Comparison of grinding media Cylpebs versus balls

Transcription

1 Minerals Engineering 17 (4) This article is also available online at: Comparison of grinding media Cylpebs versus balls F. Shi * Julius Kruttschnitt Mineral Research Centre, The University of Queensland, Isles Road, Indooroopilly, Brisbane 68, Australia Received 31 March 4; accepted 5 May 4 Abstract Cylpebs are slightly tapered cylindrical grinding media with a ratio of length to diameter of unity. The manufactures have made conflicting claims regarding the milling performance of Cylpebs in comparison with balls. One major point of interest is which one grinds finer at the same operating conditions. The difficulty in comparison is due to the shape difference. The two grinding media have different surface area, bulk density and contact mechanisms in grinding action. Comparative tests were conducted using the two types of grinding media in a laboratory Bond ball mill at various conditions of equality such as media mass, size distribution, surface area and input specific energy. The laboratory results indicate that at the same specific energy input level the Cylpebs produce a product with slightly less oversize due to their greater surface area, but essentially the same sizing at the fine end as that produced with the balls. The reason may be that the advantage of greater surface area is balanced by the line contact and area contact grinding actions with the Cylpebs. A new ball mill scale-up procedure [Man, Y.T., 1. Model-based procedure for scale-up of wet, overflow ball mills, Part I: outline of the methodology. Minerals Engineering 14 (), ] was employed to predict grinding performance of an industrial mill from the laboratory test results. The predicted full scale operation was compared with the plant survey data. Some problems in the original scale-up procedures were identified. The scale-up procedure was therefore modified to allow the predicted ball mill performance to match the observed one. The calibrated scale-up procedure was used to predict the Cylpebs performance in the full scale industrial mill using the laboratory tests results. Ó 4 Elsevier Ltd. All rights reserved. Keywords: Comminution; Grinding 1. Introduction Grinding media exert a significant influence on milling performance in terms of product size, energy consumption and grinding costs due to media wear. Grinding balls and rods are traditionally used as grinding media in the mineral industry. In the recent years other grinding media with different shapes have appeared in the market, such as Cylpebs developed by the Doering International, Powerpebs by the Donhad and Millpebs by the Wheelabrator Allevard Enterprise. * Tel.: ; fax: address: f.shi@uq.edu.au. For example, the Doering Minipebs were used in a / m tower mill with a charge of 183 tonnes operated at the Aberfoyle Hellyer lead/zinc mine in Tasmania, Australia, to replace the previously used high-chrome balls. It was reported that these allowed free flowing media to take advantage of a greater surface area and eliminate large pockets of locked charge that occurred using balls (World Mining Equipment, September 3). Discrepancies were found in the literature regarding grinding efficiency using Cylpebs. It was claimed by the manufacture (Doering International) that for a given charge volume, Cylpebs provide 25% more grinding media surface area for size reduction. The grinding performance of the Cylpebs should then be correspondingly /$ - see front matter Ó 4 Elsevier Ltd. All rights reserved. doi:.16/j.mineng

2 12 F. Shi / Minerals Engineering 17 (4) higher compared with the steel balls (Doering website: However, the Donhad conducted a grinding test using its own product Powerpebs (with length being 1.5 times diameter) to compare with balls and Cylpebs on the same ore type. The Donhad claimed that the product size distribution using Cylpebs at the identical energy input levels was significantly coarser than that produced with balls or the Powerpebs, indicating a 21% greater energy consumption using Cylpebs than that using balls or the Powerpebs to achieve the equivalent product grind size (Technical Bulletin Donhad website: These conflicting conclusions are confusing to mill operators regarding whether the Cylpebs would produce a coarser or a finer product in a grinding mill. Several questions may be asked regarding the manufacturesõ claims. The comparative tests conducted by Donhad were made at an equal specific energy input level, which implied an identical media mass when the grinding time and feed were kept constant. As the shapes of the three grinding media are different, in order to keep the same mass different media sizes must be used. Thus the results may only reflect the media size effect. Unfortunately the actual media sizes used for the comparison were not mentioned. While the Donhad result reflected the media size effect, DoeringÕs claim emphasized the surface area effect, but the increased surface area is at a cost of increased power and wear consumption. It is also noticed that the Donhad tests were undertaken in a stirred mill using single size media. In a real milling operation, the grinding media are seasoned. This is particularly true in ball milling. Therefore it is open to question whether the indications from the laboratory result using single media size are valid for the full-scale operation. In a research project conducted by the Julius Kruttschnitt Mineral Research Centre (JKMRC) in association with the WMC Fertilizers it was proposed to investigate grinding media options for fines minimisation in milling circuit at the Phosphate Hill Operation. Realising the complexity of the interrelationship of grinding media mass, size distribution and surface area due to their shape, detailed laboratory tests with Cylpebs and balls were carefully designed to distinguish the individual effects of the type of media. The experiment was carried out at the JKMRC Pilot Plant. One major challenge in this study was to predict the full-scale operation from the laboratory results. This has been achieved using a modified ball mill scale-up procedure published by Man (1999, 1). 2. Physical properties of Cylpebs Cylpebs are slightly tapered cylindrical grinding media with length equalling diameter, and all the edges being radiused (Fig. 1). Samples of Cylpebs used for the experiment were manufactured in Brazil, provided by Doering International. Doering Cylpebs are made from low-alloy chilled cast iron. The molten metal leaves the furnace at approximately C and is transferred to a continuous casting machine where the selected size Cylpebs are created; by changing the moulds the full range of cylindrical media can be manufactured via one simple process. The Cylpebs are demoulded while still red hot and placed in a cooling section for several hours to relieve internal stress. Solidification takes place in seconds and is formed from the external surface inward to the Fig. 1. Doering Cylpebs used for the comparative tests. Table 1 Comparative data of physical properties for Cylpebs and balls Cylpebs (c = 7.65kg/l) Dimension (mm) Mass (g) Surface area (cm 2 ) Specific surface (cm 2 /g) Bulk density (t/m 3 ) Balls (c = 7.85kg/l) Dimension (mm) Mass (g) Surface area (cm 2 ) Specific surface (cm 2 /g) / / / / / / / / / / / / / / / / / / / / / / Bulk density (t/m 3 )

3 F. Shi / Minerals Engineering 17 (4) centre of the media. It has been claimed that this manufacturing process contributes to the cost effectiveness of the Doering media, by being more efficient and requiring less energy than the conventional forging method. Because of their cylindrical geometry, Cylpebs have greater surface area and higher bulk density compared with balls of similar mass and size. Table 1 shows the comparative physical properties for Cylpebs and balls, given by Doering International. Cylpebs of equal diameter and length have 14.5% greater surface area than balls of the same mass, and 9% higher bulk density than steel balls, or 12% higher than cast balls. As a result, for a given charge volume, about 25% more grinding media surface area is available for size reduction when charged with Cylpebs, but the mill would also draw more power. 3. Laboratory tests As the objective of this study was to predict the grinding performance of Cylpebs in a full-scale ball mill, a model-based procedure for scale-up of ball mill was adopted, which involved laboratory tests using the Bond ball mill, and a set of model parameter scale-up criteria to simulate the steady state performance of full-scale mill circuit from the laboratory results (Man, 1999, 1). Accordingly all the laboratory tests were conducted in a standard Bond ball mill loaded with various grinding media to treat the same feed ore. Ore samples used for the laboratory tests were collected from the ball mill feed conveyor in the Phosphate Hill Beneficiation Plant during a grinding circuit survey. The ore is locally known as soft ore, with a Bond ball mill Work Index of 14.3kWh/t, and the breakage characteristic parameter A = 63.9 and b =.93 determined with a drop weight tester (Napier-Munn et al., 1996). The ore sample was prepared using a jaw crusher and a rotary divider to obtain the identical sub-samples for each test Media charge conditions Three media conditions were considered as the major factors affecting size reduction performance, viz. media mass media surface area, and media size distribution. For the comparative tests the standard Bond ball charge was set as the base case. The Cylpebs charge was controlled to match two of the three media conditions of the balls, with the third one being different, in order to distinguish one effect in each test. Table 2 (Panels A D) present the details of the media charge conditions. Charge 1 is the standard Bond ball mill charge condition which was used as the base case for the comparison. For some ball size fractions, however, no equivalent size of Cylpebs was available. Cylpebs of the nearest two sizes were therefore combined in various ratios (as indicated in the first columns) and their weighted average size was taken as the equivalent size to the balls. The combined ratio was calculated to match the two required ball conditions. Charge 2 (Table 2, Panel B) is the Cylpebs charge with similar mean size and mass in each size fraction (hence the total mass), but more than cm 2 surface area than that of balls due to their shape. Charge 3 (Table 2, Panel C) was to ensure that the Cylpebs have equal surface area and similar size, but significantly smaller mass. Table 2D presents Charge 4 in which fine Cylpebs were omitted and only two top size fractions were used to make equal mass and surface area to the balls Experimental program Two types of comparative tests were conducted: Locked-cycle (multi-stage) grinding tests, and Open circuit single-stage batch grinding tests. The locked-cycle (multi-stage) grinding test was developed by Man (1999) as one major component in the new ball mill scale-up procedure. The test procedure is the same as the well-known Bond ball mill grindability test (Bond, 1961). In the Bond ball milling test, the fresh feed to the test is crushed down to % passing 3.35mm. The mill grinds a constant ml of ore. After each grind the mill contents are screened to remove undersize which is replenished with an equal mass of new feed. The length of grinding time for each run is adjusted until the mass of the oversize fraction is consistently 2.5 times greater than the undersize. Under these conditions, the test approximates the steady-state performance of a closed circuit continuous mill with a circulating load of 2%. There are two major differences between the lockedcycle test and the standard Bond test. The first is the selection of circulating load. In the Bond ball mill test, 2% circulating load is used. In the locked-cycle tests, the circulating load is selected according to the feed condition. The plant used DSM screens in a closed-circuit ball milling operation. This resulted in a very low circulating load to minimise fines generation (Nott et al., 3). Therefore a 1.18 mm-aperture screen was used in the locked-cycle laboratory tests to mimic the DSM screen in the full-scale operation, and a % circulating load was selected to simulate the closed-circuit operation. The second difference is that at the end of the locked-cycle test, the size distribution of the final screen oversize is also obtained, while only the underscreen

4 1262 F. Shi / Minerals Engineering 17 (4) Table 2 Charge 1: standard Bond balls charge conditions (Panel A); Charge 2: Cylpebs with equal mass and equivalent size (Panel B); Charge 3: Cylpebs with equal surface area and equivalent size (Panel C); Charge 4: Fine Cylpebs truncated charge with equal mass and surface area (Panel D) Number of balls Mass (kg) Surface area (cm 2 ) Panel A Total Size fraction (mm) Number of Cylpebs Mean size (mm) Mass (kg) Surface area (cm 2 ) Panel B 9 / + 29 / / / / + 24 / / / Total Panel C 7 / + 22 / / / / + 19 / / / Total Panel D 18 / + 38 / / Total product is required in the Bond test. The locked-cycle test therefore provides the size distributions of the fresh feed, the final screen undersize and final screen oversize that is, the total mill charge is sized. The data of mill feed and product of the last run in the locked-cycle grinding test are used for fitting the breakage rate parameters of a ball mill model (Napier- Munn et al., 1996), which are then scaled-up to simulate the full scale operation. However, some problems associated with the locked-cycle test were identified (refer to the next section), and the conventional batch grinding in open circuit was then employed. The test conditions for both the locked-cycle and the single-stage open circuit tests are summarised in Table 3. In the batch grinding tests 338 mill revolutions was selected in order to be consistent with the locked-cycle Cylpebs test. This grinding time was kept constant across all the single-stage open circuit tests. Table 3 Comparative tests conditions Test type Charge type Test conditions Locked-cycle test Charge 1 Standard Bond balls charge; 1.18mm limiting screen; % circulating load Locked-cycle test Charge 2 Cylpebs of the same mass and similar size as balls, but greater surface area; 1.18mm limiting screen; % circulating load Single-stage test Charge 1 Standard Bond balls charge; batch grinding for 338 revolutions; open circuit Single-stage test Charge 2 Cylpebs of the same mass and similar size as balls, greater surface area; batch grinding for 338 revolutions; open circuit Single-stage test Charge 3 Cylpebs of the same surface area and similar size as balls, smaller mass; batch grinding for 338 revolutions; open circuit Single-stage test Charge 4 Cylpebs of the same mass and surface area; fine media omitted; batch grinding for 338 revolutions; open circuit

5 F. Shi / Minerals Engineering 17 (4) Results and analysis 4.1. Locked-cycle ball milling results New feed Charge 1, balls Charge 2, Cylpebs Fig. 2. Size distributions of the final-run mill content (screen undersize plus oversize) and new feed from the locked-cycle milling tests using balls and Cylpebs of the same mass and similar charge size distributions. Two locked-cycle milling tests using balls and cylpebs charges at the same mass and similar size distribution were conducted respectively. The results are summarised in Table 4. The standard Bond ball mill test at 2% circulating load treating the same ore sample is also included in Table 4 for comparison. The limiting screen aperture and the circulating load exert a significant impact on the Bond Work Indices. Comparing Test 1 with the standard Bond ball mill test (both tests using exactly the same balls charge and the same feed ore), the Bond WI increases from 14.3 kwh/t (for.3 mm screen, 2% circulating load) to 17.8 kwh/t (for 1.18 mm screen, % circulating load). This indicates that in the full-scale operation with DSM in the place of other classifier such as hydrocyclones that normally have around 2% circulating load in closed-circuit, the ball mill installed power requirement would be higher. The size distributions of the final-run mill content and the new feed are shown in Fig. 2. It appears that the Cylpebs grind coarser for particles below 1mm, and slightly finer above 1 mm. However, the effect of Cylpebs could not be justified from the locked-cycle tests. As shown in Table 4, Test 1 (balls) took 11 runs to get into steady-state, while Test 2 (Cylpebs) only took 8 runs. As a result, 4 mill revolutions were applied in the ball charge locked-cycle test, and only 338 revolutions for the Cylpebs charge. This indicated that grinding time in the two tests was significantly different, thus the energy input for the ball charge test was greater than for the cylpebs charge test. This is a problem in the design of the locked-cycle test. The test procedure requires to stop the test when the grindability reach equilibrium, which is normally assumed within less than 3% variation between the last two runs, or reverse its direction of increase or decrease. This is similar to the standard Bond test procedure. It has been found that in some locked-cycle tests, particularly with low circulating load, it takes long runs to stabilise and requires more mill revolutions (implying more energy consumption), sometimes never reaches equilibrium. Thus in the locked-cycle test, the energy input is not controlled. Although in the Bond test the mill revolution in the final run varies for different ores, it has been factored in the calculation of grindability (net grams of screen undersize produced per mill revolution), and it is the major variable to determine the Bond Work Index (Bond, 1961). However, in the locked-cycle test, the grindability is not incorporated in the procedure at all, and only the feed and final product sizes are used to fit the breakage rate parameters of the ball mill model. These parameters are then scaled-up for simulations. Since the locked-cycle test does not take the actual energy usage into account, the difference in the product size distributions as shown in Fig. 2 may not be the result of the inherent properties of grinding media. It may reflect the difference in the mill revolutions (4 versus 338, see Table 4), and hence the energy input, between the two locked-cycle tests. Clearly the locked-cycle test is not suitable for this study to determine the effect of grinding media on product size. Therefore the conventional single-stage batch grinding test was utilised. Table 4 Locked-cycle milling results Test Charge a Screen (lm) Recirculation (%) Runs Final revolution Grams produced per revolution b WI c (kwh/t) U/S P (lm) 1 Charge Charge Bond Charge a Refer to Table 2. b The net grams of screen undersize produced per mill revolution, which is defined as grindability in the Bond method. c Work Index.

6 1264 F. Shi / Minerals Engineering 17 (4) Single-stage batch grinding results Four batch dry grinding tests were undertaken using the Bond ball mill with various media charge conditions. In the four tests the mill run at the identical 338 revolutions and treated the same mass of the same ore. Results of the single-stage batch tests are given in Table 5 in terms of percent passing two size fractions (1.18mm representing the coarse end and.38 mm the fine end). The full size distributions of feed and product are given in Figs. 3, 5 and 6. In order to perform statistical analysis on the comparative test results, additional one single-stage batch test at Charge 1 condition (Bond balls) was conducted. A standard deviation of 5% relative was estimated on the base of percent retained size distribution. It is accepted that in such a well controlled open-circuit batch test running at sufficiently long grinding time, the experimental error is small. Fig. 3 demonstrates that at the same charge mass and similar charge size distribution, the products below.425 mm are essentially the same using the ordinary balls and the cylpebs, but the cylpebs produce a product with less oversize. The specific energy utilised in the two tests was assumed to be the same, as the mill power draw is basically determined by the charge mass (Bond, 1961; Morrell, 1996). Thus Fig. 3 is deemed to be a comparison at the same specific energy input level. This observation may reflect the fact that the cylpebs have a greater surface area than the ordinary balls at the same charge mass and similar size distribution (Table 2 Panel A versus B). It is generally agreed that more surface area would provide more opportunity for breakage. However, this general agreement is based on tests mostly using ball charges, that is, spheres, and may or may not be true of other shapes. The trend in the coarse end size fractions of the product shown in Fig. 3 seems support the surface area hypothesis, and agrees with the DoeringÕs claim. This trend, however, does not extend to the full size range as expected by the Doering. The reason why at the coarse end the Cylpebs generate finer product while at the fine end they do not was Charge 1, Bond balls Charge 2, Cylpebs Feed Fig. 3. Ball charge compared with Cylpebs of the same mass and the same media size distribution, larger surface area. sought from the contact mechanism in grinding action. Beside the point contact action similar to the balls, Cylpebs have other grinding actions resulted from line contact along the cylindrical section and area contact between the end faces on the ore particles. As illustrated in Fig. 4, the line contact and area contact increase the tendency for grinding to take place preferentially on the larger particles. Once the large particles are caught on the line or between the face areas, this prevents the smaller particles being broken further, which is similar Table 5 Single-stage batch grinding results Charge Conditions % Pass 1.18 mm Charge 1 Standard Bond balls Charge 2 Cylpebs, same mass and size, larger area Charge 3 Cylpebs, same area and size, less mass Charge 4 Cylpebs, same mass and area, coarser size (fine media omitted) % Pass.38 mm Fig. 4. Grinding action of linear contact of Cylpebs increases the tendency for grinding to take place preferentially on the larger particles.

7 F. Shi / Minerals Engineering 17 (4) to the rod mill practice. As a result, the advantage of a greater surface area of Cylpebs is cancelled out by the grinding action of line contact and area contact, and the product is almost identical at the fine end to that ground by balls at the same specific energy and same media size distribution. The benefit of the greater surface area of Cylpebs can only be realised at the coarse end and the difference is marginal. Fig. 5 shows the effect of energy input. Although having the same surface area and similar media size, the smaller charge mass of Cylpebs means a smaller energy input. The difference in product sizes above 1.7mm is not significant between Cylpebs and balls, showing that the effect of surface area may, again, dominate the grinding action in the coarse size fractions of particles. The energy input effect appears in sizes smaller than 1.7 mm, which results in almost parallel lines, with a smaller input energy producing a consistently coarser product. This is as expected. An interesting observation in Fig. 6 is that when the Cylpebs smaller than 28 mm were omitted from the charge, the size distribution curve of the ground product became steeper, i.e. producing less fines (.38 mm) and less oversize, compared with the same charge mass and the same surface area. This phenomenon may indicate that large grinding media offer a larger probability for impact breakage, but less abrasion breakage. Such a breakage mechanism would lead to rapid disappearance of the coarser particles in the feed and avoid excessive fines generation. A similar observation was reported previously in a separate research project at the JKTech using a laboratory ball mill at various media size distributions to grind silicon metal. Charge 1, Bond balls Charge 3, Cylpebs Feed Fig. 5. Ball charge compared with Cylpebs of the same media size and the same surface area, smaller mass. Charge 1, Bond balls Charge 4, Cylpebs Feed Fig. 6. Ball charge compared with Cylpebs of the same mass and the same surface area, fine Cylpebs omitted. 5. Scale-up from the laboratory milling tests 5.1. Calibration of the scale-up procedure A ball mill scale-up procedure (Man, 1) was employed to scale-up the breakage parameters determined from the laboratory milling tests to estimate the product using Cylpebs in full-scale industrial operation comprising a 4.7m diameter 9.5m length mill and DSM screens in a closed circuit. The procedure involves a number of criteria to scaleup the appearance function, breakage rate distribution and the discharge function. The variables taken into consideration in the scale-up procedure include ball size, mill diameter, feed size distribution, particle segregation and classification effect. The scale-up procedure has been coded by the author using the JKSimMet Software Development Kit (SDK) and has been tested with ManÕs original data (Man, 1999). In order to validate the scale-up procedure, the data of locked-cycle ball milling Test 1 (Table 4) was used to predict the full-scale ball mill operation. The predicted size distribution of ball mill discharge is compared with the survey data as shown in Fig. 7. This is probably the first time the independent data have been used to test the ball mill scale-up procedure. The discrepancy is obvious. Interestingly the predicted and measured % passing sizes are very similar. However, the rest of the size distribution is very different. This emphasises the need to present full size distribution for comparison, rather than a single size such as the P. Analysis of the scale-up procedures identified that the plant feed was much coarser than the database that Man

8 1266 F. Shi / Minerals Engineering 17 (4) Survey Predicted Fig. 7. Comparison of ball mill discharge size distributions between the measured and the predicted by the ball mill scale-up procedure (Man, 1). used to develop the scale-up criteria (Man, 1999). Thus the scale-up criteria for the feed size distribution effect are not valid in this case. The objective of this study was not to further develop the procedure from the first principles it aimed to use the procedure as a tool for scale-up. Therefore a simple step was taken by adjusting the criteria for the feed size effect and for the classifier factor to force the prediction to match the measured data. The calibrated model works reasonably well. Fig. 8 shows the ball mill product size distribution predicted by the calibrated scale-up procedure compared with the plant survey data Predictions of Cylpebs in the full-scale operation A similar approach was taken to calibrate the scaleup procedures from the single-stage batch grinding with balls to match the full-scale ball mill operation. The calibrated scale-up criteria were then applied to the singlestage Cylpebs grinding data. In the simulations, the charge was set at 23% volume as in the plant survey. As the bulk density of Cylpebs is higher, the model predicted that the Cylpebs mill would draw more power. Another comparison was made at the same mill power draw by reducing the Cylpebs charge volume. Two types of make up Cylpebs were used for the simulations (/ 85 85mm and /65 65mm). Comparisons of the predictions in mill product for a number of Cylpebs charge conditions are given in Figs The simulation results are also summarised in Table 6. The following trends are found in the simulations for full-scale operations from the laboratory single-stage grinding tests: With the calibrated scale-up parameters, the model predicts that the Cylpebs mill with a make-up size of /85 85 mm produces a product slightly finer around 1 mm size, compared with the ball mill charged with mm top size balls at the same load volume. Under such conditions, the Cylpebs mill will not improve the fines generated in the grinding circuit while drawing 156 kw more power. By reducing the charge volume to allow the Cylpebs mill to draw the same amount of power as the ball mill, the mill charged with /65 65 mm Cylpebs will grind a product with slightly less fines in both mill discharge and the DSM underscreen Survey Predicted Fig. 8. The calibrated model prediction of the full-scale ball mill product from the locked-cycle laboratory test Pred. mm balls 85mm Cylp. same volume Fig. 9. Predictions for full-scale mill product using mm balls versus /85 85mm Cylpebs at the identical charge volume from the singlestage batch grinding tests (the Cylpebs mill drawing more power).

9 F. Shi / Minerals Engineering 17 (4) Pred. mm balls 65mm Cylp. same power Pred. mm balls Fine Cylp. truncated, same power Fig.. Predictions for full-scale mill product using mm balls versus /65 65mm Cylpebs at the identical energy consumption from the single-stage batch grinding tests. Fig. 11. Predictions for full-scale mill product using mm balls versus /65 65mm Cylpebs at the identical energy consumption from the fine Cylpebs omitted test. Table 6 Comparison of the predicted milling performance Milling performance Ball mill survey Predicted mm ball mill 85mm Cylpebs 65mm Cylpebs Fine-Cylpebs omitted Charge volume (%) Make-up media (mm) Mill power draw (kw) Fresh feed F (mm) Fresh feed rate (tph) Combined mill feed F (mm) Mill throughput (tph) Mill discharge P (mm) Mill discharge (%.38mm) DSM U/S P (mm) DSM U/S (%.38mm) DSM O/S P (mm) Circulating load (%) A truncated Cylpebs distribution (no fine Cylpebs) will produce a product with about 5% less.38 mm fines at the same power draw. The simulations indicate that the media size distribution exerts a more significant influence on fines generation than the media type. 6. Conclusions Laboratory tests were conducted using a standard Bond ball mill to compare the milling performance of Cylpebs against balls. Effects of the three charge conditions mass, size distribution and surface area were investigated. Single-stage batch grinding tests indicated that the ground product using the Cylpebs of the same mass and the same size distribution as the balls contained slightly less oversize. This may be due to the greater surface area of the Cylpebs. This advantage, however, may be balanced by the line contact and area contact grinding action of the Cylpebs. As a result, Cylpebs produce a similar product at the fine end compared with the balls at identical charge mass, and hence at the identical specific energy input level. When compared at the same surface area and a similar size distribution, the Cylpebs produced a much coarser product, because their specific energy level was smaller than for the balls. When compared at the same mass and the same surface area, the fine-cylpebs truncated media produced a product with significantly less fines (.38 mm) and slightly less oversize. These results have emphasised that due to the difference in the shape, the two grinding media have different surface area, bulk density (implicitly the charge mass or volume when one of them is fixed) and contact mechanism in their grinding actions. All these factors need to be taken into account when comparing the

10 1268 F. Shi / Minerals Engineering 17 (4) milling performance using the two types of grinding media. In order to predict the Cylpebs performance in an industrial mill from the laboratory results, a ball mill scale-up procedure was employed. A problem associated with the locked-cycle test, one of the major components in the procedure, has been identified: the grinding time in the locked-cycle test is not controlled, and the energy usage in the test is not factored in the scale-up procedure. Thus variation in the comminution energy input may overcast the effect of different media on grinding performance. It was also found that the original ball mill scale-up procedure could not predict the Phosphate Hill ball mill operation correctly. One contributing factor is the feed size being much coarser than the database used by Man to develop the scale-up criteria. Modifications on the scale-up criteria were made and model using the calibrated scale-up criteria was able to predict the plant operation reasonably well. With the calibrated scale-up criteria and the breakage parameters determined from the single-stage grinding data, the model predicts that there is no significant difference in fines generation between using the mm balls and the /85 85mm Cylpebs at the same charge volume while the Cylpebs mill (4.7 m diameter 9.5 m length) drawing 156 kw more power. Using the / Cylpebs and reducing the charge volume to allow the mill to draw the same amount of power as the current ball mill operation resulted in the ground product contained about 2% less fines. The simulations indicate that the media size distribution exerts a more significant influence on fines generation than the media type. Acknowledgments This work was conducted as part of a research project sponsored by WMC Fertilisers Pty. Ltd. the financial support and the permission to publish this paper are greatly appreciated. The comminution circuit surveys were conducted at the Phosphate Hill Beneficiation Plant, Queensland, Australia. Dr. Mark Nott and the other plant staff in organising the surveys are acknowledged. The Cylpebs used for the comparative tests were kindly provided by Doering International. This research project was initiated by Dr. Walter Valery. Useful discussions were held with Dr. Rob Morrison. References Bond, F.C., Crushing and Grinding Calculations, Parts I and II, British Chemical Engineering, 6, No. 6 and 8. Man, Y.T., A Model-Based Scale-up Procedure for Wet, Overflow Ball Mills. Ph.D. Thesis. University of Queensland (JKMRC). Man, Y.T., 1. Model-based procedure for scale-up of wet, overflow ball mills, Part I: outline of the methodology. Minerals Engineering 14 (), Morrell, S., Power draw of wet tumbling mills and its relationship to charge dynamics Part 1: a continuum approach to mathematical modelling of mill power draw. Transaction of the Institution of Mining and Metallurgy (Section C: Mineral Processing and Extractive Metallurgy) 5, C42 C53. Napier-Munn, T.J., Morrell, S., Morrison, R.D., Kojovic, T., Mineral comminution circuits: their operation and optimisation. ISBN x. Julius Kruttschnitt Mineral ResearchCentre. Nott, M., Shi, F., Valery, W., 3. Fines minimization in the comminution of phosphate rock. In: Proceedings of Eighth Mill OperatorsÕ Conference, Townsville, Queensland AusIMM Publication Series No. 4/3.