University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | It is also critical that the spherical agglomerates are cleaned before undergoing
centrifugation. With a homogenous coal slurry, the light weight ultrafines can produce a slime
coating on the inside of the cake. In the case of pentane-agglomerated coal, the slimes form on
the outside of the cake since the agglomerates float. This impermeable layer prevents the
agglomerates from drying correctly, so it is essential that the agglomerates are thoroughly
cleaned.
There are several problems with centrifugation. First, though the centrifuge was meant to
be sealed, small amounts of pentane leaked from the top bearing under high pressures. Also,
some pentane evaporated when the sample was open to atmosphere during pouring. Pentane
evaporation allows coal to readsorb water and weakens the agglomerate structure. Without
pentane, the agglomerates will powderize at high speeds. This powder is more difficult to
dewater and may be lost between the centrifuge bars.
Second, in order to produce a relatively even cake, an insert is required in the middle of
the centrifuge. The blank forced the agglomerates into a trough. Without the insert, the
agglomerates would form a wide wedge at the base. Water drains too quickly, and the
agglomerates have no time to travel up the sides of the chamber on the water. Increasing the
speed did not make a difference. Since thin, even cakes traditionally provide lower moistures, the
insert was added and the centrifuge was fed while still. It was fed while off to provide an
opportunity for the agglomerates to fill the mold.
5.3.5. Filtering
Filtration of spherical agglomerates was meant to provide a baseline of moistures to
compare centrifugation against. It resulted in moistures varying from 20.1 to 32%. Moisture was
reduced by removing ultrafines and preventing oxidation (Table 4-12). No material was lost to
tailings with this dewatering technique since pre-cleaned agglomerates were utilized.
5.3.6. Displacement
Dewatering by displacement does not utilize oil agglomeration; however, it is based on a
similar principle in which oil rejects water. In each test, a coal slurry was filtered with a top layer
of pentane. In theory, pentane should displace the last droplets of water as it filters through the
64 |
Virginia Tech | cake; however, it did not work as intended, and resulted in moistures of 20-35%. This is a
standard range of moistures for filtration; therefore, pentane did not appear to have any impact.
There were a couple reasons why displacement did not work as intended. First, even
when the pentane was poured slowly, coal-coated droplets of water tended to form and sit on top
of slurry-oil interface. As filtration occurred, these droplets came to rest on the cake surface
while the pentane filtered around them. They slowly filtered after the pentane was gone. Second,
pentane likely filtered through the path of least resistance. While it is true that pentane rejects
water, the solids interfere with the interface. Instead of a flat interface filtering down, the solids
provide voids for droplets of water to exist. Once a path has been made, it is easier for the
pentane to filter around these droplets than to displace them all.
5.4. Conventional Cleaning and Dewatering
Conventional cleaning and dewatering were performed to baseline all of the oil
agglomeration testing. They were only performed on the original 100 mesh x 0 sample. The
release analysis indicated that Tomβs Creek is an easy-to-clean, black and white coal with good
liberation; however, it was not as easy to dewater. Even with a dewatering aid, the lowest
obtained moisture was 24.8%.
Conventional vacuum filtration was also used to calculate the void space in each of the
three main coal samples. The cake simulated the compact coal composing the agglomerates, and
the calculated void space was used to determine the pentane to void space ratios.
5.5. General Comparison of Methods
Of the six dewatering methods considered, centrifugation consistently provided the
lowest moistures and highest recoveries. It achieved the goal of single digit moisture, unlike the
other methods. Centrifugation produced a 7.5% moisture and 70.9% combustible recovery with
the 80 mesh x 0 sample. Though the moisture increased with the other samples, recovery was
generally in the high eighties and nineties. Hand shakingβs lowest moisture was 16.2% with a
recovery of only 53.3%. The lowest moistures occurred at a dip in the recovery curve, unlike
centrifugation which had a relatively consistent recovery. Most of the variations in centrifuge
recovery were due to poor feeding; the centrifuge feed tube became blocked if it was fed too
65 |
Virginia Tech | quickly. These variations could easily be eliminated to maintain high recovery. Unfortunately,
low recovery cannot be eliminated for hand shaking.
Filtration and displacement were the next methods. They both produced moistures from
20-35%. These are standard vacuum filtration values and do not represent an improvement from
conventional methods. Screening was the least practical of the five tested methods. Though it
provided low, single-digit moisture, the recoveries were also single digit. The low recoveries and
equipment difficulties prevent it from being a viable option. Finally, air classification never
underwent quantitative testing. It needed special equipment and was deemed impractical in light
of the centrifugation option.
Though moisture varied between the methods, they share some common points. In
general it appears as though spherical agglomeration is preferable because it enables fine coal to
be treated as coarse. Oxidation plays a key role in moisture elevation. Fresh samples contained
significantly less water. Finally, size distribution also impacted the moisture; an increase in
ultrafines resulted in higher moistures. However, oxidation was much more important. For
example, in centrifugation the original 100 mesh x 0 sample was cleaned from 72% to 49%
minus 325 mesh, the screened 100 mesh x 0 cleaned from 28% to 19% minus 325 mesh, and the
pulverized 80 mesh x 0 from 34% to 32% ultrafines. The screened sample had the least
ultrafines, yet the unoxidized 80 mesh x 0 provided the lowest moistures.
Ash content did not appear to have an impact on moisture, which is reasonable
considering it was being cleaned before dewatering. The dry pulverized sample had a feed ash of
19% and a concentrate of 8%. It had the highest concentrate ash yet the lowest moisture. In
contrast, the original 100 mesh x 0 had a feed of 38% and concentrate of 4% ash while the
screened 100 mesh x 0 had a feed of 19% and similar concentrate ash of 3.9%.
The ability to achieve single digit moisture with centrifugation indicates that oil, not
water is the bridging liquid; therefore, there is not an inherent amount of water needed for
agglomeration to take place. The majority of the moisture occurring in the processes originated
as free water droplets trapped between the agglomerates or as small droplets attached to the
outside of the agglomerates. However, some moisture may remain in the agglomerates since the
coal-to-void space volume ratios were only .50 to .65 for spherical agglomeration. This assumes
the agglomerated coal was not packed more tightly than the vacuum cake. Unfortunately, the oil
dosage cannot be increased to fill this void space because the agglomerates become soft and trap
66 |
Virginia Tech | large globules of water between themselves. It should be noted that some of the volume ratios
are higher than reasonable; therefore, the void space determination may be inaccurate and should
be re-examined.
The original thermodynamics calculated for this project indicated that dewatering should
be spontaneous for any contact angle greater than 90 degrees (Sohn, 1997). Accordingly, the
concentrates should contain no moisture since pentaneβs contact angle with coal in water is about
106 degrees. Though the thermodynamics examining the end states of the process are correct,
they fail to take into consideration contact angle hysteresis. Hysteresis refers to the range of
contact angles existing between the retreating and advancing angles. The measured contact angle
likely represents an advancing angle; however, during draining a receding contact angle will
occur. This angle may be lower than 90 degrees and would account for the inability of pentane to
remove all water. With this lower contact angle, the process does not produce negative free
energy. It is no longer thermodynamically favorable and spontaneous.
67 |
Virginia Tech | 6. GENERAL SUMMARY
This purpose of this project was to seek alternatives to conventional fine to ultrafine coal
cleaning and dewatering using hydrophobic displacement by pentane to produce moistures less
than ten percent. Six dewatering methods were examined: hand shaking, screening, air
classification, centrifugation, filtration, and displacement. The first five methods utilized oil
agglomeration in order to perform hydrophobic displacement during the cleaning stage. The
dewatering stage focused on removing the resulting oil-covered agglomerates from bulk water.
These methods sought to use oil agglomeration as a basis for a single solid-solid, solid-liquid
process which combined cleaning and dewatering. The last method, displacement, was the only
process which performed hydrophobic displacement during the dewatering stage.
The desired bench-scale and batch testing were completed and succeeded in meeting the
moisture requirement. Several pertinent conclusions are included below:
1. Spherical agglomeration provided most durable agglomerates for dewatering. They
essentially enabled fine coal to be treated as coarse coal. Spherical agglomerates of 1-2
mm in diameter were produced with ratios of .21-.34 M :M , .42-.70 V :V ,
pentane coal pentane coal
and .50-.67 V :V .
pentane void
2. A weighing platform provided a flexible way to determine water-only moisture when
evaporation times varied.
3. Cleaning with spherical oil agglomeration and dewatering by centrifugation produced the
best moistures and recoveries. Centrifugation with the pulverized 80 mesh x 0 Tomβs
Creek clean coal produced a moisture 7.5% with a combustible recovery of 70.9%,
rejection of 76.9%, and concentrate ash of 7.71%. Recoveries were usually in the mid
eighties to nineties, regardless of oil dosage.
4. Hand shaking produced moistures as low at 16.2%; however, low moistures were
associated with a dip in recovery. There is a great deal of error in this process due to its
arbitrary nature.
5. Filtration and displacement produced moisture of 20-35%, and they did not show an
improvement in conventional methods.
6. Though screening could produce single-digit moistures, it was not a practical process.
The recoveries were too low and the screens risked blinding.
68 |
Virginia Tech | 7. RECOMMENDATIONS FOR FUTURE WORK
There are five recommended trends for future work on this project. First, given the
promising single-digit moistures produced by centrifugation with spherical agglomeration, a
continuous version of this method should be explored. This would involve creating a closed
system mixing chamber, cleaning chamber, and sealed centrifuge. Due to the high consumption
of pentane in such a system, oil recovery by condensation should also be employed.
Second, considering the dependence of centrifuge moisture on void space, the void space
determination method should be re-evaluated. Proctor testing should be performed to determine
optimum void space and packing. Also, packing in pentane, not just water, should be examined.
Third, cleaning and dewatering by displacement should be re-examined. The theory is
sound though the displacement-filtering method attempted here did not work. Instead, large
amounts of oil should be used to create separate oil and water phases. Coal will gravitate to the
interface, and if the surface area of this crowded interface is suddenly decreased, by jigging with
inverse triangular bars for example, the extra coal particles will be ejected into the pentane phase
where they may be removed. This material should contain little to no water since each particle is
completely enveloped in oil. Unfortunately, it takes very little agitation to create an emulsion or
coal-coated water droplets in the pentane phase. The key will be to produce a gentle mechanical
motion that can still handle large coal throughputs. Though a similar test tube method was
attempted with dry coal in water and pentane and produced promising results, nothing was
attempted with a coal slurry.
Fourth, multi-stage dewatering should be examined. This would provide more time for
pentane to displace the remaining water. Also, as the coal is reworked, small water droplets may
coalesce and become easier to remove.
Finally, it is critical that sample oxidization is minimized in order to produce low
moistures. Fresh samples are also more representative of a preparation plant. Ideally, an
enclosed pilot-scale process would be run at a site. If this proves difficult, coal may be ground in
the lab, stored in a freezer, and mixed before each test. The samples should soak for a short time
to encourage wetting before dewatering takes place.
70 |
Virginia Tech | Table BΒ6b. Detailed centrifuge data for 100 mesh x 0 continued
Feeding Drying Ash Content (%)
(sec) (sec) (rpm) (gβs) T1 C
21 30 3280 447 79.70 4.11
22 30 3280 447 81.20 4.11
53 30 3280 447 77.69 4.00
29 60 3280 447 81.41 4.08
20 60 3280 447 79.12 3.93
22 60 3280 447 76.93 4.07
150 60 3280 447 75.73 3.74
26 60 3280 447 73.78 3.82
22 15 880 32 75.65 3.95
23 30 880 32 83.41 4.15
64 60 880 32 81.98 4.05
55 120 880 32 84.94 4.15
90 15 3280 447 82.05 4.00
90 30 3280 447 80.17 3.87
70 60 3280 447 81.04 3.89
80 120 3280 447 81.06 3.92
70 15 2040 173 79.00 0.90
58 30 2040 173 80.76 3.92
58 60 2040 173 85.15 4.04
120 120 2040 173 82.44 4.08
*Note: Each test was conducted with 900 ml and 6.0% solids. During feeding, the centrifuge was
still except for the 8th test which ran at 880 rpm. L2 and L1 are assumed to have the same ash as C.
Abbreviations: T1βcleaning tailings, L1βlost while feeding the centrifuge, L2βcake lost to the centrifuge
basket and not used in the moisture calculation, and Cβcentrifuge concentrate.
102 |
Virginia Tech | Table BΒ7c. Detailed centrifuge data for screened 100 mesh x 0 continued
Dry Weight (%)
T1 L1 L2 C Total
5.72 0.18 0.70 26.96 33.56
5.73 0.25 1.01 25.61 32.60
5.77 0.21 1.07 26.13 33.18
5.69 2.98 1.44 21.68 31.79
*Note: Each test was conducted with 600 ml and 6.1% solids. No centrifuge tailings (T2) were
produced. Abbreviations: T1βcleaning tailings, L1βlost while feeding the centrifuge, L2βcake lost to
the centrifuge basket and not used in the moisture calculation, and Cβcentrifuge concentrate.
Table BΒ7d. Detailed centrifuge data for screened 100 mesh x 0 continued
Yield (%) Recovery (%) Ash Rejection Concentrate Moisture
Cleaning Centrifuge Cleaning Centrifuge (%) (g) (%)
81.7 100 97.5 100 83.6 3.9 12.6
81.6 100 97.4 100 83.6 3.9 13.3
81.7 100 97.5 100 83.5 4.0 13.2
81.7 100 97.5 100 83.6 3.5 14.0
*Note: Each test was conducted with 600 ml and 6.1% solids. The yields were calculated based on
ash content. The centrifuge yield and combustible recovery are 100% because there were no
measureable centrifuge tailings. The ash rejection refers to the cleaning stage since no tailings were
produced in the centrifuge. The concentrate for the cleaning stage calculations include the
centrifuge concentrate and the material lost during centrifugation.
106 |
Virginia Tech | Table BΒ8a. Detailed centrifuge data for pulverized 80 mesh x 0 clean coal
Mass Ratio Volume Ratio
Run Test
(Pentane:Solids) (Pentane:Coal) (Pentane:Coal) (Pentane:Voids)
c5 21 0.17 0.21 0.43 0.67
22 0.17 0.21 0.43 0.67
23 0.17 0.21 0.43 0.67
24 0.17 0.21 0.43 0.67
c6 25 0.17 0.21 0.42 0.65
26 0.17 0.21 0.42 0.65
27 0.17 0.21 0.42 0.65
28 0.17 0.21 0.42 0.65
*Note: Each test was conducted with 900 ml and 6.0% solids. Spherical agglomeration was performed
with 14.8 ml, and no pentane was added after cleaning. Coal in c5 soaked 1 min, and coal in c6 soaked
3 days.
Table BΒ8b. Detailed centrifuge data for pulverized 80 mesh x 0 clean coal continued
Feeding Drying Ash Content (%)
(sec) (sec) (rpm) (gβs) T1 C
35 5 3280 447 76.36 8.61
30 15 3280 447 79.30 8.02
32 30 3280 447 60.67 7.78
25 60 3280 447 40.45 7.71
30 5 3280 447 77.08 6.59
25 15 3280 447 79.73 7.30
25 30 3280 447 84.69 7.21
30 60 3280 447 79.21 7.33
*Note: Each test was conducted with 900 ml and 6.0% solids. During feeding, the centrifuge was
still. L2 and L1 are assumed to have the same ash as C. Abbreviations: T1βcleaning tailings, L1βlost
while feeding the centrifuge, L2βcake lost to the centrifuge basket and not used in the moisture
calculation, and Cβcentrifuge concentrate.
107 |
Virginia Tech | Discussion
Several trends may be identified from the previous 32 figures. They are numbered below
for ease of reading.
1. As the amount of water increases, the evaporation rate of pentane increases (Figures C-21
through C-24). In other words, the starting evaporation rate is higher, and the pentane
tends to finish evaporating in a shorter amount of time. Pentane has no affinity for water,
and the increased presence of water causes the pentane to evaporate faster in order to
decrease the amount of liquid in contact with water.
2. As the amount of pentane increases, the starting evaporation rate of pentane decreases
and the duration of the gently sloping, moderate evaporate rate of the first leg increases
(Figures C-17 through C-20). In other words, changes in pentaneβs evaporation rate
occur more slowly as the volume of pentane increases since pentane has an affinity for
itself. The increase in pentane also minimizes the amount of pentane in contact with
water; water tends to reject pentane and cause it to evaporate faster.
3. As the amount of coal increases, the evaporation rates of both water and pentane increase
(Figures C-27 and C-28). The liquid is spread thinly over the large surface area which
aids in evaporation and decreases the time needed for liquid removal.
4. Unlike water, pentane is not characterized by a constant evaporation rate (Figure C-21).
Waterβs evaporation rate tends to remain relatively constant or displays small constant
changes. Small variations may be noted at the beginning and likely the end (not seen
here) of water evaporation (Figure C-17). This is consistent with pentaneβs trends;
however, the change in pentaneβs evaporation rate vary more widely.
Pentaneβs weight versus time curves display a distinctive flattening of the slope as the last
of the pentane evaporates. This trend is present when only pentane or pentane mixed in
coal is being tested. Therefore, it is reasonable to assume that a similar flattening of
pentaneβs weight curve occurs when water is present.
130 |
Virginia Tech | Reasons for Initial Weight Choice
The initial weight used in moisture determination was taken after the evaporation
rate became constant in the second leg of the weight curve (Figure 4-4). Traditionally,
this value would be taken at the intersection of the two legs of the graph instead of after
the elbow. Several reasons motivated this choice:
1. Reason 4 in the previous section explained that pentane shows a distinctive
flattening of its evaporation curve when by itself or mixed with coal. It is
assumed that a similar trend occurs when water is present; therefore, pentane is
still present until the elbow of the evaporation curve is complete.
2. Moisture balance testing was done with known amounts of liquids. The known
masses of pentane tended to be associated with the end of the elbow in the
evaporation curve.
3. Due to the expense of pentane, it would need to be recycled in a commercial
setting. Plants would conservatively evaporate a little extra water to ensure that
all pentane was recovered; therefore, it is reasonable that they would evaporate, at
a minimum, to the end of the elbow in the evaporation curve.
131 |
Virginia Tech | PARAMETER EVALUATION AND MODELING OF A FINE COAL
DEWATERING SCREEN-BOWL CENTRIFUGE
Ian Michael Sherrell
(ABSTRACT)
A vast majority of coal and mineral cleaning and upgrading processes involve the
addition of water. The water allows the movement of particles throughout the processing
plant and the upgrading of the material. When the process is complete the finished
product must be dewatered. This is due to storage concerns, in which the water takes up
a majority of the space, and high transportation costs, in which no compensation is
obtained from the buyer for the shipment of the liquid. Dewatering is accomplished by
many devices, with the two most common pieces of equipment being the screen-bowl
centrifuge and disk filter.
This thesis tests and compares the effect of reagents on dewatering using the
screen-bowl centrifuge and disk filter. Coal was obtained from the Upper Banner,
Pittsburgh No. 8, Taggart, and Dorchester seams, crushed and ground to the desired size,
and run through the dewatering circuits. The results showed that the moisture content of
the product can be greatly reduced in the disk filter while being only slightly reduced in
the screen-bowl centrifuge. It was also shown that the recovery can be slightly increased
in the screen-bowl centrifuge. Overall, with the addition of reagents, the disk filter
outperformed the centrifuge in both recovery and moisture content.
A model was also developed for the screen-bowl centrifuge. The results from the
screen-bowl tests helped in the development of this model. This model can be used to
predict the moisture content of the product, the recovery, particle size distribution of the
effluent and particle size distribution of the product. The model also predicted how the
product moisture and recovery were affected by changing the feed flow rate, feed percent
solids, centrifuge speed, and particle size distribution. |
Virginia Tech | ACKNOWLEDGEMENTS
The author would like to express his deepest appreciation to Dr. Gerald H.
Luttrell. His guidance and support throughout this thesis work were incalculable. The
author would also like to thank Dr. Roe-Hoan Yoon for his words of wisdom and
support. The author also thanks Dr. Greg T. Adel for his advice.
The author would like to thank Consol Coal Company, Pittston Coal Company,
Red River Coal Company, and the Department of Energy for their support.
Sincere appreciation is extended to everyone at Plantation Road, including
Ramazan Asmatulu, Matt Eisenmann, Jaisen Kohmuench, Chris Barbee, Shane Bomar,
and Boldo Luvsansambuu. A special thanks to Wayne and Billy Slusser. Work went
smoothly due to their knowledge and assistance. A heart-felt thank you to Tim and Cathy
Mckeon. With their support, both emotional and physical, this project was made a
pleasure.
The author is also grateful to all of his other friends for their love and support. A
special thank you to Cam Schini and David Gray. They both provided great love,
understanding, sympathy, and tremendous encouragement.
Lastly, the author would like to express his deepest gratitude and appreciation to
his family. Their love, support, and understanding made everything possible.
II I |
Virginia Tech | CHAPTER 1
INTRODUCTION
1.1 Background
1.1.1 Dewatering
Almost all coal processing techniques require the addition of water to form a
slurry. This allows for easy transportation within the processing plant. It also allows for
certain processes, such as flotation, to beneficiate the coal depending on certain
properties of that coal, such as specific gravity. After the coal has been upgraded it is
dewatered. There are a few reasons coal processing plants do this. Due to the way many
coal contracts with power plants are written, water that is sent to the power plant is
effectively considered an inert material (no BTUβs). Therefore, power plants do not pay
for these inert materials although coal companies still pay transportation costs for them.
This is not economically advantageous. Another reason is due to the handling of wet
material. Handling of wet material is much more difficult the finer the coal becomes. In
most instances, the drier the product is, the easier it is to handle. This is true unless it is
in a slurry form. When in this state, transportation costs are increased, which as already
stated is not advantageous, and storage space is infeasible.
There is a variety of equipment in use today to dewater fine coal. Fine coal in this
instance is considered to be below 0.589 mm (-28 mesh). This equipment includes the
screen-bowl centrifuge, disk filter, horizontal belt filter, and drum filter. The most
commonly used today, in the eastern U.S., is the screen-bowl centrifuge followed by the
disk filter.
1.1.2 Cake Beds
Cake beds made from filters and centrifuges are commonly modeled as
completely packed solids with parallel capillaries throughout. These capillaries all vary
in size. This is shown in Figure 1.
1 |
Virginia Tech | Solids Capillary
Figure 1 - Capillaries Within Cake
The LaPlace equation is used to model the dewatering within these capillaries.
This is given by Equation 1.1
( )
2gCos q
D P = Equation 1.1
r
where g is the surface tension of the liquid, J is the contact angle between the solids
liquid and air, r is the radius of the capillary, and D P is the pressure needed to sustain the
height of liquid in the capillary such that
D P = hgr Equation 1.2
where h is the height of the liquid, g is gravity, and r is the density of the liquid.
Therefore, if the pressure difference given by the LaPlace equation can be
lowered then the height of the liquid within the capillary will also be lowered. If the
actual pressure difference, given by operating conditions, is lower than the needed
pressure given by the LaPlace equation then no dewatering will occur. If the actual
pressure difference is higher than the LaPlace equation then dewatering will occur. With
a given pressure difference, there is a limit as to how small the capillaries can be and still
be dewatered. Big capillaries can be dewatered completely but will also lower the
pressure by allowing air to easily pass through the empty capillary.
The way to dewater more at a given pressure is to lower the pressure needed to
dewater given by the LaPlace equation. There are three ways to lower the pressure
needed to dewater. The first is to increase the radius of the capillary. This is very
difficult to achieve and is dependent on particle size distribution. For fine coal, there will
still be many small capillaries that will not be dewatered. The two other factors can be
2 |
Virginia Tech | altered by the addition of reagents. Certain reagents will lower the surface tension of the
liquid and thereby allow the capillaries to be more easily dewatered. Other reagents can
increase the contact angle between the solids, liquid, and gas, which will also lower the
pressure needed to dewater. The relevant contact angle for dewatering is the receding
contact angle.
The equipment used can increase the pressure difference, which aids in
dewatering. Filters do this by making a vacuum on one side of the cake, which increases
the pressure difference between that side of the cake and the atmosphere on the other side
of the cake. Centrifuges increase the pressure difference by increasing the acceleration of
the liquid shown in Equation 1.2. In this instance, g is replaced by B, where B is the
acceleration within the centrifugal field. This higher acceleration increases the pressure
difference, which dewaters the capillary. When the capillary is being dewatered, the
height within the capillary is lowering and therefore the pressure difference is also
lowering. An equilibrium between the pressure created by the centrifugal field and that
given by the LaPlace equation is achieved within the capillary at a lower height. When
this equilibrium is reached no more dewatering occurs.
1.2 Literature Review
There are many factors that can affect the dewatering of minerals in a centrifugal
field. Factors that affect dewatering can differ between different material and even
between different size distributions of the same material.
For coarse coal, the factors that influence product moisture after centrifugation
include the internal moisture of the coal, the specific surface area of the feed, the rank of
the coal, and the amount of ultra-fine particles (Firth et al., 1996). Internal moisture is
the amount of water held in pores inside of the coal structure. All other water adheres
externally to the surface of the coal. The specific surface area is the surface area of the
coal divided by the volume of the coal. Smaller particles will have a greater specific
surface area than larger particles. The rank of the coal is determined by the reflectance of
the coal. It has been shown that with increasing rank there is an associated increase in
hydrophobicity. The association is a near parabolic shaped curve.
3 |
Virginia Tech | Due to the large size difference between coarse and fine coal, the factors that
affect dewatering vary. These factors come into affect due to particle size. For different
particle sizes, different factors have more of an impact on the outcome of dewatering than
other factors. This is not to say that the factors that differ have no impact and are
completely irrelevant to the different sizes, but they have a very diminished impact.
Fine coal factors include specific surface area, total pore volume, quantity of
superfine material, water conductivity, and total dissolved solids (Rong and Hitchin,
1995). Since the specific surface area is a measure of particle size, and it has already
been stated as a factor in coarse coal dewatering, it would be assumed to have an affect
on fine coal dewatering. If it did not, then the change in particle size from coarse to fine,
and hence the change in specific surface area, would have no affect on any particle size.
The total pore volume has more of an effect due to the fact that there is more pore volume
compared to entire volume for fine coals. This allows for more water to βhideβ from
dewatering forces. In a vacuum filter, the water is hidden from the airflow, while in a
centrifuge the water is more easily trapped within the pore. Both water conductivity and
total dissolved solids have been known to have a significant effect on electro-osmotic
dewatering and electro-filtration techniques. It has now been βconcluded that ions in the
process water alter surface properties of the suspended particles and their flocculation
characteristicsβ (Rong and Hitchin, 1995).
Another factor, which affects both fine and coarse particles, is the size distribution
of pores in a packed bed (Hogg, 1995). Within a packed bed is where surface properties
such as contact angle and surface tension affect dewatering. The size distribution of the
pores in a packed bed is directly related to the size distribution of particles within the
feed. The most common approach for the correlation between cake permeability and
cake structure is the use of the Carman-Kozeny model. In this model, the bed is seen as a
bundle of similar capillaries. The model assumes that the pores are discreet, uniform
channels, that each pore has the same effective radius, and for non-uniform pores, the
mean-hydraulic radius can be used. Although the model is commonly used, it does not
account for pore radius distribution. As stated in Section 1.1.2, while bigger pores may be
dewatered, smaller pores may be unaffected. This may lead to a big difference between
the stated model (Carman-Kozeny) and the experimental data. The model also does not
4 |
Virginia Tech | account for moisture trapped in the corners of βbent trianglesβ. These bent triangles are
formed when three solid particles are in contact. The intersection of three different sized
particles can be seen in Figure 2. The resulting area between the particles is far from
circular. The corners of these triangles may trap water even with a low surface tension
and high contact angle. The resulting cake moisture is strongly dependent on both pore
size distribution and pore shape (Ranjan and Hogg, 1996).
Figure 2 - Intersection of Different Sized Particles
Temperature can also have an effect on the dewatering ability of fine coal. This is
accomplished by the reduction of viscosity of water at high temperatures. Forces acting
on the water can more effectively move the water at lower viscosities. The temperature
can also continue to dewater coal after it has been processed by the dewatering
equipment. Adiabatic cooling can effectively use the latent heat of the product to further
dewater the coal (Policow and Orphanos, 1983).
The use of surfactants, as stated earlier, can be used to enhance the dewatering
ability of certain pieces of equipment. Flocculants have been used in vacuum filtration to
further decrease the moisture of the product cake. This has been tried in centrifuges with
minimal results, due to the weakness of the flocculants and the strong forces occurring in
a centrifuge (Mishra, 1988). These strong forces break apart flocculants before they can
aid in dewatering. Even though flocculants cannot be effectively used in centrifuges, fuel
oil has been well known to aid in dewatering. Fuel oil will adsorb onto the coal surface.
This results in a much lower interfacial tension than at the coal-water interface (Mishra,
1988). This lower interfacial tension results in lower pressure differences needed to
dewater.
Some of these factors were incorporated into models that dealt with dewatering in
a centrifuge. One model was specifically intended for the use of dewatering by screen-
5 |
Virginia Tech | bowl centrifuge (Tierney et al., 1983). It depended on many factors, including the size,
shape, and physical properties of the particles, and the characteristics of the equipment.
The model accounted for mechanical degradation within the centrifuge, loss of fines,
changes in size and gravity distribution, changes in ash and sulfur content, and final
product moisture. To obtain all of these values, factors must be fit through experimental
data. This requires that samples must be taken from a centrifuge with the correct
geometry and coal being dewatered. This is no problem for an operating preparation
plant, but for planning purposes assumptions must be made.
Another model, not based on empirical data, was intended for a batch centrifuge.
It included many of the factors that can affect the dewatering of fine coal, including
surface tension and contact angle. This model is from Zeitsch and the formulation can be
found in Solid-Liquid Separation, edited by Svarovsky. The model that was used can be
found in Section 3.3.2.
Zeitschβs model deals with filtration and drainage in a centrifugal field. Filtration
occurs when there is flow of liquid through the cake while the cake is submerged in the
liquid. Drainage occurs when liquid leaves the cake and air replaces the liquid in the
cake. This only occurs when there is no liquid above the cake. The model has been
slightly modified to be used in a continuous centrifuge as seen in Section 3.3.2. In this
section, the filtration model is not being used. The water has been drained away by the
centrifugal forces when the scroll carried the solids up the beach section of the centrifuge.
The model produces a final cake moisture, which is the product from the centrifuge.
1.3 Objectives
The main objective in this project was to determine how different reagents affect
the dewatering capability of the screen-bowl centrifuge. The most important property of
concern was the moisture content of the product, but the effect on recovery was also
considered. When a screen-bowl centrifuge is tested in the laboratory it is commonly
compared against a disk filter, since these two types of equipment make up the dominant
share of fine coal dewatering equipment. Due to this reason, a disk filter was also used
for comparative reasons.
6 |
Virginia Tech | There was also a secondary objective to this project. This included the
development of a population balance model for a screen-bowl centrifuge. No reliable
models could be found within the literature and a model could be helpful in the design
process to determine how variables affect the dewatering process. Also a comparison
was made between the disk filter and the centrifuge, with and without reagents. This was
to determine if a disk filter with the reagents has a greater dewatering ability than a
screen-bowl centrifuge with the reagents.
1.4 Organization
The information in this thesis has been organized into 6 chapters. Chapter 1
includes the background, literature review, objectives, and organization of the thesis.
Chapter 2 is the experimental section. This includes a description of the coal samples and
their preparations, the reagents used, the circuits and sampling processes of both the
centrifuge and disk filter, the sample analysis, and mass balancing of the data. Chapter 3
deals with the screen-bowl centrifuge model. This includes the background to the model,
a description of the process, a description of the model and a simulation of the model.
The results of the centrifuge and disk filter and their comparison are shown and discussed
in Chapter 4. Chapter 4 also compares the model with the laboratory data collected. The
conclusions to this project are stated in Chapter 5 and Chapter 6 contains the
recommendations for future work that might result from this thesis.
7 |
Virginia Tech | CHAPTER 2
EXPERIMENTAL
2.1 Coal Samples
Four samples were obtained from three different coal companies. A dense
medium cyclone product was obtained from Moss 3 preparation plant owned by Pittston
Coal Company. This was from the Upper Banner seam. Another dense medium cyclone
product was obtained from Consol Coal Company. This sample was from the Pittsburgh
No.8 seam. The last samples obtained came from Red River Coal Company. These two
samples were from the cyclone overflow to the thickener feed. The first sample had a top
size of 100 mesh (0.147 mm) and was from the Taggart seam. The second sample had a
top size of 65 mesh (0.208 mm) and was from the Dorchester seam.
2.2 Reagents
Two different reagents were used in these tests. The chemical names of these
reagents were not given, due to the possibility of patenting, but were only referred to as U
and W. All coals were tested with reagent U except for the second sample from Red
River Coal Company (Dorchester seam), which used reagent W. Both reagents increase
the contact angle of the coal.
2.3 Sample Preparation
All of the dense medium cyclone samples needed to be ground down before they
could be further used in the fine coal dewatering equipment. The grinding circuit is
shown in Figure 3.
8 |
Virginia Tech | Feed
Hammer Screen
Mill Water
Product
Hopper
Conveyor Ball
Mill
Sand Pump
Figure 3 - Grinding Circuit
The coal was first crushed in a hammer mill to allow easy feeding into the ball mill.
After crushing the coal was fed onto a conveyor, which emptied into a hopper. The
hopper fed an 18β ball mill at a constant rate. Water was added at the head of the ball
mill. The product from the ball mill was funneled into a sand pump, which fed a Sweco
vibratory screen on the second floor. A 28-mesh screen was used. The undersize from
the screen fell into a drum on the first floor, which was the product from the grinding
circuit. The drums that contained the undersize were as free of rust as possible since it is
thought that rust can inhibit the dewatering ability of these reagents. The oversize was
fed back into the ball mill for further grinding.
After grinding, all samples were at the desired size for dewatering. All samples,
except for the Upper Banner sample were then floated using either conventional cell or
9 |
Virginia Tech | Collector &
Frother Feed
Flotation Cells
Tailings
Large Sump
Figure 5 - Conventional Cell Flotation Circuit
The column flotation occurred in an 8-inch unit. Feed was approximately 2.5% to
3% solids. Washwater was added at approximately 1.7 gallons per minute. Airflow was
approximately 34.5 L/min. The launder to the unit was vastly undersized. To account for
this problem, the product was run very wet to allow it to flow into the launder. The
product was also run into the same sump as the conventional product. The very wet
product ended up with approximately 5% solids. All of the dewatering units were run at,
or as close as possible to, 15% solids. To dewater the column product before it entered
the unit, water was siphoned off. Some fines were lost due to this.
Each time the column was run, samples were taken. They were taken of the
product, tailings, and feed, in that order. This was to determine the efficiency of the
column. Three samples were taken during each run to be statistically viable and to
determine scattering. Washwater was also recorded for mass balancing reasons (see
Chapter 2.7.3)
A picture of the unit is shown in Figure 6. A diagram of the circuit is shown in
Figure 7.
11 |
Virginia Tech | All feed to the circuit is first put in the large sump. This feeds the circuit for all tests. A
mixer on the sump keeps the feed mixed and approximately the same throughout the test.
Feed from the sump is pumped up to a head tank where it either recycles back to the
sump or is diverted by a peristaltic pump to a conditioning tank. The peristaltic pump has
a variable speed drive, which controls the feed rate to the centrifuge. The feed is kept
mixed in the conditioning tank by a mixer. The feed overflows the conditioning tank
around the 3-gallon level and flows into a sand pump.
No conditioning occurs within the conditioning tank. All conditioning is done
beforehand in the large sump. This is due to the fact that all products from the centrifuge
are recycled into the large sump due to low sample volume. There was not enough
sample to discard the products for all reagent dosages. Due to recycling, if all products
were not already conditioned, at one initial time, there would be varying degrees of
reagent dosage throughout the test. Conditioned sample would be mixing with
unconditioned sample and would be reconditioned again at the same dosage. This would
increase the overall dosage and alter the test. This problem was handled by conditioning
the entire sample at one time, in which case recycled sample could not alter the sample
dosage within the large sump. Conditioning time could not be monitored with this set-up,
but a conditioning time of at least five minutes was allowed before the first test occurred
to make sure all of the sample was conditioned properly.
The sand pump then fed the centrifuge. The centrifuge is a 4.5β x 6β laboratory
screen-bowl centrifuge operating at 431 Gβs. The product and the effluent from the
centrifuge were both combined and dropped back into the large sump. The screen
product was a recycling load that combined with the feed prior to entering the sand pump.
There were five sampling ports around the circuit of the centrifuge. These are
shown in Figure 9.
14 |
Virginia Tech | Feed
Effluent
Screen
Product
Sampling Position
Figure 9 - Sampling Positions Around the Screen-Bowl Centrifuge
The first sample taken was the effluent. This did not quickly affect the sample
conditions (i.e. % solids) due to the fact that it entered the large sump and was such a
small amount compared to the entire sample. The second taken was the product. The
third taken was the feed and screen combination. This was taken prior to entering the
sand pump. This had a much quicker effect on the sample conditions entering the unit.
This only affected the flow rate of the sample. To not greatly affect the unit operating
conditions, a very small (~200ml) sample was taken. The fourth taken was the screen
sample. This also had an effect on the unit. This affected the particle size distribution
and the feed rate to the unit. Since this was the last sample being taken from a product of
the unit, it did not affect any samples taken after it. This was taken approximately 0.5 to
1 minute after the feed and screen sample was taken to allow the feed rate to stabilize
again. The last sample taken was the feed to the unit. This sample also affects the unit
but is not affected by the unit itself. The only way to affect this stream is to have one or
all of the recycling streams to the large sump interrupted. Since the sump had a large
sample, and small amounts of samples were taken from the recycling streams, the feed
stream was not affected noticeably.
The entire circuit, including grinding, is shown is Figure 10. Pictures of the
centrifuge are shown in Figure 11 and Figure 12.
15 |
Virginia Tech | Figure 12 - Screen-Bowl Centrifuge Picture 2
2.4.2 Testing
Six tests were run with the screen-bowl centrifuge. These included dual tests with
the Upper Banner sample, dual tests with the Pittsburgh sample, and one test each with
the Taggart and Dorchester samples.
Reagent dosages were varied throughout the test to determine their effect. A
baseline was always run, as a control, with no reagents in the system. The reagent was
then added at 1 lb/ton and doubled up to 8 lb/ton. The reagent was added at the
beginning of each test and was allowed to condition at least 5 minutes.
All equipment was then turned on. Once feed entered the centrifuge, it was given
5 minutes to reach steady state. This was to allow buildup of product within the cover at
the product discharge end and to allow proper heating, which plays a part in the
dewatering process, to occur within this buildup.
After steady state was reached, samples were then taken in the order of effluent,
product, feed and screen, screen, and feed. They were taken as quickly as possible to
allow their use as samples taken at one point in time. After they were taken, the product
was quickly weighed so that the least amount of moisture could evaporate from the
surface of the coal.
17 |
Virginia Tech | Three different complete sets of samples were taken for each reagent dosage to
allow the numbers to be statistically viable and to see the scatter within the system. After
all samples were taken, for each reagent dosage, they were weighed and recorded. The
times for each sample were also taken and recorded during the sampling process. The
tares for the containers had been taken and recorded prior to testing.
2.5 Disk Filter
Only five of the tests included the disk filter. One of the five tests was performed
by Ramazan Asmatulu (Test 3). The reason for the two tests not being performed during
the time when the centrifuge was being tested was low sample volume.
2.5.1 Circuit Description and Sampling Process
A diagram of the disk filter circuit is shown in Figure 13.
Feed
Disk Filter Effluent
Mixing
Tank Vacuum
Tanks
Filter Tub
Product Overflow
Figure 13 - Disk Filter Circuit
All feed to this circuit is also fed from the same large sump that the centrifuge
uses. The peristaltic pump feeds the same conditioning tank. The feed overflows at the
same 3-gallon level so particle size distribution is not affected between disk filter and
centrifuge tests. The feed enters a mixing tank on the unit, which overflows into the filter
tub.
18 |
Virginia Tech | No conditioning occurs within the conditioning tank for these tests. For
comparison reasons, the same procedure was done for the disk filter as for the centrifuge.
All conditioning was done in the large sump prior to any test being run. Within these
tests, recycling is not a problem. All exiting streams are either collected or discarded into
a thickener system for further disposal. This could be done due to the fact that not as
much sample is used in these tests as in the centrifuge tests.
There were three sampling ports around the disk filter circuit. These are shown in
Figure 14.
Sampling Position
Figure 14 - Sampling Positions Around the Disk Filter
The first sample taken was the feed. This was taken prior to entering the
conditioning tank. A sample closer to the filter would have been more useful but was
impossible due to the filter setup. The sample did not affect the unit operation due to the
fact that it was prior to the conditioning tank, which fluctuated in volume due to the
mixing, and such a small sample was taken. The second sample was the dry cake. The
third sample was the effluent through the filter. This was done either at the end of each
individual sample or the end of the three samples at each dosage. The effluent volume
19 |
Virginia Tech | Figure 16 - Picture of the Disk Filter
2.5.2 Testing
Four tests were run with the disk filter. These include dual tests with the Upper
Banner sample, and one test each with the Taggart and Dorchester samples.
Reagent dosages were varied throughout the test to determine their effect. The
dosages corresponded with the centrifuge dosages. A baseline was run, and then reagent
was added at 1 lb/ton and doubled up to 8 lb/ton. Sometimes dosages could not reach 8
lb/ton due to loss of sample. The dosages were added at the beginning of each test and
allowed to condition for 5 minutes.
All equipment was then turned on. The filter was allowed to get to steady state.
This was determined by cake thickness at the discharge end of the filter. Once cake
thickness was constant, the filter was at steady state.
After steady state was reached, samples were taken in the order of feed, product,
and effluent. They were taken quickly to be as close to one point in time as possible.
After they were taken, the product was quickly weighed to lessen the effect of
evaporation on the product moisture.
21 |
Virginia Tech | Three different sets of samples were taken for each reagent dosage, to correspond
with the three different sets taken with the centrifuge. After all samples were taken, they
were weighed and recorded. The times for the feed and some effluents were also taken
and recorded. Filter speed and cake thickness were also recorded to calculate cake flow
rates. Pressures were recorded to determine consistency between tests. The tares for
containers had been taken and recorded prior to testing.
2.6 Sample Analysis
After the sample weights had been taken, the wet samples were filtered and dried
in an oven at approximately 80(cid:176) C. The filter paper weights had been taken and recorded
prior to filtering. The combined weights had also been taken and recorded. The product
samples were placed into plastic containers and also put in the oven to dry. The empty
plastic containers and the combined weights were also taken and recorded. The oven
temperature was set as high as the plastic containers would allow. Any higher and they
would melt.
After the samples were allowed to dry for approximately a day (or at least 4
hours), the dry weights were taken. They were then put into sample bags. Any
conglomerates were crushed by finger to the coalβs individual particle size (or as close as
possible) and mixed thoroughly. This allowed 1-gram samples taken from the entire
sample to be more representative of the whole. The one-gram samples were taken when
the samples were to be ashed.
The samples were ashed in a LECO MAC400. One-gram samples were used in
the ash analyzer. Two ashes were run for each sample taken. This allowed verification
of the numbers. More samples would be necessary for statistical analysis of the numbers
from the ash analyzer but time and sample volume did not permit this. If the samples
were close to each other, as determined by the researcher, they were taken and averaged
to get a single percent ash number. Close is taken to be within 2% ash. If the numbers
were not close then two more ashes were run for that sample and the first two numbers
were discarded. After the ashes were taken, all numbers from the tests were obtained.
22 |
Virginia Tech | 2.7 Mass Balancing
Mass balancing is used to lessen the errors that occur when taking samples. This
could be normal scattering, research error, or any of a number of other errors. The
procedure alters the data to balance the flows around a circuit, at all nodes, and to
minimize the error with the original numbers.
The experimental flow is input around all nodes in the system. This is the base
point for calculating all the flows that are being used in the calculation. At least one
other variable must be input to determine the rate of another independent flow from the
first given rate. This could be percent ash to determine ash flow rate from the solids flow
rate. The more variables and flows given, the more accurate the mass balancing is going
to be. At least one flow and one variable must be given, to determine two independent
flows for calculation reasons.
The flows can be visualized around the node given in Figure 17.
F,f P,p
T,t
Figure 17 - Mass Balancing Node
The flow rates are given by F, P, and T, where F is the feed, P is the product, and T is the
tailings. The variables are given by f, p, and t, where they are some property of the
material, such as percent ash, corresponding with the above stated flow rates.
Mass balancing makes certain that the following equations (Equation 2.1 and 2.2)
are adhered to, such as a real process would provide. This is only pertinent to numbers
taken at one point in time and not over time.
F = P+T Equation 2.1
Ff = Pp +Tt Equation 2.2
These equations are used to alter the original data. When altering, the least amount of
error is wanted. This is calculated using Equation 2.3. The relative error in the formula
is used to show which values are more reliable. A more reliable data point would have a
23 |
Virginia Tech | low relative error. Given this, the error calculated would be much bigger for a given data
change with a smaller relative error.
(cid:230) Data - Data (cid:246) 2
Error =(cid:231) original altered (cid:247) Equation 2.3
(cid:231) Ε Data Β· RelativeError(cid:247) Ε
original
All data points that are entered have this error calculated for it. These errors are
summed up over the entire data set. An iterative process is used to first calculate new
data points using Equations 2.1 and 2.2. The errors are then calculated using Equation
2.3, and are summed up. New values are calculated from Equations 2.1 and 2.2 to try and
lessen the summed error from Equation 2.3. This process is repeated until the summed
error is as low as possible.
When the process is finished, the data is checked for anything out of place. If
everything looks in order, the calculated data is then used for results from the experiment.
2.7.1 Screen-Bowl Centrifuge
The screen bowl centrifuge used the given data of solids rate, ash content, and
percent solids. The rates that were calculated and used from this were the solid rate, ash
rate, slurry rate, and liquid rate, which should all be balanced around each node. The
nodes for the centrifuge are shown in Figure 18.
Screen
Feed Feed and Screen Product
Effluent
Figure 18 - Centrifuge Nodes
There are two nodes used for the centrifuge. The first node includes the entire unit,
which balances the feed coming in with the product and effluent going out. The second
node is internal to the circuit and balances the feed coming in and the screen coming in
with the feed and screen going out. The third node seen is just a combination of the two
24 |
Virginia Tech | CHAPTER 3
MODEL
3.1 Introduction
Mathematical models for processes can be very beneficial. They allow users
familiar with the model to determine how efficient the process is, how efficient the
process can be, what are the effects of changing different variables, etc. This allows the
user to get the most out of a process as possible.
This is the same with all preparation plant processes, including dewatering. The
screen-bowl centrifuge is the most common equipment used in the Eastern U.S. to
dewater fine coal. This device dewaters coal and other minerals after they have been
upgraded in the plant. There can be more than one benefit to increasing the efficiency of
this process. There is the obvious benefit of the lower moisture content of the product
but there is also the benefit of increased yield since some of the product is lost through
the effluent.
The model must take into account all variables that affect processes throughout
the unit. Since this unit has many different areas of concern, and is a very complex
process to model, there are fundamental as well as empirical equations contained within
it.
3.2 Description of Process
The process of dewatering in a screen-bowl centrifuge is a very simple process as
seen after it has been broken into its different sections. These can be seen in Figure 21.
Feed is first inserted into the bowl of the centrifuge by way of a feed tube. The feed tube
can be placed on either side of the machine. On the laboratory screen-bowl centrifuge, it
was designed on the product discharge side. The feed is placed into the bowl near the
beach section. This allows the most time for water clarification as the water moves
toward the effluent discharge. Both the outer section and inner screw are traveling at
high rpmβs, accounting for 431 Gβs in the laboratory centrifuge used. The inner screw
will be traveling at a slightly higher or lower speed depending on the left or right-
handedness of the screw and the rotation of the bowl. This difference in speed allows the
27 |
Virginia Tech | screw to carry the solids along the bowl toward the product discharge end. The solids in
the slurry feeding the centrifuge are thrown to the outside of the bowl by the high
centrifugal forces. This is accomplished due to the fact that the solids have a higher
density than water. The solids are carried up the beach by the screw, where they leave
the slurry behind, and are taken over the screen section. The screen section allows
further dewatering where water and solids are taken from the product. The dewatered
coal or minerals are finally discharged out of the product discharge end.
Screw Screen
Feed tube
Feed
Product
Effluent Bowl
Beach
Slurry level
Figure 21 - Sections Within a Screen-Bowl Centrifuge
There are approximately five streams that make up a typical screen-bowl
centrifuge flowsheet. These are shown in Figure 22. The feed coming into the entire
circuit, which is usually between 15% and 20% solids, combines with the material
coming through the screen to feed the machine. The effluent, or clarified water, comes
out of one side of the machine and is discarded from the circuit to be used in other plant
processes, after it has been further clarified (i.e. by a thickener). The product, dewatered
minerals or coal, comes out of the other end of the machine and is, in most instances, the
final product of the entire plant process for that size fraction. The material coming
through the screen combines with the material from the feed, as stated earlier.
28 |
Virginia Tech | With this model all desired output properties can be found. These include flow
out of the effluent, distribution of particles lost through the effluent, product flow,
product particle distribution, percent moisture of the product, and recovery.
3.3.1 Population Model
Division 1 is set-up as a combination between microscopic and macroscopic
population balance models. The centrifuge is divided up into columns and rows. The
columns and rows create individual cells. These individual cells have inflows and
outflows. These flows keep a balance within the cell. There is no tracking of particles
within space within each cell (macroscopic model) but all cells taken together can show
the flow of particles through the model (microscopic model).
3.3.1.1 Model set-up
The division of columns and rows can be seen in Figure 24. These cells represent
a slice taken through the axis of the centrifuge. Flow of material is assumed to be parallel
with this axis, although flows within a centrifuge are probably helical due to the
movement of the inner screw. This assumption was used to simplify calculations and is
not considered to have an adverse effect on the model.
y
x
Figure 24 - Divisions within Population Model
These divisions, as stated earlier, create individual cells. There are flows into and out of
these cells. Since volume of the cells can not be changed (steady state model), the flows
into the cell must equal the flows out of the cell. These flows are shown in Figure 25.
30 |
Virginia Tech | Out3 In1
In4 Out4
Out2 In2
In3 Out1
Figure 25 - Generic Inflows and Outflows of a Cell
This leads to
In1+In2+ In3+In4= Out1+Out2+Out3+Out4 Equation 3.1
which must be true for all cells within the model when the model reaches steady state.
All inflows equal the outflows from adjoining cells except where the feed enters
the system. This cell has the feed inflow, which is not equal to the outflow from any
other cell. These outflows come from the movement of particles due to the centrifugal
forces within the centrifuge (terminal velocity), movement of particles due to the
differential rpmβs between the scroll and outer bowl (scroll movement), movement of
water (and entrapped particles) due to the fluid flow over the weir (fluid flow), and an
excess flow which travels within a column of cells that is needed to obtain the constant
volume of each cell. It is assumed that there is no backflow component that needs to be
calculated that would take particles from the cake and put them into the slurry section.
This could be caused by particles settling into the cake and disturbing other particles
within that cake. This is only being accounted for within the excess flow calculations,
which allows upward movement between cells. All of these flows are shown in Figure
26.
31 |
Virginia Tech | calculated. Once this is known, it is distributed among the cells. Since the only flows
into and out of a column of cells are the fluid and scroll flows then a balance around the
column can be used to obtain the entire fluid flow from the adjoining column. This is
shown in Figure 28.
Feed
Fluid Fluid Fluid Fluid
Scroll Scroll Scroll Scroll
Figure 28 - Column Balancing
This can be continued along the length of the centrifuge to obtain the entire fluid
flows of all columns. These flows are obtained until the feed column is reached. It is
assumed that there is no movement due to fluid flow between the feed and the product
discharge. Once these are obtained, all flows are known and the iteration of the model
can continue.
3.3.1.2 In-depth Model
The flows that are used in this model are obtained from empirical as well as
fundamental equations. These flows include the scroll movement, fluid flow, terminal
velocity, and excess flow.
3.3.1.2.1 Scroll Movement
The assumptions made in this section are listed below.
1) Movement is along the axis of centrifuge instead of along a helical path that is
actually taken.
2) Movement in x-direction is not dependent on x. Movement is a function of the
differential rpm between the bowl and the screw.
33 |
Virginia Tech | 3) No velocity in the y-direction due to scroll. This only occurs in the beach
section where there must be a y-velocity due to the outer bowl geometry.
4) Movement is a function of the percent solids within the cell.
Variables:
q = distance between threads
n = number of spirals
D rpm = differential rpm between the bowl and scroll
A = side area of cell
s
V = volumetric flow rate due to the scroll
s
a = solids power
F = total percent solids within a cell
T
F = percent solids within a cell within the ith size class
i
V = solids flow rate due to the scroll within the ith size class
si
Velocity of the scroll is
V = qΒ· D rpmΒ· n=V Equation 3.2
x- c
where
D rpm= rpm - rpm Equation 3.3
bowl screw
The volumetric flow rate would equal to
V =V Β· A Β· F a Equation 3.4
s s T
where F a is a retarding factor due to the fact that the scroll is moving a slurry. The
T
thicker the slurry, the higher the percent solids and therefore the faster the flow.
Equation 3.4 obtains the entire flow due to the scroll but solids flow is obtained by
Equation 3.5.
V =V Β· F Equation 3.5
si s i
This is used to acquire the flow of particles within a certain size class due to the scroll
movement.
34 |
Virginia Tech | 3.3.1.2.2 Fluid Flow
The assumptions made in this section are listed below.
1) Flow is a function of percent solids.
2) Laminar flow.
3) Velocity Profile similar to velocity profile within pipe.
Variables:
V = Fluid velocity
F
V = maximum fluid velocity (at the surface of the liquid)
Max
d = diameter of particle
y = the height at which the velocity is being calculated
y = the height at the top of the flow (slurry)
top
F = total percent solids within a cell
T
c = solids power
A = side area of the ith cell
i
Q = total fluid flow out of a column
Tf
V = volumetric flow rate due to the fluid flow
F
F = percent solids within a cell within the ith size class
i
V = solids flow rate due to the fluid flow within the ith size class
Fi
The total fluid flow out of a column of cells is determined by all other flows being
known and a constant volume within that column being maintained. This was stated
earlier. With the total fluid flow out of a column known, the individual cell flows must
be obtained from this known flow. The velocity, as a function of y, is given by Equation
3.6, which is a laminar flow profile in a pipe (Roberson and Crowe, 1993). It is modified
according to the coordinate system being used. Using a pipe profile is an assumption that
is being made, since the entire volume of the centrifuge is not filled up and there is an air
layer within the centrifuge. This does not occur within a true pipe flow and adds another
level of difficulty, which is not needed, since the pipe profile should be close enough to
the true profile to cause no adverse effects on the model.
35 |
Virginia Tech | V ( ( ) )
V = Max d2 - y- y 2 Equation 3.6
F d2 top
The actual flow is a hindered flow of the same form as the terminal velocity, where
( )
1- F cis the retarding factor. This is shown in Equation 3.7.
T
V ( ( ) ) ( )
V = Max d2 - y- y 2 1- F c Equation 3.7
F d2 top T
The total flow can be a summation of the individual velocity flows. Since the flows are
discretized, the total flow can be written as
Q = (cid:229) V A Equation 3.8
Tf Fi i
i
Substituting Equation 3.7 into Equation 3.8 yields
(cid:229) V ( ( ) ) ( )
Q = Max d2 - y- y 2 1- F cA Equation 3.9
Tf d2 top T i
i
V
Since Max is a constant it can be taken out of the summation and solved for. The results
d2
of this are shown in Equation 3.10.
V Q
Max = (cid:229) ( ( Tf ) ) ( ) Equation 3.10
d2 d2 - y- y 2 1- F cA
top T i
i
Substituting this back into Equation 3.7 yields
( ( ) ) ( )
Q d2 - y- y 2 1- F c
V = T(f ( to )p ) T Equation 3.11
F (cid:229) d2 - y- y 2 ( 1- F ) c A
top T i
i
This equation (3.11) allows the individual cell velocities to be calculated from only the
total flow out of a column of cells, the position of the cells and the side area of the cells.
It effectively divides the flow into the cells so that the summation of all cell flows is still
equal to the entire column flow. The volumetric flow rate out of the cell would be equal
to
V =V Β· A Equation 3.12
F F i
and the solids flow would be equal to
V =V Β· F Equation 3.13
Fi F i
36 |
Virginia Tech | m = mass of particle
p
m = mass of water
w
A = bottom area of cell
b
C = coefficient of drag
d
V = volumetric flow rate due to the terminal velocity
T
b = solids power
F = total percent solids within a cell
T
F = percent solids within a cell within the ith size class
i
V = solids flow rate due to the terminal velocity within the ith size class
Ti
Sum of the forces acting on the particle is given by
(cid:229)
F =G- B = D Equation 3.14
where
G = m u Equation 3.15
p
B = m u Equation 3.16
w
Centrifugal acceleration, u, is given by
u =wy2 Equation 3.17
The volume of the particle is given by
V = k d3 Equation 3.18
v
while mass of the particle and mass of the water is given by
m = r V Equation 3.19
p p
m = r V Equation 3.20
w w
Drag force is given by Equation 3.21 (Roberson and Crowe, 1993),
C A r V 2
D = d p w 0 Equation 3.21
2
where A is given by
p
A = k d2 Equation 3.22
p a
38 |
Virginia Tech | Substituting Equations 3.15, 3.16, 3.17, 3.21, and 3.22 into Equation 3.14 gives the
following
C k d 2r V2
m wy2 - m wy2 = d a w 0 Equation 3.23
p w 2
Rearranging for V gives the final velocity equation (3.24).
T
( )
2wy2k d r - r ( )
V = v p w =V y Equation 3.24
T C k r y- w
d a w
The volumetric flow rate would equal to
( )
V =V Β· A Β· 1- F b Equation 3.25
T T b T
( )
where 1- F b is a retarding factor due to hindered settling (Heiskanen, 1993). The
T
greater the percent solids, the more the individual particles interact and therefore the
slower the velocities of the particles will be.
Equation 3.25 obtains the entire flow due to the terminal velocity but solids flow
is obtained by Equation 3.26.
V =V Β· F Equation 3.26
Ti T i
This is used to acquire the flow of particles within a certain size class due to the terminal
velocity movement.
3.3.1.2.4 Excess Flow
The assumptions made in this section are listed below.
1) Volume of each individual cell (Figure 25) does not change.
This flow is used to keep a constant volume entering and leaving a cell. Since
there is nothing tying the other three flows (scroll, terminal, and fluid) together such that
the flows out equaled the flows in, another flow is needed. This flow is calculated by
adding all of the inflows into a cell and subtracting all of the outflows.
Excess =V =V +V +V +V - V - V - V Equation 3.27
Eout Tin Fin Sin Ein Tout Fout Sout
This lets the excess flow be an outflow from the cell. The model is setup such that the
flows are calculated starting from the bottom and working their way up. Therefore, the
39 |
Virginia Tech | flow is an outflow upward. This is done in all cells except for the topmost cells within
the model. All flows into and out of those cells should be equal.
Equation 3.27 obtains the entire flow due to the excess flow but solids flow is
obtained by Equation 3.28.
V =V Β· F Equation 3.28
Ei E i
This is used to acquire the flow of particles within a certain size class due to the excess
flow movement.
3.3.2 Centrifugal Cake Model
Assumptions made in this section:
1) Completely mixed in the x, y, and z directions. No distributions of particles.
2) Dewatering according to Zeitsch.
Within this section, division 2 of the screen-bowl centrifuge, Zeitschβs model can
be used to determine the percent moisture of the product. A partition curve was fit by
experimental data to determine particles loss through the screen. These particles are
recycled to the feed. With these values known, the product particle distribution and
percent moisture of the recycle can be found.
3.3.2.1 Zeitschβs model
This model was obtained from Zeitsch (1990).
3.3.2.1.1 Input variables:
e β porosity of cake
r β density of solids
S
r β density of liquid
L
n β weight fraction of solids dissolved in liquid
k β permeability of cake
r β inner radius of basket
B
w β angular velocity
H β height of cake
40 |
Virginia Tech | rate, feed percent solids, centrifuge speed, and particle size distribution. When each of
the variables were tested, they were the only parameters that were changed.
The results from varying the feed rate, percent solids, and centrifuge speed are
shown in Table 2. Figure 31, Figure 32, and Figure 33 are graphical representations of
the results.
Table 2 - Model Testing Results for Feed Rate, Percent Solids, and Centrifuge Speed
Feed Rate Percent Solids Centrifuge speed
Run g/s Recovery Moist Run % Recovery Moist Run rpm Gβs Recovery Moist
0 116.8 94.33 20.69 0 13.99 94.33 20.69 0 2247 431 94.33 20.69
1 100 95.48 23.93 1 12.00 95.64 23.93 1 2000 341 92.24 21.49
2 130 92.09 24.00 2 16.00 92.32 23.98 2 2500 533 95.85 21.58
3 50 98.83 24.14 3 8.00 99.86 26.18 3 1000 85.3 79.09 24.88
4 180 89.59 21.43 4 20.00 87.79 23.20 4 3500 1045 99.07 20.54
The data shows, for feed rate, that as the rate is increased the recovery decreases.
The relationship is essentially linear. This is due to the fact that as the feed increases the
particles can not settle out as quickly and an increased number are taken out with the
effluent. Since the scroll rate is not increasing, only so many solids can be moved at a
certain rate. This leads to an increase in the percent solids throughout the model. Due to
the increased percent solids, there is an increased effect of the hindered settling. This
greater hindered settling allows more particles to stay near the top of the slurry and
eventually be carried out through the effluent.
The relationship between moisture and feed rate is closer to a third order function.
Therefore, there is at least one minimum product moisture for each set of conditions,
dependent on the feed rate. A lower feed rate will create a wetter cake while a higher
feed rate will initially produce a wetter cake but will eventually produce a drier cake.
Depending on the desired output, it might be more beneficial to obtain a minimum
moisture while losing a portion of the product (first local minimum β medium feed rate)
or to recover all of the product while raising the moisture (very slow feed rate). If neither
of these can be achieved then as the feed rate is increased, the product moisture will
eventually go down along with the recovery.
44 |
Virginia Tech | 100 30
95 25
90 20
85 15
Centrifuge speed model test
80 10
Recovery
Moisture
75 5
0 1000 2000 3000 4000
Centrifuge Speed (rpm)
Figure 33 - Centrifuge Speed Model Test
With percent solids, the recovery again has a linear relationship. For a higher
percent solids there is a lower recovery. The cause of this is similar to the relationship
between feed rate and recovery. The higher percent solids in the feed leads to a higher
percent solids throughout the model. The scroll, again, can only take out so many solids.
The extra solids are distributed throughout the model depending on the flows. This
increase in percent solids leads to a greater hindered settling effect, which leads to more
loss of solids through the effluent.
The relationship between percent solids and moisture is not directly apparent. It
again appears to be a third order relationship with the same effects as the feed rate. In the
percent solids and feed rate relationship there is a deviation from a smooth relationship at
the initial conditions setting. This is most likely caused by the fact that the model was set
using those conditions and therefore is optimized for those conditions. This results in a
much lower moisture than the surrounding percent solids and feed rates. Besides that
point (on both sets of data), the trend is still towards a third order relationship.
The relationship between centrifuge speed and recovery is not linear as feed rate
and percent solids were. The relationship is closer to a second order relationship. As the
46 |
Virginia Tech | centrifuge speed decreases, the recovery also decreases. It is thought that as the speed
increases, the recovery will increase up to a maximum of 100% and will continue to stay
there. This is due to the fact that as the speed increases, the centrifugal acceleration
increases. This increase moves particles to the outer wall much quicker. This quicker
movement allows less particles to exit the effluent. As the centrifuge speed decreases,
there is a corresponding decrease in centrifugal acceleration and particles move toward
the outer wall slower. This allows more particles to exit through the effluent, which
accounts for the decrease in recovery at lower speeds.
Percent moisture is again a third order function. It is much closer to the actual
data than percent solids and feed rate, with no great deviation with the original settings,
although the settings still produced a moisture content that is below the surrounding
centrifuge speeds. The third order relationship produces a local minimum moisture
content for the product at approximately 2200 rpm's for these settings. To decrease the
moisture, the rpm's must be increased. To benefit the most, the centrifuge speed must be
increased as high as possible. This produces the driest product and the highest recovery.
The industry is tending towards this but is hampered by material constraints (i.e. steel
tensile strength).
The last variable that was tested was particle size distribution. It was decided to
test four different distributions. The first two would keep the same d as the original that
50
the model was set to and only change the steepness of the partition curve. The last two
would move the d higher and lower by modifying the distributions. These distributions
50
are shown in Table 3 with their corresponding recoveries and moistures. The
distributions are also shown in Figure 34, Figure 35, Figure 36, Figure 37, and Figure 38.
Figure 34 shows the original distribution that the model was set to.
Table 3 - Model Testing Results for Particle Size Distribution
Run Recovery Moist 28x48 48x100 100x200 200x325 -325 Sum d
50
0 94.33 20.69 8.60 25.37 22.36 12.48 31.20 100 0.09
1 99.65 21.99 2.00 16.00 45.00 32.00 5.00 100 0.09
2 91.89 20.51 25.00 18.00 10.00 7.00 40.00 100 0.09
3 99.32 21.75 14.00 33.00 31.00 14.00 8.00 100 0.135
4 90.73 20.55 5.00 22.00 18.00 12.00 43.00 100 0.06
47 |
Virginia Tech | 100
80
60
40
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Particle Size (mm)
50
)%(
woleB
tnecreP Model test - distribution 4
Figure 38 - Distribution 4 Model Test
With a steeper partition curve (distribution 1) the model reports a dramatic
increase in recovery with a small increase in the moisture. The recovery is most likely
due to the fact that there are fewer fines. The fines have a tendency to be lost with the
effluent because they settle much slower than bigger particles. The increase in moisture
is opposite from what was assumed. It was thought that with less fines, the cake would
have a greater ability to be dewatered over the screen section. The increase in moisture
might be due to the high recovery, in which almost all of the product is recovered. Even
with a lower percentage of fines, the greater flow rate of material over the screen section
produced a wetter cake. The greater flow rate over the screen produced a higher cake
which, in turn, held more water. The greater thickness of the cake makes the dewatering
much more difficult and with a greater volume of moisture in the cake, the dewatering
ability was lowered and the centrifuge produced a wetter product.
A much flatter partition curve (distribution 2) had a decrease in both recovery and
moisture. The decrease in recovery can be due to the increase in fines, which have a |
Virginia Tech | greater tendency to leave through the effluent. The moisture response to this distribution
is also opposite what was originally thought. With a higher fines content, the cake was
assumed to be more difficult to dewater. This low moisture content might also be due to
the recovery. Since the recovery was low, less product was available to be dewatered
over the screen section of the centrifuge. This resulted in a thinner cake and a lower
volume of water. Both of these contributed to a lower moisture product.
With the d increased (distribution 3), the model reports a dramatic increase in
50
recovery with a small increase in the moisture. Since a lower percentage of fines were
needed to increase the d , the recovery is most likely due to this fact. The increase in
50
moisture is again attributed to the amount of product being dewatered across the screen
section. The results are very similar to distribution 1βs results. With a lower percentage
of fines, a higher recovery and moisture were obtained.
With the d decreased (distribution 4), the model reports a decrease in both
50
recovery and moisture. The decrease in recovery is most likely due to the fact that a
higher percentage of fines were needed to decrease the d . This results in a lower
50
recovery, which, in turn, results in less material being dewatered across the screen
section. These results are similar to distribution 2βs results. With a higher percentage of
fines, a lower recovery and moisture were obtained.
51 |
Virginia Tech | CHAPTER 4
RESULTS AND DISCUSSION
4.1 Flotation
Two different types of flotation were used in these tests: conventional cell
flotation and column flotation. The procedures for these were discussed in section 2.3.
As stated before, the conventional cell flotation was used for the Pittsburgh No.8 sample
while column flotation was used for the Taggart and Dorchester samples. The Upper
Banner sample was not floated.
The conventional cell produced an ash content of the feed, clean coal, and reject
products of 6.30%, 4.30%, and 42.10% respectively. These ash contents give a total
mass yield of 94.7% and a combustible recovery of 96.7%. Percent ash is given on a dry
basis while percent moisture is given on an as-received basis and is calculated by mass.
This is the format for all assay data throughout this thesis. These results show a very
good recovery for this coal using conventional cells. The product from these flotation
cells, which feeds the dewatering equipment, can be considered to be fairly representative
of a βreal-worldβ feed.
The column flotation results are shown in Table 4 and Table 5. The Dorchester
seam resulted in a mass yield between 61.90% and 63.46% and a combustible recovery of
between 91.47% and 92.94%. This is usual for a high ash feed (37.57%-38.13%) such as
was used. This resulted in a product ash between 7.94% and 8.83%. The Taggart seam
faired better with a mass yield of between 73.97% and 77.34% and a combustible
recovery between 96.68% and 97.55%. The product ash (3.93%-5.65%) was also much
lower than the Dorchester seam. The better performance of the Taggart seam is probably
due to the lower feed ash of the Taggart seam (between 25.20% and 26.79%) and the
smaller size of the particles. Bigger particles in a column are apt to fall off of the bubbles
before they reach the froth stage. Clays in these coals, which tend to make up a high
percent of the ash, typically stay within the same size range, which is smaller than most
coal, and are not dependent on the coal size. Since the Dorchester seam (-65 mesh) was
slightly bigger than the Taggart seam (-100 mesh), the big particles in the Dorchester
seam (mainly coal), were not recovered in the flotation, and resulted in a lower mass
52 |
Virginia Tech | yield and combustible recovery. Both of these sample results are typical of a high ash,
fine coal using a column for flotation.
Table 4 - Column Flotation Results from the Dorchester Seam
Ash Content Mass Recovery Combustible
Sample # Stream
(%) (%) Recovery (%)
Feed 37.69
1 Tails 86.05 61.92 91.47
Product 7.94
Feed 38.13
2 Tails 88.02 63.00 92.84
Product 8.83
Feed 37.57
3 Tails 87.94 63.46 92.94
Product 8.58
Table 5 - Column Flotation Results from the Taggart Seam
Ash Content Mass Recovery Combustible
Sample # Stream
(%) (%) Recovery (%)
Feed 26.81
1 Tails 91.51 73.88 96.97
Product 3.93
Feed 26.26
2 Tails 92.85 75.35 97.61
Product 4.47
Feed 25.86
3 Tails 90.46 74.72 96.75
Product 4.00
4.2 Screen-Bowl Centrifuge
There were two different types of samples that were tested using dewatering
agents. These included a dense medium cyclone (DMC) sample and a cyclone overflow
sample. The DMC sample was approximately a minus 1/2 inch sample of clean coal that
was crushed to the desired size. Pittsburgh No. 8 was floated while Upper Banner was
not. Since the sample was of clean coal there was not much ash, in the form of clays, in
the sample before or after cleaning as compared to the cyclone overflow sample. The
cyclone overflow samples were already of a fine size (-65 mesh) and a high ash content
(~30%-40%) before cleaning. Since these two types of samples were very different in
characteristics they are being considered separately.
53 |
Virginia Tech | In Test 1, the reagent had an adverse effect on the moisture of the product. It
continually rose from 19.5% to 42.0%. Although there was an increase in yield at 2 lbs
of reagent, there was an increase in moisture content at the same dosage and the yield
drastically fell at 4 lbs. This mild increase in the yield is encouraging for the reagent and
shows some mild dewatering yield improvement, although there is no dewatering
improvement in this sample. This is most likely due to oxidation of the coal. The
dewatering agents have been known to make no improvement or have a detrimental
effect on the dewatering of oxidized coal.
Tests 2, 3, and 4 had very similar results. They faired much better overall, which
is most likely due to the sample being run much quicker. This resulted in less oxidation.
All of these tests showed a drop in both the moisture of the product and yield between the
1 and 2 lb dosage. After this low point was reached, both the moisture and yield
increased. The moisture of the product increased at least to the baseline and most of the
time above this. The yield increased dramatically at high dosages (8 lb/ton) to within
99.0% recovery, well above the baseline of approximately 93.0%. The increase in yield
is most likely due to coagulation of fine coal due to hydrophobic forces. This would
allow the centrifugal forces to have an effect on the fines, by increasing their size,
without the water carrying the particles out through the effluent, but would also trap
water within the structure of the conglomerate. This entrapment of water would also
account for the increase in product moisture. The high yield can be very beneficial to
low yield coals that are being dewatered using a screen-bowl centrifuge, although the
benefits in yield might be outweighed by the increased moisture.
4.2.2 Cyclone Overflow
The cyclone overflow samples include the Taggart and Dorchester samples.
Results from these tests (5 & 6) are shown in Table 10 and Table 11 and Figure 43 and
Figure 44.
58 |
Virginia Tech | Since Test 5 ran out of sample much sooner than the other tests, which resulted in
fewer dosages being tested, it is harder to determine the effect that the reagent had on the
dewatering of this coal. This test did have the same drop in moisture at the 2 lb/ton
dosage as Tests 2-4 had and also had a decrease in yield at this same dosage. Since only
two dosage levels were run, it is impossible to determine the full effect of the reagent,
although it is surmised that the effects are similar to the effects seen in Tests 2, 3, and 4.
Test 6 had twice as many data points as Test 5 due to doubling of the sample
volume. This allowed the full effect of the reagent to be seen. The results for this test are
similar to the results from Test 2, 3,and 4 with a drop in both the moisture and yield at the
2 lb/ton level. There is also an upswing in both the moisture and yield after this dosage.
There is one deviation from previous tests, in that the yield does not reach or exceed the
baseline. This could be due to the very high baseline recovery (~98.0%) or the high fine
content of the coal. Overall, the reagent did not have a beneficial effect on the
dewatering of this coal type in a screen bowl centrifuge.
4.3 Disk Filter
The same sample types were tested in the disk filter as in the centrifuge. The
same division of samples is also observed due to comparison reasons. Only 4 tests were
performed using the disk filter. This is due to low sample volume. A fifth test (Test3)
was performed by Ramazan Asmatulu.
4.3.1 DMC
The DMC tests include Test 1, 2, and 3. Test 3 was performed by Ramazan
Asmatulu. Results from these tests are given in Table 12, Table 13, and Table 14 and
Figure 45, Figure 46, and Figure 47.
61 |
Virginia Tech | In Test 1 the reagent had a beneficial effect at the 1 lb/ton dosage, dropping the
product moisture from 30.0% to 22.0%, while the yield stayed approximately the same.
At higher dosages there was an adverse effect of the reagent on dewatering, with the
product moisture increasing above the baseline and the yield dropping from 99.5% to
99.0%. This effect was most likely due to oxidation of the coal as seen in Centrifuge
Test 1.
In Test 2 and 3 the samples were run much quicker to reduce the effect of
oxidation. Both of these tests showed a dramatic improvement in the product moisture.
The moisture dropped to approximately half of the baseline value at dosages between 2
and 4 lb/ton. The yield stayed approximately the same throughout Test 2, with not much
room for improvement from the baseline of 99.6% while Test 3 did not account for the
yield of the filter.
4.3.2 Cyclone Overflow
The cyclone overflow samples include the Taggart and Dorchester samples.
Results from these tests (5 & 6) are shown in Table 15 and Table 16 and Figure 48 and
Figure 49.
Table 15 - Results from Disk Filter Test 5 Using Taggart Sample
Dosage Cake Moisture Mass Recovery Combustible
(lb/ton) (%) (%) Recovery (%)
0 46.65 99.56 99.60
0 30.81 99.52 99.56
0 31.83 99.52 99.56
2 26.09 99.38 99.66
2 26.32 99.34 99.65
2 24.57 99.24 99.56
65 |
Virginia Tech | 50 100
40
90
30
20
80
Test 6 - Dorchester
10 Moisture
Yield
0 70
0 2 4 6 8 10
Reagent Dosage (lb/ton)
Figure 49 - Disk Filter Data from Test 6
As with the centrifuge test, Test 5 ran out of sample before a full array of dosages
could be tested. Only two dosage levels could be tested. There was an improvement in
the product moisture between these two dosage levels from approximately 31.0% at
baseline to 26.0% at 2 lb/ton. This is not as much of an improvement as was seen in
Tests 1, 2, and 3 and might be due to the increased fines. It also should be remembered
that only two dosage levels were tested, which tells very little about the effect that the
reagent had on this particular coal.
Test 6 faired much better with three dosage levels, which does show a trend. It
shows a drop from 34.0% to 23.0% at 2lb/ton which is similar to the DMC tests but not
as dramatic. There was an increase in yield from 98.9% to 99.7%. The low initial yield
is probably due to increased fines. These increased fines also allowed less improvement
to occur with the dewatering agent.
67 |
Virginia Tech | 4.4 Equipment Comparison
With no reagent, the centrifuge outperforms the disk filter on cake moisture, while
under performing on recoveries. This is true for all samples and tests run. With the
addition of reagents the moisture improved minimally, up to a point (~2 lb/ton dosage),
and then returned to normal using the centrifuge. With the disk filter, the moisture
reduction was over half of the baseline at the 2 lb/ton dosage level and continued to
improve after that dosage. The yield did increase at high dosages for the centrifuge, but
only in Test 2 and 4 did it reach the 99th percentile in which the disk filter continuously
ran. The disk filter is designed to perform at high efficiencies and therefore no room is
left for improvement of the yield. In real world situations, the centrifuge would run at
higher throughputs which would result in lower yields (~80% baseline compared to
93%). With the addition of reagents the yield would increase, but most likely not to the
values obtained in these tests. Overall, with the addition of reagents, the disk filter
outperformed the centrifuge on product recovery and yield. It has been shown that the
product moisture and recovery obtained from the centrifuge can both be improved with
the addition of the dewatering aids, although not at the same time. This can be beneficial
to processing plants with existing screen-bowl centrifuges.
4.5 Model Comparison
The performance of the model was compared against three experimental tests run
on the centrifuge. These tests are shown in Table 17. Run 0 was used to fit the
parameters of the model. These tests (Run1 β Run3) were picked to determine the
performance of the centrifuge model with varying degrees of feed rate, percent solids,
and particle size distribution. Feed rate and percent solids continuously decreased
throughout the runs. The particle size distribution started out very similar to Run 0 and
then deviated in Runs 2 and 3. This is shown in Figure 50, Figure 51, and Figure 52.
The original distribution that the model was set to (Run 0) can be found in Figure 34.
These samples were also picked because no reagents had been added to the feed. If the
reagents had been added, the surface tension and/or contact angle would have to be
determined. This would have had to be done during the testing period. This was not
done at that time. For these samples, the surface tension and contact angle should be very
68 |
Virginia Tech | Table 18 - Model Test: Experimental Results
Exp. Exp. % Model Model % % Error
Run % Error Moist
Recovery Moist Recovery Moist Recovery
0 94.33 20.69 94.33 20.69 0.00 0.00
1 93.77 22.98 95.26 23.05 1.59 0.29
2 94.23 33.76 90.13 24.81 4.35 26.50
3 97.64 29.54 99.51 25.19 1.92 14.71
At a very similar feed rate, percent solids, and particle size distribution, the model
performed well, with a percent error for the recovery and moisture of 1.59% and 0.29%
respectively. The recovery error is a little more than was desired (~1.00%) but is fine for
a first run of the model. The errors increased with the next two runs, escalating to a high
of 4.35% for the recovery and 26.50% for the moisture. These increasing errors are most
likely due to a combination of feed rate, percent solids, and particle size distribution
changes. Since the errors for Run 2 are greater than the errors for Run 3, it is assumed
that the greater cause for these errors is the particle size distribution. This is due to the
fact that the feed rate and percent solids for Run 2 are closer to Run 0 than Run 3 is to
Run 0 while the particle size distribution is farther from Run 0 than Run 3 is from Run 0.
There are a few areas to improve on the model that would most likely reduce
these errors. These include a more thorough understanding of fluid flow towards the
effluent of the model and a greater understanding of particle loss through the screen of
the centrifuge. The fluid flow towards the effluent was assumed to be linear to the axis of
the centrifuge until the back wall of the centrifuge was reached. The flow would then
travel up the wall and exit out over the weir. The actual flow would more likely be
helical (due to the scroll movement) and travel downward and then upward gradually
(more parabolic) to eventually exit over the weir. This is similar to fluid flow patterns
encountered in fluid dynamics. Particle loss through the screen would be much more
difficult to model. Particle loss for the current model was found by knowing the flows
from all of the outputs of the centrifuge. This allowed a partition curve to be determined
for the loss of the particles through the screen. This partition curve was then used for all
model tests. Instead of being constant, a more accurate partition curve would most likely
be a function of the particle size distribution, solids content, feed rate, scroll speed, and
centrifugal acceleration at the screen.
71 |
Virginia Tech | Chapter 5
CONCLUSION
The primary objective of this thesis was to determine the effect that reagents had
on the dewatering ability of the screen-bowl centrifuge. The effects that the reagents had
on the disk filter were also determined. The reagents tested altered the contact angle of
the coal to improve dewatering. The coal tested came from four seams including Upper
Banner, Pittsburgh No. 8, Taggart, and Dorchester. Several results were obtained from
evaluating these reagents:
β’ The reagents improved the dewatering ability of the disk filter by
approximately half at the 2 lb/ton dosage. An increase in reagent dosage
above this resulted in a further decrease in the moisture content of the product.
β’ The reagents kept the recovery of the disk filter at approximately 99.5% or
improved on this slightly. There was little room for improvement due to the
initial high efficiency of the disk filter.
β’ The reagents improved the dewatering ability of the centrifuge slightly at the 2
lb/ton dosage. Anything above this brought the moisture content of the coal
back to the baseline or increased the moisture above the baseline.
β’ The reagents initially decreased the recovery of the centrifuge at the 2 lb/ton
dosage. An increase in recovery was obtained at a higher dosage level
resulting in recoveries above the baseline.
β’ Overall, it was determined that the centrifuge produces a drier product than
the disk filter with no reagent but produces a lower recovery. With reagents
the disk filter outperformed the centrifuge, producing a drier product and a
higher recovery.
Along with the experimental testing, a model of the centrifuge was developed.
This included fundamental, as well as empirical relationships to determine product
moisture, particle size distribution for the effluent and product, recovery and flow rates
throughout the centrifuge. The model was tested against experimental data. The effects
72 |
Virginia Tech | of changing the feed rate, percent solids, centrifuge speed, and particle size distribution
were also determined.
β’ The relationship between feed rate and recovery is essentially linear. As feed
rate increased, recovery decreases, which is most likely due to an increase in
particles within the centrifuge. This increase in particles increases the
hindered settling effect, which allows fewer particles to travel towards the
outer bowl. This leads to a higher particle flow through the effluent.
β’ Percent solids and recovery has the same relationship as feed rate and
recovery.
β’ The relationship between centrifuge speed and recovery is a second order
function. As centrifuge speed is increased the recovery is increased, due to a
greater centrifugal acceleration. This is assumed to reach 100% and stay
there.
β’ The relationship between feed rate and moisture content is a third order
function. There is a local minimum at a medium feed rate. As the feed rate
decreases the moisture increases and as the feed rate increases the moisture
decreases.
β’ Percent solids and moisture has the same relationship as feed rate and
moisture.
β’ Centrifuge speed and moisture has the same relationship as feed rate and
moisture. It is thought that the moisture content will approach 0% at very
high centrifuge speeds.
β’ The particle distribution study determined that the percentage of fines had a
great impact on the recovery and moisture content. As the fines increased, by
decreasing the d or making the partition curve shallower, the recovery and
50
moisture decreased. As the fines decreased, by increasing the d or making
50
the partition curve steeper, the recovery and moisture increased.
β’ The model predicted the moisture and recovery at a similar particle size
distribution. As the particle size distribution, feed rate, and percent solids
were altered, the error between true recovery and predicted recovery and true
moisture and predicted moisture was increased. It is thought that the particle
73 |
Virginia Tech | Chapter 6
RECOMMENDATIONS FOR FUTURE WORK
Based upon the knowledge gained from the work on this thesis the following are
considered excellent areas for further research.
β’ A more thorough understanding of the effects that contact angle-altering
reagents have on coal being dewatered in a screen-bowl centrifuge. Due
to low sample volume an entire array of coal types and surfactants could not
be used. This could be very beneficial to industry and also to the
understanding of how coal with reagents reacts in a centrifugal field.
β’ Determining the effect that surface tension-altering reagents have on coal
being dewatered in a screen-bowl centrifuge. It is known that lowering the
surface tension can also decrease the cake moisture of coals. This could
greatly increase the understanding of dewatering in a centrifugal field. The
design of a spray system for continuous screen-bowl centrifuge use would
also be beneficial. A study of this was started, with promising results, but was
not followed through.
β’ The incorporation of true fluid flow in a centrifuge model. This includes a
more thorough understanding of fluid dynamics within a semi-open pipe
under great centrifugal force with an inner turning screw. This is not a simple
problem and might require simplifying assumptions. This might include
taking out the inner screw, which might not affect the results adversely. With
this understanding a more predictable model can be obtained.
β’ An understanding of particle and fluid flow through a screen. This would
take into account the centrifugal forces inside a centrifuge along with the
screen type and size, the particle size distribution, the percent moisture, and
the particle shape. This is definitely needed for the screen section of the
model for the model to have more predictive ability. As of now it is purely an
empirical relationship and might be machine or material specific.
75 |
Virginia Tech | Desalination of Produced Water via Gas Hydrate Formation and Post Treatment
by
Jing Niu
Abstract
This study presents a two-step desalination process, in which produced water is
cleaned by forming gas hydrate in it and subsequently dewatering the hydrate to remove
the residual produced water trapped in between the hydrate crystals. All experiments
were performed with pressure in the range of 450 to 800psi and temperature in the range
of -1 to 1Β°C using CO as guest molecule for the hydrate crystals. The experiments were
2
conducted using artificial produced waters containing different amounts of NaCl, CaCl
2
and MgCl at varying temperature (T) and pressure (P). The results are presented as
2
functions of %Reduction of difference chemical elements, CO requirements and applied
2
T and P conditions.
The impact of dewatering techniques, including centrifuge and filtration process, on
gas hydrate solid product is studied. The results showed that over 99% of dissolved NaCl
and MgCl can be removed from artificial saline water in laboratory experiments. This
2
was achieved in a process involving a single-stage hydrate formation step, followed by a
single-step solid-liquid separation (or dewatering). The results also show that the
%Reduction (percentage of the concentration decrease) of artificial produced water
increases with centrifugation time and rotational speed (rpm). The %Reduction was
increased considerably after hydrate crystals were crushed and filtered, indicating that the
artificial process water was entrapped in between the hydrate crystals. It was found also
that the finer the particle size, the higher the extent of salt removal. In general, filtration
was a better than centrifugation for the removal of TDS (Total Dissolved Solids). |
Virginia Tech | Master Thesis of Jing Niu
Acknowledgement
First and foremost, I would like to acknowledge my advisor, Dr. Roe-Hoan Yoon, for
his support and guidance throughout my two years study in Virginia Tech. I would like to
express my appreciation to him for his help both for my research life and my personal life
in Blacksburg. During this period, I have learned a lot from Dr. Yoon about how to
conduct research projects, how to think, how to write. I also thank Dr. Gregory Adel, Dr.
Gerald Luttrell and Dr. Gandour who served as my thesis committee member for their
suggestions on my research works.
I sincerely appreciate to Dr. Jialin Wang for his guidance and instruction in gas
hydrate formation and Lei Pan for his patient instruction and guidance in experimental
methods. Especially, I would like to thank Mr. Zuoli Li and Juan Ma who solved my
titration problem in testing ion concentration.
I would like to express my appreciation to Yongkoo Seol for his discussion on
technical problems on filtration process. Also I want to express my sincere gratitude to
Dr. Jongho Cha for his discussion on gas hydrate formation kinetic and equilibrium
process.
I would like to thank past and present members in Center of Advanced Separation
Techniques (CAST), Chris Hull, Kirsten Titland and Dr. Jinming Zhang for their support
and help.
Last but not least, I would like to express my deepest gratitude to my parents for their
selfless love and support. Without their love and support, I would not have my
accomplishment.
iii |
Virginia Tech | Master Thesis of Jing Niu
Chapter 1 Introduction
1.1 General Background
The 20th century has experienced a rapid global population increase and along with it,
an explosive growth of potential water shortage problems. This shortage can be partially
attributed to limited natural resources and increased industrial activities. To resolve the
water scarcity problems in many regions, seawater is no longer merely a marginal water
resource but a commercial option for securing water supplies. Desalination applications
are not limited to seawater; brackish water, river water and wastewater also apply.
Desalination of saline water must be regarded seriously in the above circumstances.
Desalination technologies can be classified by their separation mechanism into membrane
based and thermal desalination. In membrane desalination, water diffuses through a
membrane, while salts are almost completely retained, i.e., reverse osmosis. Thermal
desalination separates salt from water by evaporation and condensation, i.e., distillation.
Reverse osmosis (RO) membrane desalination is a well-established membrane-based
water desalination technology. It has the advantages of high efficiency and low energy
cost, however, the process consumes a large amount of fresh water due to the low
backpressure. As a result, only 5-15 percent of the water entering the system can be
recovered. 7 Due to its membrane construction of fine pores, reverse osmosis not only
removes contaminants that may be present in the water, it strips many of the good,
healthy minerals from the water as well. Reverse Osmosis water is so chemically unstable
and acidic that many countries restrict water filtered via reverse osmosis from being
reintroduced into copper pipe due to its corrosiveness of the copper. This also has
implications for reverse osmosis filtration systems that use steel storage tanks, as the
acidity of the water can lead to the steel rusting over time and contaminating the post-
filter water.
Another widely used desalination technology is the distillation method. All distillation
techniques are based on a similar working principle. The water and the dissolved gases in
it are volatile when boiling saline water, while the minerals and dissolved salts do not
1 |
Virginia Tech | Master Thesis of Jing Niu
evaporate easily unless with boiling temperatures are above 300 Β°C. Single stage
distillation, multi-stage distillation and other kinds of new distillation techniques have
been developed. 8 The techniques above both address the problems in the treatment of
high salinity water, like produced water from shale gas industry. This way, clathrate
hydrate crystallization has been reconsidered as a potential unit operation for high salinity
water desalination.
Gas hydrates (clathrate hydrates) are crystalline solid structures consisting of water
and small molecules, gas or liquid, such as CO , N , CH , et al. which are formed under
2 2 4
conditions of specific temperature and pressure. It is now obvious that clathrate hydrate
are non-stoichiometric compounds and different from ice which has a hexagonal
structure. The majority of gas hydrates are known to form three typical hydrate crystal
structures, structure I (sI), structure II (sII), and structure H (sH) 9 . Figure 1.1 shows the
different hydrate structures and their associated cage type: sI comprises two different
cage types, a small pentagonal dodecahedral cage, denoted 512 (contains 12 pentagonal
faces on the cage), and a large tetrakaidecahedral cage, denoted 51262 contains 12
pentagonal and 2 hexagonal faces on the cage). sII also includes the small 512 cage in
addition to a large hexacaidecahedral cage, denoted 51264 (contains 12 pentagonal and 4
hexagonal faces on the cage). Structure H is composed of the small 512 cage, a mid-sized
435663 cage (contains 3 square, 6 pentagonal, and 3 hexagonal faces on the cage), and a
large icosahedral cage, denoted 51268 (contains 12 pentagonal and 8 hexagonal faces on
the cage). The type of structure formed depends primarily on the size of the guest
molecule; i.e., methane fits into both the small and large cages of sI, whereas propane is
too large to fit into the large cage of sI but can fit into the large cage of sII and therefore
forms sII. Gas hydrates found in oil and gas pipelines are mainly sII because natural gas
contains methane with small amounts of larger hydrocarbon molecules such as propane
and isobutane. Conversely, the majority of naturally occurring deposits of gas hydrates
are sI because they are composed of methane (from biogenic sources) and do not contain
heavier hydrocarbons. Exceptions are thermogenic gas hydrate deposits that contain
heavier hydrocarbons and therefore are formed from sII and in rarer cases sH. In a gas
hydrate structure, water molecules consist of cavity, and usually each cavity includes one
gas or liquid molecule inside.
2 |
Virginia Tech | Master Thesis of Jing Niu
Desalination by Freezing
During the freezing process, only water molecules can construct ice crystal and salts
are excluded under controlled conditions. Before the entire mass of water has been frozen,
the mixture is usually washed and rinsed to remove salts in the remaining water or
adhering to the ice. 10
Youssef Mandri et al. 11 studied the sweating step in the seawater desalination process
by indirect freezing. The whole process of desalination involves a freezing step, followed
by a purification of the ice layer by sweating. They successfully reduced an ice layer of
salinity from 35g/kg to lower than 0.5g/kg within 23h, obtaining a desalination efficiency
of 86%. They concluded that progressive increase of the sweating temperature does not
seem to bring more salinity away from ice layer.
Cong-shuang Luo et al.12 worked on factors affecting the quality of ice crystal during
the freezing concentration for the brackish water by using the method of crushing ice and
centrifugation. From their experiments, it was almost impossible to obtain pure ice
crystals with the freezing method only, because during the growth that solution was
trapped in ice in the form of brine pockets according to Ebert et al.13 . Crushing ice can
lead to the breakage of this structure and release brine.
Lubos Vrbka and Pavel Jungwirth14 studied the molecular dynamics of brine rejection
from freezing salt solutions. They observed that the system freezes as neat ice by
expelling ions into a higher brine concentration region, where ions can sometimes be
trapped inside the ice crystal and incorporated further water molecules into the ice lattice
slower.
Khhoudir Medjani15 believed in the existence of brine pocket during the freezing of
NaCl-H O compound. They observed the interface between the ice and liquid, revealing
2
that the liquid is trapped inside the mushy zone (a place where liquid and ice coexist) in
the form of pockets as a result of necking of the interfacial tubes as the ice grows.
Freeze-melting process requires much less energy than evaporation/distillation process
and minimizes the capital cost because many inexpensive plastics or low cost materials
4 |
Virginia Tech | Master Thesis of Jing Niu
can be utilized at low temperature. 16 However, It has not achieved commercial success
because of the complexity of the process and higher cost in total.
Phase Equilibrium in Gas Hydrate Formation
Usually, three major methods are used in gas hydrate phase equilibrium measurement,
Gibbs phase rule, Duhemβs theorem and Van Poolen. Juan G. Beltran, Hallvard and
Phillip 17 summarized the Gibbs phase rule in measuring techniques in gas hydrate
equilibrium. Pure hydrate system and divariant or multicomponent hydrate system are
studied.
In the pure hydrate system 18 , only one variable needs to be controlled to search for
equilibrium. Some laboratories prefer to manipulate temperature while others tend to
control pressures. In a divariant hydrate system, three variables including pressure and
temperature need to been provided and plotted, and verification need to be done through
comparison 19.
J.- M. Herri, A. Bouchemoua, et al.20 studied the gas hydrate equilibrium for CO -N
2 2
and CO -CH mixtures. They analyzed the gas mixtures with a gas chromatography to
2 4
obtain the exact composition of each gas and calculated gas composition in different
phases through the gaseous component mass balance
Xiao-Sen Li et al. 21-23 studied the effect of tetrabutyl ammonium bromide (TBAB) on
equilibrium of CO -H mixture hydrate formation. They confirmed as Hashimoto et al.
2 2
and Aladko et al. reported that type B TBAB hydrate formed preferentially with mole
fraction less than 1.4%, while type A TBAB hydrate is more thermodynamically stable
with mole fraction larger than 1.4%. From experiments they found that the addition of
TBAB reduces pressure in hydrate formation considerably, TBAB additive make a larger
reduction of the equilibrium formation pressure comparing to THF and CP. They
obtained gas separation efficiency by reducing CO mole fraction from 39.2 to 18%.
2
Kinetic Formation of Gas hydrate
Junshe Zhang and Jae W. Lee 24 worked on enhancing kinetic formation of CO
2
hydrate by adding cyclopentane (CP). They found that increasing the amount of CP leads
5 |
Virginia Tech | Master Thesis of Jing Niu
to an increase of both the water conversion and CO recovery, while decreasing the ratio
2
of CO entrapped in sI hydrate to that in sII hydrate 25 . When they increased the
2
formation water supply from 100 to 200cm3, the water conversion and CO growth rate
2
both decreased for the first charge of CO , and with more gas hydrate formed. They also
2
considered that the driving force of hydrate formation is the highest near vessel wall
which cause the gas hydrate to grow along the wall, and more hydrate accumulate as a
porous structure at this position and draw water from bulk water by the capillary force.
Matthew A. Clarke and P.R. Bishnio 28 worked on the intrinsic kinetics of CO
2
hydrate formation by using in situ particle size analysis with a focused beam reflection
method (FBRM). The starting point of nucleation was observed when the small particles
number decreased immediately while the large particles number began to grow.
J. S. Zhang, et al. 29 found there is no clear relation between SDS addition amount and
induction time. The promoting effects of SDS on hydrate nucleation was confirmed,
though increasing SDS may inhibit initial hydrate nucleation due to the fact adding
electrolyte concentration decrease water activity. They got that SDS can prevent the
agglomeration of hydrate particles, which contribute to increasing particle numbers over
time for a given nucleation rate at the start of hydrate growth.
Seong-Pil Kang and Yutaek Seo 30 formed methane and carbon dioxide hydrate in
silica gel pores instead of traditional mechanical stirring method. They figured out that
methane hydrate formation rate is not dependent on the pore size, and for carbon dioxide
hydrate there exists an optimal pore size for the hydrate formation rate. Though hydrate
formed from porous silica takes longer formation time and has lower initial formation
rate than traditional methods, they accomplished remarkable cage occupancy and helped
to remove expensive mechanical agitation.
Masayoshi Takahashi et al. 31 found the possibility of applying shrinking microbubble
on the gas hydrate formation due to their large surface area, very long stagnation and a
pressurized interior gas. They pointed out that microbubble can change the nucleation
conditions because of its excellent gas-dissolving capacity. Hideo Tajima, et al. 32
6 |
Virginia Tech | Master Thesis of Jing Niu
thought about the possibility to replace traditional stirring vessels with a kinetic-type
static mixer in formation of CO hydrate continuously.
2
Thermodynamic Properties Improvement in Gas Hydrate Formation
Atle Svandal, et al. 32 focused on thermodynamic properties of the H O/CO /CH
2 2 4
system, especially the interface thermodynamics. They considered interfacial thickness
and free energy to derive kinetic and thermodynamic properties. Their investigation into
bulk CO hydrate and water system showed the hydrogen bonds per water molecule is
2
1.65 for hydrate water, while 1.39 for liquid water, and thus they expected that the
interface Gibbs free energy for this hydrate system is close to that of ice-water system
(Ξ³=29.1Β±0.8 mJ/m2).
Angel Martin and Cor J. Peters 2 modified the fugacity-based Van der Waals-
Platteeuw model of Klauda and Sandler 33 . From substitution of thermodynamic model
they modified, they figured out that the lattice distortion depends only on the occupancy
of the large cavity by THF. Also they got that neither large nor small cavity occupancies
has variations with the initial concentration of THF in the mixture, changing the initial
addition of THF can only lead to the amount of hydrate formed.
Instrumental Techniques in Analyzing Gas Hydrate Formation
A series of advanced instrumental techniques are needed to observe the gas hydrate
formation, as well as the structures and properties of different kinds gas hydrate34 .
Y. Rojas and X. Lou 35 summarized some modern instrumental analyses and their
corresponding basic physical science principles and the gas hydrate properties that each
method can characterize. Robert W. Hennig, Arthur J. Schultz, et al. 36 investigated the
formation of CO hydrate in the use of time-of-flight neutron powder diffraction at
2
temperature range of 230 to 290K, and a pressure of 900psi. They found that peaks
corresponding to ice slowly disappeared, while hydrate peaks intensity increased when
the sample reached 275.6K. Diffraction peaks of indicating ice disappeared completely at
276.4K, while the diffraction peaks of hydrate number appeared and began to increase.
They also certificated the phenomenon that the rate-limiting step of the process after
7 |
Virginia Tech | Master Thesis of Jing Niu
approximately 20% conversion is the diffusion of CO molecules through the layer of
2
hydrate 37 , in agreement with the assumption 1 that the slow transport of CO molecules
2
across the growing solid hydrate layer may cause the slower CO hydrate growth rate in a
2
water drop covered with a thin film of CO hydrate 1 .
2
Igor L, Christopher I. Ratcliffe, et al. 2 used 1H NMR (Nuclear Magnetic Resonance)
microimaging techniques to examine hydrate-coated ice particle as a function of
temperature from 250 to 276K. From images captured by NMR, they figured out that ice
inside the hydrate, though may not be easily observed by monitoring temperature and
pressure, are able to melt. They also observed that conversion from ice to hydrate is
stopped where the volume required inside the shell increased as shown in Figure 1.2,
which in agreement with the Pilling-Bedworth rule believing that the surface reaction
Figure 1.2 Transformation of D O ice to CO hydrate: (a) ice at 263 K before
2 2
adding CO ; (b) mixture of ice and CO hydrate at 263 K and 900
2 2
psi; (c) CO hydrate at 276.4 K and 900 psi. The noisy background
2
is due to the short collecting times of 15 min per histogram. 6
8 |
Virginia Tech | Master Thesis of Jing Niu
should result in a dense surface-layer that ultimately limits the reaction.
Carolyn A. Koh and Jeffery L. Savidge 3 used energy-dispersive diffraction in
capturing the crystalline structural dynamics of carbon dioxide and propane hydrate
structures during formation. They confirmed the sI structure of CO hydrate and sII of
2
propane hydrate.
T Komai, T Kawamura, et al. observed in situ gas hydrate behavior under high
pressure with Raman spectroscopy. At first, they tested CH hydrate and got the hydrate
4
number in the range of 6.4-6.0, the occupancy ratio in small cages was estimated to be
approximately 85-95%, meanwhile, presence of guest gases in large and small cages are
both measured and showed by peaks in Raman spectra from Figure 1.3. In measurement
of CO gas hydrate, they found that gas hydrate formation rate change greatly with
2
temperature, especially in the range between 269 and 275K, suggesting the existence of
quasi-liquid water, a type of slightly melting ice that help to promote gas hydrate
formation in the gas-solid interface.
Figure 1.3 Raman shifts in the disassociation of CH gas hydrate 3
4
Hydration Energy of Selected Ions
Enthalpy of hydration, which is the amount of energy released when a mole of the ion
dissolves in a large amount of water forming a hydrated ion. Hydration energy of selected
ions is listed in Figure 1.4.
9 |
Virginia Tech | Master Thesis of Jing Niu
Chapter 2 Single Step Desalination %Reduction through
Gas Hydrate Formation Process
Abstract
Gas hydrate was formed to clean up artificial produced water, all experiments were
performed under condition of pressure in the range of 450 to 800psi and temperature in
the range of -1 to 1Β°C. The artificial produced water contains various dissolved ions,
including Na+, Ca2+, Mg2+, Cl- and other ions, and the total dissolved solids (TDS) are in
the range of 40,000 to 450. NaCl had the weakest inhibiting effect on gas hydrate
formation and consumed the largest amount of CO . Higher NaCl concentration leads to
2
longer induction time. The optimal incipient pressure of CO consumption was obtained
2
for both NaCl and CaCl . The MgCl had the lowest desalination %Reduction while
2 2
CaCl obtained the highest by gas hydrate formation. The present study illustrated the
2
optimal control condition during gas hydrate formation and CO consumption;
2
meanwhile it certificated the possibility of applying gas hydrate formation process in
desalination.
2.1 Introduction
Gas hydrate belongs to a group of compound known as clathrate, in which water
molecules are linked through hydrogen bonding and construct cavities (host lattice) that
can enclose a large variety of molecules (guest). No chemical bonding exists between the
host molecules and the enclosed guest molecules 4 . Clathrate formation causes a series of
problems in oil and gas industry during the production, transportation processes. Large
hydrated masses occurring in natural gas pipelines, for example, in Arctic regions and in
the sea, can slow or completely obstruct flow. Also there has been a recent resurgence in
developing methods to harvest the huge amount of methane present in natural gas hydrate,
such as methane hydrate stored in the sediment in the deep ocean that could be regarded
as a promising new energy supply, and natural gas stored in the earth crust. A recent
estimate of methane amount trapped in hydrate is as much as 300 times that in
conventional U. S. reserves 5 . Widely interests have been presented on the clathrate
14 |
Virginia Tech | Master Thesis of Jing Niu
hydrate properties, structure, and formation process during the last decades. Gas hydrates
are thermodynamically stable under high pressure and near freezing temperature of pure
water 6. Although different gas has different thermodynamic stable pressure and
temperature conditions, most of gas can form gas hydrate in specific conditions,
including methane, carbon dioxide, hydrogen and some gas mixture.
Tsutomu Uchida et al. 8 worked on the kinetic and stability of CO hydrate comparing
2
to methane hydrate by using chromatography and Raman Spectroscopy. From their
thermal calculation, they observed that methane molecules are preferentially crystallized
in the early stages of hydrate formation when the initial methane concentration is much
less than that of CO , and pure methane hydrate was first formed when existing together
2
with CO . However in the reverse process, CO is much easier to be released from CH -
2 2 4
CO mixture under the same condition.
2
Susan Circone et al. 9-11 searched on synthesis, composition and dissociation behavior
of CO hydrate in comparison with methane hydrate. They divided the dissociation
2
behavior of CO hydrate into two parts, one is when the temperature under 273 K, sI CO
2 2
hydrate dissociation is dominated by one characteristic: bulk of hydrate is released base
on a time scale regardless of the pressure-temperature pathway the sample follows, which
is in contrast with CH that is highly temperature-pressure path dependent; another
4
situation happens above 273 K, the observed dissociation behavior of CO hydrate is
2
consistent to that of CH hydrate in that the internal temperature decreased below the
4
H O melting point by following the pressure release until dissociation completed and
2
then climb to the H O melting point.
2
Gang Li et al. studied on effect factors for CO hydrate rapid formation in a water-
2
spraying apparatus. They found that oscillating gas supply mode leads to larger gas
consumption in the condition of continuous gas supply mode, due to large mass-transfer
driving force. CO has a larger solubility in gas hydrate formation, and thus larger
2
volume of water in the vessel cause a large amount of gas consumption. In their spraying
system, nozzle atomizing angle can be different and has various effects on the gas
hydrate formation, they acquired that a larger nozzle atomizing angel cause shorter
15 |
Virginia Tech | Master Thesis of Jing Niu
induction time, explaining by the reason that making liquid water into spray micro fog
drops or tiny particles can enhance the interface and contact areas significantly.
2.2 Experimental
Materials
Airgas carbon dioxide with a certificated purity of bone dry grade. Sodium Chloride of
certificated ACS crystalline supplied by Fisher science Co.; Calcium Chloride supplied
by Fisher Science Education with 4-20 mesh; and Magnesium Chloride supplied by Alfa
Aesar Co. with purity of 99% were used.
EDTA (Ethylene diamineteraacetic acid disodium salt dehydrate) with a purity of 99+%
supplied by Alfa Aesar Co. was used. Sodium Hydroxide with a purity of β₯98% supplied
by Sigma-Aldrich Co. was used. Eriochrome Black T with pure indicator grade supplied
by Acros Organic Co. and Patton & Reedersβ (Calconcarboxylic acid) for complexometry
were used as indicator in titration.
Analysis of Produced Water from Shale Gas Industry
Produced water (PWs) is generated during processes such as fossil fuel extraction,
fossil fuel energy production, and industrial operations. In the continental US, major
resources of PWs include conventional natural gas PWs, conventional oil PWs, coal-bed
methane PWs, shale gas PWs and tight gas sands PWs 12 . Some produced water are
released during or after fracturing of the source formation and gas recovery, such as coal-
bed methane, shale gas and tight gas sands produced water 13 . Bethany Alley et al. 16
studied about chemical and physical characterization of produced water, they got the fact
that significant difference were observed of calcium, potassium, magnesium, sodium iron,
manganese, zinc and chloride concentrations between different kinds of produced water.
In our sample data of shale gas onsite from NETL, we found that usually produced water
includes bicarb, calcium, chloride, magnesium, potassium, potassium, sodium and many
other organic components.
16 |
Virginia Tech | Master Thesis of Jing Niu
Table 2.1 Average concentrations of the various electrolytes present in produced water
Table 2.2 Major ion concentration in produced water from onsite data
However, through the calculation and analysis, we acknowledged that major salt ions
are chloride, magnesium, sodium and calcium and their corresponding concentration was
shown in table 2.1 and table 2.2.
Experimental Apparatus
The main component of the apparatus consists of a stainless steel volume constant
equilibrium vessel with a maximum volume of 100mL. The vessel is machined from a
solid 316 stainless steel and equipped with a magnetic stirrer drive with a maximum rpm
of 683.10. The maximum working pressure of the vessel is 3000psi, and the maximum
working temperature is 350Β°C. The vessel is sealed with PTFE Gaskets as well as O-rings.
The temperature inside is measured with a type J (iron constant) Omega thermocouple.
The accuracy of the thermocouple measurement is believed to be Β±0.10K. The pressure is
measured by a 0-3000psi stainless steel Bourdon tube and 1/4" NPT male connections
pressure gage from Parr Instrument Company. The coolant bath is Endocal RTE-series
refrigerated circulating bath supplied by Neslab Co.
Experiment Procedures
Weighed 50g deionized H O and specific weight of salt on an electronic balance with
2
a readability of Β±0.1mg, put them all into a 200mL beaker, and kept magnetic stirring for
5min to prepare feed saline water, after the solution turning to clear and all solute
17 |
Virginia Tech | Master Thesis of Jing Niu
dissolved, finally discharged well stirred solution into the vessel. The uncertainty of the
solution concentration was less than 0.2%. The computer-based data-acquisition system
can automatically record real-time changes of pressure and temperature in the vessel
every one second.
Figure 2.1 Schematic of the experimental apparatus in gas hydrate formation
The reaction vessel was rinsed three times with deionized water at the beginning.
Sealed the reaction system and pumped CO gas in a low pressure and room temperature
2
in order to evacuate air or other gas in the gas system. Kept flowing CO for around
2
15min under 25-35psi and room temperature to flush all the other gases with CO in the
2
vessel, then we stopped CO gas flow and let the inner pressure decrease slowly and
2
continuously to 0psi to make sure CO was the only gas fulfilled the vessel. Pumped CO
2 2
gas once again into the vessel until the inside pressure reached 60psi, turned off the valve
and put the vessel into coolant bath. Set the coolant bath temperature as 1Β°C, cooling
down the vessel until the temperature inside stayed stable in 1Β°C for 3-4 hours. Started
experiment by imputing CO , after the vessel was pressurized to an expected starting
2
value; CO was stopped and agitation was started at a speed of 683.10 rpm at the same
2
time. As long as the reaction started, the system temperature was lowered, first frequently
and then slowly, accompanying with the same trend of pressure to form hydrate. Each
experiment acquired ending point of gas hydrate formation for 20min. Recorded the
18 |
Virginia Tech | Master Thesis of Jing Niu
temperature and pressure both at the starting and ending point, starting point was
regarded as the point when we stopped imputing CO and started agitation. If there was
2
gas hydrate formed in the equilibrium vessel at the end of experiment run when
temperature and pressure stayed in constant, then we counted this point right before the
vessel was opened as ending point. Stopped agitation and opened the valve to discharge
residual CO inside the vessel until the inner pressure decrease to 0psi. Opened the vessel
2
in room temperature when two parts were obtained, CO hydrate in solid phase and dirty
2
water in liquid phase. We moved gas hydrate and dirty water parts into two beakers
separately, tested the concentration of each part with correct ion-test techniques. The
schematic of the experimental apparatus for gas hydrate formation is shown in Figure 2.1.
Measurement of Na+ Ion Concentration
CON 510 Bench Conductivity/TDS Conductivity Meter was used to test sodium ion
concentration. Calibrated the ion meter with standard solution before measurement.
Prepared NaCl solutions with weight concentrations range from 0.01 to 0.1wt.%,
recorded corresponding conductivity every 0.01wt.%, and plotted the relationship
between concentrations and conductivity. We get linearly relation between concentration
and conductivity through following equation:
πΆππππππ‘πππ‘πππ = 5Γ10!!πΆππππ’ππ‘ππ£ππ‘π¦β0.0034 (2.1)
in the equation, concentration unit is wt.%, and conductivity unit is Β΅S/cm. From the
above equation, we can calculate NaCl concentration according to conductivity.
Measurement of Ca2+ Ion Concentration
EDTA titration was used to detect Ca concentration. Prepared a 0.005mol Lβ1 EDTA
solution and 8mol L-1 NaOH solution. EDTA (ethylenediaminetetraacetic acid) forms a
complex with calcium ions. Used Patton-Reeder indicator triturate as indicator of end
point.
For the titration, the indicator is added to the sample solution containing the calcium
ions and forms the pink/red calcium ion-indicator complex (Ca-PR). This solution is then
titrated with EDTA. The endpoint occurs when the solution turns blue, indicating that the
19 |
Virginia Tech | Master Thesis of Jing Niu
Ca-PR complex has been completely replaced by the calcium ion-EDTA complex and the
PR indicator reverts to its blue color.
The reaction is:
πΆπβππ
+πΈπ·ππ΄!! β ππ
+ πΆπβπΈπ·ππ΄ !! (2.2)
Pipetted a 10mL of the unknown solution into a 250ml conical flask. Added 40mL
distilled water and 4mL sodium hydroxide solution of 8mol Lβ1, and allowed solution to
stand for about 5min with occasional swirling. Added the indicator after you have given
this precipitate a chance to form. Added 0.1g of Patton-Reeder indicator and swirled the
solution to dissolve the indicator. Titrated the sample with the EDTA solution. Recorded
volume use of EDTA solution and calculated with following equation to get Ca2+
concentration:
πΆπ!! πΆππππππ‘πππ‘πππ ππ π’πππππ€π π πππ’π‘πππ = !(!"#$)Γ!(!"#$) (2.3)
!(!"#$%&)
unit of Ca2+ concentration is mol L-1.
Measurement of Mg2+ Ion Concentration
EDTA titration method was used to detect Mg2+ concentration. The reaction of Mg2+
with EDTA (Ethylenediaminetetraacetic acid) may be expressed as:
ππ!! +π» π!! β πππ!! +2π»! (2.4)
!
Chose unknown solution, pipetted 3ml of unknown solution into a conical flask, added
approximately 10ml of buffer and 50ml deionized water into the flask; added 4 drops of
Eriochrome Black T indicator (aq.) and we got a light, wine-red color. Titrated solution
with standardized EDTA solution to a clear blue color. Recorded volume of EDTA used
in titration. Use following equation to calculate the concentration of unknown solution:
Mg!!Concentration of unknown solution = ! !"#$ Γ!(!"#$) (2.5)
!(!"#$%&)
unit of Mg2+ concentration we get from above equation is mol L-1.
20 |
Virginia Tech | Master Thesis of Jing Niu
Estimation of Carbon Dioxide Consumption
Gas consumption was calculated in the base of ideal gas law modification with
compressibility factor, pressure and temperature in both starting point and ending point
when gas hydrate formed, which was also used by Hyun Ju Lee. et al. 17. Following
equation was used to calculate the moles of CO consumed.
2
!" !"
βπ = β (2.6)
!"# ! !"# !
where compressibility factor, z, is calculated by CO critical point and present point with
2
temperature and pressure; P and T are from read of reactor; while V is approximately
equal to volume of reactor minus volume of water in the reactor.
Calculation of %Reduction of Salts
We defined desalination %Reduction with following equation:
! !!
%π
πππ’ππ‘πππ = ! !"#$%&βΓ100% (2.7)
!!
in which c is feed water concentration, and c is concentration in final gas hydrate
0 product
solid product.
2.3 Results and Discussion
Effect of Feed Water Concentration on the CO Consumption
2
Table 2.3 Experimental data of %Reduction of salts with different TDS.
21 |
Virginia Tech | Master Thesis of Jing Niu
Figure 2.4 Effect of Mg2+ concentration in feed water on the CO consumption,
2
with initial pressure as 650psi and temperature as 1Β°C.
For NaCl solution, in high concentration range from 10 to 20wt.% as shown in Figure
2.2, with initial condition of 800psi and 1Β°C, CO consumption increased from 0.0428 to
2
0.0444moles as concentration increased. This implies CO solubility increases in gas
2
hydrate formed in higher concentration feed water.
Figure 2.3 shows CO2 consumption decreases from 0.0528 to 0.0394moles as
concentration increases in low concentration range, which is in agreement with Pankaij D.
Dholabhai 1 research that they saw clear inhibiting effect of NaCl with concentration
between 3 to 5wt.%. Higher concentration tends to consume more CO2 in average,
indicating stronger inhibiting effect on gas hydrate formation and CO2 consumption.
Under the gas hydrate formation conditions of 650psi and 1Β°C, the mole consumption of
both MgCl2 and CaCl2 are much lower than that of NaCl solution, indicating CaCl2 and
MgCl2 own stronger inhibiting effect in gas hydrate formation.
24 |
Virginia Tech | Master Thesis of Jing Niu
to 1.0wt.%. The first part is in agreement with Seong-Pil Kang et al. 1 theory about the
MgCl2 inhibiting effects on CO2 hydrate formation and thus encapsulate less gas. We try
to explain the second part for the reason that CO2 solubility in concentration higher than
0.5wt.% cannot be ignored any more.
Figure 2.5 shows the CO capture ability in concentration range of 1.5 to 5.5wt.%
2
CaCl , concentration point is chosen every 0.5 wt.% under same initial condition of
2
700psi and 1Β°C. In the first region of 1.5 to 2.5wt.%, CaCl concentration growth leads to
2
less CO consumed smoothly, in the second region of 2.5 to 3.5wt.%, the CO
2 2
consumption decreases rapidly; in the third region, CO consumption stayed almost the
2
same. In the first two regions, the inhibiting effect of CaCl is observed obviously.
2
In low concentration range shown in Figure 2.6, addition of CaCl leads to CO
2 2
consumption reduction. Overall, it seems like that CaCl owns a stronger inhibiting effect
2
on gas consumption in low concentration than that in high concentration range, which is
opposite to general acknowledgement. We explain this phenomenon by different pressure
manipulated.
Effect of Initial Pressure on the CO Consumption
2
We carried out CO hydrate formation under different incipient pressure varied from
2
450 to 700psi by controlling temperature in a specific value of 1Β°C. 1.08wt.% NaCl and
0.015wt.% CaCl were chosen as supply water separately, with pressure point gathered
2
every 50psi, and corresponding CO consumption were calculated.
2
Figure 2.7 shows that when we increase pressure from 450 to 650psi, gas
consumption rise gradually and reaches the largest mole consumption of 0.066 moles at
650psi, but if pressure is further improved to 700psi, a decrease of consumed CO is
2
observed. It implies that an optimal pressure on gas consumption is obtained and as
higher pressure leads to a larger driving force, and larger driving force promotes CO
2
sequestration.
In 0.015wt.% CaCl solution as shown in Figure 2.8, an optimal pressure of 720psi on
2
gas consumption is noted, which is in consistent with NaCl. Overall, larger driving force
26 |
Virginia Tech | Master Thesis of Jing Niu
Effect of Feed Water Concentration on the Induction Time
Artificial brine solutions in various TDS were prepared and their corresponding curves
of pressure-reaction time are plotted. In curve of pressure-reaction time, a spike
indicating the starting point of gas hydrate formation is observed and time period from
beginning to the spike is defined as induction time. According to Rajnish Kumar et al. 2
visual observation of hydrate growthοΌ extensive hydrate formation and crystal
agglomeration results in accumulation of crystal as a stagnant layer at the gas/water
interface, which prevents more gas from contacting with water.
In Figure 2.9, pressure-reaction time curves of NaCl in concentration from 10 to
18wt.% are presented. Induction time increase from 3 to 15min as concentration
improves from 10 to 18wt.%, which confirms what we obtained earlier that higher NaCl
concentration owns stronger inhibiting effect on gas hydrate formation. In general, the
induction time is shorter as the driving force increases. 3 From above-mentioned,
increment of concentration leads to weaker driving force, and longer time for nucleation
in high concentration range.
Induction time in low NaCl concentration range are shown in Figure 2.10, indicating
that under incipient formation condition of 650 psi and 1 Β°C, 1.08wt.% results in the
shortest induction time of 1min, while 2.0wt.% has the longest. We are able to figure out
that there is an optimal concentration of 1.08wt.% on induction time. Overall, brine
concentration increment leads to stronger inhibiting effect though it is not obvious in low
concentration range.
For MgCl , it is can be seen in Figure 2.11 that the shortest induction time is obtained
2
in 1wt.%, while the longest time happened in 0.5wt.%. In this way, it is reasonable to
conclude that either increasing or decreasing concentration from 0.5wt.%, can we start
gas hydrate formation in a shorter time and inhibiting effect of Mg concentration is not
significant in this region either.
28 |
Virginia Tech | Master Thesis of Jing Niu
Figure 2.13 Effect comparison of different ions on the induction time, with pressure
as 650 psi and temperature as 1 Β°C.
No obvious impact of Ca concentration on induction time is observed in Figure 2.12,
more tests need to be operated if we wanted to explore the correct relation between
concentration and induction time in CaCl .
2
We produced gas hydrate formation with different brine and results are plotted in
Figure 2.13. It indicates that NaCl obtains the shortest induction time while 0.5 CaCl
2
and MgCl has almost the same induction time. No obvious difference of induction time
2
between brine water is observed from Figure 2.13.
Effect of Initial Pressure on the Single-Stage Desalination %Reduction
All experimental data were collected and shown in table 3 and plotted in Figure 2.14,
it indicates that %Reduction improved from 11.62 to 38.44% and decreases after as the
increase of initial pressure from 450 to 650psi. We figured out an optimal formation
pressure of 650psi.
31 |
Virginia Tech | Master Thesis of Jing Niu
In higher concentration region (>1.0 wt.%), %Reduction of NaCl in various
concentrations were presented in Figure 2..It was observed the optimal concentration of
10 wt.%. Overall, the average %Reduction was as low as 20 %.
Experimental data of CaCl were shown in Figure 2.15, it could be seen that increase
2
of Ca concentration from 1.5 to 4.5wt.% brings %Reduction of CaCl from 47.87 down
2
to 10.03%, and %Reduction stays almost the same in concentration higher than 4.5wt.%.
Overall, we found that higher concentration doesnβt result in lower %Reduction and vice
versa for both salts and higher %Reduction of CaCl than that of NaCl.
2
Figure 2.17 Effect comparison of different TDS in feed water on the %Reduction,
with initial pressure as 650psi and temperature as 1Β°C.
In lower concentration region (<1.0 wt.%), all of experiments performed under
condition of 650 psi and 1 Β°C. Figure 2.16 implies that increment of Mg concentration
leads to %Reduction fell from 7.13 down to 1.15%.
In Figure 2.17 we compared three brine feed water concentration effects
on %Reduction; we figured out that an optimal concentration of CaCl happened on
2
0.05wt.% while NaCl appeared on 1.08wt.%. We found that decreasing concentration
doesnβt cause better %Reduction for NaCl and CaCl solution, however, for MgCl
2 2
34 |
Virginia Tech | Master Thesis of Jing Niu
Chapter 3 Dewatering of Gas Hydrate Formed from
Artificial Produced Water
Abstract
The results obtained in this study showed that over 99% of dissolved NaCl and MgCl
2
can be removed from artificial saline water in a process involving a single-stage hydrate
formation step, followed by a single-step dewatering. The hydrate crystals were cut into
small pieces and placed in vials and centrifuged. The results show that the %Reduction
(percentage of the concentration decrease) of NaCl increases with centrifugation time and
rotational speed (rpm) after the centrifugation. In filtration, artificial produced water
containing 0.8 to 2.0wt.% NaCl was used as feed water. The resulting hydrate crystals
were crushed using a mortar and pestle and subjected to vacuum filtration. As shown, the
values of %Reduction of NaCl were substantially higher than obtained using the
centrifuge. The %Reduction increased with increasing filtration time as anticipated. Also,
the process is effective over a wide range of salinity, which is an advantage over reverse
osmosis. After reducing the particle size further by grinding, %Reduction reached 90%,
indicating that the finer the particle size, the higher the extent of cleaning.
3.1 Introduction
Gas hydrates are ice-like crystalline compounds that are formed by the combination of
water with gas molecules under suitable pressure and temperature conditions. The water
molecules through hydrogen bonding are capable of forming a three-dimensional lattice
structure containing cavities. These cavities can be occupied by molecules of gases and
volatile liquids whose molecular diameter is smaller than that of the cavity. By the
consumption of gas or liquid molecules, the structure, by itself thermodynamically
unstable, is stabilized. Each cavity structure may contain one guest molecule.
Based on the fact that gas hydrate is capable to form a solid substance consisting of
clean water and gas molecules only, the possibility of using gas hydrate for water
desalination has been investigated during the last decades 4,5 and is actively being
developed at industrial scale at present. The process of gas hydrate formation and
37 |
Virginia Tech | Master Thesis of Jing Niu
decomposition are exothermal and endothermic respectively, which can cause their
mutual inhibition during phase changes. Meanwhile, we can achieve energy consumption
enhancement by optimizing the exothermal and endothermic processes by utilizing heat
produced in hydrate formation process on the melting of hydrate in further stage of the
desalination. The objective of modern desalination industry field requires producing a
large amount of clean water in a short period and with the minimum energy
consumptions. In this way, optimizations of the gas hydrate formation and further stages
in the desalination process are urgently needed 6 .
CO hydrate forms at very high pressure but the low cost of CO , the ease of stripping
2 2
it from solution, its low corrosive impact on the apparatus, lack of toxicity and
flammability 7 put it in a promising place in desalination industry.
M. D. Jager and E. D. Sloan 8 worked on the effect of pressure on methane hydrate in
sodium chloride solutions and pure water. Djurdjica Corak et al. 6 explored the effect of
subcooling and amount of formation of cyclopentane hydrate in brine and obtained the
desalination efficiency through the gas hydrate formation process. The removal of NaCl
varied from 89 to 73% with the best value obtained under temperature condition of 5.6 K
and 3.6 K in atmosphere pressure. Larger amount of gas hydrate formed under 3.6 K
series and the residual salinity is likely to be strongly related to the procedure for
separating the solid hydrate phase from residual water. From the relationship between
adhesion forces and temperature concluded by A.G. Speight 9 , they also obtained that the
attractive force between hydrate particles was stronger and thus hydrate crystal adhered
each other in a more tight framework formed at 3.6 K than the one at 5.6 K, and that
means more brine will be entrapped between the crystals; extraction procedure like
centrifuge was performed to remove this part of brine.
Zadjia Atik et al. 10 found that gas hydrate dissociation pressure for pure methane in
different electrolyte solutions. They compared the effect of MgCl and CaCl on the
2 2
methane hydrate formation, MgCl solution had a stronger ability to lower water activity
2
and thus a higher prevention capability, and it also causes a larger freezing point
depression than CaCl does. P. R. Bishnoi and P. D. Dholabhai 11 tried to acquire the
2
equilibrium conditions of propane hydrate in aqueous electrolyte solutions. They used
38 |
Virginia Tech | Master Thesis of Jing Niu
two methods, one is keeping temperature in constant and increased the system pressure,
while another was setting both temperature and pressure as constant and then decreases
temperature, to determine the gas hydrate equilibrium conditions.
A. M. Aliev et al. 12 developed a mathematical model of seawater desalination through
gas hydrate method, they calculated the optimal feed rate of gas hydrate formation
apparatus and total height of working part if their apparatus applied in industry field.
3.2 Experimental
Materials
Airgas carbon dioxide with a certificated purity of bone dry grade. Sodium Chloride of
certificated ACS crystalline supplied by Fisher science Co.; Calcium Chloride supplied
by Fisher Science Education with 4-20 mesh; and Magnesium Chloride supplied by Alfa
Aesar Co. with purity of 99% were used.
EDTA (Ethylene diamineteraacetic acid disodium salt dehydrate) with a purity of 99+%
supplied by Alfa Aesar Co. was used. Sodium Hydroxide with a purity of β₯98% supplied
by Sigma-Aldrich Co. was used. Eriochrome Black T with pure indicator grade supplied
by Acros Organic Co. and Patton & Reedersβ (Calconcarboxylic acid) for complexometry
was used as indicator in titration.
Analysis of Produced Water from Shale Gas Industry
Produced water (PWs) is generated during processes such as fossil fuel extraction,
fossil fuel energy production, and industrial operations. 13 In the continental US, major
resources of PWs include conventional natural gas PWs, conventional oil PWs, coal-bed
methane PWs, shale gas PWs and tight gas sands PWs 14 . Some produced water are
released during or after fracturing of the source formation and gas recovery, such as coal-
bed methane, shale gas and tight gas sands produced water 15-17 . Bethany Alley et al. 18
studied about chemical and physical characterization of produced water, they got the fact
that significant difference were observed of calcium, potassium, magnesium, sodium iron,
manganese, zinc and chloride concentrations between different kinds of produced water.
In our sample data of shale gas onsite from NETL, we found that usually produced water
39 |
Virginia Tech | Master Thesis of Jing Niu
Measurement of Ca2+ Ion Concentration
EDTA titration was used to detect Ca concentration. Prepared a 0.005mol Lβ1 EDTA
solution and 8mol L-1 NaOH solution. EDTA (ethylenediaminetetraacetic acid) forms a
complex with calcium ions. Used Patton-Reeder indicator triturate as indicator of end
point.
For the titration, the indicator is added to the sample solution containing the calcium
ions and forms the pink/red calcium ion-indicator complex (Ca-PR). This solution is then
titrated with EDTA. The endpoint occurs when the solution turns blue, indicating that the
Ca-PR complex has been completely replaced by the calcium ion-EDTA complex and the
PR indicator reverts to its blue color.
The reaction is:
πΆπβππ
+πΈπ·ππ΄!! β ππ
+ πΆπβπΈπ·ππ΄ !! (3.2)
Pipetted a 10mL of the unknown solution into a 250ml conical flask. Added 40mL
distilled water and 4mL sodium hydroxide solution of 8mol Lβ1, and allowed solution to
stand for about 5min with occasional swirling. Add the indicator after you have given this
precipitate a chance to form. Added 0.1g of Patton-Reeder indicator and swirled the
solution to dissolve the indicator. Titrated the sample with the EDTA solution. Recorded
volume use of EDTA solution and calculated with following equation to get Ca2+
concentration:
πΆπ!! πΆππππππ‘πππ‘πππ ππ π’πππππ€π π πππ’π‘πππ = !(!"#$)Γ!(!"#$) (3.3)
!(!"#$%&)
unit of Ca2+ concentration is mol L-1.
Measurement of Mg2+ Ion Concentration
EDTA titration method was used to detect Mg2+ concentration. The reaction of Mg2+
with EDTA (Ethylenediaminetetraacetic acid) may be expressed as:
ππ!! +π» π!! β πππ!! +2π»! (3.4)
!
41 |
Virginia Tech | Master Thesis of Jing Niu
Chose unknown solution, pipetted 3ml of unknown solution into a conical flask, added
approximately 10ml of buffer and 50ml deionized water into the flask; added 4 drops of
Eriochrome Black T indicator (aq.) and we got a light, wine-red color. Titrated solution
with standardized EDTA solution to a clear blue color. Recorded EDTA volume use in
titration. Use following equation to calculate the concentration of unknown solution:
Mg!!Concentration of unknown solution = ! !"#$ Γ!(!"#$) (3.5)
!(!"#$%&)
unit of Mg2+ concentration we get from above equation is mol L-1.
Measurement of Total Mg2+ and Ca2+ Ions Concentration
Using EDTA titration and a blue dye called Eriochrome Black T (ErioT) is used as the
indicator. This blue dye forms a complex with the calcium and magnesium ions,
changing color from blue to pink in the process. The dyeβmetal ion complex is less stable
than the EDTAβmetal ion complex. For the titration, the sample solution containing the
calcium and magnesium ions is reacted with an excess of EDTA. The indicator is added
and remains blue as all the Ca2+ and Mg2+ ions present are complexes with the EDTA.
A back titration is carried out using a solution of magnesium chloride. This forms a
complex with the excess EDTA molecules until the end-point, when all the excess EDTA
has been a complex. The remaining magnesium ions of the magnesium chloride solution
then start to complex with ErioT indicator, immediately changing its color from blue to
pink.
The main reactions are:
πΆπ!! +πΈπ·ππ΄!! β πΆπβπΈπ·ππ΄ !! (3.6)
ππ!! +πΈπ·ππ΄!! β ππβπΈπ·ππ΄ !! (3.7)
πΈππππ+ππ!! β πΈππππβππ (3.8)
Note: ErioT is blue and ErioT-Mg is pink.
42 |
Virginia Tech | Master Thesis of Jing Niu
We can use following equation to calculate the final total magnesium and calcium ion
concentration:
πππ‘ππ π +π =
! !"#$ Γ! !"#$ !!(!"!"!)Γ!(!"#$!)
(3.9)
!" !"
!(!"#$%&)
Unit of total concentration we got from the above equation was mol L-1
.
Experimental Procedure
Weighed 50g deionized H O and specific weight of salt on an electronic balance with
2
a readability of Β±0.1mg, put them all into a 200mL beaker, and kept magnetic stirring for
5min to prepare feed saline water, after the solution turning to clear and all solute
dissolved, finally discharged well stirred solution into the vessel. The uncertainty of the
solution concentration was less than 0.2%. The computer-based data-acquisition system
can automatically record real-time changes of pressure and temperature in the vessel
every one second.
Figure 3.1 Schematic of the experimental apparatus in gas hydrate formation
The reaction vessel was rinsed three times with deionized water at the beginning.
Sealed the reaction system and pumped CO gas in a low pressure and room temperature
2
in order to evacuate air or other gas in the gas system. Kept flowing CO for around
2
15min under 25-35psi and room temperature to flush all the other gases with CO in the
2
43 |
Virginia Tech | Master Thesis of Jing Niu
vessel, then we stopped CO gas flow and let the inner pressure decrease slowly and
2
continuously to 0psi to make sure CO was the only gas fulfilled the vessel. Pumped CO
2 2
gas once again into the vessel until the inside pressure reached 60psi, turned off the valve
and put the vessel into coolant bath. Set the coolant bath temperature as 1Β°C, cooling
down the vessel until the temperature inside stayed stable in 1Β°C for 3-4 hours. Started
experiment by imputing CO , after the vessel was pressurized to an expected starting
2
value; CO was stopped and agitation was started at a speed of 683.10 rpm at the same
2
time. As long as the reaction started, the system temperature was lowered, first frequently
and then slowly, accompanying with the same trend of pressure to form hydrate. Each
experiment acquired ending point of gas hydrate formation for 20 min. Recorded the
temperature and pressure both at the starting and ending point, starting point was
regarded as the point when we stopped imputing CO and started agitation. If there was
2
gas hydrate formed in the equilibrium vessel at the end of experiment run when
temperature and pressure stayed in constant, then we counted this point right before the
vessel was opened as ending point. Stopped agitation and opened the valve to discharge
residual CO inside the vessel until the inner pressure decrease to 0psi. Opened the vessel
2
in room temperature when two parts were obtained, CO hydrate in solid phase and dirty
2
water in liquid phase. We moved gas hydrate and dirty water parts into two beakers
separately, tested the concentration of each part with correct ion-test techniques.
The experimental apparatus for gas hydrate formation was shown in Figure 3.1. Two
parts were obtained after gas hydrate formation: CO hydrate in solid phase and dirty
2
water in liquid phase. In principle, CO hydrate is formed with pure water, thereby
2
excluding dissolved impurities. However, the residual water containing high-salinity
brines are trapped in between the hydrate crystals. Therefore, it is necessary to go through
a solid-liquid separation (or dewatering) process before melting the hydrate crystals to
obtain pure water. We chose two different separation methods.
1) Centrifuge
2) Filtration with vacuum pump
44 |
Virginia Tech | Master Thesis of Jing Niu
Centrifugation Dewatering Procedure
Figure 3.2 Flow sheet of desalination process involving gas hydrate formation and
centrifugation.
A Centrifuge (IEC CENTRA CL2) from Thermo ELECTRON COPORATION In.
with the maximum rpm of 8500 and the minimum of 3900 was used in the centrifuge
process. We used electrolyte with selected TDS as feed water to form gas hydrate in a
series of experiments. Two parts were obtained after gas hydrate formation, one is named
as dirty water in liquid phase and the other part is gas hydrate product in solid phase.
Separated gas hydrate from dirty water and put both parts into beakers. We crushed solid
gas hydrate into finer particles with spatula, so that we were able to put small particles
into 6 tubes separately, guaranteed each tube had equal amount of gas hydrate in order to
keep the balance of centrifuge machine. Put 6 tubes into centrifuge machine, set different
centrifuge times while keeping rpm in a specific value, then set different rpms and keep
centrifuge time as constant to study effects of time and rpm on desalination reduction
separately. After centrifugation, we obtained two components in the tubes, melting gas
hydrate in liquid phase and solid gas hydrate, poured melting part into one beaker and left
solid gas hydrate in the tube until it melted completely in the room temperature, then
poured it into the third beaker. Finally we got three parts: dirty water from first stage,
final gas hydrate product and second stage dirty water from centrifuge process. Tested
and recorded specific ion concentration of three parts separately, then calculated the mass
balance of the whole process to improve the degree of confidence. The saline component
of final hydrate product was measured and compared to supply water salinity to define
45 |
Virginia Tech | Master Thesis of Jing Niu
were cleaned with deionized water twice to keep the clean of equipment. We used
electrolyte in selected concentration as feed water in a consequence of experiment. Two
parts were obtained after gas hydrate formation: one part is called dirty water in liquid
phase and another part was gas hydrate product in solid phase. Separated gas hydrate
product from dirty water and treated solid gas hydrate in two ways: 1οΌCrushed solid
hydrate with spatula into small pieces; 2οΌGround solid hydrate with mortar and pestle
into finer particles. Then we discharged liquid and solid gas hydrate mixture into a funnel
covered by filter paper on the top. Started the vacuum pump and recorded filtration time
with stopwatch. After the filtration, gas hydrate was separated into two parts, one is final
product in solid phase which left on the top of filter paper, and another part is dirty water
went through filter paper in filtration. Measured and recorded ion concentration and
volume of three parts in order to improve the degree of confidence through mass balance
calculation. The saline concentration of final hydrate product was measured and
compared to supply water salinity to define the desalination %Reduction of two-stage
separation. The flow sheet of desalination with gas hydrate formation-filtration process is
presented in
Figure 3.3 and experimental set-up of filtration was shown in Figure 3.4.
3.3 Results and Discussion
Table 3.3 Data obtained in centrifuge, including the centrifuge time, rpm, single-
stage %Reduction, two-stage %Reduction and volume yield
47 |
Virginia Tech | Master Thesis of Jing Niu
At the beginning, we need to clarify the definition of desalination %Reduction in the
process with following equation:
! !!
π·ππ ππππππ‘ππ %ππππ’ππ‘πππ = ! !"#$%&βΓ100 % (3.10)
!!
in which c is the centrifuge feed water concentration; c is the final solid hydrate
1 product
product concentration.
Djurdjica Corak et al 6 obtained 87 to 67% NaCl removal of 3.5wt.% NaCl solution
under 3.6K and atmosphere pressure, while under 5.6K and atmosphere pressure from 89
to 73% NaCl was removed after involving centrifuge process.
We used 1.08wt.% NaCl solution as feed water and in order to guarantee the
consistence and reproducibility of centrifuge samples, consistent formation condition of
650psi and 1Β°C were controlled, each experiment was carried out twice. After
centrifugation, the water released from the small crystals was decanted off, and the
remaining crystals were allowed to dissociate.
Figure 3.5 Effect of centrifuge time on %Reduction, with initial pressure as 650
psi and temperature as 1Β°C; 1.08wt.% NaCl is feed water and rpm is fixed as 3000
49 |
Virginia Tech | Master Thesis of Jing Niu
Figure 3.5 and Figure 3.6 show results about effects of centrifuge time and rpm on
%Reduction; it is obviously that the %Reduction of NaCl increased with centrifugation
time and rotational speed (rpm). We obtained that the %reduction relates to centrifuge
time linearly by following equation:
%π
πππ’ππ‘πππ = 0.487Γ πΆπππ‘ππππ’ππ π‘πππ+2.763 (3.11)
and %Reduction relates linearly with centrifuge rpm by following equation:
%π
πππ’ππ‘πππ = 0.00664ΓπΆπππ‘ππππ’ππ π
ππ+41.05 (3.12)
The increase in efficiency with increasing rpm was due to the increase in the G-force
that removes the entrapped water. Note, however, that the hydrate crystals still contained
entrapped water, which may be attributed to two reasons. First, the entrapped water
removed by centrifugation was removed by decantation rather than filtration; therefore,
part of the residual water still remained with the crystals. Centrifugal filtration would
have given better results. Second, the particle size of the broken hydrate crystals was too
large. Crushing the crystals to finer sizes would have given better results. In Figure 3.7,
gas hydrate formation centrifuge process does enhance the desalination %reduction.
But the volume yield goes into the opposite direction as %reduction. We define
volume yield by following equation:
! (!"#$% !"#$% !"#$%&β)
ππππ’ππ πππππ = Γ100 % (3.13)
! (!"##$% !"#$%)
In Figure 3.8 and
Figure 3.9, when centrifuge time and rpm increase, the volume yield is brought down
sharply. This is explained that longer centrifuge time leads to gas hydrate exposed to
room temperature for longer period and more contacts with room temperature tube, thus
higher fractions of gas hydrate melts in the room temperature.
51 |
Virginia Tech | Master Thesis of Jing Niu
Dewatering by Filtration
We define the %Reduction of salt by following equation:
! !!
π·ππ ππππππ‘πππ % ππππ’πππ‘ππ = ! !"#$%&βΓ100 % (3.14)
!!
in which c is the filtration feed water concentration; c is the final solid hydrate
1 product
product concentration.
Figure 3.10 shows effect of filtration time on %Reduction of NaCl. CO hydrates
2
were made at 1Β°C and 650psi. As shown, the %Reduction increased with increasing
filtration time as anticipated. Also, the process is effective over a wide range of salinity,
which is an advantage over reverse osmosis. We found that in the first region from 120 to
300s, %Reduction enhances rapidly from 52.53 to 97.14%. In the second region from 300
to 420s, %Reduction stays almost the same and changes between 97.14 and 98.43%. In
the first region, longer filtration time results in more trapped in brine ions are released by
melting, however, in the second region, we suggest that an attraction force may exist
between salt ions and gas hydrate and begin to play the main role to obstruct removal of
Figure 3.10 Effect of filtration time on %Reduction, with initial pressure as 650psi
and temperature as 1Β°C; 1.08wt.% NaCl is feed water.
53 |
Virginia Tech | Master Thesis of Jing Niu
In industrial reverse osmosis desalination process, wash water composed of clean
water is used to create high driving force. The impact of wash water was shown in
Figure 3.12, adding wash water for 30 seconds leads to average %Reduction improved
from 75.86 to 93.77%. But no proficient improvement of %Reduction with longer
washing time was obtained. However, only 30s addition of wash water consumed 100
mL clean water, comparing to the small enhancement on the %Reduction, a more
efficient and economic way need to be explored.
Figure 3.13 shows the results obtained with feed water as MgCl . Excellent results
2
were obtained, which is showing over 99 %Reduction in the dissolved species. The effect
of MgCl concentrations is not obvious on the %Reduction, but an optimal concentration
2
of 0.5 wt.% is obtained.
Comparing to the low %Reduction before filtration of 7.13%, it shows positive effect
of filtration in desalination from Figure 3.14. We conclude that most part of salinity is
removed in the filtration stage and gas hydrate formation prepares excellent gas hydrate
feed to filtration in MgCl solution desalination.
2
Figure 3.14 Comparison of Mg2+ concentration effect on %Reduction before and
after filtration, with initial pressure as 650psi and temperature as 1Β°C.
56 |
Virginia Tech | Master Thesis of Jing Niu
In Figure 3.15, it indicates higher initial pressure doesnβt remove more salinity with
temperature as 1Β°C and 0.014wt.% CaCl as feed water. So we control pressure as 650
2
psi in CaCl desalination.
2
As shown in Figure 3.16 and Figure 3.17, the %Reductions achieved without
filtration were in the range of less than 20%. In the next series of experiments, the
hydrate crystals were crushed and filtered. The %Reduction was increased considerably.
After reducing the particle size further by grinding, %Reduction reached 90%, indicating
that the finer the particle size, the higher the extent of cleaning.
In Figure 3.18, the result obtained with artificial produced waters containing CaCl ,
2
MgCl , and NaCl are plotted. These results demonstrate that the process we have
2
developed could remove in excess of 90% of the dissolved solids. The highest
average %Reduction of 99.23% was obtained in MgCl , the lowest was obtained in CaCl .
2 2
We explained that Mg is easier to be sucked out of clathrate structure than other two ions
due to the weakest connecting force with gas hydrate.
Figure 3.19 Effect comparison of artificial produced water of different TDS on
desalination volume yield in filtration, with initial pressure as 650psi
and temperature as 1Β°C, and filtration time is 360s.
59 |
Virginia Tech | Master Thesis of Jing Niu
hydrate and CaCl hydrate may have same freezing point that are both higher than that of
2
MgCl hydrate in experiments, and thus MgCl hydrate melts in a faster speed. The
2 2
volume yield relates to the %Reduction because of faster melting speed causes more
brine ions to be released from clathrate structure.
In Ca-Mg mixture, Ca concentration was controlled as 0.015wt.% and Mg
concentrations was controlled as 0.13 wt.% separately with results shown in Figure 3.20
and Figure 3.21. It is obviously that addition of both MgCl and CaCl
2 2
decrease %Reduction of salt comparing to corresponding single ion solution. The
inhibiting effect of Ca is much stronger than that of Mg in mixture salinity removal. We
assumed that calcium ions may connect with magnesium ion or with themselves more
tight than magnesium ions connected with themselves.
Comparison between Centrifuge and Filtration
Comparison of %Reduction of salt between centrifuge and filtration is shown in
Figure 3.22, it is easily to find out that filtration methods produces much
higher %Reduction of 90 % than centrifugation does when 1.08 wt.% NaCl served as
Figure 3.22 Effect comparison of centrifuge and filtration on %Reduction, with initial
pressure as 650psi and temperature as 1Β°C, 1.08wt.% NaCl is feed water.
61 |
Virginia Tech | Master Thesis of Jing Niu
Process Flow Sheet
Based on the results presented above, we have developed a flow sheet for desalination
of produced water, as shown in Figure 3.24. High salinity producer water is fed to a
hydrate reactor using pressurized CO . The hydrate crystals will be fed to a screen to
2
remove the water that has not been incorporated into hydrate as screen underflow. This
water should have high concentrations of dissolved species such as Ca2+ and Mg2+ ions.
The hydrate crystals are recovered as screen overflow. These crystals are formed by
pure water but contain some entrapped produced water in between the fine-grained
crystals. In order to remove the entrapped produced water, the screen overflow is crushed,
ground, and then subsequently dewatered using a centrifuge (or a filter). The dewatered
CO hydrate is melted down to clean liquid water. The clean water obtained in this
2
manner should be free of dissolved species.
The filtrate from the dewatering centrifuge and the screen underflow are contacted
with CO to precipitate calcium and magnesium carbonate, which can be a byproduct that
2
can be marketed as building materials, pigments, fillers, etc. The residual water from the
CO mineralization reactor will be fed to a flotation cell, in which air bubbles remove
2
organic impurities. The underflow from the flotation column may be combined with the
Figure 3.24 Process flow sheet for produced water desalination
63 |
Virginia Tech | Master Thesis of Jing Niu
Table 3.8 Cost of reverse osmosis process for produced water desalination
(ΓakmakcΔ±, et al. 4)
results of our comparative analysis as shown below.
Tables 3.5 to 3.7 show the costs of desalinating seawater suing three different
processes, namely, reverse osmosis, distillation, and hydrate formation methods. The cost
for desalinating 1,000 gallon of seawater decreases from $4.84 for multi-stage flash
distillation, $2.98 for reverse osmosis, and $2.57 for hydrate formation. One can
understand the fact that distillation is the most expensive of the three. However, it was
surprising that the hydrate formation process was less costly than the reverse osmosis
process despite the fact that Bradshaw et al. 20 used three-step hydrate formation process.
The reason that they used the three-step process was because a single-step process was
unable to remove sufficient amounts of salts due to the entrapment problem we have
discussed in the present work.
We have shown in this report that the process of crushing and grinding followed by
dewatering can reduce the multi-step hydrate formation process to a single-step process.
Therefore, the process developed in the present work should substantially lower the cost
of desalination below that reported by Bradshaw et al. 20
As has already been noted, the reverse osmosis process has a serious problem in its
application to desalinating produced water, particularly the frac water containing very
high levels of TDS. For this reason, the cost of reverse osmosis increases sharply up to
$19.6 per 1,000 gallons at high salinity as shown in Table 3.8. The process developed in
the present work should be substantially lower than this figure. However, it is necessary
to conduct a series of pilot-scale tests to carryout reliable cost analysis. The authors
65 |
Virginia Tech | Master Thesis of Jing Niu
believe that the process of desalinating produced water as described in this report is
patentable.
3.4 Conclusions
The results obtained in this study showed that over 99% of dissolved NaCl and MgCl
2
can be removed from artificial saline water in a process involving a single-stage hydrate
formation step, followed by a single-step dewatering (centrifugation or filtration). The
results show that the %Reduction of NaCl increases with centrifugation time and
rotational speed (rpm). Centrifugal filtration would have given better results for two
reasons: first, the entrapped water removed by centrifugation was removed by
decantation; therefore, part of the residual water still remained with the crystals. Second,
the particle size of the broken hydrate crystals was too large. Crushing the crystals to
finer sizes would have given better results.
The values of %Reduction of NaCl in filtration were substantially higher than
obtained using the centrifugation. The % Reduction increased with increasing filtration
time as anticipated. Also, the process is effective over a wide range of salinity, which is
an advantage over reverse osmosis. Excellent results were obtained in MgCl , which is
2
showing over 99% reduction in the dissolved species. In Ca-Mg mixture, Ca has stronger
inhibiting effect than Mg on %Reduction.
In the next series of experiments, the hydrate crystals were crushed and filtered. The
%Reduction was increased considerably. After reducing the particle size further by
grinding, %Reduction reached 90%, indicating that the finer the particle size, the higher
the extent of cleaning. These results demonstrate that the process we have developed can
remove in excess of 90 % of the dissolved solids. The use of this new process should help
minimize the steps involved in cleaning produced water.
3.5 References
1 El-Dessouky, H. T., Ettouney, H. M. . Fundamentals of salt water desalination.
Elsevier Science: Amsterdam, Netherlands (2002).
2 ΓakmakcΔ±, M., Kayaalp, N., Koyuncu, I. . Desalination of produced water from
oil production fields by membrane processes. Desalination Vol. 222, 176-186
((2008)).
66 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.