Abstract Over the years, screen bowl centrifuges have been widely used for dewateringfine coal in coal preparation plants in the United States and elsewhere. Its popularity is attributed to its relatively low cost, its high capacity of providing low moisture content product and its relative ease of operation and maintenance. It is generally recognized in the engineering and scientific communities that screen bowl centrifuges provide some degree of particle size separation while dewatering fine coal in a common application. However, the extent of differential partitioning of coarse and fine particles achievable by a screen bowl centrifuge has not been systematically studied in the past. The present investigation was aimed at conducting a parametric study using a statistically designed experimental program to better understand and optimize the size classification performance of a screen bowl centrifuge. A continuously operating screen bowl centrifuge having a bowl diameter of 0.5 m was used for this study at the Illinois Coal Development Park. Three key operating parameters, Le., feed flow rate, feed solid content and pool depth, were varied to conduct a total of 17 experiments using a three-level factorial test matrix. Some of the best size separation performances achieved in this study may be described as having an imperfection value of 0.13 at an effective separation size (d^sub 50c^) of 38 um and an imperfection value of 0.27 at an effective separation size (d50c) of 2.8 um. Due to an effective separation of ultrafine high ash materials, the ash content of the screen bowl feed was reduced from 22.3% to a minimum of 8.84% with a combustible recovery of 84.1% and an ash rejection of 71.6%. A higher combustible recovery of 92.1% was achieved at a product ash content of 12.5% with a d^sub 50c^ of 2.8 um and imperfection of 0.27.

Key words: Coal, Size classification, Parametric study, Screen bowl centrifuge

Introduction

The current trend in many U.S. coal preparation plants treating steam (non-coking) coal is to classify its minus 45-[mu]m size fraction, which is typically concentrated with a majority of the ash- forming minerals. The nominally -1500-[mu]m particle size fraction of raw coal is deslimed, and only the nominally 150 x 45-[mu]m size fraction is cleaned using froth flotation. Unfortunately, the classifying cyclones commonly used for this size separation allow misplacements of up to 30% of the high-ash undersize materials to the cyclone underflow and subsequently to the flotation cells. The conventional froth flotation banks used in many preparation plants are not efficient in treating ultrafine high-ash particles, and, as a result, flotation product quality is diluted. The detrimental effects of fine-particle classification on coal cleaning performance have been reported in greater detail by many past investigators (Firth et al., 1995, 1997; Firth et al., 1998).

Several studies have been conducted in the past to improve size classification performance at the ultrafine particle size range using apex water-injection systems with classifying cyclone (Dahlstorm, 1954; Kelsall and Holmes, 1960; Firth et al., 1997; Patil and Rao, 1999; Honaker et al., 2001a; 2001b; Mohanty et al., 2002a). These studies indicate that undersize misplacement to the cyclone underflow can be minimized by the injection of a sufficient amount of elutriation water at the apex of the cyclone. However, the separation size is increased, and, as one of the aforementioned studies (Mohanty et al., 2002) indicated, misplacement of oversize materials to the cyclone overflow is also increased due to water injection at the cyclone apex. In addition, none of these studies indicated any significant improvement in the overall efficiency of size separation. Several other studies (Firth and O'Brian, 2003; Honaker at al., 2006) reported improved efficiency obtained from cyclone circuits using more than one stage of cyclone operation. Rong and Napier-Munn (2003) developed a new cyclone design to improve the efficiency of size classification. Mohanty (2003) reported significant improvement in size separation efficiency using a linear screen.

Figure 1 - Size-by-size analysis of the feed, product, centrate (main effluent) and filtrate (screen-drain) streams of a screen bowl centrifuge.

Table 1 - The particle size distribution and the corresponding ash and sulfur contents for each size fraction of the original dewatering feed slurry sample used in the parametric study.

A more recent study conducted by Honaker et al. (2006) emphasized the importance of achieving a high-efficiency size separation at particle size range of 25 to 50 [mu]m. As reported, the ash content of the 150- x 25-[mu]m size fraction of the finecoal circuit feed materials in many coal preparation plants in the United States is less than 10%. In other words, an efficient ultrafine particle size separation process could produce a high-quality (low-ash) product, obviating the need for any froth flotation cleaning. In addition, an efficient ultrafine size-separation process would enable enhanced gravity separators to be used in coal preparation plants for cleaning the especially high-sulfur, fine coal commonly found in the Midwestern and Eastern regions of the United States.

The screen bowl centrifuge has found widespread application in coal preparation plants in the United States and elsewhere for dewatering fine coal. It is generally recognized in the scientific and engineering communities that the screen bowl centrifuge provides some degree of particle size separation in its common de watering application. A simple examination of size analysis data for all four streams (feed, product, centrale and filtrate) of a screen bowl centrifuge indicates the significance of differences in the individual size distributions, as illustrated in Fig. 1. Some past studies (High, 1977; Gallagher et al., 1981) indicated that the majority of the particles finer than 40 [mu]m in the feed report to the centrale (main effluent) stream of a screen bowl centrifuge. However, only a few past studies (Meyers el al., 2002; Mohanty el al., 2002) reported the exlenl of differential partitioning of coarse and fine particles achieved by a screen bowl centrifuge. A study conducted by Meyers el al. (2002) investigated the performance of a full-scale screen bowl cenlrifuge operating in a coal preparation plant in which the filtrate (screen-drain) stream was constantly recirculated to the feed slream. As their dala indicated, particle size distribution of the centrale slream was much finer than those of the feed and product streams. However, there was not much difference between the particle size distribution of the feed and product streams.

None of the past investigations included a systematic study to investigate the size separation potential of a screen bowl centrifuge in a controlled environment. Therefore, the present investigation conducted a parametric study using a statistically designed experimental program to better understand and optimize the size classification performance of a screen bowl centrifuge. A continuously operating screen bowl centrifuge was used in this study, which was conducted in a pilot-scale research facility. The results of the size classification parametric sludy and the fine- coal cleaning due to the high-efficiency size separation are discussed in this publication.

Experimental

Sample. A bulk coal slurry sample was collected from the feed stream of raw coal classifying cyclones at a coal preparation plant operating in the Midwestern United Slates. The bulk sample was homogenized in an 8,000-L-capacity tank to be used as the feed material for the screen bowl centrifuge test program. The size-by- size analysis resulls of a representative sample, listed in Table 1, indicate the presence of 27.4% minus 45-[mu]m size fraction in the feed to the screen bowl cenlrifuge. The ash content of the minus 45- [mu]m size fraction of the sample is 61.4%, which is considered high but very typical of the bituminous non-coking coal found in the United Slates.

Experimental layout and procedure. A0.5-m-diameter screen bowl cenlrifuge was used in this investigation. The experimental layout included the screen bowl centrifuge, two centrifugal pumps, a 4,000- L capacity feed sump, an intermediate tank and a product collection tank, as shown in Fig. 2. The feed slurry was pumped to an intermediate lank to gravity feed the screen bowl centrifuge at a constant desired flow rate monitored by an online magnetic slurry- flow meter. The samples of feed slurry, dry coal product, centrale (main effluent) and filtrate (screendrain) were collected for each test. All product streams were mixed together in the product collection tank and recirculated to the feed tank to perform a series of tests using a limited amount of slurry sample. Because significant particle size degradation was expected from recirculating slurry through a screen bowl centrifuge over a long period, the feed slurry was introduced to the centrifuge only for the few minutes required for the collection of samples from the three product streams after adjusting all the operating parameters. During the remainder of the experimental period, the feed slurry was merely recirculated in the circuit to keep the solids in suspension, as shown in Fig. 2. A specific batch of sample was discarded after being used for 4 to 5 hours to avoid any significant particle size degradation due to continuous pumping. Figure 2 - A schematic layout of the experimental setup used in this investigation.

The rotational speed of a screen bowl centrifuge is intuitively an important parameter that may affect its size separation performance. High rotational speed, resulting in a high g-force inside the bowl, would force finer particles to the wall, thus providing a finer cut-size. However, excessively high rotational speed also causes more wear and tear. Therefore, the rotational speed of the screen bowl centrifuge was maintained constant at the manufacturer's recommended level of 1,400 rpm during the entire experimental program to produce a centrifugal field g-force of nearly 500 g. The pool depth was varied by suitably adjusting the height of the weir plate at the feed end of the screen bowl centrifuge. After running a series of exploratory tests, a total of 17 experiments were conducted using a f actorially designed experimental program to investigate the main and interaction effects of feed volumetric flow rate, feed solid content and pool depth on particle size classification performance of the screen bowl centrifuge. A completely randomized test matrix was developed using the Design Expert statistical software package by varying feed flow rate in the range of 75 to 190 L/min), feed solid content in the range of 9% to 27% and pool depth from O to 38.1 mm. The complete test matrix, showing the operating condition for each test is shown in Table 2.

Results and discussion

Mass yields to individual streams (product, centrale and filtrate) were calculated utilizing the well-known three-product formula (Weiss ,1985) and using two different assays (solid content and ash content) of each of the four process streams, including the feed. The samples collected from the feed and product streams for each test were subjected to a detailed size-by-size analysis utilizing wet sieving for plus 75-[mu]m size fraction and a laserparticle size analyzer for the minus 75-[mu] size fraction. The solid partitioning to each stream (product, centrale and filtrate) is shown in Fig. 3 for a specific test (listed as Test 12 in Tables 2 and 3). The mass splits to the individual process streams for this test were 81.83%, 16.14% and 2.03%, respectively. As clearly shown, the majority of the fine particles reported to the centrale, whereas the coarse particles reported to the product streams providing an effective separation size (d^sub 50c^) of 2.8 [mu]m. The separation efficiency was determined by a measure of the imperfection value (0.27) as follows

Table 2 - A list of operating parameter values utilized for the parametric study conducted to evaluate the size classification potential of a screen bowl centrifuge

Figure 3 - The general extent of coarse and fine particle partitioning achieved in a screen bowl centrifuge.

Tables-Size classification performance obtained from a screen bowl centrifuge.

The imperfection values and the effective separation size (d^sub 50c^) were determined for each of the 17 tests. As listed in Table 3, not all tests provided the excellent size separation efficiency that was achieved in Test 12. The imperfection values varied over a range of 0.13 to 1.10 and the d^sub 50c^ values varied between 2.8 and 38 [mu]m. Although, only one test was conducted at each operating condition, the repeatability and, thus, the reliability of the test results can be measured by comparing the results obtained from Tests 1, 5, 10, 12 and 15, all of which were conducted at the same operating condition (medium level of each operating parameter). The d^sub 50c^ values obtained from all five repeat tests were either exactly equal or were very close to one another. The imperfection values were also very close to one another except for Test 10.

The ash cleaning achievable due to high-efficiency ultrafine particle size separation is illustrated by the selected test data presented in Table 4. As presented in Table 1, the majority of the ash-forming minerals in the given coal is present in the minus 45- [mu]m size fraction. Screen bowl centrifuge Tests 1,4,8,12 and 15 produced significant reductions in ash content, from nearly 22% in the feed to a range of 8.84% to 12.7% in the product due to the highly efficient size classification performances as indicated by the imperfection values in the range of 0.13 to 0.30 and d^sub 50^ values of 2.8 to 38 [mu]m. The highest combustible recovery of 92.1% was obtained with an ash rejection of 53.4% in Test 12. Similarly, the highest ash rejection of 71.6% was obtained in Test 4 with a reasonably high combustible recovery value of 84.1%. Because the distribution of ash forming minerals in the given coal is quite similar to many other North American steam coals, the use of the screen bowl centrifuge could potentially be extended to ash cleaning and particle size separation from its conventional use for dewatering of fine coal in North America. However, it must be realized that the majority of the coal pyrites would report to the screen bowl dewatered product stream, and hence its sulfur content would be significantly higher than that of the screen bowl feed.

Parametric study. The experimental program was conducted using a three-level, three-factor factorial design to be able to study, not only the main effects, but also the interaction effects of the key process parameters on individual process response. Initial data analysis concentrated on developing empirical models for both responses, i.e., imperfection and d^sub 50c^ as functions of A (coded feed flow rate), B (coded feed solid content) and C (coded pool depth). Coding is a process of normalizing the actual values used for individual variables so that each variable is varied over the same range of -1 to +1 during the entire experimental program. The empirical models written using the coded variables are thus easier to interpret while investigating the relative effects of individual process parameters on a specific response.

Figure 4 - Statistical perturbation plots illustrating the main effects of the key process parameters on the size classification performance of a screen bowl centrifuge.

Table 4 - Ash cleaning performance obtained from the screen bowl centrifuge for a selected number of tests those produced excellent size separation.

Clearly, it is not only the main parameters, but also the parameter interactions, represented by the terms AB, AC, etc., that play significant roles in affecting the size classification performance of a screen bowl centrifuge. It may be noted that all three interactions, i.e., AB (feed flow rate and feed solid content), BC (feed solid content and pool depth), as well as AC (feed flow rate and pool depth), were significant at a level (a value) of 0.05 and, hence, included in the equation for the imperfection response. For the d^sub 50c^ response, however, only AB interaction was significant. All main factor effects were significant for both responses. In the absence of the parameter interaction terms, it would be easy to study the parametric effects by interpreting Eqs. (2) and (3), because these are written using the coded variables. However, because all of the parameter interactions play significant roles, graphical analyses were conducted to further investigate the parametric effects.

To illustrate the main effect of each process parameter on the two responses, the statistical perturbation plots for both responses are shown in Fig. 4. Steep slope or curvature in the plots shows the sensitivity of a response to a particular factor, whereas a relatively flat line is indicative of the relative insignificance of the factor being investigated. The plots, shown in Fig. 4, reveal that the classification efficiency (measured by imperfection) is more affected by feed solid content and pool depth than volumetric feed flow rate. However, the separation size (d^sub 50c^) appears to be equally affected by all three process parameters. High volumetric flow rate reduces the residence time for the coal slurry inside the screen bowl centrifuge and, hence, tends to provide a coarser separation size (increased d^sub 50c^). On the other hand, increased pool depth would tend to increase the available volume inside the bowl and, hence, would tend to increase the residence time resulting in a finer size-cut. Increase in feed solid content may change the particle settling environment from free settling type to hindered settling type inside the bowl. This may alter the settling rate of particles under a constant g-force and thus would tend to alter the cut-size of separation. To a large extend, the plots in Fig. 4 support these explanations,.

The interaction effects of the process parameters on imperfection are illustrated in Fig. 5. As shown in Fig. 5 (a), the increasing feed flow rate at low feed solid content tends to make the size classification process increasingly inefficient possibly due to a decreasing residence time. However, imperfection does not change that much at high feed solid content. As shown in Fig. 5 (b), the size separation efficiency deteriorates quite significantly by increasing the feed flow rate. At low pool depth, due to limited slurry holding capacity inside the screen bowl centrifuge, the effect of decreasing retention time as a function of increasing feed flow rate is quite significant. At high pool depth, the separation efficiency improves significantly at high feed flow rates. As shown in Fig. 5 (c), there is an increasing trend in imperfection as a function of feed solid content at high pool depth. This phenomenon may be caused by the fact that increasing feed solid content may create a hindered settling environment inside the bowl, which affects the settling kinetics of the solid particles and thus tends to result in a less efficient size classification. However, this trend switches at low pool depth, and the classification process becomes more efficient with increasing feed solid content. Figure 5 - Illustration of parameter interaction effects on classification efficiency (imperfection) obtained by a screen bowl centrifuge.

Figure 6 - Illustration of a parameter interaction effect on the effective separation size (d^sub 50^) obtained by a screen bowl centrifuge.

The only interaction effect found statistically significant for separation size (d^sub 50c^) response was AC, i.e., feed flow rate and pool depth interaction, as illustrated in Fig. 6. Coarser separation size results at high feed flow rate due to retention time limitation. The coarsest separation size resulting at the lowest pool depth and the highest feed flow rate is explained by the fact that the slurry residence time is the minimum at this operating condition, which may not allow sufficient time for finer particles to settle to the bowl wall.

Classification performance optimization. Performance optimization may have many different interpretations; however, one of the interpretations relates to the identification of an appropriate experimental region, i.e., a specific set of experimental conditions, to achieve a desired combination of process responses. Two such optimized experimental regions have been identified for the screen bowl classification process by overlaying the individual response surface contours generated for both d^sub 50c^ and imperfection, as shown in Fig. 7 (a) and (b). The dark area in Fig. 7 (a) defines a set of experimental conditions that would provide a d^sub 50c^ value in the range of 15 and 25 [mu]m at a minimum imperfection value of 0.20. As indicated in the plot, a feed solid content in the range of 20% to 27%, a very shallow pool depth of 0 to 5 mm and a feed flow rate (constant parameter) of 132.5 L/m would result in the aforementioned excellent size separation performance. The optimized region shown in Fig. 7 (b) would provide a precise size-cut (d^sub 50c^) at 13 [mu]m at the same high efficiency level defined by a maximum imperfection of 0.2.

Conclusions

The extent of particle size classification achievable from the screen bowl centrifuge and its impact on fine-coal cleaning was successfully evaluated in this study using a 0.5-m-diameter screen bowl centrifuge. The important findings of this study are summarized as follows:

Figure 7 - Optimization of process parameters to achieve desired size of separation at high efficiency from a screen bowl centrifuge.

* High classification efficiencies of close to 0.2 (imperfection value) at ultrafine particle sizes of separation (^50) in the range of 3 to 38 [mu]m were achieved utilizing specific test conditions.

* Due to an effective removal of ultrafine particles (indicated by an imperfection value of 0.25 and d^sub 50c^ of 20 [mu]m) the ash content of the screen bowl feed was reduced from 22.3% to 8.84% with a combustible recovery of 84.1% and ash rejection of 71.6%. Higher combustible recovery of 92.1% was achieved at a product ash content of 12.5% with a d^sub 50c^ of 2.8 [mu]m and imperfection of 0.27.

* A factorial-design was conducted varying three key process parameters, i.e., feed flow rate, feed solid content and pool depth. The empirical models developed for imperfection and separation size indicated the high significance of parameter interaction effects on the size classification performance achievable from a screen bowl centrifuge.

Based on the above findings, the use of the screen bowl centrifuge could potentially be extended to ash cleaning and particle size separation from its conventional use for dewatering of fine coal. However, it must be realized that the majority of the coal pyrites would report to the screen bowl dewatered product stream, and hence its sulfur content would be significantly higher than that of the screen bowl feed.

Acknowledgments

The authors sincerely acknowledge the funds provided by the Illinois Clean Coal Review Board for a part of this investigation. In addition, the authors express their special gratitude to Mr. Ken Robinette and Mr. Ron Jahnings of the Decanter Machine Inc., USA, for their technical guidance and equipment support, without which this test program would not have been possible.

Paper number MMP-07-012. Original manuscript submitted March 2007. Revised manuscript accepted for publication July 2007. Discussion of this peer-reviewed and approved paper is invited and must be submitted to SME Publications Dept. prior to Nov. 30, 2008. Copyright 2008, Society for Mining, Metallurgy, and Exploration, Inc.

References

Dahlstorm, D.A., 1954, "Fundamentals and applications of the liquid cyclone," Chemical Engineering Progress Symposium Series, American Institute of Chemical Engineers, Vol. 15, pp. 41-45.

Firth, B., Edward, D, Clarkson, CJ. and O'Brien, M., 1995. "The impact of fine classification on coal preparation performance," Proceedings 7th Australian Coal Preparation Conference, Paper E2, Australian Coal Preparation Society, pp. 250-276.

Firth, B., Edward, D, Clarkson, C. J. and O'Brien, M., 1997, "Role of fine-coal classification efficiency on coal preparation plant performance, " Minerals & Metallurgical Processing, February, pp. 1-6.

Firth, B., Hu, S., O'Brien, M., Clarkson, CJ., and Edward, D., 1998, "Some aspects of hydraulic size classification," Proceedings XIII International Coal Prep. Congress, A.C. Partridge and LR. Partridge, eds., Australian Coal Prep. Society, Vol. 1, pp. 236- 244.

Firth, B., and O'Brien, M., 2003, "Hydrocyclone circuits," Coal Preparation, Vol. 23, pp. 167-183.

Gallagher, E.L., Post, J., Swanson, A., and Armstrong, L., 1981, "Dewatering of fine coal by screen bowl centrifuges, " Proceedings of the 1st Australian Coal Preparation Conference, A.R. Swanson, ed., October, pp. 134-154.

High, R.E., 1977, "Decanter centrifuges in Australian coal industry," Presentation notes, Australian Coal Preparation Society, Queensland Branch Technical Meeting, September.

Honaker, R.Q., Ozsever, A.V., Singh, N., and Parekh, B.K., 2001a, "Improved classification efficiency using the Krebs Cyclowash System," Proceedings 18th International Coal Preparation Conference, Lexington, Kentucky, pp. 101-111.

Honaker, R.Q., Ozsever, A.V., Singh, N., and Parekh, B.K., 2001b, "Apex water injection for improved hydrocyclone classification efficiency," Minerals Engineering, Vol. 14, No. 11, pp. 1445-1457.

Honaker, R.Q., Boaten, F., Luttrell, G.H., Moorhead, R., and Coker, T., 2006, "Ultrafine coal classification using 6-inch diameter gMax cyclone circuits," Coal Prep, Lexington, Kentucky, 2006, pp. 21-36.

Kelsall, D.F, and Holmes, J.A., 1960, "Improvement in classification efficiency in hydraulic cyclones by water injection, " Proceedings, 5th Mineral Processing Congress, lnst. of Mining and Metallurgy Paper 9, pp. 159-170.

Meyers, A.D., Wex, T., and Leach, K.R., 2002, "Solids partitioning in screen bowl centrifuges, Proceedings of the 9th Australian Coal Preparation Conference, B. Firth, ed., October, pp. 209-221.

Mohanty, M.K., 2003, "Fine coal screening performance enhancement using the Pansep screen," International Journal of Mineral Processing, Vol.69, pp. 205-220.

Mohanty, M.K., Palit, A., and Dube, B., 2002a, "A comparative evaluation of new fine particle size separation technologies," Minerals Engineering, Vol. 15, pp. 727-736.

Mohanty, M.K., Palit, A., and Dube, B., 2002b, "Enhanced fine coal dewatering performance from the screen bowl centrifuge," Presentation at the Solid-Liquid Separation-02, Falmouth, United Kingdom, June 18-20.

Patil, D.D., and Rao, T.C., 1999, "Classification evaluation of water-injection hydrocyclone," Minerals Engineering, Vol. 12, pp. 1427-1432.

Rong, R., and Napier-Munn, TJ., 2003, Development of a more efficient classifying cyclone," Coal Preparation, Vol. 23, pp. 149- 165.

Weiss, N., 1985, "Metallurgical accounting and mill reports," SME Mineral Processing Handbook, Vol. 2, pp. 32.15.

M.K. Mohanty, B. Zhang, N. Khanna, A. Palit and B. Dube

Associate professor and graduate students, respectively, Department of Mining and Mineral

Resources Engineering, Southern Illinois University at Carbondale, Carbondale, Illinois.

Copyright Society for Mining, Metallurgy, and Exploration, Inc. May 2008

(c) 2008 Minerals & Metallurgical Processing. Provided by ProQuest Information and Learning. All rights Reserved.

Source: Minerals & Metallurgical Processing