and the original background velocity
model when interpolated in Voxler 2. There are several low and high velocity changes between
49 |
Virginia Tech | the two velocity models. The effect of these zones should be mitigated if the control is
subtracted from the synthetic tests, however, additional scrutiny should be given to any velocity
anomalies detected in these zones. The velocity scale chosen for this comparison and subsequent
comparisons is intended to highlight areas of decreasing velocity, increasing velocity, and areas
with negligible change. Regions colored in black represent a change towards a lower velocity,
while regions colored in white represent a change towards a higher velocity. The magnitude of
these changes is not indicated on the comparison tomograms.
Figure 3.19: Velocity change between the control test and the background velocity model at -300 meters
northing
The application of double-difference tomography to this data set has generated an image
showing velocity discontinuities within homogeneous velocity layers. Some velocity anomalies
are expected, due to the travel time calculation method and slight velocity model differences
between applications. The node-based calculation method of the travel time calculator will
produce travel-times with inaccuracies as a function of the number of nodes used, which could
account for these anomalies. The background velocity model used in the travel-time calculator
also differs slightly from the background velocity model used in tomoDD, in that layer
boundaries may be misplaced by several meters due to the method with which the travel-time
50 |
Virginia Tech | calculator generates a velocity model. The purpose of running a simulation with no low velocity
zone is to help reduce the impact of these velocity anomalies on future tests by providing a
baseline velocity model which can be subtracted from the remaining synthetic tests. The
difference between the “No Low Velocity Zone” test and test with a low velocity zone should
highlight the impact that low velocity zone has on the velocity model, while minimizing the
impact of experimental errors.
3.4.2 Differing Synthetic Plume Locations
The first set of synthetic simulations varies the location of the plume. The first of these locations
is Location A. A 500x500x100 meter low velocity zone is placed in Location A, centered at 600
meters easting, -200 meters northing, and 1,600 meters in depth. Three tomograms are presented
for this test. Figure 3.20 shows the isolated-layer velocity structure of the target area. Figure
3.21 shows the difference between Location A and the baseline tomogram without a low velocity
zone.
Figure 3.20: Tomograms of the 1,790 layer (left) and 1,400 layer (right) for Location A at -200 meters
northing
51 |
Virginia Tech | Figure 3.21: Difference between Location A and the control at -200 meters northing
Figure 3.21 shows a low velocity zone near the appropriate location with a slightly larger size.
The location of the synthetic plume was purposefully designed to be easy to detect due to its
placement directly between the events and geophones. With the given event-receiver
arrangement, it was expected that a low velocity zone would be detected between the events and
receivers, but that the low velocity zone would be stretched between those two locations. These
results are encouraging, as the low velocity zone was located more accurately than anticipated,
and would suggest that the given event-receiver arrangement can determine the plume location in
the easting direction better than anticipated. The low velocity zone detected here exhibits a large
amount of smearing, spanning the entire diagonal, northing extents between the events and
receivers.
The second location is Location B. A 500x500x100 meter low velocity zone is placed in
Location B, centered at 800 meters easting, -300 meters northing, and 1,700 meters in depth.
Three tomograms are presented for this test. Figure 3.22 shows the isolated-layer velocity
structure of the target area. Figure 3.23 shows the difference between Location A and the
baseline tomogram without a low velocity zone.
52 |
Virginia Tech | Location C is fairly well represented in Figure 3.24, as it is the only of the three different
locations to have a significant low velocity presence in the 1,790 layer. The easting and depth
position of this low velocity zone was slightly shifted, as could be expected, but the dimensions
are approximately correct. The presence of a low velocity zone in these tomograms suggest that
even in the presence of a changing velocity layer, the velocity reduction associated with a CO
2
plume may still be detectable. Similar to Location A and B, significant vertical smearing was
detected between the events and receivers for Location C.
3.4.3 Different Sized Synthetic Plumes
The second set of synthetic tests involves performing double-difference tomography with various
plumes of different sizes. The first low velocity zone, simulating a 250 meter radius plume
placed at the likely CO2 accumulation point, was unable to be detected. The travel times
associated with the 250 meter radius plume are identical to those of the control test, suggesting
that a plume of this size cannot be detected in its current location. The second low velocity zone,
simulating a plume with a 500 meter radius, and vertical extent of 100 meters is centered at
1,800 meters easting, -400 meters northing, and 1,770 meters in depth. Figure 3.26 shows the
isolated-layer velocity structure of the target area. Figure 3.27 shows the velocity difference
between the 500 meter radius plume and the baseline tomogram not containing a low velocity
zone.
55 |
Virginia Tech | Figure 3.31: Difference between the Desert Creek saturation test and the control at -300 meters northing
A segmented low velocity zone can be seen. The similarities between the Desert Creek
saturation simulation and the 1,000 meter plume radius simulation should be noted. The 1,000
meter plume radius simulation looks very similar to the Desert Creek saturation simulation
through the extents of the 1,000 meter plume radius. Outside of the 1,000 meter radius, the
tomogram from the Desert Creek saturation simulation shows another low velocity segment,
while the 1,000 meter radius simulation does not. Both contain an area with seemingly little
velocity change where a low velocity is expected. A lack of ray path coverage here due to ray
bending could explain why the velocity remains relatively unchanged.
The synthetic tests help to highlight several key factors when analyzing the Aneth Unit data with
the same event-receiver arrangement and background velocity model. The first of these factors
is the importance of a control or baseline which can be subtracted from the other simulations to
isolate velocity changes. There are low and high velocity zones in every synthetic test that are a
direct result of improper velocity model adjustments, due to incorrect travel times or incorrect
tomoDD adjustments. If not for a baseline model, these low and high velocity zones could be
assumed to represent something geophysically significant. The second factor is the ability to
detect a plume located on the boundary between low velocity and high velocity geologic
formations. Location C, the 1,000 meter radius plume, and the Desert Creek saturation tests all
showed low velocity zones of the proper sizes in the approximate areas of plume placement. It
59 |
Virginia Tech | should therefore not be assumed that the absence of a low velocity zone in the Aneth Unit data
are a result of a low velocity zone “getting lost” among changing geologic strata. The third
factor to be considered is potential stretching or shifting of low velocity zones. Location A,
Location B, Location C, the 1,000 meter radius plume, and the Desert Creek saturation
simulations all had some form of stretching or shifting in both the vertical and lateral
dimensions. Very little confidence can be placed in the location or quantity of sequestered CO
2
based on the location and extent of a low velocity zone in this region. The fourth factor to
consider is the size of the plume. Neither the 250 meter radius plume nor 500 meter radius
plume were able to be detected with the given event-receiver arrangement. This is important to
consider because the highest elevation of the Desert Creek reservoir lies where these two plumes
were centered, and buoyantly driven flow could lead to CO to accumulate in this region. A
2
plume residing in this location would likely need to be of a diameter greater than 500 meters to
be accurately detected.
3.4.4 Aneth Unit DWS Analysis
The Aneth data are presented in the same way as the synthetic data. In order to perform a time-
lapse analysis, the data were divided into four time periods, which are presented sequentially.
Increased ray path coverage should lead to a higher accuracy characterization of the reservoir, so
only tomograms with high DWS values at the Desert Creek reservoir depth will be presented.
Figure 3.32 shows the DWS values at northings of -500 meters, -300 meters, and -50 meters.
60 |
Virginia Tech | Figure 3.32: DWS levels at -500 meters northing (top left), -300 meters northing (top right), and -50 meters
northing (bottom right)
These tomograms are important for several reasons. First, they represent the northing extents of
good Desert Creek reservoir ray path coverage. No information should be inferred from velocity
tomograms representing layers outside of these extents. Second, they provide an easting range in
which the ray path coverage is acceptable. The easting range varies with the northing slice.
Easting values, from which a confidence in results can roughly be assumed, range from 200
meters to 1,500 meters. All reported tomograms will be taken at the -300 meter northing slice
due to its high DWS values and large range of confidence.
3.4.5 Aneth Unit Analysis by Layer
Beginning with the first time period, tomograms are presented of the lower velocity layer that
underlies the Desert Creek reservoir, Figure 3.33, and the upper velocity layer that contains the
Desert Creek reservoir, Figure 3.34, referred to as the 1,790 layer and 1,400 layer respectively.
61 |
Virginia Tech | Figure 3.33: Tomograms of the 1,790 layer for Time 1 (top left), Time 2 (top right), Time 3 (bottom left), and
Time 4 (bottom right) at -300 meters northing
A low velocity zone can be seen, in these tomograms, that correlates well to the DWS for this
region. The easting range with confidence at this northing coordinate is roughly 400 meters to
1,200 meters, which is the general region exhibiting a low velocity for all time periods. This low
velocity zone, while not located in the Desert Creek reservoir, could be the result of stretching of
the plume. If this low velocity zone is a result of CO2 concentrations in the region, then there
also appears to be a reduction in the levels of CO2 from Time 2 to Time 3. The other time
periods follow an expected trend of decreasing velocity as the quantity of injected CO2
increases, with Time 1 showing the highest velocity, and Time 4 showing the lowest velocity.
62 |
Virginia Tech | No absolute conclusion can be made about which, if any, low velocity zone represents a CO
2
plume and to what extent. However, the effect that different plumes have on the velocity model
can be seen through the use of synthetic data tests. The synthetic simulation that most closely
resembles the Aneth Unit results is the saturation of the Desert Creek reservoir. This simulation
makes the assumption that from a depth of 1,720 meters to 1,820 meters, the velocity model is 2
km/s lower than the background for every northing and easting. The result of subtracting the
synthetic control from the Desert Creek saturation test at a northing of -300 meters is shown in
Figure 3.39. Subtracting the Aneth Time 3 results from Time 1 results at a northing of -300
meters is shown in Figure 3.40, and subtracting the Aneth Time 3 results from the synthetic
control is shown in Figure 3.41.
1
2
5
4 3
6
Figure 3.39: Difference between Desert Creek saturation test and the synthetic control at -300 meters
northing with low velocity zones highlighted
68 |
Virginia Tech | There are six boxes highlighting the low velocity areas of the Desert Creek saturation synthetic
simulation. All six of these low velocity zones are represented in Figure 3.40 and four of them
are present in Figure 3.41. These low velocity zones are not the same size, but appear in the
same location. Other synthetic tests show low velocity zones in similar locations, but only the
Desert Creek saturation test can account for all the low velocity zones. The likely explanation
for this is not the concentration of CO or necessarily its placement, but rather that all ray paths
2
are traversing the plume. As a result, it would be difficult to determine the amount of CO
2
sequestered, or preciesely where it is located, because the plume itself is not being analyzed but
rather the behavior of tomoDD under certain synthetic conditions. If the tomogram is very
similar to the Desert Creek saturation synthetic test, it may then be concluded that the Desert
Creek reservoir is experiencing some form of uniform CO inundation throughout the 300 meters
2
to 1,500 meters easting range and -600 meters northing to -50 meters northing range.
It may also be concluded that no significant leakage is occurring within the observable region.
The synthetic simulation results for Location A, Location B, and Location C, all show clear,
unique, low velocity zones above the Desert Creek reservoir where an artificial plume was
placed, yet no clear low velocity zone is shown in any of the Aneth Unit time periods that would
indicate there is an accumulation of CO where there should not be.
2
3.4.7 Aneth Unit DWS Threshold
Lastly, a DWS threshold analysis is performed. The acceptable DWS threshold is set to the top
75% for the first test, top 50% for the second test, and top 25% for the third test. In each test, the
first time period is used as a baseline and is subtracted from the remaining time periods. The
results for time period 2 are given in Figure 3.42, time period 3 in Figure 3.43, and time period 4
in Figure 3.44. Time period 2 is reported at the -550 meter northing slice, while time periods 3
and 4 are reported at the -300 meter slice.
70 |
Virginia Tech | All three time periods show a very distinct trend of the target low velocity zone growing smaller
as the DWS threshold is decreased. The DWS threshold test containing the data with the highest
confidence shows a low velocity zone stretching from approximately 1,700 meters to 1,850
meters in depth, encompassing the entirety of the Desert Creek reservoir. It should also be noted
that this low velocity zone extends across the entire easting extent, and continues for every
northing value at the same depth. Average velocities for each time period and DWS threshold
are reported from the 1,720 meter to 1,790 meter depth in Table 3.2. This is the assumed depth
of the Desert Creek reservoir.
Table 3.2: Average velocities at different DWS thresholds and time periods
Top 25% DWS Top 50% DWS Top 75% DWS
Time
Average Velocity Average Velocity Average Velocity
Period
(km/s) (km/s) (km/s)
1 5.222 5.220 5.219
2 5.222 5.219 5.218
3 5.217 5.216 5.216
4 5.226 5.223 5.221
No trend exists for average velocity per time period. Any trend in velocity change could be used
to indicate plume behavior, such as migration or an increasing concentration. It is therefore
concluded that any migration of CO or small increases in CO concentration cannot be detected
2 2
with the current resolution of the CO plume.
2
3.4.8 Discussion of Results
There are two explanations for the cause of a low velocity zone in the reported tomograms,
neglecting uncertainties associated with this tomography application. The first is the presence of
a CO plume within the Desert Creek reservoir. The second is boundary changes throughout the
2
velocity model, most notably the 0.88 km/s change from the 1,400 layer to the 1,789 layer in the
vicinity of the Desert Creek reservoir.
74 |
Virginia Tech | Low velocity zones are only expected to be caused by CO2 concentrations and geologic
anomalies not included in the initial one-dimensional velocity model. Repeated small low
velocity zones are found throughout nearly every Aneth Unit test near the approximate 1,720
meter to 1,780 meter depth of the Desert Creek reservoir. Figures 3.35 through 3.37 show a
migration of one of these low velocity zones, which is potential behavior of a CO2 plume. The
direction of this plume’s migration, however, is towards a lower elevation, which is unlikely
given the buoyancy of CO2. The arrangement of small low velocity zones in the Aneth Unit data
resemble the arrangement of small low velocity zones in the synthetic Desert Creek saturation
simulation, shown in Figures 3.39 through 3.41. Only the Desert Creek saturation simulation
contains low velocity zones in the same locations as the Aneth Unit data while remaining the
same approximate size. Given the exaggerated drop in velocity assigned to the Desert Creek
saturation simulation, these similar results are likely due to ray paths being forced to travel
through the plume rather than an indication of the amount of CO contained in this region. More
2
evidence to suggest the existence of a CO plume is the appearance of a large, distinct low
2
velocity zone surrounding the Desert Creek reservoir when confidence limits are applied to the
data. Figures 3.42 through 3.44 show an increasingly clear low velocity zone 1,700 meter to
1,850 meter depth. Removing voxels with a low DWS values creates an interpolation region
showing a consistent low velocity that continues through the northing extents of the observable
region. This low velocity zone does not show the smearing along the ray path direction
exhibited in the synthetic tests varying plume location, but neither does the Desert Creek
saturation simulation or the 1,000 meter radius plume simulation, suggesting that the smearing
was related to plume positioning precisely between the events and receivers.
The positioning of both the Desert Creek reservoir and the event cluster near a velocity layer
boundary in the velocity model increases the difficulty with which anomalous stretching of
velocity layers and actual low velocity zones can be differentiated. Velocity layer boundaries
exist at depths of 1,050 meters, 1,400 meters, 1,790 meters, and 2,050 meters. Low or high
velocity anomalies are present at each of these depths in every test of both Aneth Unit data and
synthetic data. However, low velocity anomalies at the 1,790 meter depth velocity boundary are
small in all tests that do not include a plume in the immediate vicinity. While small velocity
perturbations may be expected as a result of velocity layer boundaries, the low velocity region
75 |
Virginia Tech | detected in the high DWS Aneth Unit data tests are too large to be the result of velocity layer
boundaries alone.
Determining the degree to which MVA can be performed on this reservoir using double-
difference seismic tomography is a major objective of this project. The three components of
MVA are defined as follows:
- Monitoring - monitor both the location and impact of sequestered CO .
2
- Verifying - verify movement of CO and ensure that sequestered CO is not
2 2
permeating the sealing mechanism, or migrating to an unsealed area.
- Accounting - account for the amount of injected CO , by comparing it to the amount
2
of CO estimated to be in place, through the chosen monitoring method.
2
The results obtained show potential for monitoring applications. The location of a plume can be
determined in two ways. The first is through establishing confidence thresholds for voxels that
will exclude data points with low ray path coverage. The second is by comparing supposed
plume locations with actual data to find results that match in velocity zone placement.
Verification of CO in the Aneth Unit is a substantial challenge given the event-receiver
2
arrangement. None of the tomograms show a low velocity zone confined to the Desert Creek
reservoir. Nearly all of the tomograms show significant stretching of low velocity zones up to
hundreds of meters in either the positive or negative depth direction. The exaggerated size of
these low velocity zones makes assessing leakage and migration very difficult without significant
changes to the plume location.
Accounting for the amount of injected CO requires both accurate imaging of the extents of a
2
CO plume and the concentrations therein. The plume located in the Desert Creek reservoir
2
suffers from two flaws that prevent accounting for the amount of CO it contains. The first flaw
2
is the same exaggerated size of the plume that prevented accurate verification. The effect of a
smeared plume size can be mitigated to an extent because the same concentration of CO is
2
likely spread over a larger area. A reduced concentration over a larger area may yield the same
amount of CO as a normal concentration over a smaller area. The second flaw is the lateral
2
76 |
Virginia Tech | Chapter 4: Summary of Results and Conclusion
Double-difference tomography was performed on a carbon sequestration project in the Aneth
Unit of the Aneth Oil Field in southeast Utah. Location data for 1,211 passively induced seismic
events was provided as well as travel times between the seismic events and a vertical borehole
arrangement of 22 geophones. The provided data were split into four sequential time periods for
a time-lapse tomographic analysis of subsurface conditions. Simulations using synthetic data
were also conducted to understand the behavior and sensitivity of tomoDD as well as to compare
expected results with actual results.
Locating a CO plume in the Aneth Unit is made difficult due to a narrow ray path distribution,
2
generated by a borehole array of geophones and single cluster of seismic events. A low velocity
zone was detected, however, in the Desert Creek reservoir among tomograms excluding low
DWS values. This low velocity zone could be the result of an increasing CO concentration
2
within the Desert Creek reservoir. It is distributed across the extents of the observable region,
which would be expected in a region experiencing injection from multiple wells. This low
velocity zone begins to deteriorate with decreased DWS, highlighting the importance of ray path
coverage.
Also reinforcing the existence of a plume within the Desert Creek reservoir is the comparison of
a synthetic test simulating the complete saturation of the Desert Creek reservoir with the results
obtained from Aneth Unit data. The synthetic simulation shows low velocity zones in the same
place and of the approximate size as in the Aneth Unit data. These low velocity zones
themselves may not represent CO plumes, but the behavior associated with a Desert Creek
2
saturation synthetic test being very similar to the Aneth Unit results reinforces the conclusion
that there may have been a uniformly distributed CO inundation across the reservoir detected by
2
double-difference tomography.
With the event and receiver arrangements, double-difference seismic tomography is a poor
choice as a method of MVA for the Aneth Unit. The resolution of CO detected is too low to
2
79 |
Virginia Tech | make informed verification or accounting decisions. Monitoring is possible to a limited extent,
but precise and accurate locations for a CO plume cannot be determined. A more
2
comprehensive event-receiver arrangement is required for accurate tomographic imaging of the
Desert Creek reservoir.
4.1 Possible Sources of Error
The possible sources of error within the Aneth Unit data include the event-receiver arrangement,
the methods by which the event locations and travel times were determined, and a one-
dimensional initial velocity model. The event-receiver arrangement in the Aneth Unit results in
an undesirable ray path distribution in the Desert Creek reservoir. It is unknown the extent to
which the event-receiver arrangement affects the results, but the accuracy of tomography
increases with ray paths traversing the body-of-interest from as many angles as possible. The
events recorded at the Aneth Unit formed two relatively narrow clusters, forcing all ray paths
along two narrow paths. With the northern cluster having been removed in data pre-processing
due to an insufficient number of events, this reduces the ray path coverage crossing the Desert
Creek reservoir even further, as shown in Figure 4.1 and Figure 4.2.
80 |
Virginia Tech | It is difficult to obtain accurate, passive, seismic event locations. Double-difference tomography
should assist in precise relocation of events, but if arrival times were determined by inaccurately
located events, then the relocation may not improve event locations to within an acceptable level.
The accuracy of event location is a function of the receiver arrangement. The single column of
events located 200 meters to 1,100 meters above the events may provide a better estimation of
depth than a varied surface array, but likely results in a poor characterization of the events’
lateral positions.
The accuracy of the initial velocity model affects the accuracy of the final velocity model. It is
better to use the one-dimensional velocity model provided, than to assume a constant background
velocity, but this velocity model could be further improved by the use of an accurate two or
three-dimensional velocity model. In the current velocity model there is assumed to be no lateral
variation in geology. It is unknown the extent to which lateral heterogeneities in the geologic
structure would affect the results of tomography.
A possible source of error in the synthetic data is the means by which travel times were
calculated. The travel time calculator is node-based, meaning each event and receiver will have
to draw a travel path to the nearest node before calculating an event-receiver travel path. This
additional travel distance is greatest for events furthest from a node. Creating an event-node or
station-node travel path artificially increases the total travel time. For example, two events
equidistant from a receiver in a constant velocity model should have identical travel times;
however, if one event is closer to a node than another, the closer event will have a shorter travel
time. This has the potential to be especially problematic when applying double-difference
tomography, as the assumption will be made that the difference in travel times of two closely
spaced events is due to a velocity anomaly in the immediate event region rather than an extended
node-event travel path. The effect this has on travel paths is a function of the number of nodes
used, which must be kept small due to processing restrictions.
82 |
Virginia Tech | 4.2 Future Work
Several additional steps can be taken to improve the accuracy of tomographic imaging of carbon
sequestration in the Aneth Unit. The first step is the improvement of the event-receiver
arrangement. While drilling more boreholes for increased ray path coverage would be
expensive, the addition of a surface geophone array, when used in conjunction with the current
borehole array could provide significantly better ray path coverage.
Another step to improving the results that can be taken is to improve the synthetic travel-time
calculation method. The current method of calculating travel times is either inaccurate or
resource intensive. A progressive travel-time calculator that does not need to individually check
every segment combination in a node-based grid could provide better results faster.
A third step is imposing a constraint on the velocity model that would prevent changes from
being made to certain regions. Within the Aneth Unit, the only major change occurring during
the experiment is within the Desert Creek reservoir. If the velocity model were not allowed to
change from one time period to the next except within the immediate vicinity of the Desert Creek
reservoir, all changes reflected in travel times would be appropriately applied to this region.
While verification of CO leakage could be affected by this approach, if CO concentrations are
2 2
detected in a region immediately above the Desert Creek reservoir, the velocity model could be
later extended to account for the extent of CO leakage.
2
83 |
Virginia Tech | * OBSCT: min # of obs/pair for network data (0= no clustering)
* OBSCC OBSCT CC_format
0 0 2
*
*--- solution control:
* ISTART: 1 = from single source; 2 = from network sources
* ISOLV: 1 = SVD, 2=lsqr
* NSET: number of sets of iteration with specifications following
* ISTART ISOLV NSET weight1 weight weight3 air_depth
2 2 16 100 100 100 0
* i3D delt1 ndip iskip scale1 scale2 iusep iuses iuseq
2 1 9 1 1.0 2.0 1 1 0
* invdel ifixl xfac tlim nitpb(1) nitpb(2) stepl
1 0 1.3 0.002 10 10 1.0
* lat_Orig lon_Orig Z_Orig iorig rota
0.00001 0.00001 0 1 0
*
*--- data weighting and re-weighting:
* NITER: last iteration to used the following weights
* WTCCP, WTCCS: weight cross P, S
* WTCTP, WTCTS: weight catalog P, S
* WRCC, WRCT: residual threshold in sec for cross, catalog data
* WDCC, WDCT: max dist [km] between cross, catalog linked pairs
* DAMP: damping (for lsqr only)
* --- CROSS DATA ----- ----CATALOG DATA ----
* NITER WTCCP WTCCS WRCC WDCC WTCTP WTCTS WRCT WDCT WTCD DAMP JOINT THRES
3 0.01 0.01 -9 -9 1.0 0 -9 -9 -9 80 1 2
3 0.01 0.01 -9 -9 1.0 0 -9 -9 -9 80 0 2
3 0.01 0.01 -9 -9 1.0 0 -9 10 10 80 1 2
3 0.01 0.01 -9 -9 1.0 0 -9 10 10 80 0 2
3 0.01 0.01 -9 -9 1.0 0 6 10 10 80 1 2
3 0.01 0.01 -9 -9 1.0 0 6 10 10 80 0 2
3 0.01 0.01 -9 -9 1.0 0 6 10 10 80 1 2
3 0.01 0.01 -9 -9 1.0 0 6 10 10 80 0 2
3 1 0.5 -9 -9 1.0 0 6 10 10 80 1 2
3 1 0.5 -9 -9 1.0 0 6 5 10 80 0 2
3 1 0.5 -9 -9 1.0 0 6 5 5 80 1 2
3 1 0.5 -9 -9 1.0 0 6 5 5 80 0 2
3 1 0.5 8 -9 1.0 0 6 5 5 80 1 2
3 1 0.5 8 -9 1.0 0 6 2 2 80 0 2
3 1 0.5 8 -9 1.0 0 6 2 2 80 1 2
3 1 0.5 8 -9 1.0 0 6 2 2 80 0 2
*
*--- event selection:
* CID: cluster to be relocated (0 = all)
* ID: cuspids of event to be relocated (8 per line)
* CID
0
* ID
85 |
Virginia Tech | LOW TEMPERATURE DRYING OF ULTRAFINE COAL
by
Chad L. Freeland
ABSTRACT
A new dewatering technology, called low temperature drying, has been developed to
remove water from ultrafine (minus 325 mesh) coal particles. The process subjects partially
dewatered solids to intense mechanical shearing in the presence of unsaturated air. Theoretical
analysis of the thermodynamic properties of water indicates residual surface moisture should
spontaneously evaporate under these conditions. This is contingent on the large surface area of
these fine particles being adequately exposed to an unsaturated stream of air. To demonstrate this
process, three dispersion methods were selected for bench-scale testing; the static breaker, air jet
conveyor, and centrifugal fan. Each of these devices was chosen for its ability to fully disperse
and pneumatically convey the feed cake. The moisture content of the feed cake, and the
temperature and relative humidity of the process air were the key parameters that most
significantly affected dryer performance. Of the three methods tested, the centrifugal fan
produced the best results. The fan was capable of handling feeds as wet as 21.5% and
consistently dried the coal fines below 2% moisture. The cost of the air and heat required to
provide good drying performance was modeled to explore the practicality of the drying process.
Modeling was accomplished by modifying equations developed for thermal dryers. The
modeling results indicate, if good exposure of the fine particle surface area is achieved, dryers
operating with either heated or unheated (ambient) air can be used for drying ultrafine coal
ii |
Virginia Tech | 1.0 INTRODUCTION
1.1 Preamble
According to the Energy Information Administration (EIA 2010), coal is used to generate
nearly half of the electrical power consumed in the United States (Figure 1-1). However, in
recent years, the environmental impact of coal mining, processing, and power generation has
come under increasing scrutiny. There are even initiatives pushing for the burning of coal
burning to be stopped entirely. Yet, coal remains America’s most affordable source of electricity
despite recent advancements in renewable energy and increasing interest in nuclear power,
(Wald 2009). With 27% of the world’s recoverable reserves, the United States has the largest
coal reserves of any nation on earth (EIA 2009). Such a plentiful and cost effective resource is
difficult to ignore and warrants great efforts to mitigate its environmental shortcomings.
One of the most prominent environmental issues facing the coal industry today is slurry
impoundments and the water quality and flooding risks associated with them. Impoundments are
Petroleum &
others
Hydroelectric 4.5%
6.0%
Nuclear
19.6% Coal
48.5%
Natural Gas
21.3%
Figure 1-1. Electric power generation by energy source in the U.S. (EIA, 2010).
1 |
Virginia Tech | manmade ponds or lakes used to settle out the fine waste produced by coal preparation plants. It
is possible for this waste to seep into the surrounding water table, potentially compromising the
drinking water of local communities. There is also the more remote risk of this material escaping
en masse in the form of a damn failure, like the one at Buffalo Creek WV in 1972, or a blowout
through old works, as occurred in Martin County KY in 2000.
Ultrafine coal (minus 325 mesh) is one of the primary components of the fine waste
found in impoundments. Ultrafine coal is discarded because the high moisture content such
material makes this size category uneconomical to recover (DOE 2009). The loss of the minus
325 mesh material is especially tragic because coal particles that small are extremely well
liberated. This allows lower ash values that would be impossible with larger particle sizes. If this
material could be recovered and dewatered, impoundments could be made smaller and cleaner
while simultaneously increasing coal production and lowering the ash content of processed coal.
In a conventional coal preparation plant all material under 100 mesh is typically
recovered by froth flotation and sent on to screenbowl centrifuges for dewatering. This process
produces a clean dewatered product, but as screenbowls remove water from the coal
approximately half of the ultrafine feed solids are discarded (Luttrell et al., 2006). Some of the
newest preparation plants pre-empt this loss by desliming the flotation feed with 6-inch diameter
cyclones. This cuts out the minus 325 mesh material so that flotation capacity is not wasted on
coal that will be discarded. This practice of desliming is even applied when fine coal is remined
from closed impoundments. In effect, high-grading this supply of potentially marketable coal
(Manlapig et al. 2001).
Processing plants that produce coal for the metallurgical market work somewhat
differently. The amount of material lost from a screenbowl typically is not acceptable for this
2 |
Virginia Tech | market, since coking coal commands a price that can be substantially higher than that of steam
coal. Instead, vacuum filters may be employed to achieve higher recoveries of the ultrafine coal.
This process gives near perfect recoveries, but also produces much higher moistures.
Unfortunately, while coking coal contracts pay a premium, they also typically include strict
moisture limits. Therefore, wherever possible, the wet filter cake is sent on to a thermal dryer
where the water is forcibly boiled away via intense heating. Thermal drying can produce bone
dry coal, but this process also presents a unique set of challenges. Exposing coal to high
temperatures without combustion causes the coal to release volatile organic compounds (VOCs)
which are an environmental concern. In the process of generating their heat, thermal dryers can
also emit SOx and NOx. As potentially large sources of air pollution, new thermal dryers can be
extremely difficult to permit.
1.2 Low Temperature Drying Process
In light of the problems associated with removing water from fine coal, a new process
known as “low temperature drying” has been under development at Virginia Tech. Low
temperature drying dewaters coal by taking advantage of liquid water’s natural propensity to
evaporate. A liquid’s evaporation rate is directly related to its surface area and water exposed to
air at less than 100% relative humidity is always evaporating, albeit slowly. Therefore, if a fixed
mass of water is spread out over a large surface area and the air above that water is continually
replaced to keep the air unsaturated, that water would evaporate very quickly.
Minus 325 mesh particles have an exceptionally large surface area per unit mass (e.g.,
108.6 m2/kg or 530 ft2/lb). So, if the water that remains with the ultrafines after mechanical
dewatering was to be spread evenly over that surface and the air around the water were
continuously replaced, that water would swiftly evaporate. This is exactly how low temperature
3 |
Virginia Tech | drying works. This process seeks to expose the entire surface of the ultrafines by using a
shearing device to break up the filter cake into its constituent particles and prevents saturation of
the air by pneumatically conveying the particles through the shearer.
1.3 Objectives
The goal of this project is to take slurry composed primarily of minus 325 mesh material
and produce a clean product containing less than 10% moisture. The slurry may originate from a
plant’s deslime cyclones or an actual impoundment. Two cleaning paths and three drying devices
were tested. The first path was cleaning by oil agglomeration followed by centrifugal
dewatering. The second path was cleaning by froth flotation and dewatering by vacuum
filtration. The three dryers tested were the static breaker, air jet conveyor and centrifugal fan. All
three drying devices utilized the newly developed concept of low temperature drying where
surface moisture is absorbed and carried away by ambient or slightly heated air. This project also
explored the economic feasibility of low temperature drying via mathematical modeling of the
process.
4 |
Virginia Tech | 2.0 LITERATURE REVIEW
2.1 Fine Coal Cleaning
When attempting to dewatering fines, it is important to understand the preceding steps.
Several factors determined by cleaning influence the final product moisture. These include, but
are not limited to: percent solids, what sizes report to the clean product, and whether those sizes
are ash or coal.
2.1.1 Froth Flotation
Froth flotation is the most widespread method to clean minus 100 mesh (minus 150 µm)
coal particles and some sort of flotation circuit is present in all but the oldest of preparation
plants (Aplan 1991). Flotation is a process where hydrophobic particles, like coal, are separated
from hydrophilic particles such as those clays or silicas that constitute ash. This is accomplished
by sparging air through dilute slurry in a mixing cell. The hydrophobic coal particles stick to the
air bubbles and rise to the top forming a froth while the ash particles remain in solution with the
water. The coal rich froth overflows the cell edges while the ash laden water is drained from the
bottom of the cell and into the thickener.
Although conventional flotation sends the majority of the water to the thickener, there is
still a significant volume of water that is required to form the froth that carries the clean coal
particles away. This leads to an effect called hydraulic entrainment where the finer ash particles,
especially clays, stay with the water wherever that water goes. Thus the ash split is almost
identical to the water split. This problem can be rectified by desliming, multistage flotation or
column flotation. Desliming is accomplished by sending the fine coal stream through 6-inch
5 |
Virginia Tech | diameter cyclones, cutting out the minus 325 mesh (minus 44 µm) material. This is the size
fraction that contains the most entrainable clays, has the highest feed ash, and is the hardest to
dewater (Manlapig et al. 2001). Multistage flotation, the solution favored by the minerals
industry, is to send the product froth through one or more “cleaner” flotation cells. This starts an
infinite dilution loop where less and less waste reports to the product stream. Column flotation
takes a different approach by introducing “wash water” into the froth section of the cell. The
wash water flows down through the thin films of the foam displacing the original ash bearing
water with clean water. The low unit value of coal makes multistage flotation cost prohibitive,
leaving desliming or column flotation as the favored industry solutions (Davis et al. 1995).
2.1.2 Oil Agglomeration
Although oil agglomeration is far less commonly used than froth flotation, agglomeration
does represent an alternative method of effectively cleaning fine coal while at the same time
lowering its moisture content. As with flotation, oil agglomeration depends on the
hydrophobicity of coal particles to differentiate them from the hydrophyllic ash particles. When
oil is mixed with coal slurry, oil droplets stick to coal particles and oil coated coal particles stick
to each other, eventually forming agglomerates of various shapes and sizes. Small amounts of
water are trapped within these agglomerates so hydraulic entrainment still occurs. However, as
moisture levels are much lower in the agglomerates than in flotation froth, so is the level of
entrainment. Traditionally, the more oil used, the larger the agglomerates and the lower the
moisture content.
Oil agglomeration made its debut during World War I as the Trent process where coal
slurry and approximately 30% of the raw coal’s weight in oil were mixed together for 15 minutes
(Capes 1991). By the end of the mixing time, large nodules of clean coal had formed which
6 |
Virginia Tech | could then be screened out from the ash bearing water. The Convertol process, developed by the
West Germans in the 1950s, was able to reduce the amount of oil and mix time required. This
was achieved through the use of high sheer mills that increased the collision rate between oil and
coal. Another advantage Convertol had over Trent was the use of a screen centrifuge to separate
the agglomerate from the bulk water instead of a static screen (Nicol and Swanson 1980). This
lead to a scant difference in product moisture under normal operating conditions but when given
exceptionally fine feed, Convertol vastly outperforms Trent (Table 2.1-1).
Many other techniques followed. Some of the more notable methods are the Spherical
Agglomeration process developed by the National Research Council of Canada, the Shell
Pelletizing Separator, the Olifloc process and the Selective Agglomeration process from BHP.
Each of these processes had their own strengths but all possessed the same weakness; the high
price of oil. In 1991, Capes noted that oil made up half to two thirds of the product cost for these
methods. Between that year and 2008, the price for a barrel of oil had more than quadrupled
(EIA 2008). When these processes were new, controlling oil costs was important. Now
controlling oil costs is absolutely critical.
One potential solution was utilized in the Otisca process that used pentane, a light
Table 2.1-1. Comparison of Trent and Convertol processes ( Nicol and Swanson 1980).
Used under fair use guidelines, 2010
Property Trent Convertol
Oil (% wt of raw coal) 30 10-15
Mix Time (Min) 15 0.25-0.5
Typical Moisture (% wt) 8-12 10-15
Fine Moisture (% wt) 40 20-25
7 |
Virginia Tech | paraffin, as an agglomerating agent. The advantage of using pentane over other hydrocarbons is
pentane has the exceptionally low boiling point of 36.1ºC (96.7ºF). This allows comparatively
large amounts be used because pentane can be recovered later through mild heating and
condensation. Unfortunately, this process was targeted at producing ultraclean (<1% ash) coal
water mixtures called “CWM” (Keller 1995). As the name would suggest, a CWM is coal slurry
containing 50-80% solids, which could be pumped and burned in the same manner as fuel oil
(Yoon et al., 1991). When the CWM market soured in 1989, Otisca Ltd. ceased production,
never to resume (Keller 1995).
2.2 Fine Coal Dewatering
Most coal cleaning operations are wet operations and thus their products are water laden.
The water left on coal after processing is known as “free” or “surface” moisture but there are
also coals that contain “inherent” or “chemically bound” moisture that is inside the coal or part
of the coal structure. Either of these is undesirable as the vaporization of water steals heat energy
during combustion, raises transportation costs, and creates handling problems when wet coal
freezes in a stockpile or railcar. Only thermal drying can remove inherent moisture, but there are
a number of steps that can be taken to remove surface water. For the larger size fractions,
dewatering is accomplished with relative ease by high speed vibrators, basket centrifuges, and
screenbowl centrifuges, as show in . But below 44 µm (325 mesh) it becomes extremely
challenging to keep moisture to a reasonable level. These ultrafines are so difficult and by
extension expensive to dewater, that it is a popular practice to simply discard them without
processing (Manlapig et al. 2001).
There are a number of different ways to overcome this problem. In the United States, the
most popular way to dewater fines is the screenbowl centrifuge. This method strikes a balance
8 |
Virginia Tech | Figure 2-1 Common dewatering equipment (Parenkh et al. 1991)
Used under fair use guidelines, 2010
between moisture and recovery, maintaining reasonable moistures by discarding a fraction of the
ultrafines (Luttrell et al., 2006). Vacuum filters achieve high recoveries but suffer from
correspondingly higher moistures. Hyperbaric filters try to use higher than atmospheric pressures
to attain low moistures while maintaining the same recovery as vacuum filters. Thermal dryers
can achieve both near perfect recovery and a bone dry product but can also be a significant
source of air pollution and are extremely difficult to permit.
2.2.1 Centrifugation
Screenbowl centrifuges were introduced to the coal industry in 1969 and, as the name
implies, they are a hybrid of screen and bowl centrifuges. The feed end of the device is a solid
bowl where the enhanced gravity of the centrifuge pulls any solids in the slurry to the wall.
Concurrently, the clarified water escapes over a weir in the rear of the unit. A screw conveyor
running the length of centrifuge turns slightly faster than the bowl pulling solids along the wall
9 |
Virginia Tech | Figure 2-2. Cutaway view of a screenbowl centrifuge.
and towards the discharge. The walls of the centrifuge slope to the center as they approach the
discharge, thus the screw drags material up and out of the bulk water. In a solid bowl centrifuge
dewatering would be complete at this point but a screenbowl incorporates a screen section just
prior to the discharge. The screen permits further dewatering of the cake and allows screenbowls
to achieve approximately 12% moisture on a typical fine feed (Luttrell et al., 2006). Any solids
lost through the screen section are recycled back into the original feed. A diagram of a
screenbowl is shown in Figure 2-2.
Unfortunately, this performance comes at the expense of recovery, as a screenbowl can
only achieve 80-90% recovery on most fine feeds with approximately 50% of the minus 44 µm
(325 mesh) material lost with the main effluent (Luttrell et al, 2006). When fed straight
ultrafines with an approximate mean size of 25 µm, as might be found in a pond reclaim
operation, recovery can drop to 65% while cake moisture climbs to 30% (Parekh et al., 1999).
That is both a poor recovery and a high moisture making screenbowl centrifuges an inappropriate
choice for dewatering feeds containing any substantial amount of ultrafine material.
10 |
Virginia Tech | off the filter. The more popular disc filter operates in the same manner, but instead of a drum
with the filter cloth covering its perimeter, the disc filter is made up of a series of vertical discs
with the filter cloth on their sides.
Vacuum filters enjoy near perfect recoveries (99%), but by capturing all the minus 44 µm
(325 mesh) particles, they also capture a substantial amount of excess water. Depending on the
size distribution of the slurry, vacuum filters can produce moistures anywhere between 15 and
30% (Aplan 1991). With a feed of straight ultrafines (approximate mean size of 25 µm), the cake
moisture approaches 30%. This is still a very high moisture and no improvement over the
screenbowl. Fortunately, vacuum filters respond well to dewatering aids, and with the addition of
surfactants, the same material can be dewatered to as low as 24% moisture (Parekh et al., 1999).
2.2.3 Hyperbaric Filtration
Hyperbaric filters resemble vacuum filters that have been enclosed within pressure
vessels. This is done so that they can achieve pressure differences across the cake that exceeds 1
atmosphere. Due to their high costs these units are quite rare in the coal industry, but they also
achieve the best dewatering results. When fed straight ultrafines (approximate mean size of 25
µm), hyperbaric filters have achieved moistures below 24%. With the addition of cationic
surfactants, product moistures can approach 19%. These moistures are attained while still
maintaining the 99% recovery of a vacuum filter (Parekh et al., 1999).
2.3 Thermal Drying
Thermal drying was once the process engineer’s trump card against moisture. After
mechanical dewatering had brought moistures as low as possible and the coal was still too wet, a
thermal dryer could always finish the job. Thermal drying is the only dewatering method that can
reduce moisture content to 1% regardless of feed size. This is accompanied with high costs and a
12 |
Virginia Tech | normally accomplished with wide diameter cyclones.
Screen and flash dryers operate in a similar manner. A screen dryer uses less airflow than
the fluidized bed dryer and coal slides across the grate without fluidization. In contrast a flash
dryer uses enough air to carry all the coal with the airflow, allowing the coal to dry during its trip
to the cyclone.
One of the hazards associated with thermal dryers is filter cake sticking together and
forming lumps inside the drying chamber which will subsequently catch fire. To avoid this, it is a
popular practice to mix 6.35mm x 0.595mm (1/4 inch x 30 mesh) coarse coal in with the cake in
ratios from 2:1 to 4:1 (Luckie 1991). This practice prevents lumps from sticking inside the dryer,
but also results in most of the coal going to the dryer being coarse coal that doesn’t need to be
thermally dried in the first place.
An even worse problem with thermal dryers is their environmental impact. Exposing coal
to high temperatures without burning the coal leads to the release of volatile organic compounds
(VOCs), several of which are categorized as carcinogens by the EPA (EPA 2009). The
temperatures that cause these unwanted emissions are summarized in Figure 2-5, which shows
650
600 1132
550 1032
500 932
450 832
400 732
350 632
300
532
250
432
200
332
150
100 232
50 132
0 32
Lowest in Industry Highest in
use Average use
Used under fair use guidelines, 2010
14
suisleC
tiehnerhaF
Undergrate/Intake
Overgrate/Drying Chamber
Exhaust Gas
Figure 2-5. Operating temperatures of thermal dryers (Luckie 1991) |
Virginia Tech | 3.0 EXPERIMENTAL
3.1 General Procedures
The end goal of the project is to produce clean marketable product from ultrafine coal
slurry with a special emphasis on achieving low moistures (<10%). These moistures were to be
achieved through the use of the low temperature drying process. The path to this goal can be
broken into three sequential steps; cleaning, dewatering and drying.
3.2 Cleaning
Two methods of removing ash particles from the coal slurry were employed: oil
agglomeration and froth flotation. In this work, cleaning was not performed so much to create a
low ash product but to prepare the coal for dewatering by removing the clays present in the
slurry. Clays are hydrophilic, clinging tenaciously to water, and keeping that water with them
even after dewatering. Clays also fill the void space in filter cakes, thus lowering permeability
and reducing the effectiveness of centrifugation or filtration. Filtration in particular suffers due to
clays blinding the filter cloth.
3.2.1 Froth Flotation
Due to difficulty in obtaining a small-scale column flotation machine, a Dorr Oliver
laboratory batch unit was employed instead. Since the cell was conventional, a rougher and two
recleaning steps were necessary to remove all the free clays from the slurry. Prior to the rougher
stage 0.75 kg/tonne (1.5 lb/ton) of hydrocarbon collector was added. MIBC (Methylisobutyl
Carbinol) frother was added to the rougher and cleaner stages as was necessary to maintain a
16 |
Virginia Tech | pentane was utilized as an agglomerating agent. As many of the feed slurries had ash contents
near 50%, this dosage was nearly equivalent to using 10% of the solid feed weight in pentane.
Depending on the desired weight of product, anywhere from 400 to 800 ml of feed slurry was
placed in the funnel followed by the appropriate amount of pentane. Due to pentane’s fast
evaporation rate, the funnel had to be sealed quickly and tightly to avoid pentane losses. Once
sealed the funnel was shaken vigorously by hand for two minutes. Subsequently, the funnel was
placed in a ring stand over an empty beaker. Finally, the stopcock at the bottom of the funnel was
opened allowing the ash laden water to drain into the beaker. The pentane agglomerates were
both buoyant and moderately stable, thus they floated in one large mass atop of the water while
the water was draining. The agglomerants also bridged over the open valve once the water was
gone. In this way, draining effectively separated product from tails just as screening or
centrifugation had in older agglomeration processes.
3.3 Dewatering
Two different dewatering methods were employed to accommodate the unique products
of the two cleaning processes. Oil agglomerates, although still high in moisture (>40%), could be
handled as a solid and therefore put into a basket centrifuge. The flotation froth, on the other
hand, was dilute and would run through the centrifuge basket before the solids could form a
cake. Unlike the centrifuge, the vacuum filter was able to hold thin slurries in place while they
formed a cake, making the filter a better choice for dewatering the float product. Furthermore, in
order to compare the limitations of the different dewatering processes, they had to be tested on
the same coal over a wide variety of size ranges. To this end, 2.38 mm (8 mesh) clean coal
samples from the Moss No. 3 preparation plant were ground to different sizes in a laboratory ball
mill. Once the samples were prepared, they were dewatered and dried by the following methods.
18 |
Virginia Tech | brought to the centrifuge and the agglomerates were poured into the basket. The agglomerates
would then be spread evenly around the basket to ensure a uniform cake thickness and the
centrifuge sealed. Once closed, the centrifuge was spun for 1 minute at top speed. The
centrifugal force exerted on the agglomerates simultaneously compacted them and pulled out any
free water.
Some samples were degassed after draining the funnel in order to test whether or not
pentane could be recovered prior to centrifugation. This was accomplished by attaching a
vacuum to the stopcock of the funnel. The pressure in the funnel was reduced below 508mm Hg
(20 in Hg) and left that way for 1 minute. Opening the funnel revealed that the agglomerates had
transformed into a thick paste. Centrifuging this paste for the same time and at the same speed as
the agglomerates resulted in a cake with a moisture of just over 29%. This same paste occurred
when the fragile agglomerates were compacted prior to centrifugation.
Moisture determination was normally performed by weighing samples, placing them in
an oven at 80 ºC (176 ºF) for an hour or more and then weighing them again. In order to prevent
unevaporated pentane from interfering with moisture determination the time for the pentane to
completely evaporate had to be found. This was accomplished by placing a freshly agglomerated
sample on a scale and recording its percent change in mass over time. The quickly evaporating
pentane produced a high rate of mass loss and when the rate slowed all the pentane was assumed
to be gone. The evaporation rate of a sample at 26 ºC (78 ºF) is shown in Figure 3-3. Based on
this plot, an evaporation time of 6 minutes was established, i.e., all centrifuged samples were set
aside in open air for at least 6 minutes before being subjected to moisture determination.
20 |
Virginia Tech | 160
140
120
100
80
60
40
20
0
0 5 10
Time (min)
Figure 3-3. Evaporation time for pentane.
3.3.2 Filtration
Filtration was performed using a 6.35 cm (2.5 inch) diameter Peterson laboratory filter.
This filter was attached to a vacuum pump capable of maintaining 508 mm Hg (20 in Hg) of
vacuum while the filter was in use. Between the pump and filter there was a valve to control the
flow of vacuum. Once the clean froth had been collected from the float cell and poured into the
vacuum filter, the valve would be opened. This allowed the filter to go straight from atmospheric
pressure to full vacuum without waiting for the pump to come up to speed. The slurry was left in
the filter until all visible surface water was gone and the filter cake had cracked. Cake thickness
was typically 0.635 cm (¼ inch) and the pump was left on until the cake was removed from the
filter.
In cases where lower moistures were desired, a dewatering aid was added after flotation
but prior to filtration. Reagent RU, developed at Virginia Tech, was used as the dewatering aide
21
)nim/gm(
etaR
noitaropavE |
Virginia Tech | in all experiments. To be used effectively, Reagent RU had to be dissolved in 2 parts diesel fuel
before the reagent was added to the slurry. In order to properly disperse RU, the slurry was
churned in a high shear mixer for 5 minutes before the sample was poured into the vacuum filter.
3.4 Drying
All three drying methods operated by pneumatically conveying the filter cake through a
shearing device that breaks the cake up into its constituent particles. Breaking up the cake
exposes sufficient surface area for the water to evaporate and pneumatic conveyance refreshes
the saturated air quickly enough to allow evaporation to continue unhindered. Unless otherwise
noted, all drying tests were performed on samples taken from the overflow of the 6-inch diameter
deslime cyclones at Arch Coal’s Cardinal plant near Sharples, West Virginia. Once cleaned, this
coal contained 79% ultrafine (minus 325 mesh) material.
3.4.1 Static Breaker
The static breaker shown in Figure 3-4 was designed to break up filter cake by slamming
the cake into a pipe wall at high velocity. The primary component of this device was a 5.08 cm
(2 in) diameter Plexiglas pipe which had an injection port bored into its side. To capture the dried
coal, the pipe lead to an 18.93 L (5 gal) bucket lined with a plastic bag. To retrieve any undried
material, an airlock was fitted to the base of the pipe. The injection assembly shown to the right
of the pipe was a venturi tube made out of a pipe “tee” and a funnel. The injector was powered
entirely by compressed air.
3.4.2 Air Jet Conveyer
Air jet conveyors are most typically used to transport small parts or foodstuffs (Exair
2011). The air jet conveyor was selected as a drying device in an attempt to intrinsically prevent
the problem of cake sticking inside the dryer by breaking up the cake without that cake
22 |
Virginia Tech | 4.0 RESULTS AND DISCUSSION
4.1 Cleaning
Cleaning results are hard to quantify because the goal of cleaning was to remove liberated
clay particles, and an ash analysis does not differentiate between free clays and noncombustibles
bound within the coal particles. Yet, the importance of effective cleaning was still apparent in the
dewatering results. When triple floated, a sample from the Clarksburg impoundment could be
filtered down to 17.3% moisture. When the Clarksburg sample was only floated twice, the filter
cake contained a higher moisture content of 23.6%.
4.2 Dewatering
Figure 4-1 shows how each of the three dewatering methods performed based on the
fraction of ultrafine (minus 325 mesh) particles in their feed. The dashed line at 22% moisture
represents the maximum moisture that can be effectively fed to a low temperature dryer without
creating operational problems such as sticking and plugging. Inspection of the data reveals
agglomeration followed by centrifugation can dewater even the finest of feeds to the level where
they can be put through a dryer. This makes that process appropriate for pond reclamation
projects where the size distribution of the feed can vary wildly. Vacuum filters utilizing
dewatering aids can stay below 22% moisture all the way up to 60% ultrafines. Therefore, this
approach may work well on flotation product from preparation plants where they do not deslime
the flotation feed. Material that is coarse enough to be dewatered below 22% with just vacuum
filters is unlikely to be in need of drying as the more advanced dewatering methods could be
used to lower its moisture instead.
26 |
Virginia Tech | The failing of the static breaker lies in how the rejects manifest themselves. Pieces of
filter cake that fail to break apart do not simply bounce off the wall and fall to the bottom of the
pipe; instead they stick to the wall upon impact. As more and more pieces stick to the wall a
finger of rejects grows across the pipe to block the injection port. The rejects had the consistency
of packed clay and striking the pipe wall was ineffective in dislodging the blockage. This
plugging phenomenon occurred in all tests and not just those using high moisture feed. In any
case where recovery was less than 100% the device would eventually become clogged. The only
effective method of removing the blockage was to install a bolt opposite the injector which could
push the rejects off the wall. Unfortunately this solution was deemed impractical for large scale
use.
4.2.2 Air jet conveyor
Table 4.2-1 shows the test results for the air jet conveyor. Required sweep air refers to the
Table 4.2-1. Air jet conveyor results.
Temperature (ºF) 74 74 76 74
Relative Humidity (%) 55 48 50 48
Dry Time (min) 0.5 1.5 2.5 2.0
Feed Moisture (%) 22.2 27.4 27.7 20.7
Product Moisture (%) 3.4 10.7 11.6 5.4
Rejects Moisture (%) 22.6 27.2 21.7 14.8
Combined Moisture (%) 9.3 21.2 11.8 5.9
Recovery (%) 80.0 44.0 98.2 95.5
Sweep Air Required (ft3) 59.2 34.0 36.5 29.3
Sweep Air Actual (ft3) 65.2 195.6 326.0 260.8
Actual/Required Air 1.1 5.8 8.9 8.9
29 |
Virginia Tech | amount of air that should have been necessary, as calculated by the low temperature drying
model, for the moisture reduction achieved by the dryer. Inspection of data reveals that the
conveyor never reaches moistures as low as those from the static breaker, 3.4 instead of 0.5%.
Moreover, the conveyor can only attain higher recoveries when given copious amounts of extra
air. A 98.2% recovery was achieved using the air jet as opposed to the static breaker’s 97% and
this required almost 9 times as much air as should have been necessary. The higher product
moistures are likely the result of incomplete breakup of the filter cake as small pieces of intact
cake could be found in the filter bag.
The relatively similar temperatures and humidities of the four tests are a result of the
large volumes of sweep air that originated as compressed air to run the air jet conveyor. The
drying times varied between 0.5 and 2.5 minutes, first in order to try to match the calculated
amount of required air and later to see if a slower feed rate could improve product moistures.
More air did help with the feeds containing greater than 27% moisture. The high moisture test
that used only 5.8 times the required airflow had a 10.7% product moisture and a poor recovery
of 44%. When the amount of air was increased to 8.9 times the required airflow, the product
moisture did climb up to 11.6%, but recovery leapt up to 98.2%. The effect of extra air was also
apparent with the feeds closer to 22% moisture. Increasing flow from 1.1 times to 8.9 times the
required rate brought product moisture up to 5.4% from 3.4%, but increased airflow also raised
recovery from 80% to 95.5%.
What the air jet conveyor could do that was impossible for the static breaker was to
accept high moisture feeds without plugging. The device appeared to be self cleaning since the
compressed air jets blasted free any buildup inside the conveyor tube. This also led to a notable
difference in the location of the rejects. The static breaker’s rejects where located at their first
30 |
Virginia Tech | point of impact, while those produced by the air jet conveyor where found throughout the hose
that connected the conveyor to the Shop Vac. The buildup seemed especially concentrated on the
ribs of the hose. It is possible that given a smooth passage all the rejects would have passed into
the filter bag without sticking.
The row marked “combined moisture” in Table 4.2-1 represents what the moisture would
be if the rejects and product were blended together. This blending does little for the samples with
high feed moistures, since the 11.8% and 21.1% combined moistures do not meet the project
goal of less than 10% moisture. However, if the feed moisture is kept below 22%, blending can
provide a 100% recovery while providing single digit moistures.
4.2.3 Centrifugal Fan
Table 4.2-2 shows fan performance when operating under a variety of weather
conditions. Although the centrifugal fan produced moistures slightly higher than the static
Table 4.2-2. Centrifugal fan dryer results.
Moisture (%) Sweep Air
Temp. RH Recovery Required Actual Actual/
Feed Product Rejects
(ºF) (%) (%) (ft3) (ft3) Required
61 79 19.1 1.0 16.0 72.3 163 80.4 0.49
58 92 21.7 2.1 16.0 75.9 523.6 321.6 0.61
60 42 20.1 1.1 4.1 92.4 61.1 80.4 1.32
63 94 21.4 1.1 4.0 94.9 415.5 937.5 2.26
70 84 21.5 0.9 3.2 95.2 238.2 522.6 2.19
79 29 19.4 1.3 2.7 96.3 34.3 32.2 0.94
65 42 19.5 1.6 0 100 42.5 80.4 1.89
31 |
Virginia Tech | breaker, this device suffered from few of the same blockage problems. As long as the feed
moisture did not significantly exceed 22%, the fan would not plug. The buildup which formed
under normal operating conditions was low in moisture (below 4.1%). This indicates the material
had dried and was similar to buildup which could be found on any fan with a large amount of
particulate in the air flow. However, if the fan was fed unevenly, given feed moistures exceeding
22% or used with significantly less than the required amount of air, then stable deposits of high
moisture (>16%) material would build up on the blades and the fan would plug. Choke feeding
was distinctly unsuccessful as there was little to no increase in product moisture (2.1% and
1.0%) and recovery dropped significantly (75.9% and 72.3%). In contrast, the two samples with
feed moistures under 20% produced the highest recoveries (96.3% and 100%) and in one case no
detectable amount of coal remained on the blades. The fan also made very efficient use of its air.
The highest successful ratio of actual-to-required air was 2.26 and the lowest was 0.94. For these
ranges, and all the conditions in between, recoveries above 92% and product moistures below
1.6% were readily attainable. The process was even successful when tested in 94% humidity air
with results of a 1.1% product moisture and a quite respectable 94.9% recovery.
The centrifugal fan was promising enough to be tested on more than the samples from the
Cardinal cyclone overflow. Additional samples tested included material from three
impoundments that were under consideration for pond reclaim operations. These samples were
from Trans Alta WA, Alden KY, and Clarksburg WV. The fan’s cutoff feed moisture was 22%
and these samples were significantly coarser than the Cardinal overflow, making them excellent
candidates to test whether feeds prepared without an oil agglomeration step could be dried
effectively using low temperature drying. To investigate this possibility, all three of the pond
samples were prepared both by agglomeration-centrifugation and by flotation-filtration. The test
32 |
Virginia Tech | Table 4.2-3. Pond sample fan dryer results.
Agglomerated & Centrifuged Floated & Filtered
Trans Clarks- Trans Clarks-
Pond Alden Alden
Alta burg Alta burg
% Ultrafine
(Minus325 69 27 28 69 27 28
mesh)
Cake
Moisture 21.4 21.6 22.3 23.0 14.4 17.3
(%)
Product
Moisture 0.9 1.6 1.2 0.5 1.5 1.0
(%)
results are shown in Table 4.2-3. The Trans Alta sample was treated with 2.5 kg/tonne (5 lb/ton)
of Reagent RU dewatering aid, Alden was treated with 1.5 kg/tonne (3 lb/ton) and Clarksburg
received none at all.
Both sets of samples were effectively dried and none of them caused buildup on the
fan blades with the exception of the filtered Trans Alta. A minor amount of that sample stuck to
the fan blades. Unfortunately the actual weight of those rejects is unknown due to the small
quantity of buildup and difficulties associated with retrieving rejects from the device without
turning the fan on. This buildup was likely due to the filtered Trans Alta containing feed
moisture just beyond acceptable limits. Interestingly, the filtered samples all dried to a lower
moisture than their centrifuged equivalents. This could be a result of filter cake being more
fragile than centrifuge cake and thereby being much more thoroughly broken up in the dryer.
Oddly enough, filtration also outperformed agglomeration and centrifugation in dewatering by
more than 5% in both coarse pond samples. This unexpected result may be because
33 |
Virginia Tech | agglomeration does a relatively poor job of recovering coarse particles. If agglomeration only
recovered the minus 149 µm (100 mesh) solids, then centrifugation would not receive the same
dewatering benefits that coarser particles provided to filtration.
4.3 Conclusions
While the static breaker produces the lowest moistures, that device also plugs itself on a
regular basis and is therefore inappropriate for use on a commercial scale. The air jet conveyor
excels at remaining unplugged, but does so at the expense of moisture and air consumption.
That60% of final airflow from the air jet conveyor came into the system as compressed air may
also prove an obstacle for scaling-up as an industrial dryer. Low temperature drying requires
large amounts of airflow and compressing air is an expensive way to achieve that flow.
However, the centrifugal fan is efficient in its use of air and produces low moistures without
clogging as long as its feed contains less than 22% moisture. That moisture level can be readily
achieved with oil agglomeration and centrifugation or, in many cases, with flotation and vacuum
filtration (possibly with the addition of dewatering aids).
4.4 Recommendations
The static breaker may be worthy of further experimentation. Its plugging problem may
be overcome by changing the shape of the breaker or by injecting coal at an angle against the
pipe wall instead of straight at the wall. The air conveyor fails to thoroughly break up the cake
and consumes too much air. Therefore, future work with the conveyor should focus on finding an
air jet arrangement that can better shred the cake while leaving the generation of sweep air up to
an exhaust fan.
The centrifugal fan used in this paper was equipped with forward curved fan wheels. That
type of wheel is the most sensitive to particulate buildup, so much so that forward curved fans
34 |
Virginia Tech | 5.0 MODELING AND SIMULATION
5.1 Low Temperature Drying
The new process of low temperature drying dries coal by creating optimum conditions to
maximize the rate of natural evaporation of surface water. At terrestrial temperatures and
pressures, liquid water is always, however slowly, evaporating. The rate of this evaporation is
governed by Langmuir’s expression of vapor pressure (Zemansky 1968), which is given by:
&
M
P = 2πRT/M [5.1]
A
where P is the difference between partial and saturated vapor pressures, M(cid:2) is the rate of mass
loss, A is surface area, R is the gas constant, T is the absolute temperature and M is the
molecular weight. The expression is more useful for determining evaporation rate when it is
rewritten in the following form:
&
M = AP M /(2πRT) [5.2]
Table 5.1-1. Calculated surface area of coal fines.
U.S. Sieve Size 30 50 100 200 325 500
Surface Area (ft2/lb) 38 77 153 310 530 914
Diameter (µm) 595 297 149 74 44 25
Surface Area (m2/kg) 7.9 15.8 31.4 63.3 106.4 187.3
Hence, this expression shows that water with a higher exposed surface area will evaporate faster.
36 |
Virginia Tech | 5.2 Model Development
Low temperature drying uses natural evaporation instead of forced boiling like thermal
drying. As a result, the process depends heavily on the water carrying capacity of the available
air. Once the air is saturated, no further water can be removed. This in turn means that the
process is highly sensitive to local weather conditions like temperature and relative humidity.
These factors together determine how much water can be carried per unit of air. If conditions are
unfavorable, more air can be used to carry away the same amount of water. However, at some
critical point, conditions are so poor that the air has no appreciable carrying capacity for
moisture.
In order to determine the cutoff weather conditions where low temperature drying
becomes impractical, a mathematical model needed to be developed. One that would reveal how
much air is necessary to remove a fixed amount of water under different temperature and
humidity conditions. The foundation for such a model has already been developed by Luckie for
thermal drying. (Luckie 1991) However, some modifications are necessary in order for this
model to accurately describe the conditions relevant to the low temperature drying process.
The first modification was to calculate the entering humidity of the air based on weather
conditions instead of using a constant 0.01 kg water per kg air (0.01 lb H O/lb air) that Luckie
2
assumed. Entering humidity is only of minor concern when designing thermal dryers because the
carrying capacity generated by heating the air dwarfs any naturally occurring humidity.
However, in low temperature drying, the entering humidity makes up a large fraction of the total
carrying capacity of the air; therefore, accounting for the impact of this factor cannot be left to an
estimate. The actual entering humidity can be found using relative humidity (%RH) and
maximum humidity. Maximum humidity can be calculated from air temperature.
40 |
Virginia Tech | ambient air. Inspection of plot reveals that there is a steep increase in the required amount of
sweep air when there is a confluence of high humidity and extreme cold. This makes sense in
that cool wet air has the smallest water carrying capacity. The range above 80% humidity and
below -6.67 ºC (20ºF) is especially bad as the air requirements leaps from less than 14158
m3/min (0.5 MCFM) to almost 84,950 m3/min (3.0 MCFM). Moreover, no data is shown for a
relative humidity greater than 90% due to the fact that the amount of sweep air required quickly
approaches infinity as the moisture in the air nears maximum humidity.
In an effort to compare the low temperature drying method to those that employ heat, the
cost of electricity to generate the sweep air, is shown in Figure 5-4. Under most weather
conditions, unheated drying uses less than $8/ton worth of energy to dewater the coal. Sweep air
costs actually drop below $1/ton of dried product under the ideal conditions of less than 30%
humidity and temperatures higher than 21.11 ºC (70 ºF). Considering that without drying, the
ultrafine coal is a waste product with zero dollar value. This operating cost would be quite
tolerable for most coal operations. However, at a higher humidity and a lower temperature, the
cost spirals out of control passing $19/ton at 80% humidity and -12.22 ºC (10 ºF) and
culminating in an energy bill of $84/ton at 90% humidity and -17.77 ºC (0ºF). The former cost is
likely to be expensive enough to start making the process unfeasible for steam contracts and the
later price could only be supported on the most lucrative of metallurgical contracts.
Consequently, a dryer using ambient air would not operate dependably during the cold winter
months or in extremely humid climates. In most of the U.S. coal fields, this would limit the use
of unheated air to specific seasons or periods of fair weather.
43 |
Virginia Tech | 5.2.2 Low Heat Model
The results presented in the previous section indicate that drying with ambient air is not
possible under all weather conditions and, therefore, the low temperature drying process may not
be suitable for year round industrial use. A high relative humidity renders the dryer inoperable
because air that is already saturated cannot carry away moisture. The only way to make low
temperature drying an all-weather process is to increase the carrying capacity of the air. The least
costly way to increase the moisture carrying capacity of air is to heat it. Heating does not actually
remove any of the moisture from the air, but simply increases the air’s ability to hold more
moisture. The maximum humidity of air is exponentially related to its temperature. Each degree
the temperature of air is raised adds more water carrying capacity than the last degree the air was
raised. This approach is still different from thermal drying in that mild heating is only increasing
the amount of water the sweep air can carry and is not raising the water to its boiling point.
Figure 5-5 shows the amount of sweep air necessary if the overgrate/drying chamber is
operating at 48.89 ºC (120ºF). The addition of heat reduces the average sweep air required by
two orders of magnitude and shifts the high air requirements away from the cold humid to the
hot humid intake air. The latter is a result of the hot air undergoing a small temperature rise
compared to the cold air and thus has a correspondingly low rise in carrying capacity. While the
air requirements do rise sharply in the range above 80% humidity and 32.22 ºC (90 ºF) the
increase is not that much more air in absolute terms. To operate at 100% humidity and 37.78 ºC
(100ºF) the low heat dryer takes 424.8 m3/min (15,000 CFM) more air than the dryer does on
average. This is compared to an increase of 79,290 m3/min (2.8 MCFM) for the unheated dryer
to run under equivalently poor conditions. But this reduction in air flow is achieved at the
expense of heat to raise the air temperature.
44 |
Virginia Tech | 5.3 Model Validation
To demonstrate that an expanded surface area is necessary to facilitate effective drying an
experiment was performed between broken and unbroken wet coal samples. The first sample
was put through the centrifugal fan in 1 minute thus shredding the filter cake and exposing coal
particles to 2.28 m3 (80.4 ft3) of ambient air at 15.56 ºC (60ºF) and 42% relative humidity. The
second sample was held in a pan and placed under a 2.209 m3/min (78 CFM) flow of dry air for
10 min. This second flow of air was both generated and dehumidified by passing that air through
a compressor. Because the sample sizes and feed moistures did not match exactly, the total
weights of water removed were compared. The first test where the filter cake was thoroughly
comminuted lost 9.35g (0.33 oz) of its original 9.64g (0.34 oz) of water content. As a rate of loss
this would be 9.35 g/min (0.33 oz/min) or 4.10 g/m3of air (0.00410 oz/ft3). The second intact
sample lost 3.4 g (0.12 oz) of its original 5.95g (0.21 oz) of water and its rate of loss would be
0.34 g/min (0.012 oz/min) or 0.15 g/m3 of air (0.00015 oz/ft3). Ergo the increase in surface area
that results from the comminution of the filter cake allows water to be removed at almost
threefold the rate of an intact sample, while using just over a tenth of the air. This is the case
even if the broken filter cake is dried with wet air and the intact sample with dry air.
Other than increased surface area another possible driving force behind the swift
evaporation of water from the fines’ surface is the Gibbs Thomson effect in which small droplets
of liquid exhibit vapor pressures higher than they would as flat films. When a droplet is
sufficiently small, the surface tension of the liquid squeezes the whole droplet, effectively raising
its vapor pressure. Artificially higher vapor pressures result in higher rates of evaporation or
evaporation at humidites where evaporation would otherwise be impossible. How small a
droplet needs to be to take advantage of this effect can be calculated using the equation:
49 |
Virginia Tech | area to accommodate 22% of its weight in moisture in the form of 0.1 µm droplets without them
coalescing into an encompassing film. If water droplets are being shorn from the coal’s surface
during the drying process they may be small enough to benefit from the Gibbs Thomson effect
but no experiments were performed to detect such droplets.
The effect may explain why, even under the most ideal conditions there are minute
amounts of surface water remaining after low temperature drying. When Gibbs Thomson is
applied to a convex surface, like that of a droplet, the effect increases vapor pressure. Yet when
applied to a concave surface, like that which forms inside a capillary, the effect decreases vapor
pressure. So, if two coal particles failed to separate during the during the shearing step as the
water evaporated, the water would recede farther and farther in to the space between particles
and a concave surface with a smaller and smaller radius would be exposed to the air until the
vapor pressure was so low evaporation stopped. In the same way water could become trapped in
sub micron fissures in the surface of the coal particles. An illustration of how concave water
surfaces could occur between particles or in fissures is shown in Figure 5-10.
Figure 5-10. Concave Water Surfaces
51 |
Virginia Tech | 5.4 Conclusions
Based on the energy consumption predicted by these models low temperature drying with
unheated air is only economical in warm dry climates due to the process’ vulnerability to
changing weather conditions. Yet when accompanied with mild heating, low temperature drying
becomes an economical method of dewatering fine coal under all weather conditions. Thermal
dryers are a significant source of air pollution because they expose coal to high temperatures
without combusting it. The average overgrate temperature of a thermal dryer is 171ºF and they
commonly operate with intake temperatures in excess of 900ºF. However, a low temperature
dryer can achieve a low moisture product without exceeding an overgrate of 120ºF or an
undergrate of 467ºF. These drastically lower temperatures should allow the low temperature
dryer to function with few of the environmental problems associated with thermal dryers.
5.5 Recommendations
The higher the overgrate temperature, the lower the operating costs for the dryer but at
some temperature volatile organic compounds (VOCs) will start to be released from the coal.
The modeled temperature of 48.89 ºC (120 ºF) was selected in an attempt to run as cold as
possible while maintaining all-weather capabilities. If that temperature is raised to 61.67 ºC
(143ºF) the low temperature dryer has the same energy costs as the thermal dryer operating at
61.11 ºC (151ºF) but then the dryer is no longer operating at a “low” temperature. Furthermore, it
is not readily apparent whether it is the extended exposure to the high temperatures in the drying
chamber or the brief exposure to the extremely high temperatures at the intake that cause the coal
to release pollutants. Future work will have to determine what conditions lead to the creation
52 |
Virginia Tech | 6.0 SUMMARY AND CONCLUSIONS
The project goal was to dewater ultrafine coal particles below 10% moisture. This was
successfully achieved through a new process called low temperature drying. Low temperature
drying evaporates water from coal fines by exposing the entire surface area of those fines to air
that is constantly replaced as that air becomes saturated with water vapor. A number of relevant
discoveries were made about low temperature drying. Revelations about how the air carries away
the moisture include:
• Pneumatically conveying the fines is an effective method of continually refreshing the
saturated air around the water’s surface.
• The volume of air required can be accurately predicted using the modified thermal dryer
equations.
• The process works poorly with cold, humid air.
• Warming the air can make the process work well under any reasonable weather
conditions. Drying chamber temperatures as low as 48.89 ºC (120 ºF) should be effective.
• Warming the air also drastically reduces the required volume of sweep air and lowers
overall energy costs.
Discoveries about how to fully expose the surface of the fines include:
• The static breaker design produces moistures below 1% but this design is also prone to
plugging.
• The air jet conveyor excels at remaining unplugged but produces moistures greater than
3%. Moreover, to achieve high recoveries the air jet must consume almost ninefold the
volume of air that should be required to dry the fines.
54 |
Virginia Tech | With fine particles the thickness of the surface water significantly contributes to the final
diameter of the combined particle and therefore the surface area. To correct this, the volume of
water was calculated from 22% of the coal’s weight and a density of 62.4 lb/ft3 (1000 kg/m3).
The thickness of the water film was then found by dividing the water volume by the surface area
of the coal. Twice this thickness was added to the original coal particle’s diameter to find the
diameter for the combined particle.
7 = (cid:29) /
8),9* 8),9* :(cid:5))4
This combined diameter was then used to calculate the surface area an the combined particle and
that area was multiplied by the original number of original coal particles to find the total exposed
surface area of the water.
= 6
,(cid:5),)4 :()*, ()*,
The evaporation rate can then be found by reentering the total area into Langmuir’s evaporation
expression. Finally the dry time can be solved for by dividing the mass of water (22% mass of
coal) by the evaporation rate. The surface area does change as water evaporates but this change
had little impact on evaporation time in absolute terms.
61 |
Virginia Tech | To determine how much air will be necessary to carry away the water the maximum humidity of
the air must be know. Maximum humidity is measured in lbs H2O per lbs of dry air.
Temperature is in ºF and atm is atmospheric pressure in psi.
18 JKLM.LOLPJ
@A(cid:7)B@C@ ℎC@B0BFG =
/(AF@(cid:9) (cid:14)9Q(R1SP.MTLM.1KPOL)(cid:28)
−1)
29
The carrying capacity of the water is the difference between this maximum humidity and
whatever water the sweep air entered the dryer with. Luckie assumes that the sweep air enters
with a humidity of 0.01 lbs H O per lb of dry air. This is a safe assumption because thermal
2
dryers elevate the air temperature enough that variations in this entering humidity are dwarfed by
the raise in the max humidity created by heating the air. Low temperature dryers by definition
don’t elevate the temperature nearly as high and so are much more vulnerable to the entering
humidity of the sweep air. To get the actual entering humidity the max humidity is multiplied by
the relative humidity of the ambient air that the dryer is operating in.
ℎC@ = ℎC@ ×ℎC@
9U, *94 Q)V
Now in reality a dryer cannot fully saturate its sweep air but near complete saturation is possible.
According to Luckie thermal dryer designers like to achieve 90-95% relative humidity (RH) in
their exhaust air. In this paper the relative humidity of the exhaust air was taken to be 95%. So
the difference in humidities of the entering and exiting air is:
WℎC@ ×%'/100X−ℎC@
Q)V 9U,
And the tph of sweep air equals
'/WℎC@ ×%'/100X−ℎC@
Q)V 9U,
63 |
Virginia Tech | 5% was used for all cases in this paper
W. Btu/hr required
X. Fan Capacity SCFM
The 0.0351 accounts for the volume of water vapor being created inside the dryer and SCFM
assume that the air is at standard temperature and pressure(60 ºF 14.696psi). But the gases are
not at STP so Luckie employed two correction factors to turn SCFM into an actual flow rate.
The calculations in this paper used an altitude of 1000ft. Luckie also included a minor correction
factor for static fan pressure.
But without knowing details of the fan being utilized this correction can’t be used.
There are no modifications to the final value calculations for either of the dryers models
that use heat but some changes are required for the dryer that uses no heat at all.
R . MAX HUM is the max humidity of the air based on the exhaust/product temperature E. If
a
the water vapor isn’t heated the vapor can’t cool and condensate rendering the -15 ºF
unnecessary.
T . Btus released per lb of dry air
a
66 |
Virginia Tech | Table C-4 Unheated Dryer Simulation Results Part 4
Sweep
Intake Carrying
Atmospheric Required Air Undergrate ACFM Btu/hr $/Ton
Temp (0F) Relative Capacity (lb Air (tph) Heat Temp (0F) (cfm) required Product
Hum (%) H O/lb air)
2 (Btu)
0 0 0.05022 5.36 321861 467 2983 980998 2.01
0 10 0.05013 5.37 322426 467 2988 981604 2.01
0 20 0.05004 5.38 322992 467 2993 982212 2.01
0 30 0.04996 5.39 323560 466 2998 982823 2.01
0 40 0.04987 5.40 324131 466 3003 983435 2.02
0 50 0.04978 5.41 324703 465 3008 984050 2.02
0 60 0.04969 5.42 325277 465 3013 984666 2.02
0 70 0.04960 5.43 325854 465 3018 985285 2.02
0 80 0.04952 5.44 326432 464 3022 985906 2.02
0 90 0.04943 5.45 327013 464 3027 986530 2.02
0 100 0.04934 5.46 327596 464 3032 987155 2.02
10 0 0.05022 5.36 295039 457 2983 952200 1.95
10 10 0.05008 5.38 295869 457 2991 953091 1.96
10 20 0.04994 5.39 296703 456 2999 953987 1.96
10 30 0.04980 5.41 297543 455 3007 954888 1.96
10 40 0.04966 5.43 298387 455 3015 955794 1.96
10 50 0.04952 5.44 299235 454 3023 956705 1.96
10 60 0.04937 5.46 300089 453 3031 957622 1.97
10 70 0.04923 5.47 300947 453 3039 958543 1.97
10 80 0.04909 5.49 301810 452 3047 959470 1.97
10 90 0.04895 5.50 302679 452 3055 960403 1.97
10 100 0.04881 5.52 303552 451 3063 961340 1.97
20 0 0.05022 5.36 268218 447 2983 923402 1.90
20 10 0.05000 5.39 269400 446 2995 924672 1.90
20 20 0.04978 5.41 270593 445 3008 925953 1.90
20 30 0.04956 5.44 271797 444 3020 927245 1.91
20 40 0.04934 5.46 273011 443 3033 928549 1.91
20 50 0.04912 5.48 274237 442 3045 929865 1.91
20 60 0.04890 5.51 275473 441 3058 931192 1.91
20 70 0.04868 5.53 276721 440 3071 932532 1.92
20 80 0.04846 5.56 277980 439 3084 933884 1.92
20 90 0.04824 5.59 279250 438 3097 935248 1.92
20 100 0.04801 5.61 280532 437 3110 936624 1.93
30 0 0.05022 5.36 241396 437 2983 894604 1.84
30 10 0.04988 5.40 243031 435 3002 896359 1.84
30 20 0.04954 5.44 244688 434 3021 898139 1.85
30 30 0.04921 5.47 246368 432 3040 899942 1.85
30 40 0.04887 5.51 248071 431 3060 901771 1.86
30 50 0.04853 5.55 249798 429 3079 903625 1.86
30 60 0.04819 5.59 251549 428 3099 905505 1.87
30 70 0.04785 5.63 253324 426 3120 907412 1.87
30 80 0.04752 5.67 255125 425 3140 909345 1.87
30 90 0.04718 5.71 256952 423 3161 911307 1.88
30 100 0.04684 5.75 258805 422 3183 913296 1.88
72 |
Virginia Tech | Table C-5 Unheated Dryer Simulation Results Part 5
Intake Sweep
Carrying
Atmospheric Relative Required Air Undergrate ACFM Btu/hr $/Ton
Temp (0F) Hum Capacity (lb Air (tph) Heat Temp (0F) (cfm) required Product
H O/lb air)
(%) 2 (Btu)
40 0 0.05022 5.36 214574 427 2983 865806 1.78
40 10 0.04971 5.42 216765 424 3011 868159 1.79
40 20 0.04920 5.48 219002 422 3040 870560 1.80
40 30 0.04870 5.53 221285 420 3070 873012 1.80
40 40 0.04819 5.59 223617 418 3100 875515 1.81
40 50 0.04768 5.65 225998 415 3130 878071 1.81
40 60 0.04717 5.71 228430 413 3162 880683 1.82
40 70 0.04667 5.77 230915 411 3194 883351 1.82
40 80 0.04616 5.84 233455 408 3226 886078 1.83
40 90 0.04565 5.90 236051 406 3260 888865 1.84
40 100 0.04514 5.97 238706 404 3294 891716 1.84
50 0 0.05022 5.36 187752 416 2983 837008 1.73
50 10 0.04947 5.45 190597 413 3025 840062 1.73
50 20 0.04872 5.53 193529 410 3068 843211 1.74
50 30 0.04797 5.62 196553 406 3113 846457 1.75
50 40 0.04722 5.70 199673 403 3159 849807 1.76
50 50 0.04647 5.80 202894 400 3206 853265 1.77
50 60 0.04572 5.89 206220 396 3255 856836 1.77
50 70 0.04497 5.99 209657 393 3306 860526 1.78
50 80 0.04422 6.09 213210 390 3358 864342 1.79
50 90 0.04347 6.20 216886 386 3412 868289 1.80
50 100 0.04272 6.31 220691 383 3468 872374 1.81
60 0 0.05022 5.36 160931 406 2983 808210 1.67
60 10 0.04913 5.48 164498 401 3044 812040 1.68
60 20 0.04804 5.61 168226 396 3109 816043 1.69
60 30 0.04695 5.74 172128 392 3176 820232 1.70
60 40 0.04586 5.87 176215 387 3246 824621 1.71
60 50 0.04477 6.02 180501 382 3319 829222 1.72
60 60 0.04369 6.17 185000 377 3397 834053 1.73
60 70 0.04260 6.32 189730 372 3478 839131 1.75
60 80 0.04151 6.49 194708 367 3563 844476 1.76
60 90 0.04042 6.67 199954 362 3653 850108 1.77
60 100 0.03933 6.85 205490 357 3749 856053 1.79
70 0 0.05022 5.36 134109 396 2983 779412 1.61
70 10 0.04866 5.54 138407 389 3072 784027 1.63
70 20 0.04710 5.72 142990 382 3166 788947 1.64
70 30 0.04554 5.92 147886 375 3267 794205 1.65
70 40 0.04398 6.13 153130 368 3375 799835 1.67
70 50 0.04242 6.35 158760 361 3491 805879 1.68
70 60 0.04086 6.59 164819 354 3616 812385 1.70
70 70 0.03930 6.85 171359 347 3751 819407 1.72
70 80 0.03774 7.14 178440 340 3897 827009 1.74
70 90 0.03618 7.45 186131 333 4055 835267 1.76
70 100 0.03462 7.78 194514 326 4228 844268 1.78
73 |
Virginia Tech | Table C-6 Unheated Dryer Simulation Results Part 6
Carrying
Sweep
Intake Capacity
Atmospheric Required Air Undergrate ACFM Btu/hr $/Ton
Relative (lb
Temp (0F) Air (tph) Heat Temp (0F) (cfm) required Product
Hum (%) H O/lb
2 (Btu)
air)
80 0 0.05022 5.36 107287 386 2983 750614 1.56
80 10 0.04801 5.61 112215 376 3110 755904 1.57
80 20 0.04581 5.88 117617 366 3249 761705 1.59
80 30 0.04360 6.18 123566 356 3403 768092 1.61
80 40 0.04140 6.51 130148 346 3572 775159 1.62
80 50 0.03919 6.87 137471 336 3761 783022 1.64
80 60 0.03699 7.28 145668 327 3972 791822 1.67
80 70 0.03478 7.75 154904 317 4210 801739 1.69
80 80 0.03258 8.27 165390 307 4480 812998 1.72
80 90 0.03037 8.87 177399 297 4789 825892 1.76
80 100 0.02817 9.56 191289 287 5147 840805 1.80
90 0 0.05022 5.36 80465 376 2983 721815 1.50
90 10 0.04713 5.72 85731 362 3164 727470 1.52
90 20 0.04405 6.12 91735 348 3370 733916 1.54
90 30 0.04097 6.58 98643 334 3608 741333 1.56
90 40 0.03788 7.11 106676 320 3884 749957 1.58
90 50 0.03480 7.74 116133 307 4208 760111 1.61
90 60 0.03171 8.50 127430 293 4596 772241 1.65
90 70 0.02863 9.41 141162 279 5068 786985 1.69
90 80 0.02554 10.55 158211 265 5654 805290 1.74
90 90 0.02246 12.00 179944 251 6400 828624 1.81
90 100 0.01937 13.91 208598 237 7385 859390 1.90
100 0 0.05022 5.36 53644 365 2983 693017 1.45
100 10 0.04594 5.86 58637 346 3241 698379 1.46
100 20 0.04167 6.47 64655 327 3551 704841 1.49
100 30 0.03739 7.21 72051 308 3932 712781 1.51
100 40 0.03311 8.14 81356 289 4411 722772 1.55
100 50 0.02884 9.34 93421 270 5033 735726 1.59
100 60 0.02456 10.97 109689 250 5871 753192 1.65
100 70 0.02028 13.28 132816 231 7063 778024 1.73
100 80 0.01601 16.83 168301 212 8891 816123 1.86
100 90 0.01173 22.97 229660 193 12052 882004 2.08
100 100 0.00745 36.14 361431 174 18842 1023484 2.56
74 |
Virginia Tech | Table C-7 Unheated Dryer Simulation Results Part 7
Carrying
Intake Sweep
Atmospheric Capacity Required Undergrate ACFM Btu/hr $/Ton
Temp (0F) Relative (lb H O/lb Air (tph) Air Heat Temp (0F) (cfm) required Product
Hum (%) 2 (Btu)
air)
0 0 0.13371 2.01 152117 958 1325 856184 1.72
0 10 0.13362 2.02 152217 958 1325 856292 1.72
0 20 0.13353 2.02 152317 958 1326 856399 1.72
0 30 0.13344 2.02 152418 957 1327 856507 1.72
0 40 0.13336 2.02 152518 957 1327 856615 1.72
0 50 0.13327 2.02 152619 956 1328 856723 1.72
0 60 0.13318 2.02 152720 956 1329 856831 1.72
0 70 0.13309 2.02 152820 955 1330 856939 1.72
0 80 0.13300 2.03 152921 955 1330 857048 1.72
0 90 0.13292 2.03 153023 955 1331 857156 1.72
0 100 0.13283 2.03 153124 954 1332 857265 1.72
10 0 0.13371 2.01 142043 948 1325 845368 1.70
10 10 0.13357 2.02 142193 948 1326 845529 1.70
10 20 0.13343 2.02 142343 947 1327 845690 1.70
10 30 0.13329 2.02 142493 946 1328 845851 1.70
10 40 0.13314 2.02 142644 945 1329 846013 1.70
10 50 0.13300 2.03 142795 945 1330 846175 1.70
10 60 0.13286 2.03 142947 944 1332 846338 1.70
10 70 0.13272 2.03 143098 943 1333 846501 1.70
10 80 0.13258 2.03 143250 943 1334 846664 1.70
10 90 0.13244 2.03 143403 942 1335 846828 1.70
10 100 0.13230 2.04 143555 941 1336 846992 1.70
20 0 0.13371 2.01 131969 938 1325 834552 1.67
20 10 0.13349 2.02 132187 937 1326 834786 1.67
20 20 0.13327 2.02 132406 936 1328 835021 1.67
20 30 0.13305 2.02 132625 935 1330 835256 1.68
20 40 0.13283 2.03 132845 934 1332 835492 1.68
20 50 0.13261 2.03 133066 933 1334 835730 1.68
20 60 0.13239 2.03 133288 932 1336 835967 1.68
20 70 0.13216 2.04 133510 930 1337 836206 1.68
20 80 0.13194 2.04 133733 929 1339 836446 1.68
20 90 0.13172 2.05 133957 928 1341 836686 1.68
20 100 0.13150 2.05 134181 927 1343 836927 1.68
30 0 0.13371 2.01 121895 928 1325 823736 1.65
30 10 0.13337 2.02 122204 926 1327 824067 1.65
30 20 0.13303 2.03 122514 925 1330 824400 1.65
30 30 0.13269 2.03 122826 923 1333 824735 1.65
30 40 0.13236 2.04 123140 921 1336 825072 1.66
30 50 0.13202 2.04 123455 920 1339 825410 1.66
30 60 0.13168 2.05 123771 918 1341 825750 1.66
30 70 0.13134 2.05 124090 916 1344 826092 1.66
30 80 0.13101 2.06 124410 915 1347 826435 1.66
30 90 0.13067 2.06 124731 913 1350 826781 1.66
30 100 0.13033 2.07 125055 911 1353 827128 1.66
75 |
Virginia Tech | Table C-8 Unheated Dryer Simulation Results Part 8
Carrying
Intake Sweep
Atmospheric Capacity Required Undergrate ACFM Btu/hr $/Ton
Temp (0F) Relative (lb H O/lb Air (tph) Air Heat Temp (0F) (cfm) required Product
Hum (%) 2 (Btu)
air)
40 0 0.13371 2.01 111821 918 1325 812919 1.63
40 10 0.13320 2.02 112247 915 1329 813377 1.63
40 20 0.13269 2.03 112677 913 1333 813838 1.63
40 30 0.13218 2.04 113110 910 1337 814303 1.63
40 40 0.13168 2.05 113546 908 1341 814771 1.64
40 50 0.13117 2.05 113985 905 1346 815243 1.64
40 60 0.13066 2.06 114428 903 1350 815718 1.64
40 70 0.13015 2.07 114874 900 1354 816197 1.64
40 80 0.12965 2.08 115324 898 1359 816680 1.64
40 90 0.12914 2.09 115778 895 1363 817167 1.64
40 100 0.12863 2.09 116235 893 1368 817658 1.64
50 0 0.13371 2.01 101747 907 1325 802103 1.61
50 10 0.13296 2.03 102321 904 1331 802719 1.61
50 20 0.13221 2.04 102901 900 1337 803342 1.61
50 30 0.13146 2.05 103488 896 1343 803972 1.61
50 40 0.13071 2.06 104081 893 1350 804609 1.62
50 50 0.12996 2.07 104681 889 1356 805253 1.62
50 60 0.12921 2.08 105289 885 1363 805905 1.62
50 70 0.12846 2.10 105903 882 1369 806565 1.62
50 80 0.12771 2.11 106524 878 1376 807232 1.62
50 90 0.12696 2.12 107153 874 1383 807907 1.62
50 100 0.12621 2.13 107790 871 1390 808591 1.62
60 0 0.13371 2.01 91673 897 1325 791287 1.59
60 10 0.13262 2.03 92426 892 1334 792095 1.59
60 20 0.13153 2.05 93191 887 1343 792917 1.59
60 30 0.13044 2.07 93969 881 1352 793752 1.59
60 40 0.12935 2.08 94760 876 1361 794601 1.60
60 50 0.12826 2.10 95565 871 1371 795465 1.60
60 60 0.12717 2.12 96383 865 1381 796344 1.60
60 70 0.12608 2.14 97216 860 1391 797237 1.60
60 80 0.12500 2.16 98063 855 1401 798147 1.60
60 90 0.12391 2.17 98924 849 1411 799072 1.61
60 100 0.12282 2.19 99802 844 1422 800014 1.61
70 0 0.13371 2.01 81599 887 1325 780470 1.57
70 10 0.13215 2.04 82562 879 1337 781504 1.57
70 20 0.13059 2.06 83548 872 1351 782563 1.57
70 30 0.12903 2.09 84558 864 1364 783647 1.58
70 40 0.12747 2.11 85593 857 1378 784758 1.58
70 50 0.12591 2.14 86653 849 1392 785896 1.58
70 60 0.12435 2.17 87739 841 1407 787063 1.58
70 70 0.12279 2.19 88854 834 1422 788260 1.59
70 80 0.12123 2.22 89997 826 1437 789487 1.59
70 90 0.11967 2.25 91170 818 1453 790746 1.59
70 100 0.11811 2.28 92373 811 1469 792039 1.59
76 |
Virginia Tech | Table C-9 Unheated Dryer Simulation Results Part 9
Carrying
Sweep
Intake Capacity
Atmospheric Required Air Undergrate ACFM Btu/hr $/Ton
Relative (lb
Temp (0F) Air (tph) Heat Temp (0F) (cfm) required Product
Hum (%) H O/lb
2 (Btu)
air)
80 0 0.13371 2.01 71525 877 1325 769654 1.55
80 10 0.13150 2.05 72725 866 1343 770942 1.55
80 20 0.12930 2.08 73965 855 1362 772274 1.55
80 30 0.12709 2.12 75249 844 1382 773652 1.56
80 40 0.12489 2.16 76577 834 1402 775079 1.56
80 50 0.12268 2.20 77954 823 1423 776556 1.56
80 60 0.12048 2.24 79381 812 1445 778089 1.57
80 70 0.11827 2.28 80861 801 1467 779678 1.57
80 80 0.11607 2.32 82397 791 1491 781327 1.57
80 90 0.11386 2.37 83993 780 1515 783041 1.58
80 100 0.11165 2.41 85652 769 1541 784822 1.58
90 0 0.13371 2.01 61451 867 1325 758838 1.53
90 10 0.13062 2.06 62902 852 1350 760396 1.53
90 20 0.12754 2.11 64424 836 1377 762030 1.53
90 30 0.12445 2.16 66021 821 1406 763744 1.54
90 40 0.12137 2.22 67699 806 1436 765546 1.54
90 50 0.11828 2.28 69464 791 1467 767441 1.55
90 60 0.11520 2.34 71324 776 1500 769438 1.55
90 70 0.11211 2.40 73287 761 1535 771545 1.56
90 80 0.10903 2.47 75360 746 1572 773772 1.56
90 90 0.10595 2.54 77555 731 1611 776128 1.57
90 100 0.10286 2.62 79880 716 1653 778625 1.57
100 0 0.13371 2.01 51377 856 1325 748022 1.50
100 10 0.12943 2.08 53075 836 1361 749844 1.51
100 20 0.12515 2.15 54888 815 1399 751791 1.51
100 30 0.12088 2.23 56830 794 1441 753876 1.52
100 40 0.11660 2.31 58915 773 1485 756114 1.52
100 50 0.11232 2.40 61158 752 1533 758523 1.53
100 60 0.10805 2.49 63578 731 1584 761122 1.54
100 70 0.10377 2.60 66199 710 1640 763935 1.54
100 80 0.09950 2.71 69044 689 1701 766990 1.55
100 90 0.09522 2.83 72145 668 1767 770320 1.56
100 100 0.09094 2.96 75538 647 1839 773962 1.57
77 |
Virginia Tech | Direct Force Measurements between Surfaces Coated with
Hydrophobic Polymers in Aqueous Solutions and
the Separation of Mixed Plastics by Flotation
Nini Ma
ABSTRACT
Froth floatation is an important process used in the mining industry for separating
minerals from each other. The separation process is based on rendering a selected mineral
hydrophobic using an appropriate hydrophobizing reagent (collector), so that it can
selectively attach onto the surfaces of a rising stream of air bubbles. Thus, controlling the
hydrophobicity of the minerals to be separated from each other is of critical importance in
flotation. If one wishes to separate plastics from each other by flotation, however, it
would be necessary to render a selected plastic hydrophilic and leave the others
hydrophobic. In the present work, the possibility of separating common plastics from
each other by flotation has been explored.
While water contact angle is the most widely used measure of the hydrophobicity of a
solid, it does not give the information on the kinetics of flotation. Therefore, the forces
acting between the surfaces coated with different hydrophobic polymers (or plastics) in
water were measured using the Atomic Force Microscope (AFM). The results obtained
with polystyrene, poly(methyl methacrylate) (PMMA), polypropylene (PP), and Teflon
showed the existence of long-range attractive forces (or hydrophobic force) that cannot
be explained by the classical DLVO theory. The surface force measurements were
conducted in pure water and in solutions of surfactant (alkyltrimethylammonium chloride)
and a salt (NaCl). In pure water, the attractive forces were much stronger than van der
Waals force. In the presence of the surfactant and NaCl, the long-range attraction
decreased with increasing concentration and the alkyl chain length. |
Virginia Tech | I - INTRODUCTION
1.1 Hydrophobic Forces
In the mining industry, froth flotation is the most widely used method for separating
fine particles. In this process, selected hydrophobic minerals are attached to air bubbles to
be separated from hydrophilic particles. The control of particle hydrophobicity is thus of
crucial importance for improving flotation performance. However, the exact nature of the
mechanisms involved in hydrophobicity, especially the origin of forces between two
hydrophobic surfaces in water, is still not completely understood.
In the early days, classical DLVO theory was used to predict the kinetics of
coagulation and to describe interactions between particles. The DLVO theory considers
two surface forces: the double-layer force and the van der Waals force. In 1961,
Derjaguin and Dukhin [1] modeled flotation using these two surface forces. However,
both the double-layer and the van der Waals forces are repulsive in most of the conditions
encountered in flotation, which made it difficult to model fast and spontaneous flotation.
The presence of an additional non-DLVO interaction was first recognized by Laskowski
and Kitchener in 1969 [2]. They found that methylated and pure silica particles had
practically the same ξ-potentials, and yet only the former floated while the latter did not,
which led to the suggestion that a long-range “hydrophobic influence” may be
responsible for the rupture of the wetting film, and thus, for the bubble-particle
interactions in flotation. The authors suggested that the hydrophobic influence could be
caused by the instability of the water structure in the vicinity of a hydrophobic surface.
Later, Blake and Kitchener [3] showed that the thin film of water between a hydrophobic
solid and an air bubble ruptures fast and spontaneously and that it does so at separation
distances of about 64nm, which was much larger than predicted by the classical DLVO
theory. These investigators suggested the presence of a “hydrophobic force” in wetting
film between bubble-particle. The term hydrophobic force is now widely used to describe
the long-range, non-DLVO attractive forces measured between macroscopic hydrophobic
solid surfaces immersed in water.
1 |
Virginia Tech | The non-DLVO hydrophobic force was first measured experimentally in 1982 using
the Surface Force Apparatus (SFA) by Israelachvili and Pashley [4]. The force was
shown to decay exponentially in the 0-10 nm range with a decay length of 1 nm. The
authors postulated that the force was related to the hydrophobic effect, which is widely
used to explain mutual attraction between hydrophobic solutes (e.g. hydrocarbons) in
water. Many other investigators conducted follow-up experiments and reported even
longer-range of hydrophobic forces, whose decay lengths were shown to vary in the
range of 10-30 nm.
When measuring hydrophobic forces, the adsorption of surfactants at the solid-liquid
interface involves complex mechanisms and may introduce artifacts. Since the 1990’s,
direct force measurement between naturally hydrophobic polymer surfaces, without the
use of surfactants, has become popular in the study of hydrophobic forces. However, the
existence of hydrophobic forces remains controversial, since previous studies provide
contradictory results. In 1993, Karaman et al. [5] measured the forces between
polystyrene surfaces in water. The result showed that attractive forces pulled the surfaces
into contact, from an initial separation distance as large as 30 nm. No further details
regarding the nature of this attractive force were presented. In the same year, Li et al. [6]
measured forces between 2 µm diameter polystyrene spheres in solutions using a
Scanning Force Microscope. The results showed no long-range attraction, but only the
existence of a repulsive electrostatic force. In 1994, Meagher and Craig [7] studied the
effect of degassing on the surface force between polypropylene surfaces (θ = 90°-111°) in
NaCl solution, using an AFM. The authors postulated the existence an attractive force
stronger than the van der Waals force both with and without degassing the solution. The
jump distance, particularly significant before degassing, was measured to be as large as
21.0+5.2 nm, indicating the presence of the hydrophobic force. In 1999, no long-range
hydrophobic attraction was found in the measurement performed by Schmitt et al. [8] on
the surface between two fused polystyrene surfaces. In 2001, Vinogradova et al. [9]
measured the interaction force between polystyrene surfaces and discovered a long-range
attraction inconsistent with the DLVO theory. Considine et al. [10] investigated the force
between polystyrene sphere and plate in aqueous media using an AFM. It was reported
2 |
Virginia Tech | that an attraction, much stronger than the van der Waals force, pulled the two spheres into
contact with the initial separation distance from a quite long range (20-400 nm). The
range of this strong attraction was found to decrease significantly after degassing.
Although much attention has been paid to the long-range attraction between
hydrophobic surfaces in aqueous solutions, its origin still remains unclear. Two
mechanisms have been proposed to explain the attractive forces between hydrophobic
surfaces, including the popular nano-bubble hypothesis, which attributes the attraction to
the capillary force caused by the coalescence of preexisting nano-bubbles as surfaces
approach each other, and the charged-patch model, which uses the change of water
structure near hydrophobic surfaces to account for the attractive force. Zhang et al. [11]
observed and measured the presence of attractive forces between a glass sphere and a
fused-silica plate in aqueous C TACl solutions using AFM. The magnitude of the
18
attractive force was found to be much larger than the magnitude of the Van der Waals
force, both in air-saturated and degassed solutions. The effect of dissolved gas on the
hydrophobic attraction between surfactant-coated mica surfaces was studied by Meyer et
al. [12] using SFA. Degassing was found to reduce the attraction range and to have
virtually no effect within short range.
Both the above studies and the fact that no steps were detected on the surface force
curve indicated that the nano-bubble hypothesis is inapplicable to the surfactant in situ
adsorption system. This hypothesis, however, would require further investigation in the
case of natural hydrophobic polymer surfaces, which exhibit high hydrophobicity (θ>80°).
The charged-patch model was proposed by Miklavic et al. [13], who studied long-
range forces in high-concentration electrolyte solutions. In the framework of this model,
it is suggested that “foreign” species in the solution may cause the breakdown of the
water structure and thus decrease the long-range attraction. Zhang et al. [14] reported
surface force measurements using an AFM with silica surfaces immersed in C TACl
n
solutions in the absence of salt (n=12-18). These measured forces were in agreement with
the charged-patch model. The results also showed that the long-range attractions first
3 |
Virginia Tech | increased with growing surfactant concentration, reaching a maximum at the point of
charge neutralization (p.c.n.) and then decreased.
The main objective of the present investigation was to measure the surface forces
between different plastics (hydrophobic polymers) immersed in water using an AFM. The
measurements were conducted more systematically than reported previously in literature,
and the results were analyzed quantitatively in view of the extended DLVO theory (Xu
and Yoon, 1990, [15]).
1.2 Plastics Flotation
Despite a growing interest in recycling worldwide, a relatively small amount of plastic
is recycled due to the lack of effective methods for separating different types of plastics
from each other from a mixture. If one wishes to separate them by froth flotation, proper
understanding of the surface properties of different plastics would be essential. Various
wetting agents are used to modify the surface properties (hydrophobicity) of plastics. In
the present work, water contact angles are used as the measure of surface hydrophobicity.
Our research aims at controlling the hydrophobicity by using appropriate wetting agents
and achieving the separation of mixed plastics by flotation.
During the past two decades, research in plastic flotation mainly focused on wetting
plastics selectively by using various flotation depressants. Saitoh et al. [16] measured the
water contact angles of various plastics such as polypropylene (PP), polyethylene (PE),
polystyrene (PS), and polyvinyl chlorite (PVC), at different concentrations of a wetting
agent (sodium lignosulfonate). In pure water, all plastics exhibit contact angles greater
than 80o, indicating strong hydrophobicity. The differences in hydrophobicity among
these plastics are too small to allow separation by flotation. As the concentration of the
wetting agent increases, however, the contact angles are reduced substantially and the
difference in hydrophobicity is thus magnified. This phenomenon serves as a basis for
separating plastics by flotation.
4 |
Virginia Tech | Many investigators showed that mixed plastics can be separated from each other by
flotation if the surface properties are modified appropriately. Sisson [17] showed that in
2-15% NaOH (or KOH) and 0.005-0.1% NEODOL91-6 (a non-ionic surfactant)
solutions the contact angle of PET was decreased to below 25o, while that of PVC
remained above 45o. Flotation separation under these conditions gave a 93.5% PVC
recovery and a 97.5% PET recovery. Shibata et al. [18] developed a flowsheet for the
flotation separation of a mixed plastic, in which POM and PVC were floated together by
depressing PVC with 500 mg/l sodium lignosulfonate. Th POM was then floated from the
POM/PC mixture by using 200 mg/t Saponin and 50 mg/l Aerosol OT to depress PC.
Singh [19] separated PVC and POM successfully using appropriate wetting agents. Their
results showed that sodium lignosulfonate was a good depressant for PVC while sorbitan
monolaurate was good for POM. Drelich et al. [20] found that the hydrophobicity of PET
was strongly altered by strong alkaline solutions of sodium hydroxide, while the
hydrophobicity of PVC remained the same. The PET and PVC recoveries were 95 to
100%, respectively. Le Guern et al. [21, 22] studied the adsorption mechanism of
lignosulphonate onto PVC and PET surfaces. Results showed that lignosulphonate could
selectively interact with PET and render it hydrophilic. Also, the presence of divalent
cations such as calcium could enhance the hydrophilization of PET through electrostatic
bridge action between the lignosulphonate and the plastics. Both of them were negatively
charged before the treatment. Shen et al. [23, 24] found that methyl cellulose can depress
the flotation of PVC more readily than PET. The authors also investigated the floatability
of seven plastics in the presence of an alkyl ethoxylated non-ionic surfactant (15-S-7, for
example). The floatability was shown to decrease with addition of the surfactant.
Floatability was depressed to a different extent for each plastic (POM < PVC < PMMA <
PET < PC < ABS < PS). Dodbiba et al. [25] separated PET from both PET/ PE and PET/
PP mixtures using either 0.02 kg/m3 dodecylamine acetate or polyvinyl alcohol. Basarova
et al. [26] studied the wettability and floatability of PS in the presence of terpinol,
polyethylene glycol dodecyl ether (PGDE), tannic acid, and calcium lignosulfonate. The
results showed that at concentrations above 60 mg/l, PGDE increased the wettability of
PS considerably while at concentrations below 20 mg/l, calcium lignosulfonate strongly
5 |
Virginia Tech | decreased the wettability. Tannic acid, on the other hand, with increasing concentration,
decreased the floatability linearly.
The froth flotation technique is based on the selective adherence of gas bubbles onto
the particles to be separated. For this purpose, a sufficient difference in wettability is
required between these particles. Since plastics are naturally hydrophobic, selective
wetting of each component is necessary for the separation by flotation. However, a
comprehensive understanding of the wetting phenomena involved has not been
established yet. Thus, the wetting action of a given surfactant during flotation cannot be
predicted.
1.3 Research Layout
This work was devoted to the experimental investigation of surface forces and
wettability of plastics in the presence of various wetting agents. The overall objective was
to better understand the origin of hydrophobic force and identify the conditions under
which different plastics (or hydrophobic polymers) can be separated from each other. In
Chapter 2, the results of the AFM surface force measurements conducted between PS,
PMMA, PP and Teflon surfaces in pure water and solutions of different surfactants and
electrolytes. Chapter 3 describes an experimental study aimed at finding more efficient
wetting agents for the separation of PS and PVC by flotation.
6 |
Virginia Tech | II - HYDROPHOBIC FORCES BETWEEN PLASTIC SURFACES IN
AQUEOUS SOLUTIONS
The surface forces acting between hydrophobic surfaces play an important role in
flotation. In the froth floatation process, the efficiency of separation is largely determined
by the interaction between hydrophobic air bubbles and hydrophobic particles. Therefore,
it has become increasingly important to achieve a better understanding of the origin of
hydrophobic forces. The non-DLVO theory, which takes into account the hydrophobic
force, the van der Waals force and the electrostatic force, has been recently applied to
explain the phenomena involving hydrophobic interaction. There have been more and
more evidences indicating that the non-DLVO theory is a more thorough theory than the
classical DLVO theory, for which numerous discrepancies with experiment results have
been found.
Plastics are ideal materials to study the hydrophobic force, because they are naturally
hydrophobic, and they don’t introduce the complexity arising from surfactant adsorption.
The existence of hydrophobic attraction between hydrophobic plastic surfaces has been
investigated by many researchers. In 1993, Karaman et al. [5] measured the forces
between polystyrene surfaces in water, and recorded a jump-in distance as large as 30 nm.
Li et al. [6], in 1993 also measured forces between polystyrene surfaces but observed no
long-range attraction force. In 1999, Schmitt et al. [8] performed surface force
measurement between two fused polystyrene surfaces, but their results showed no long-
range hydrophobic attraction. In the same year, Considine [10] observed a much stronger
attraction force than the van der Waals fore, with a jump-in distance from a quite long
range (20-400 nm), between different pairs of polystyrene spheres. In 2001, Vinogradoca
et al. [9] tested the same plastic, and a long-range attraction was also observed.
Force measurements have also been conducted with Teflon and other plastics. In 1994,
Meager and Craig [7], using an AFM, investigated the hydrophobic interaction between
polypropylene surfaces in NaCl solution. A 30 nm range where they observed the
attractive forces was found in solutions both with and without degassing. Nalaskowski et
al. [27] measured forces between a polyethylene sphere and a polyethylene surface in de-
7 |
Virginia Tech | aerated and aerated water. A long-range attractive force between hydrophobic
polyethylene surface and sphere was observed.
The Teflon has a very high contact angle (θ=110°). In 1999, Considine et al. [28]
measured the force between a micron-sized colloidal sphere and a flat plate, both coated
with a copolymer of perfluoro (2, 2-dimethyl-1, 3-sioxole) and Teflon AF 1600. It was
shown that the surfaces experienced a very strong attraction as two surfaces contact each
other. The jump-in distance was about 500 nm in water. In 2005, Hallam et al. [29] also
observed a long-ranged attraction between hydrophobic amorphous fluoropolymer
surfaces. They found that the range of the attraction and its attraction decreased in de-
aerated water as compared to normal, aerated water. However, the range and the strength
of the attraction in deaerated water remained significantly greater than those of the van
der Waals attraction for this system.
In this chapter, the results of the hydrophobic force measured between various
hydrophobic polymer-coated surfaces, which included PS, PMMA, PP and Teflon were
studied. The measurements were conducted using an Atomic force microscope in water
and surfactant solutions. The relationship between surface forces and contact angles has
been studied. The objective of this study was to study the nature of hydrophobic forces.
2.1 Materials and Experiments
2.1.1 Plastic Spheres
Polystyrene (PS) spheres of 10-15 μm in diameter were purchased from Polysciences,
Inc., Warrington, PA. They were soaked and settled in ethanol for 1 hour. The suspension
was withdrawn to remove possible organic contaminants. This procedure was repeated
three times. Polymethyl methacrylate (PMMA) spheres of 19.8 μm in diameter were
provided by Corpuscular Inc., New York. They were cleaned in the same manner as
described for polystyrene. PP and Teflon spheres were formed directly on the tip of a
cantilever by melting a small amount of powder at 130°C and 190°C. The spheres were
then stored in a N atmosphere until use (no longer than 3-4 days).
2
8 |
Virginia Tech | 2.1.2 Synthesis of PS, PMMA and PP Substrates Through Spin-coating
Silicon wafers were provided by Sumco, Oregon. They were boiled in a freshly
prepared solution of NH OH:H O :H O (1:1:5 by volume) for about 1.5 hour at 110±5oC
4 2 2 2
to clean their surface. The wafers were then rinsed thoroughly with a large amount of
Milli-Q water, and dried by blowing pure nitrogen over the surface. The plates were then
heated in a piranha solution, a solution of concentrated H SO and H O (7:3 by volume),
2 4 2 2
for at least 60 minutes, rinsed thoroughly with Milli-Q water, and then dried with
nitrogen again. The plates were further submerged in an HF solution for 3 minutes and in
an ammonium fluoride solution for about 1 minute. After being rinsed with water and
dried with nitrogen, the plates were immediately spin-coated.
PS, PMMA and PP powders were dissolved in toluene to get 0.3% solutions for spin
coating on the surface of silicon wafer.
Teflon substrates were prepared by melting the Teflon solid layer which was evenly
placed on the cleaned silicon wafer surface and heated to 190°C. A nitrogen stream was
used to remove remaining flakes.
2.1.3 Water and Surfactant Solutions
Nanopure water was obtained by using the Nanopure III (Barnstead IA) water
purification system. The conductivity of the water was 18.2 MΩ/cm at 25°C and the
surface tension was 72.5 mN/m at 22°C. Extra purging with ultra-pure nitrogen gas, to
exclude atmospheric CO , has not been used here. Thus, the pH of the solutions was
2
within the 5.8-6.0 range. Tetradecyltrimetylammonium chloride (C TACl),
14
hexadecyltrimetylammonium chloride (C TACl) and octadecyltrimetylammonium
16
chloride (C TACl) were provided by TCI, with purity greater than 97%. These
18
surfactants were used for surface force measurements without further purification.
Solutions of low surfactant concentrations were prepared in containers with volumes of at
least 100 mL to prevent significant depletion of surfactant via adsorption onto the glass
walls.
9 |
Virginia Tech | 2.1.4 Surface Force Measurements on PS Surface
A Digital Instrument Nanoscope III (Veeco, CA) atomic force microscope (AFM) was
used for surface force measurements [30, 31]. All experiments were performed at room
temperature (22±1oC). Silicon dLever cantilevers and silicon-nitride NP-20 cantilevers
were purchased for Veeco, CA. The cantilevers were calibrated using an adapted version
of the Resonant Frequency Technique [32], which limits the calibration errors up to ±7%.
A plastic sphere was attached on the tip of a cantilever spring so that the surface force
measurement could be measured using the sphere-plate geometry. EPON 1004 resin
(Shell Chemical Co.) was used to attach the spheres (10-15μm in diameter as measured
by a microscope) onto cantilevers. The force measurements were conducted at different
surfactant concentrations. The concentrations were controlled by adding more additional
surfactant incrementally. At a given surfactant concentration, the system was equilibrated
for 30 minutes before taking the measurements. No efforts were made to exclude the air
from dissolving into the solution.
2.1.5 Contact Angle Measurement
The hydrophobicity of plastics (or silicon wafer coated with different hydrophobic
polymers) was characterized by means of the water contact angles measured using the
cessile drop technique. For each measurement, a drop of pure water was placed onto a
plastic-coated silicon wafer, and the contact angles were measured using a goniometer.
The results reported in this work were averages of over five such measurements.
The contact angles of the plastics were also measured in solutions of surfactants and
electrolytes. In this case, the contact angles were measured using the captive bubble
technique. In this method, a solid sample was immersed in a solution, and an air bubble
formed at the end of a hypodermic needle was brought in contact with the solid surface
sample from below. The contact angle was defined as the angle between a line tangent to
the bubble and the surface of the solid sample, as shown in Figure 2.1.
10 |
Virginia Tech | 2.1.6 Zeta-potential Measurement
The plastic spheres used for the AFM force measurements were used to measure the ξ-
potentials at the solid-liquid interface. The measurements were conducted using the
Zetanano series laser meter. The reported results were averaged over six independent
measurements. 0.5g/l solution was prepared and ultrasonication was applied for 30 min to
remove air bubble and breakup agglomerates.
2.2 Results and Discussion
2.2.1 AFM Image
Figure 2.2.a-c shows the AFM images of a silicon wafer surface coated with PS,
PMMA and PP. The coating was prepared using the spin-coating as described in 2.1.2.
The image represents the magnification of a 5 x 5 µm section of the surface, and its
morphology is represented by color scaling. As shown the maximum peak-valley distance
was approximately 1.8 nm, indicating that the spin-coated surface was smooth enough for
surface force measurement.
2.2.2 Zeta-potential
The zeta-potentials of PS, PMMA and PP were obtained by means of the
electrophoresis technique. The values of these potentials were found to be -30 mV, -67
mV and -80 mV, respectively, in water at pH = 6.0. Teflon zeta-potential data was not
available due to the high hydrophobicity of Teflon and it was hard to get the suspension
solution which was required for the measurement. The zeta potentials as a function of pH
were also shown in Figure 2.3.
2.2.3 Surface Forces Measured in Nanopure Water
Figure 2.4 shows the results of the AFM force measurements conducted between
polystyrene (PS) sphere and plate in water at neutral pH. The squares represent the
experimental data, and the solid line represents a fit to the classical DLVO theory.
11 |
Virginia Tech | According to this theory, the total surface force (F) should be the sum of the electrostatic
t
force (F ) and the van der Waals dispersion force (F ) as follows:
e d
F F F , (1)
t e d
in which
RA
F 131, (2)
d 6H3
where R is the radius of the colloidal probe, A the Hamaker constant, and H is the
131
closest separation distance between two macroscopic surfaces. The fit was made with
A (Hamaker constant) =1.3x10-20 J [33] and ψ (double-layer potential) = -30mV. As
131 0
shown in Figure 2.3, two surfaces jumped into contact at 19.5 nm, which was
substantially larger than what the DLVO theory predicted. This finding suggests that
there may be additional attractive force other than the van der Waals force.
Figure 2.5 shows the same experimental results fitted to the DLVO theory (Eq. (1))
extended to include the hydrophobic force (F ), as follows:
h
F F F F , (3)
t e d h
in addition to the electrostatic and van der Waals forces. The hydrophobic force may be
represented by the following expression:
H
F RCexp , (4)
h D
where R is the radius of the sphere of the colloidal probe, H the closest separation
distance between the surface of the sphere and the flat surface, and C and D are fitting
parameters. The constant D is referred to as decay length, and the constant C is related to
the interfacial tension between solid and liquid. The hydrophobic force curve shown in
Figure 2.5 was obtained with C = 1.0 mN/m and D = 4.4 nm. The water contact angle
measured using the sessile drop technique was 76°. The figure clearly shows that the
12 |
Virginia Tech | extended DLVO theory (Eq. (3)) can predict the jump-in distance better than the classical
DLVO theory (Eq. (1)).
Figure 2.6 shows the results of the AFM force measurements conducted between a
polymethyl methacrylate (PMMA) sphere and a silicon wafer coated with PMMA in
water at the natural (i.e. unadjusted) pH. The solid line represents the DLVO fit of the
data with A = 1.05×10-20 J [24] and ψ = -68mV. The fit is not very good, as the theory
131 0
predicts that the two surfaces should jump into contact at 6 nm, while it actually occurs at
24 nm. The large discrepancy suggests that additional attractive forces must be present
between the hydrophobic surfaces. Therefore, the experimental data were fitted to the
extended DLVO theory with C = 3.0 mN/m and D = 4 nm, as shown in Figure 2.7. The
fit based on the recognition of the hydrophobic force was much better. The sessile drop
contact angle of the PMMA plate was 72°.
Figure 2.8 shows the results of the AFM force measurements conducted between a
polypropylene (PP) sphere and a PP-coated silicon wafer plate in Nanopure water at a
neutral pH. Contact angle of the PP-coated surface was 98o as measured by the sessile
bubble technique. The solid line represents the experimental data fitted to the DLVO
theory with A = 6.0×10-20 J [35] and ψ = -82.4mV. The DLVO theory predicts that the
131 0
two surfaces should jump into contact at 5 nm, whereas the jump actually occurs at 28
nm. This large discrepancy again suggests an additional attraction causing the jump at
such a large distance. Therefore, the experimental data was fitted to the extended DLVO
theory (Eq. (3)) with C = 3.1 mN/m and D = 7.6 nm, as shown in Figure 2.9.
Table 2.1 summarizes the data presented in Figures 2.4-2.9. According to the
equilibrium sessile drop contact angle () data, PP is most hydrophobic, followed by PS
and then PMMA. It is found that the decay length (D) follows the same order. The
difference in is the largest between PP and PMMA. It is also found that these two
polymers show the largest difference in D also.
13 |
Virginia Tech | Figure 2.10 shows the AFM force measurements conducted between a Teflon sphere
and a Teflon-coated plate as described in 2.1.2. in Nanopure water at a neutral pH. With
the sessile drop technique, the contact angle of the Teflon surface was measured to be
110o. In Figure 2.10, the experimental data fitted to the DLVO theory with A =
131
3.59×10-21 J [27] is represented by the solid line. As discussed above, the two surfaces are
predicted to jump into contact at 6 nm according to the DLVO theory. However, the jump
actually occurs at 30 nm as shown in Figure 2.11. This discrepancy suggests the existence
of an additional attraction between the two hydrophobic surfaces. Figure 2.11 also
demonstrates that the experimental data is better fitted to the extended DLVO theory with
C = 3.0 mN/m and D = 5.6 nm. However, although Teflon has the largest contact angle
among these four plastics, the amplitude of its hydrophobic forces is not the greatest. This
may be due to the roughness of Teflon, which could yield to an underestimation of the
measured forces. The high potential of Teflon could also be a reason that the hydrophobic
force appear to be small.
2.2.4 Surface Forces Measured in Surfactant Solutions
Figure 2.12 shows the AFM surface force measurements conducted between a PS
sphere and a PS-coated silicon wafer surface in the presence of different concentrations
of a cationic surfactant (C TACl). At low concentrations, i.e., 8x10-5 and 1x10-4 M, the
14
surface force decreased, due to both charge neutralization and increase in hydrophobic
force. Since PS is negatively charged at neutral pH, the cationic surfactant can adsorb on
the PS surface and thus increase the hydrophobicity of PS. The contact angle increased as
the concentration was raised from 8x10-5 to 1x10-4 M, as shown in Table 2.2. These
observations suggest that the cationic surfactant adsorbs with a normal mode of
orientation, i.e. the polar head group being in contact with the PS surface.
Several authors [36-39] have shown that the attractive forces measured between
hydrophobic surfaces reach a maximum as the surfactant’s concentration gets close to the
point of charge neutralization (p.c.n.). The p.c.n. is defined as the concentration, at which
the surface charge of the substrate is neutralized by the adsorption of the cationic
surfactant. For the F/R versus H curves shown in Figure 2.12, the p.c.n. corresponding to
14 |
Virginia Tech | the PS-C TACl system is approximately situated at 1×10-4 M. At this concentration, the
14
attractive force reached a maximum. At the p.c.n., most negative sites are occupied by the
C TA+ ions, forming hemi-micelles in which the hydrocarbon tails of the adsorbed
14
surfactant molecules are closely packed.
At 5×10-4M C TACl, additional C TA+ ions would adsorb on the substrate surface,
14 14
but with inverse orientation. Such an orientation is favored, because it requires the least
amount of energy to insert additional surfactant molecules into the adsorption layer with
their polar heads furthest apart from each other. The inverse orientation exposes the head
groups toward the aqueous phase and, hence, causes the surface to be less hydrophobic.
Therefore, the decrease in attractive force at the higher concentration can be attributed to
this phenomenon. The mechanism of inverse orientation is illustrated in Figure 2.19.
Table 2.2 shows that the contact angle decreases to 78o as the C TACl concentration was
14
further increased to 5x10-4 M most probably due to inverse orientation. The additional
surfactant present in the second-layer of adsorbed species not only reduce the contact
angle but also increase the negative surface charge, both of which causing the surface
force to be repulsive as shown in Figure 2.12.
Figure 2.13 shows the results obtained with a longer-chain homologue, C TACl. At
16
1×10-5M C TACl, a very long-ranged and net-attractive force was observed. The
16
strongest attractive force was measured for a concentration of 3×10-5M C TACl. Thus,
16
this concentration may be close to the p.c.n. of the PS-C TACl system. This value is
16
significantly lower than that observed for the shorter chain homologues. The most
interesting aspect of the data shown in Figure 2.13 is that the C-16 surfactant gives much
stronger and longer-range attractive forces than those observed for the shorter chain
homologues. With C TACl, net attractive forces were observed at separations as large as
16
60nm.
Figure 2.14 shows the results obtained with C TACl. The data show that the C-18
18
surfactant gave much stronger and longer-range attractive forces than those observed for
the shorter chain homologues. At 5×10-6 M C TACl, a very long-ranged and net-
18
15 |
Virginia Tech | attractive force was observed for a separation distance as large as 65 nm. When the
concentration was raised to 2×10-5 M C TACl, the attractive force became much stronger
18
at a separation distance as large as 120 nm.
Figure 2.15 shows the results of the AFM surface force measurements conducted
between a PS sphere and a PS-coated silicon wafer surface in a 0.1% sodium
lignosulfonate solution. As shown in the figure, the surface forces measured with this
anionic surfactant increased drastically and became much more repulsive in the presence
of the anionic surfactant. At this concentration, the contact angle decreased sharply from
= 74.4o to 32.2o. The decrease in contact angle and the increase in repulsive surface
forces are most probably due to the inverse orientation of the anionic surfactant, i.e.
adsorption with the hydrocarbon chains in contact with the PS surface and the anionic
polar heads heading toward the aqueous phase.
Figure 2.16 shows the AFM surface force measurements conducted between a PMMA
sphere and a PMMA-coated silicon wafer surface in the presence of different
concentrations of a cationic surfactant (C TACl). At 8×10-5M C TACl, a very long-
14 14
ranged and net-attractive force was observed. When the concentration was raised to 1×10-
4 M, the attractive force reached its maximum. Thus, the p.c.n. of the PMMA-coated
silicon wafer-C TACl system was approximately 1×10-4 M, which is the same as for the
14
PS-coated silicon wafer-C TACl system. As the concentration was further increased to
14
5x10-4 M, the cationic surfactant adsorbed with an inverse orientation. The inverse
orientation further increases the positive charge of the surface and at the same time
decreases the hydrophobicity, which is consistent with the result shown in Table 2.2,
where the contact angle is shown to decrease from 72 ° to 67o, and the results shown in
Figure 2.16, where the surface force becomes repulsive.
2.2.5 Surface Forces Measured in Salt Solution
Figure 2.17 shows the surface forces measured between a PMMA sphere and a
PMMA-coated silicon wafer substrate immersed in salt (NaCl) solutions. When the salt
concentration was increased from 1x10-3 M to 6x10-3 M, the measured forces changed
16 |
Virginia Tech | from repulsive to attractive. At 1x10-3 M, a long-range repulsive force was observed.
However, the two hydrophobic surfaces jumped into contact at approximately 9 nm,
which was considerably larger than predicted by the classical DLVO theory. Thus, the
hydrophobic force was still there despite the fact that the measured force was net
repulsive. At 3x10-3 M NaCl, the measured force was net attractive due to the reduction
of the double-layer force. At 6x10-3 M, the double-layer force was further reduced and
the two surfaces jumped into contact at a separation as large as 25 nm. These results
indicate that the hydrophobic force was still present in the presence of 1x10-3 to 6x10-3 M
NaCl. The D value was listed in Table 2.3 after eDLVO calculation.
Figure 2.18 shows the surface forces measured between a PP sphere and a PP-coated
silicon wafer plate immersed in a 10-3 M NaCl solution. The results are similar to those
obtained with the PMMA surfaces. The force measured in pure water was net repulsive
due to a high surface charge. Yet, the two surfaces jumped into contact at approximately
30 nm, due to a strong hydrophobic force with a decay length (D) of 7.6 nm. At 1x10-3 M
NaCl, the repulsive force disappeared, and the two hydrophobic surfaces jumped into
contact at approximately 13 nm, indicating the presence of a strong hydrophobic force.
The decay length was 8.1 nm, indicating that the presence of NaCl did not affect the
hydrophobic force significantly.
2.3 Summary
An atomic force microscope (AFM) was used to measure the surface forces between
PS, PMMA PP and Teflon surfaces in pure water and in different concentrations of
surfactant and salt (NaCl) solutions. The forces measured in pure water were net
repulsive due to the high surface charges, as indicated by their -potentials: -30 mV for
PS, -67 mV for PMMA, and -80 mV for PP. Nevertheless, two hydrophobic surfaces
jump into contact at separation distances much larger than predicted by the classical
DLVO theory. The discrepancy can be attributed to the presence of the long-range
hydrophobic force between the symmetric PS, PMMA and PP surfaces. In pure water, the
hydrophobic forces decrease in the order of PP, PMMA and PS. In the presence of NaCl,
17 |
Virginia Tech | the repulsive double-layer force is decreased, and net attractive forces are observed due to
the hydrophobic force.
The surface forces were also measured in the presence of a cationic (C TACl) and
14
anionic (sodium lignosulfonate) surfactant. In the presence of the former, the
hydrophobic force becomes stronger with increasing surfactant concentration due to the
adsorption of the cationic surfactant with a normal mode of orientation, i.e., the head
group in contact with the surface and the hydrocarbon tail protruding into the aqueous
phase. As the surfactant concentration was increased beyond the p.c.n., the cationic
surfactant begins to adsorb with inverse orientation and render the surface less
hydrophobic. As a result the hydrophobic force becomes weaker. The surface forces
measured with C - and C -TACl homologues showed the same trend as C TACl except
16 18 14
that the hydrophobic forces measured at a given surfactant concentration was lower. In
the presence of the anionic surfactant, the measured surface force becomes much more
repulsive due to the inverse orientation of the surfactant molecules on the hydrophobic
surface.
18 |
Virginia Tech | III – SEPARATION OF MIXED PLASTICS BY FLOTATION
This chapter is devoted to the study of plastics floatation. Plastics are naturally
hydrophobic and thus difficult to separate directly. However, adding surfactant can
change their surface property and form different hydrophilicity, which makes froth
floatation of plastics possible. As an indicator of wettability, the contact angles of two
commonly used plastics, PS and PVC, were measured in the presence of various wetting
agents. The objective of this study is to explore the efficiency of wetting agents for froth
floatation of plastics.
3.1 Materials and Experiments
3.1.1 Plastics
PS sheets were obtained from United States Plastic Corporation, Lima, Ohio. PVC
sheets were obtained from Polysciences, Inc., Warrington, PA. After being washed with
plenty of water, plastic sheets were cut into 2×2cm pieces, and each piece was tied with a
copper wire onto a glass for contact angle measurement
3.1.2 Surfactants
Five different surfactants were used wetting agents for the plastics. Dioctyl sodium
sulfosuccinate (DSS) and Bij30 were purchased from Sigma Aldrich. Methyl cellulose
(MC), sodium lignosulfonate (SL) and saponin were purchased from Fisher Scientific.
DSS, Bij30, SL and saponin solutions of the following concentrations, i.e., 0.01, 0.05,
0.1 0.15, 0.2, 0.25 and 0.3, were prepared by mixing 0.01, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3 g
of a surfactant with 100ml of water. MC solutions of concentration 0.01, 0.05, 0.1 0.15,
0.2, 0.25 and 0.3, were prepared by heating 100 ml of water to 90C° in a beaker, before
adding 0.01, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3g of solid MC. The solutions thus obtained
were then cooled down under constant stirring.
31 |
Virginia Tech | The plastic samples, prepared as described in paragraph 3.2.1, were dipped in
surfactant solutions for one hour. Contact angle measurements were then conducted using
the contact angle goniometer FTA 200.
3.1.3 Water
Pure water was obtained by using the Direct-Q 3 (Barnstead IA) water purification
system. The conductivity of the water was 18.2 MΩ/cm at 25°C and the surface tension
was 72.5 mN/m at 22°C. The pure water was not purged with nitrogen gas to exclude the
atmospheric CO Thus; the pH of the water was within the 5.6-5.8 range.
2.
3.1.4 Contact Angle Measurement and Captive Bubble Technique
A plastic sheet was immersed in a solution, and an air bubble, generated by pushing
air through a syringe with a curved needle, was brought to contact with the solid sample
from below. The contact angle was measured through the aqueous phase by means of a
goniometer. More detailed procedure already been described in Chapter II.
3.1.5 Micro-flotation
Micro-flotation tests were conducted with various plastics powder in the presence of
different wetting agents. The flotation experiments were performed on individual plastics
rather than mixtures thereof.
3.1.5.1 Reagents
A 0.01% saponin solution (1g in 100 ml water) was used as a wetting agent to control
the wettability (contact angle) of plastics, while MIBC (4-Methyl-2-pentanol)from Sigma
Aldrich (98% purity) was used as frother. Both reagents were of commercial grade and
used without further purification. Stock solutions of these reagents were prepared each
day. All flotation tests were conducted using pure water at pH 5.6.
3.1.5.2 Test Procedure:
1g of plastic powder (which was purchased from Fisher Scientific) was fed into an 80
ml flotation cell and mixed with 60 ml of water. A glass stirring bar was used to agitate
32 |
Virginia Tech | the mixture. 6 ml of collector and a drop of frother were then added. The plastic
suspension thus obtained was aerated for 3 min in order to completely mix the test
sample with the reagents. While the plastic suspension was being aerated, the flotation
froth was continuously removed. The concentrates obtained through flotation were dried
in an oven for 1 hour. The flotation recovery was calculated by comparing the weight
fraction of the plastic powder present at the beginning of the floatation test. All
experiments were carried out at the natural water pH, and at room temperature.
3.2 Results
The separation of plastics by flotation was based on hydrophobicity (or wettability)
control. Therefore, it was essential to characterize plastics in terms of their hydrophobic
properties or contact angles. In the present work, the contact angle measurements were
conducted on PS and PVC using both the sessile drop and the captive bubble techniques.
Contact angles were initially measured on virgin polymers in the absence of any wetting
agent using the sessile drop technique. The sessile drop contact angles measured on PS
and PVC plates were 74.5° and 70.8°, respectively.
3.2.1 Effect of Different Wetting Agents on Contact Angles of PS
3.2.1.1 SL, DSS, Saponin and MC
The contact angle values of PS in the presence of different wetting agents are given in
Table 3.1. In general, when using Sodium Lignolsulfonate(SL), Dioctyl sodium
sulfosuccinate (DSS), saponin and methyl cellulose (MC) the contact angles decreased
with increasing surfactant concentration. The largest change in contact angle was
obtained with DSS as a wetting agent. The contact angle dropped from 74.4º to 8.0º as
the concentration was increased from 0 to 0.3%. The case of saponin, however, it was
particularly interesting: even though the smallest achievable contact angle was higher
than with DSS, a significantly large decrease in contact angle was obtained at
concentrations as low as 0.01%. Figure 3.1 shows the variation of the PS contact angle as
a function of the surfactant concentration.
33 |
Virginia Tech | 1.0 INTRODUCTION
1.1 Background
1.1.1 Preamble
Flotation, or more specifically froth flotation, is a physiochemical method of
concentrating finely ground ores. The basis of this process is the utilization of the
differences in the natural or created hydrophobicity of the two minerals that are to be
separated. The separation involves chemical treatment of an ore pulp to create conditions
favorable for the attachment of selected mineral particles to air bubble. The combined
air/mineral particles have specific gravities less than the pulp and will float to the surface,
forming a mineral laden froth that can be skimmed off from the flotation unit while the
other minerals remain submerged in the pulp.
Flotation is the most important single mineral concentrating operation. It is also
the most efficient, most complicated, most sensitive, most challenging and least
understood mineral processing operation today. The art of flotation has been correctly
called the most impressive wide-scale technical application of classical surface
chemistry. Only in the last three decades has much progress been made in understanding
its mechanisms.
Over 95% of all base metals are concentrated by flotation. In coal preparation
plants using flotation, approximately 5-15% of the raw coal feed is processed through the
flotation unit for recovery of fine coal particles. Virtually the entire free-world supply of
copper, lead, zinc, and silver is recovered in the froth of flotation machines. As the grade
(the assay of valuable constituent) of mineral deposits decreases and coal and metal
1 |
Virginia Tech | prices rise, the importance of flotation increases. In turn, the necessity of improving
selective recoveries becomes even more critical.
1.1.2 Coal Flotation
The basis for flotation is the creation or enhancement of the differences in
hydrophobicity. By making specific mineral surfaces hydrophobic chemically, a physical
separation of different particles can be achieved when air is injected into the slurry. The
basic difference between ore and coal flotation is that for ores the entire tonnage is
ground to flotation size, whereas for coal only the fine fraction (about 60 mesh x 0) not
processed by gravity concentration is treated by flotation.
In coal flotation, the differences in the wettability of coal (hydrophobic) and rock
(hydrophilic) are exploited. Hydrophobic particles strike bubbles and attach (see Figure
1), while hydrophilic particles do not attach. This attachment occurs when air is injected
into the slurry containing coal and rock. The coal particles attach to the air bubbles and
FFFrrrooottthhh
FFFeeeeeeddd
TTTaaaiiilllsss
Figure 1. Attachment of coal particles to a bubble in froth flotation.
2 |
Virginia Tech | are buoyed to the pulp surface. The rock particles remain in the slurry and are rejected as
tailings. The pulp phase provides an opportunity for particles to selectively attach to the
air bubbles. The froth phase is used to separate bubble-particle aggregates from the pulp.
The last objective is to form at the surface of the cell a semi-stable mineral-laden
froth that is fragile enough to allow entrapped rock materials to fall and stable enough to
retain the mineral particles selectively. The froth must also have sufficient strength to
permit mechanical or natural removal from the machine. Once the froth is out of the
flotation cell, it should break down rapidly to allow pumping of the concentrate. The
mineral-laden froth is also referred to as concentrate or in coal flotation “the float
product”.
1.1.3 Flotation Machines
Froth flotation is performed in machines specifically designed for that purpose.
There are two basic types of flotation machines and adaptations of these machines.
Conventional (mechanical) and column flotation cells are the two types of flotation
machines used in flotation (Figure 2).
Most of the installations in the coal industry use mechanical (conventional)
flotation machines. Conventional cells are open top vessels connected in rows (banks)
through which the pulp flows continuously from the head (feed) end to the tail end. These
machines are typically arranged in banks of 4-6 cells. The cells are individually agitated
using an impeller. The impeller provides the energy needed to keep particles in
suspension, suck in air (a blower may also be used) and to disperse air into small air
bubbles.
3 |
Virginia Tech | attach to the air bubble well. The collector should be applied as far back in the flotation
circuit as possible because conditioning time with the flotation feed is very important.
The conditioning time becomes even more important with harder to float coal such as low
rank bituminous or oxidized coal. Conditioning time in a flotation circuit may range from
seconds to minutes.
The most common types of collectors for coal flotation are diesel fuel, kerosene,
and fuel oil. There are also environmentally friendly collectors available such as bio-
diesel and tree products. The collector dosages range from 0 to 5 lb/ton. Higher rank
bituminous coal may require no collector, low rank bituminous may require as much as 3
lb/ton of coal and weathered (oxidized) coal may require 5 lb/ton of coal.
The main purpose of the frothing agent is the creation of a froth capable of
carrying the mineral-laden bubbles until they can be removed from the flotation machine.
This objective is accomplished by imparting temporary toughness to the covering film of
the bubble. The life of the individual bubble is thus prolonged until it can be further
stabilized by adherence of mineral particles and joined with other bubbles at the pulp
surface to form a froth. Once it is withdrawn from the flotation machine, however, the
froth should break down rapidly, to prevent interference with subsequent processing
operations. The frother decreases the size of the bubble in the pulp, thus increasing the
surface area of the air introduced into the flotation machine. This enlarges the number of
collisions of the mineral particles with the bubble and chance of the mineral to be floated
to the cell surface.
The ability to lower the surface tension of water is a characteristic of all frothers.
Since most organic compounds can do so to at least a limited degree, the number of
5 |
Virginia Tech | frothers commercially available might be expected to be quite large. However, the
following requirements on the frothing reagent have restricted the frothers to only a few
chemicals. A frother must:
• be low in cost,
• be readily available,
• be effective in low concentrations,
• have no collecting properties or have collecting affinity to the valuable mineral,
• be non-toxic,
• not have a repulsive odor,
• produce a persistent froth that will last the length of the flotation bank, but that
will collapse when skimmed into the launder,
• be stable during storage even at adverse atmospheric conditions (low and high
temperatures),
• have strength enough to withstand the turbulence in the flotation cells and to carry
the valuable mineral, and
• be brittle enough to allow the rejection of unwanted gangue material.
Frothers may consist of short-chain alcohols, derivatives, glycols, and glycol mixtures.
The dosages of frother for conventional cells should typically be in the range of 7 to 15
ppm. The dosages for a column cell, which are usually higher, should typically be from
10 to 25 ppm.
1.2 Flotation Kinetics
Flotation performance is dependent on rate constant (k), retention time (τ), and
cell mixing (Pe). The interrelationship between these variables is described in greater
detail later in this document. Retention time is the amount of time the coal particle has to
stay in the pulp. In a coal flotation circuit, 3½ to 4 minutes of residence time is needed
for good recoveries. Mean retention time can be determined by using retention time
distribution (RTD) testing or by using the formula:
6 |
Virginia Tech | τ = V/Q [1]
in which V is the active volume of the cell and Q is the flow rate to the cell. The number
and size of the flotation cells should be adjusted to have the proper amount of time for
flotation. RTD testing involves using a tracer solution, such as salt, applied to the feed
end of a flotation cell and taking timed samples vs. concentration of each sample at the
tails of the cell.
Cell mixing is the last but not least of the areas affecting flotation. The best
flotation recoveries would come from a plug flow system rather than a well mixed
system. Conventional cells are set up to try and obtain plug flow using cell to cell.
Column flotation is well mixed with a greater chance of short circuiting.
The rate constant (k) indicates how fast coal floats and is dependent on coal
type/size/solids content/feed rate/froth depth, reagent type/dosage, gas flow rate and
sparging/agitating system. Some of the factors that can impact flotation kinetics are
described below.
1.2.1 Coal Rank
Floatability is very dependent on the rank of coal. Low volatile, mid volatile, high
volatile, anthracite, or oxidized (weathered) coals all float differently. Low volatile coal
is the easiest to float and oxidized coal is the hardest to float. Oxidized coal causes
problems with poor recoveries even when blended with higher rank coals.
1.2.2 Particle Size
Coal flotation is performed mainly on about 60 mesh x 0 coals. The size
consistency within the 60 mesh x 0 is very important to coal flotation. The courser the
7 |
Virginia Tech | coal the harder it is to keep attached to the air bubble. The finer the coal is will lead to
poor air bubble particle collision.
1.2.3 Solids Content
The solids content of a flotation circuit should be around 3-7% depending on the
classification system. At higher the solids contents, a trade-off between residence time
and entrainment begins to occur.
1.2.4 Froth Depth
Froth depth changes how wet or dry the froth will be for a given reagent dosage.
A deep dry froth reduces entrainment and clean coal ash. However, a very deep froth may
become unstable and will lower recovery. The type of frother should be tested to
determine the best product for each flotation circuit. Dosages of frother should be
determined to optimize the flotation circuit. The types and dosages of frother can increase
or decrease the flotation rate.
1.2.5 Aeration Rate
The gas flow rate should be sufficient to provide the bubble surface area needed
to carry out the particles. Conventional cells should use about 3.5 cfm/ft2 and columns
about 5.0 cfm/ft2. The rate the coal floats decreases with low gas flow rates.
1.2.6 Sparging System
The sparging/agitating systems should be sufficient to keep the coal particles
moving for attachment to the air bubble. The energy supplied to the flotation unit should
be sufficient to accomplish this task.
8 |
Virginia Tech | 1.3 Flotation Modeling
Despite being a complex process, flotation can often be modeled using simple
kinetic expressions. For a batch flotation cell, recovery can be predicted based on:
R = 1-exp(-kτ) [2]
in which R is the factional recovery, k is the rate constant (1/min) and τ is the mean
retention time (min). Flotation feeds are typically assumed to contain three components:
fast floating (e.g., pure coal)
slow floating (e.g., middlings)
non-floating (e.g., pure rock)
Therefore, mathematically, total recovery (R ) can be calculated from the factional
tot
recoveries of fast (R), slow (R ) and non-floating (R ) components using:
f s n
R = R + R + R [3]
tot f s n
R = φ (1-exp{-kτ}) [4]
f f f
R = φ (1-exp{k τ}) [5]
s s s
R = φ (1-exp{-k τ}) [6]
n n n
in which φ, φ and φ , are the respective mass fractions of fast, slow and non-floating
f s n
solids and k, k and k are the corresponding rate constants of fast, slow and non-floating
f s n
components. In practice, the faction of non-floating solids can usually be determined
from kinetics testing and release analysis.
Kinetics testing, or timed flotation, involves collecting froth products as a
function of time (see Figure 3(a)). The froth products collected from this procedure
include both floatable solids as well as entrained solids. Release analysis, on the other
9 |
Virginia Tech | Feed
Products Collected
Clean #1 (1/2 min) As a Function of
Cumulative Time
Clean #2 (1 min)
Clean #3 (2 min)
Clean #4 (4 min)
a) Kinetics Test Tail
Step I Step II
Feed
Separation of Separation of
Nonfloatables Clean #1 Floatables
Clean #2
Clean #3
Clean #4
Tail #1
b) Release Test
Tail #2
Figure 3. Schematic of laboratory flotation procedures used in (a) timed kinetics
testing and (b) release analysis testing.
hand, is often used to establish the best possible separation results for flotation. It is used
for plant design and for assessing flotation efficiency. As shown in Figure 3(b), the
process of running a release analysis consists of two steps. The first step is to separate the
floating from the non-floating coal. Step two the fast floating coal to slow floating coals
are separated into grades and results are calculated to give grade recovery data and plots
associated with yield and recovery. The final froth products collected from this procedure
include only floatable solids since entrained solids have been removed upstream. Table 1
shows a representative set of kinetics data for a typical coal flotation system.
10 |
Virginia Tech | Table 1. Example of kinetic parameters for a typical coal flotation system.
Parameter Fast Floating Slow Floating Non-Floating
Mass Fraction 0.30 0.40 0.30
Ash Content 3.5% 10.0% 85%
Rate Constant 0.80 min-1 0.30 min-1 0.002 min-1
1.4 Objective and Approach
Froth flotation is widely considered to be the most practical and economical
method for upgrading coal fines that are too small to be treated using density-based
separators. Most of the installations in the coal industry use mechanical (conventional)
flotation machines. The major shortcoming of conventional flotation technology is that it
allows some ultrafine impurities (e.g., clay) to be hydraulically carried into the clean coal
product in the process water that reports to the froth. This contamination substantially
reduces the overall quality of the clean coal froth product. The entrainment problem has
forced many coal operations to install column flotation cells in place of conventional
flotation machines (Davis et al., 1995). Column cells are able to reduce entrainment by
the addition of a countercurrent flow of wash water to the top of the froth. The water
rinses ultrafine impurities such as clay from the froth back into the flotation pulp. Studies
indicate that less than 1% of the feed slurry reports to the froth in a properly operated
column. However, a recent field survey showed that while column technology can indeed
improve clean coal quality, many industrial installations are unable to provide good coal
recoveries (>80-85%) when compared to their conventional counterparts.
11 |
Virginia Tech | water. Nevertheless, several operators have identified suitable niches where the
advantages of columns can be successfully exploited.
2.1.2 Commercial Installations
Table 2 provides a listing of some of the full-scale installations of columns that
are currently in commercial service in the U.S. coal industry. As shown, the most popular
brands of columns include Microcel, Eriez/CPT (CoalPro) and Jameson. Although the
Jameson cell is not really a “column”, it is included in this list since it typically uses wash
water to improve ash rejection. Details related to the specific design features of the
Table 2. Examples of U.S. commercial coal column installations.
Number Diameter Height
Company / Location Brand Units (ft) (ft) Size
ANR – White Tail Eriez/CPT 5 14 28 100 M x 0
ANR – Middle Fork Microcel 5 10 25 100 M x 0
ANR - Holston Microcel 1 14 30 100 M x 0
ANR - Roxanna Microcel 2 13 28 65 M x 0
ANR – Toms Creek Microcel 2 14 28 100 M x 0
ANR – Brooks Run Microcel 2 15 35 100 M x 0
Carbontronics Fuel - Pawney Jameson 1 16 12 100 M x 0
Carbontronics Fuel - Gibraltar Jameson 1 16 12 100 M x 0
Carbontronics Fuel - Linville Jameson 1 16 12 100 M x 0
Cline Resources – Coal Clean Eriez/CPT 3 15 28 100 M x 0
CONSOL Energy - Keystone Jameson 1 16 12 100 M x 0
CQ, Inc. – Ginger Hill Jameson 1 16 12 100 M x 0
CQ, Inc. – Pleasant Ridge Jameson 1 16 12 100 M x 0
ICG – Marrowbone Microcel 1 8 26 150 M x 0
Massey – Lady Dunn Microcel 3 14 30 100 M x 0
Massey– Power Mountain Eriez/CPT 2 13 24 100x325 M
Massey– Liberty Eriez/CPT 3 13 24 100x325 M
Massey– Bandmill EIMCO 2 13 24 100x325 M
Massey– Marfork Jameson 4 16 12 100 M x 0
Ohio Valley - Century Eriez/CPT 2 14 24 100x325 M
Peabody – Sugar Camp Jameson 1 16 12 100 M x 0
Powell Mountain – Mayflower Ken-Flote 4 8 22 100 M x 0
Sigmon Mining – Sigmon Eriez/CPT 2 14 28 100 M x 0
TECO– Clintwood Elkhorn Eriez/CPT 2 10 24 100x325 M
15 |
Virginia Tech | various column technologies are available in the literature (McKay et al., 1988; Yoon et
al., 1992; Davis, 1995; Manlapig et al., 1993; Finch and Dobby, 1990; Finch, 1995;
Rubinstein, 1995; Wyslouvil, 1997). Due to economy of scale, recent trends in the coal
industry have shifted toward the installation of smaller numbers of very large (>14 ft)
diameter units. Although most of the installations involve the treatment of minus 100
mesh feeds, some operations have been able to justify the use of columns to treat
deslimed 100x325 mesh streams. In such cases, the columns have been used to further
reject high-ash slimes that bypass to the deslime underflow.
2.2 Column Operation and Design
2.2.1 Separation Performance
The primary advantage of column cells is their ability to achieve a higher level of
separation performance compared to conventional flotation machines. For example,
Figure 3 provides a comparison of separation results obtained using both conventional
and column flotation cells at an industrial plant site. The column data were obtained
using both a laboratory (2 inch diameter) test column and a full-scale (10 ft diameter)
production column. For comparison, test data were also obtained from a single bank of
five full-scale Denver cells (300 ft3 each) already in operation at the preparation plant. A
release analysis test was also conducted on the flotation feed to provide a baseline for
comparative purposes (Dell et al., 1972). Release analysis is often described as the
flotation equivalent of the washability procedure for density separations.
Several important conclusions may be drawn from the results given in Figure 6.
First, the separation data obtained using the column cells fall along nearly identical
16 |
Virginia Tech | performance that would be difficult to achieve even after multiple stages of cleaning by
conventional machines. For this particular example, the column cells reduced the
concentrate ash by nearly 8 percentage points compared to the existing plant rougher
bank, while increasing combustible recovery by approximately 20 percentage points.
2.2.2 Slurry Residence Time
All flotation cells must have adequate slurry residence time in order to achieve an
acceptable recovery of clean coal. The mean residence time can be calculated by dividing
the active volume of the flotation pulp by the volumetric flow rate of refuse. For a typical
100 mesh x 0 feed, a residence time of 3.5-4.0 minutes is often adequate for 3-5 cell bank
of conventional flotation machines. The cell-to-cell arrangement used by conventional
machines minimizes the bypass of feed slurry to tailings. In contrast, column cells are
usually arranged in parallel, which can allow some floatable coal to bypass to tailings. As
a result, columns typically require significantly more residence time to achieve the same
level of coal recovery (Dobby and Finch, 1985; Mankosa et al., 1992). This situation is
particularly serious for the relatively low profile columns that are typically employed in
the coal industry This problem can lead to lower recoveries of coal if not accounted for
during the column design process.
The detailed procedure for estimating the column residence time required to
achieve a given recovery has been presented in the literature (Dobby and Finch, 1986a).
This procedure is complicated and well beyond the scope of this article. However, as a
rough rule-of-thumb, a column cell will typically require approximately twice the
residence time of a 4-cell bank of conventional cells and three times the residence time of
a batch laboratory flotation cell. Therefore, it would not be unusual for a column cell to
18 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.