PARAMETER EVALUATION AND MODELING OF A FINE COAL DEWATERING SCREEN-BOWL CENTRIFUGE

Transcription

1 PARAMETER EVALUATION AND MODELING OF A FINE COAL DEWATERING SCREEN-BOWL CENTRIFUGE Ian Michael Sherrell Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE In Mining and Minerals Engineering G.H. Luttrell, Chair R.H. Yoon G.T. Adel March 22, 2001 Blacksburg, Virginia Keywords: Coal, Dewatering, Reagents, Screen-Bowl Centrifuge, Disk Filter, Model

2 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.

3 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. III

4 TABLE OF CONTENTS TITLE PAGE...I ABSTRACT...II ACKNOWLEDGEMENTS...III TABLE OF CONTENTS... IV LIST OF FIGURES... VI LIST OF TABLES...VIII 1 INTRODUCTION BACKGROUND DEWATERING CAKE BEDS LITERATURE REVIEW OBJECTIVES ORGANIZATION EXPERIMENTAL COAL SAMPLES REAGENTS SAMPLE PREPARATION SCREEN-BOWL CENTRIFUGE CIRCUIT DESCRIPTION AND SAMPLING PROCESS TESTING DISK FILTER CIRCUIT DESCRIPTION AND SAMPLING PROCESS TESTING SAMPLE ANALYSIS MASS BALANCING SCREEN-BOWL CENTRIFUGE DISK FILTER COLUMN MODEL INTRODUCTION DESCRIPTION OF PROCESS MODEL FORMULATION POPULATION MODEL MODEL SET-UP IN-DEPTH MODEL SCROLL MOVEMENT IV

5 FLUID FLOW TERMINAL VELOCITY EXCESS FLOW CENTRIFUGAL CAKE MODEL ZEITSCH'S MODEL INPUT VARIABLES OUTPUT OR CALCULATION VARIABLES EQUATIONS OTHER INPUTS SIMULATION RESULTS AND DISCUSSION FLOTATION SCREEN-BOWL CENTRIFUGE D.M.C CYCLONE OVERFLOW DISK FILTER D.M.C CYCLONE OVERFLOW EQUIPMENT COMPARISON MODEL COMPARISON CONCLUSIONS RECOMMENDATIONS FOR FUTURE WORK REFERENCES APPENDIX A MASS BALANCED DATA FOR CENTRIFUGE APPENDIX B MASS BALANCED DATA FOR DISK FILTER APPENDIX C MASS BALANCED DATA FOR COLUMN APPENDIX D MODEL RUN VALUES (INITIAL SET RUN) VITA V

6 LIST OF FIGURES FIGURE 1 - CAPILLARIES WITHIN CAKE... 2 FIGURE 2 - INTERSECTION OF DIFFERENT SIZED PARTICLES... 5 FIGURE 3 - GRINDING CIRCUIT... 9 FIGURE 4 - CONVENTIONAL CELL FLOTATION UNIT FIGURE 5 - CONVENTIONAL CELL FLOTATION CIRCUIT FIGURE 6 - COLUMN FLOTATION UNIT FIGURE 7 - COLUMN FLOTATION CIRCUIT FIGURE 8 - SCREEN-BOWL CENTRIFUGE CIRCUIT FIGURE 9 - SAMPLING POSITIONS AROUND THE SCREEN-BOWL CENTRIFUGE FIGURE 10 - ENTIRE CIRCUIT FOR THE SCREEN-BOWL CENTRIFUGE FIGURE 11 - SCREEN-BOWL CENTRIFUGE PICTURE FIGURE 12 - SCREEN-BOWL CENTRIFUGE PICTURE FIGURE 13 - DISK FILTER CIRCUIT FIGURE 14 - SAMPLING POSITIONS AROUND THE DISK FILTER FIGURE 15 - ENTIRE CIRCUIT FOR THE DISK FILTER FIGURE 16 - PICTURE OF THE DISK FILTER FIGURE 17 - MASS BALANCING NODE FIGURE 18 - CENTRIFUGE NODES FIGURE 19 - DISK FILTER NODES FIGURE 20 - COLUMN NODES FIGURE 21 - SECTIONS WITHIN A SCREEN-BOWL CENTRIFUGE FIGURE 22 - TYPICAL FLOWSHEET FOR A SCREEN-BOWL CENTRIFUGE FIGURE 23 - POPULATION MODEL DIVISIONS OF A SCREEN-BOWL CENTRIFUGE FIGURE 24 - DIVISIONS WITHIN POPULATION MODEL FIGURE 25 - GENERIC INFLOWS AND OUTFLOWS OF A CELL FIGURE 26 - FLOWS INTO AND OUT OF A CELL FIGURE 27 - OVERALL FLOWS OF THE CENTRIFUGE FIGURE 28 - COLUMN BALANCING FIGURE 29 - FORCE DIAGRAM ON PARTICLE FIGURE 30 - PERCENT SOLIDS WITHIN A SCREEN-BOWL CENTRIFUGE FIGURE 31 - FEED RATE MODEL TEST FIGURE 32 - PERCENT SOLIDS MODEL TEST FIGURE 33 - CENTRIFUGE SPEED MODEL TEST FIGURE 34 - MODEL SET DISTRIBUTION (0) VI

7 FIGURE 35 - DISTRIBUTION 1 MODEL TEST FIGURE 36 - DISTRIBUTION 2 MODEL TEST FIGURE 37 - DISTRIBUTION 3 MODEL TEST FIGURE 38 - DISTRIBUTION 4 MODEL TEST FIGURE 39 - CENTRIFUGE DATA FROM TEST FIGURE 40 - CENTRIFUGE DATA FROM TEST FIGURE 41 - CENTRIFUGE DATA FROM TEST FIGURE 42 - CENTRIFUGE DATA FROM TEST FIGURE 43 - CENTRIFUGE DATA FROM TEST FIGURE 44 - CENTRIFUGE DATA FROM TEST FIGURE 45 - DISK FILTER DATA FROM TEST FIGURE 46 - DISK FILTER DATA FROM TEST FIGURE 47 - DISK FILTER DATA FROM TEST FIGURE 48 - DISK FILTER DATA FROM TEST FIGURE 49 - DISK FILTER DATA FROM TEST FIGURE 50 - MODEL TEST: EXPERIMENTAL 5/ FIGURE 51 - MODEL TEST: EXPERIMENTAL 6/ FIGURE 52 - MODEL TEST: EXPERIMENTAL 6/ VII

8 LIST OF TABLES TABLE 1 - COAL SAMPLES TABLE 2 - MODEL TESTING RESULTS FOR FEED RATE, PERCENT SOLIDS, AND CENTRIFUGE SPEED TABLE 3 - MODEL TESTING RESULTS FOR PARTICLE SIZE DISTRIBUTION.. 47 TABLE 4 - COLUMN FLOTATION RESULTS FROM THE DORCHESTER SEAM 53 TABLE 5 - COLUMN FLOTATION RESULTS FROM THE TAGGART SEAM TABLE 6 - RESULTS FROM CENTRIFUGE TEST 1 USING UPPER BANNER SAMPLE TABLE 7 - RESULTS FROM CENTRIFUGE TEST 2 USING UPPER BANNER SAMPLE TABLE 8 - RESULTS FROM CENTRIFUGE TEST 3 USING PITTSBURGH NO.8 SAMPLE TABLE 9 - RESULTS FROM CENTRIFUGE TEST 4 USING PITTSBURGH NO.8 SAMPLE TABLE 10 - RESULTS FROM CENTRIFUGE TEST 5 USING TAGGART SAMPLE TABLE 11 - RESULTS FROM CENTRIFUGE TEST 6 USING DORCHESTER SAMPLE TABLE 12 - RESULTS FROM DISK FILTER TEST 1 USING UPPER BANNER SAMPLE TABLE 13 - RESULTS FROM DISK FILTER TEST 2 USING UPPER BANNER SAMPLE TABLE 14 - RESULTS FROM DISK FILTER TEST 3 USING PITTSBURGH NO.8 SAMPLE TABLE 15 - RESULTS FROM DISK FILTER TEST 5 USING TAGGART SAMPLE65 TABLE 16 - RESULTS FROM DISK FILTER TEST 6 USING DORCHESTER SAMPLE TABLE 17 - MODEL TEST: EXPERIMENTAL VARIABLES TABLE 18 - MODEL TEST: EXPERIMENTAL RESULTS VIII

9 CHAPTER 1 INTRODUCTION 1.1 Background 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 BTU s). 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 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 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

10 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 ( θ ) 2γCos P = Equation 1.1 r where γ is the surface tension of the liquid, ϑ is the contact angle between the solids liquid and air, r is the radius of the capillary, and P is the pressure needed to sustain the height of liquid in the capillary such that P = hgρ Equation 1.2 where h is the height of the liquid, g is gravity, and ρ 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

11 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

12 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 hide 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 concluded that ions in the process water alter surface properties of the suspended particles and their flocculation characteristics (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

13 account for moisture trapped in the corners of bent triangles. 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

14 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 Zeitsch s 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 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

15 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

16 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

17 Hammer Mill Water Screen Product Hopper Conveyor Ball Mill Figure 3 - Grinding Circuit Sand Pump 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 18 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

18 column flotation. The sample that was not floated had 1 lb/ton of diesel and 100mg/ton of MIBC added, in each test, to simulate a flotation product. The conventional cell flotation occurred in a 4-cell laboratory Denver unit. was added at approximately 15% solids. To account for the high percent solids entering the unit, water was also added along with the feed. The product entered a launder, which fed into a 220-gallon sump. The product percent solids was approximately between 15% and 20%. A picture of the unit is shown in Figure 4. A diagram of the circuit is shown in Figure 5. Figure 4 - Conventional Cell Flotation Unit 10

19 Collector & Frother Flotation Cells Tailings Large Sump Figure 5 - Conventional Cell Flotation Circuit The column flotation occurred in an 8-inch unit. 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

20 Figure 6 - Column Flotation Unit Collector & Frother Large Sump Tailings Figure 7 - Column Flotation Circuit 12

21 A summary of the samples is given in Table 1. This shows what test each sample was used in, how they were prepared, and what units were used for dewatering in each test. Table 1 - Coal Samples Test # Sample Flotation Unit Frother Type Frother Dosage Collector Type Collector Dosage Reagent used Dewatering Equipment 1 U. Banner None MIBC 100 mg/ton Diesel 1 lb/ton U Cent./Disk 2 U. Banner None MIBC 100 mg/ton Diesel 1 lb/ton U Cent./Disk 3 Pitts. No.8 Conventional MIBC 100 mg/ton Diesel 1 lb/ton U Cent./Disk 4 Pitts. No.8 Conventional MIBC 100 mg/ton Diesel 1 lb/ton U Centrifuge 5 Taggart Column PPG 100 mg/ton Diesel 1 lb/ton U Cent./Disk 6 Dorchester Column PPG 100 mg/ton Diesel 0.25 lb/ton W Cent./Disk 2.4 Screen-Bowl Centrifuge The screen-bowl centrifuge was used in all tests Circuit Description and Sampling Process A diagram of the screen-bowl centrifuge circuit is shown in Figure 8. Centrifuge Effluent Sand Pump Screen Product Figure 8 - Screen-Bowl Centrifuge Circuit 13

22 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. 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 4.5 x 6 laboratory screen-bowl centrifuge operating at 431 G s. 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

23 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

24 Collector & Frother Tailings Collector & Frother Water Tailings Figure 10 - Entire Circuit for the Screen-Bowl Centrifuge Figure 11 - Screen-Bowl Centrifuge Picture 1 16

25 Figure 12 - Screen-Bowl Centrifuge Picture 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

26 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 Circuit Description and Sampling Process A diagram of the disk filter circuit is shown in Figure 13. 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

27 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

28 determined the time that the sample was taken. If the sample did not contain enough dry product it could not be analyzed for ash content. The entire circuit, including grinding, is shown in Figure 15. A picture of the disk filter is shown in Figure 16. Collector & Frother Tailings Collector & Frother Water Product To Thickener Figure 15 - Entire Circuit for the Disk Filter 20

29 Figure 16 - Picture of the Disk Filter 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

30 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 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 coal s 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

31 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

32 low relative error. Given this, the error calculated would be much bigger for a given data change with a smaller relative error. Data Error = Data original original Data altered Re lativeerror 2 Equation 2.3 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 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 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

33 other nodes with the feed and screen coming in and the screen, product, and effluent going out. The third node is not used since it gives no new information. All mass balanced data for the centrifuge is given in Appendix A Disk Filter The disk filter uses the same given data as the centrifuge. It also calculates the same rates that were balanced around the nodes. The nodes for the disk filter are shown in Figure 19. Product Effluent Figure 19 - Disk Filter Nodes There is only one node for the disk filter as compared to the centrifuge. The node includes the entire unit, which balances the feed coming in with the product and effluent going out. All mass balanced data for the disk filter is given in Appendix B Column The column uses the given data of solids rate, ash content, percent solids, and wash water rate. The rates calculated are the same as for the centrifuge and disk filter, which were solids rate, ash rate, slurry rate, and liquid rate. These were balanced around the nodes for the column. The nodes for the column are shown in Figure 20. Wash Water Product Tailings Figure 20 - Column Nodes 25

34 There is only one node for the column, which includes the entire unit. The node balances the feed coming in and wash water coming in with the product and tailings going out. All mass balanced data for the column is given in Appendix C. 26

35 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. 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 rpm s, accounting for 431 G s in the laboratory centrifuge used. The inner screw will be traveling at a slightly higher or lower speed depending on the left or righthandedness of the screw and the rotation of the bowl. This difference in speed allows the 27

36 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 tube Effluent Bowl Slurry level Beach Product 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

37 Comb. Screen Effluent Product Figure 22 - Typical Flowsheet for a Screen-Bowl Centrifuge 3.3 Model Formulation This model breaks the centrifuge into two main divisions. Division 1 includes the entire section within the bowl of the centrifuge. This accounts for all parts in the screen bowl centrifuge except for the screen section, which is included in Division 2. To get a feel for the entire process and to get information about the product, both divisions must be used. These different divisions are shown in Figure 23. y x Division 1 Division 2 Figure 23 - Population Model Divisions of a Screen-Bowl Centrifuge This model is run as a dynamic model but results can only be obtained when it reaches steady state due to assumptions made throughout the model set-up. 29

38 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 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) 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