of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  MINERALS & METALLURGICAL PROCESSING Vol. 28 No. 1 ã February 2011 1 Not another HPGR trade-off study! P. Amelunxen P. Eng., consulting metallurgist, Aminpro Chile, Santiago, Chile D. Meadows Global process engineering manager, FLSmidth Minerals, Bethlehem, PA Abstract Over the past several years, the combination of higher energy and steel costs and the recent commercial deploy-ment of high-pressure grinding rolls (HPGR) technology in hard rock mining have led many multinational mining companies to evaluate the suitability of HPGR technology for their particular comminution circuits. HPGR technology is perceived to require a higher capital investment while delivering lower long-term operating costs. This paper undertakes a critical evaluation of this perception. By evaluating the capital and life-of-mine operating costs of competing circuit designs for soft, medium and hard ores, the HPGR break-even point can be identified, at which the incremental capital cost of an HPGR circuit is equivalent to the net present value of the projected operating cost reductions. For ores with hardness parameters above the break-even point, HPGR circuits could offer economic benefits, while for softer ores, SAG mill/ball mill/pebble crushing (SABC) circuits are likely to be more economical and energy efficient. This evaluation considers both an SABC reference circuit and a traditional stage crushing configuration. Other influential factors are also discussed. Paper number MMP-10-018. Original manuscript submitted April 2010. Revised manuscript accepted for publication July 2010. Discussion of this peer-reviewed and approved paper is invited and must be submitted to SME Publications Dept. prior to August 31, 2011. Copyright 2011, Society for Mining, Metallurgy, and Exploration, Inc. Background & methodology While there is little doubt that in an open circuit situation, the energy efciency of high-pressure grind-ing rolls (HPGRs) is higher than that of a traditional cone crusher (Fuerstenau and De, 1995), grinding circuits are generally designed on the basis of cost efciency, not energy efciency. In the absence of political factors, reducing the carbon footprint of a grinding circuit is not, in the authors’ experience, a driving factor of the design, unless the cost footprint is also reduced. The operating and capital costs of a grinding circuit are specic to the ore characteristics and the local cost and availability of steel, labor, energy and other resources. For this reason, the trade-off studies performed to evaluate the costs of competing designs are also specic to the application, and caution should be used when attempting to apply the ndings of one trade-off study to a different ore body. Nevertheless, this study attempts to do just that: provide a generic comparison between three grinding circuit cost footprints for three ctional ore bodies. Analysis For this analysis, the ore characteristics have been chosen to represent a soft ore, a medium ore and a hard ore and the circuits have been sized using industry-accepted process models calibrated to the authors’ database of grinding circuit audits. The capital and operating costs have been estimated from databases of equipment costs and media consumption. In all cases, the assumptions used for this study are similar to those used by the authors for commercial project evaluation. They are deemed realistic for typical copper porphyry ores in the western hemisphere. Ore characteristics. While modern grinding circuit design is based on distributed hardness values, for simplicity this study assumes perfect homogeneity of ore hard-ness. The Bond work index values were selected to represent the 30th percentile, 60th percentile and 90th percentile of the cumulative distribution of Bond work index values from McKen et al.’s database (2006). They correspond to metric work index values of 11.8 kWh/t, 14.4 kWh/t, and 18.5 kWh/t, respectively (Fig. 1).The semi-autogenous grinding (SAG) mills were sized on the basis of the SAG power index (SPI) power model as described below. The average SPI of the ore was chosen to approximate the mean SPI for the given work index (after McKen et al., 2006), as per Fig. 2. The SPI values for the soft, medium, and hard ores are 40 minutes, 85 minutes and 160 minutes, approximately corresponding to A*b values of, respectively, 80, 42 and 25 (after Doll and Barratt, 2009).Note that the correlation between SPI and Bond work index, shown in Fig. 2, is not signicant to this work and no conclusions are derived thereof. It is only shown to illustrate that the work index values chosen for this study are reasonable approximations of the mean values one would expect of ores with hard, medium and soft competencies. Circuit sizing. The nominal plant capacity was assumed to be 100K tpd and a nominal grind of 180 microns ( P 80 ) was selected.The SAG mills were sized using the published form of the SPI calibration equation (Starkey and Dobby, 1996; Kosick and Bennett, 1999) and updated using audit data from industrial grinding circuits. The SAG mill feed size ( F  80 ) was estimated from industrial data and correlated to hardness. Mill power draws were calculated using the Morrell power draw model, updated with new grinding mill audits from the authors’ da-tabase (Morrell, 1993).The cone crushers were sized using the catalog rating for hard ore and a 15% derating was applied when fed directly Key words: HPGR technology, Comminution circuits, Mine design, Cost analysis, Hardrock mining  February 2011 ã Vol. 28 No. 1 MINERALS & METALLURGICAL PROCESSING 2 from a screen oversize chute. For the traditional stage crushing circuit, the cone crusher capacity requirements were based on open circuit secondary cone crushing with modular bin design and dedicated tertiary screening building. For the HPGR case, the secondary crushers were assumed to be in closed circuit conguration with 45 mm screens.HPGR performance was estimated after Daniel & Mor-rell (2004). The breakage and selection functions were taken from a medium ore and were not adjusted for the hard and the soft cases; rather, the specic press force was reduced for the medium and soft ore to account for changes in ore breakage characteristics. A specic press force of 3.0 N/mm 2  was used for the hard ore case, 2.5 N/mm 2  for the medium ore case and 2.0 N/mm 2  for the soft ore case. The specic energy draw of the units was adjusted accordingly.The largest HPGR units currently operating in hard rock base metal mining (2.4 by 1.65 m machines) were chosen for this study and the number of required units was determined on the basis of the ore’s specic throughput. Because the ore hardness, as dened by the Bond work index and SPI, plays a relatively small role in determining machine throughput (Klymowsky et al., 2002), the specic throughput is assumed to be constant for all cases. Its value was selected based on a simple power re-gression on specic press force, specic energy, and feed size to gap ratio (Fig. 3) and inated by 10% to account for improved performance caused by the recycle of nes in the full-scale plant. The resulting specic throughput value is 242 t-s/m 3 -hr.The ball mills were sized using the corrected Bond equation, (1)where kWh/t is the pinion power draw per metric ton of throughput, W  i  is the Bond work index (also metric),  P 80  and F  80  are the 80% passing size of the product and feed size distributions (microns), and CF  net   is the net correction factor and is determined from the authors’database of grinding circuit audits.Choosing the correct value of CF  net   is the subject of some controversy. Rowland (1973) described a list of efciency factors and their respective calculation methods that, when combined, should equal the value of the CF  net  . Rowland (2002) suggested caution when sizing ball mills that are fed by primary autogenous or semi-autogenous mills, mainly due to the dif-ference in the slope of the size distribution of the product of these mills (compared to, say, the product of rod mills). It is evident from the authors’ database that the Rowland efciency factors are not applicable, in their published form, to large-diameter ball mills that are fed by primary semiautogenous or autogenous mills. This can be seen in Fig. 4, which compares the predicted and observed CF  net   values from over 100 ball mill circuit audits (for which the ball mills were fed by autogenous or semi-autogenous mills). The correlation could, perhaps, be improved by using the work index of the primary mill product rather than the primary mill feed (as suggested by Rowland, 2002), although this becomes a somewhat difcult proposition for a greenelds project. Figure 1 — Bond work index for three ore types (after McKen et al., 2006). Figure 2  — Bond work index vs SPI™ (after McKen et. al, 2006).  MINERALS & METALLURGICAL PROCESSING Vol. 28 No. 1 ã February 2011 3 The requirement of a screen analysis correction has also been noted by Morrell (2009) for cases when the size distribution of the primary SAG/AG mills, or crushing circuit, is not parallel to that of the cyclone overow in log-log space. Given the apparent lack of a universally accepted method for scale-up, it is small wonder that most designers employ their own proprietary test meth-odologies, models and databases (the authors are no exception). This is an unfortunate circumstance, because the results of any trade-off study are particu-larly sensitive to the assumptions used for sizing the ball mill circuits, especially for the SAG/AG case in which the Rowland factors begin to break down. For this reason, it is worth discussing the authors’ method in more detail.There are two main factors that affect the slope of the size distribution of a SAG mill circuit product. They are the ore hardness and the percent crushed circulating load.For typical SAG mills, hard ores lead to longer SAG/AG mill residence times and a higher propor-tion of nes or nished product in the SAG/AG mill circuit product stream. This can be seen in Fig. 5, which shows the size distribution from three SAG mill circuit products with similar T  80  values but different ore hardness characteristics. Because the uncorrected Bond equation does not consider the nes portion of the size distribution curve, the correction must be made by adjusting the CF  net   value. As the ore gets harder, the amount of nes increases and the CF  net  for a ball mill circuit fed by SAG/AG mills drops. The effects of harder ore, however, are often miti-gated by the fact that hard ores also lead to higher circulating loads. As a larger percentage of the mill throughput is subject to cone crushing, the particle size distribution becomes coarser, with lower nes content. SAG mill circuits with a large amount of crushed circulating load tend to produce a lower proportion of nished material in the screen under-size than those with less pebble crushing. This can be observed in Fig. 6, which shows the SAG circuit product size distribution from the same grinding line for three different levels of circulating load (in the case of 0%, the pebble crusher was completely bypassed). The slope of the product size distribution also depends on the choice of comminution equipment. Figure 7 compares typical ball mill circuit feed size distributions for three circuit congurations, also with similar T  80  values. It is clear that for the same T  80 , the AG mill circuit, and to a lesser extent the SAG mill circuit, produce more nes than an HPGR-based circuit (almost double, in the case of the AG mill circuit). This must also be considered when selecting the value of the CF  net   term. Ball mill circuits that are fed by AG mill circuits tend to require less ball mill power, and hence have a lower CF  net term, than those fed by SAG mill circuits. Ball mill circuits fed by HPGR circuits tend to have higher CF  net   terms than those fed by SAG mill circuits. And ball mill circuits fed with the product of a cone crusher circuit tend to have the highest CF  net  values (the size distribution curve for a cone crusher product is omitted from Fig. 7 due to the difculties in nding a cone crusher circuit product with a comparable T  80  value). Figure 3  — Selection of 220 t-s/m 3 -hr specific throughput from the authors’ database of HPGR studies. Figure 4  — Comparison between measured and calculated Rowland efficiency factors for ball mills fed by primary SAG/AG mills. Figure 5  — SAG circuit product size distributions for soft, medium, and hard ores from three typical SAG circuits, from authors’ database (SPI values are 40, 100, and 200 minutes, respectively).  February 2011 ã Vol. 28 No. 1 MINERALS & METALLURGICAL PROCESSING 4 estimated at 132% of equipment for crushing plants and 98% of equipment for SAG mill and ball mill plants; these factors are a benchmark from recent concentrator evaluation projects and include costs associated with earthworks, foundation and concrete, structural steel, labor, and materials. Site prepara-tion and contingency are estimated at 12.9% of the equipment and installation costs. Indirect costs are estimated at 24.3% of the direct costs; they include capital spares, owners’ costs, distributable costs, a temporary construction camp, freight and insurance on capital equipment and EPCM services. Operating costs.  The estimated operating costs for each case are summarized in Table 4. The primary assumptions used to determine the value for each category are described below. Steel and wear parts. Table 5 shows prices for steel and wear materials, including balls and liners for grinding mills, Figure 6  — SAG circuit product size distributions from the same grinding line for a 0%, 19% and 32% circulating load of crushed pebbles. Note the different T  80 s (SPI values are all approximately 90 minutes). Figure 7 —   Size distributions of AG circuit product, SAG circuit product, and HPGR circuit product for three similar ores, from authors’ database (the metric Bond W  i   values are, respectively, 13.8, 13.5 and 14.9 kWh/t). Table 1  — CF  net   values for hard, medium and soft ores. OreCF net  SABCCF net HPGRCF net Cone crushingHard 0.7840.9661.030 Medium 0.8430.9661.030 Soft 0.8840.9661.030 For this study it has been assumed that the value of the CF  net  does not change with ore hardness for the SABC circuit; rather, pebble crushing capacity is added as the ore becomes harder, and the increased coarse particles generated by the crushing of pebbles offsets the additional nes caused by the harder ore. The SAG mill circulating loads assumed for this study are 10%, 18% and 25% for the soft, medium, and hard ore cases. The resulting CF  net   values are shown in Table 1. They are supported by the authors’ extensive industrial audit data.Ball mill runtime is assumed to be 93% in the case of the SABC circuit, 92% in the case of the HPGR (it is slightly lower due to interlocking with the tertiary crushing circuit via the pebble return belt), and 97.5% in the case of cone crushing (it is higher due to the use of a ne ore stockpile and the fact that it is not interlocked with the crushing circuit).Numerous studies have shown that the Bond work index is reduced by crushing in an HPGR circuit, due to micro-cracking of product particles. Examples include:1. Los Bronces study, 12%-17% reduction (Oestreicher and Spollen, 2006)2. Cerro Verde study, 10% reduction (Vanderbeek et al., 2006) 3. South American porphyry, 7.5% (unpublished)4. North American porphyry, 5.7% (unpublished)5. North American molybdenum project, 11% (unpublished)For this study, the authors assume 10% reduction in the Bond work index due to microcracking. The reduction is applied to mill sizing and power consumption calculations.Secondary or ancillary equipment sizing considerations:ã Screen sizes were estimated using the Allis Chalmers empirical model (Nichols, 1982).ã All ball mills are equipped with gearless drives.ã Surge bin capacities were selected based on current industry practice.ã HPGRs are equipped with wet dust collectors.ã Cyclones and cyclone feed pumps were sized based on 350% circulating load.The size of the major comminution equipment for each case is shown in Table 2. Mill sizes were chosen to minimize capital expenditure. For example, for the hard ore traditional crushing case, either three 8.5 x 12.4-m (28 x 40.8-ft) mills or four 7.3 x 13.9-m (24 x 45.5-ft) mills would have sufced, but four smaller mills are approximately $8 million cheaper than three larger mills. This somewhat counter-intuitive observation stems from the fact that the mill capital cost per unit length increases exponentially with the diameter of the mill. At larger mill sizes, the cost vs. diameter curve increases faster than the power vs. diameter curve, resulting in some situations for which the capital costs of a ball mill circuit can be reduced by installing a larger number of smaller mills. Capital costs. Capital cost estimates for grinding mills and gyratory crushers were provided by FLSmidth minerals and reect current pricing (Table 3). Other major equipment costs are based on recent quotes that have been adjusted to November 2009 US dollars using the producer price index for steel mill products (WPU1017), published by the United States Bureau of Labor Statistics (, 2010). Installation costs are
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks