A Pilot-Scale Examination of a Novel High Pressure Grinding Roll / Stirred Mill Comminution Circuit for Hard-Rock Mining Applications

Transcription

1 A Pilot-Scale Examination of a Novel High Pressure Grinding Roll / Stirred Mill Comminution Circuit for Hard-Rock Mining Applications by Jeffrey Adam Drozdiak B.A.Sc., The University of British Columbia, 2006 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in The Faculty of Graduate Studies (Mining Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2011 Jeffrey Adam Drozdiak, 2011

2 Abstract The mining industry will be faced with new challenges as the need to develop lower grade ore deposits expands to meet the rising demand for raw resources. Low-grade deposits require a substantially increased tonnage to achieve adequate metal production and have caused the consumption of energy in mining practices such as comminution to rise dramatically. If improvements could be made in the processes employed for metal extraction, the mining industry could remain sustainable for future generations. This research focused on the development of a novel comminution circuit design to addresses these issues. The circuit design incorporated two, known energy efficient technologies, the High Pressure Grinding Roll (HPGR) and the horizontal high-speed stirred mill, and examined the technical feasibility of a circuit operating without the need for a tumbling mill. The main objectives of this research were to setup pilot-scale research equipment and develop the design criteria necessary to operate an HPGR / stirred mill circuit. Testing consisted of using a copper-nickel sulphide ore from Teck Limited s Mesaba deposit to evaluate a circuit comprised of two stages of HPGR comminution followed by stirred mill grinding. To evaluate the potential energy benefits of this novel circuit arrangement, energy consumption related to comminution was calculated for the circuit using power draw readings off the main motor and the throughput recorded during testing. To provide a basis for comparison, the energy requirements for two conventional circuits, a cone crusher / ball mill and an HPGR / ball mill, were determined through HPGR pilot-scale testing, Bond grindability testing and JK SimMet flowsheet simulation. Results from this research showed that operating the first-stage HPGR in open circuit and the second stage in closed circuit with a 710µm screen, resulted in a circuit energy requirement of 14.85kWh/t, a reduction of 9.2 and 16.7% over the HPGR / ball mill and cone crusher / ball mill circuits, respectively. To assist in future HPGR / stirred mill studies, a refined testing procedure was developed with a reduced sample commitment and the ability to perform an energy comparison with a Semi-Autogenous Grinding (SAG) mill / ball mill circuit. ii

3 Preface The work presented in the following thesis was performed solely by the author. Some of the information provided in Chapter 4 was published in the following paper: Drozdiak, J., Nadolski, S., Bamber, A., Klein, B., & Wilson, S. (2010). A Comparison of the energy requirements of an HPGR / stirred mill circuit and conventional grinding circuits for the comminution of mesaba ore. 42 nd Annual Meeting of the Canadian Mineral Processors, Ottawa, ON, Canada. The co-authors made contributions regarding the structure and layout of the paper and all experimental testing was carried out by the principle author. My research committee, consisting of Dr. Bern Klein, Dr. Andrew Bamber, Dr. Marek Pawlik, Josh Rubenstein, and Mike Larson, provided input into design of the test program and provided advice on the research project. Stefan Nadolski, of Koeppern Machinery Australia, provided assistance during HPGR pilot-scale testwork and provided a summary for HPGR operating data. iii

4 Table of Contents Abstract... ii Preface... iii Table of Contents... iv List of Tables... vi List of Figures... viii Acknowledgements... xii 1 Introduction Literature Review High Pressure Grinding Rolls Background Technology Overview HPGR Operating Parameters Energy Efficient Comminution HPGR Flowsheet Considerations Advantages and Disadvantages Stirred Media Mills Background Vertical Stirred Mill Technology Horizontal Stirred Mill Technology Horizontal Stirred Mill Operating Parameters Energy Efficiency for Stirred Media Mills Horizontal Stirred Mill Flowsheet Options Process Benefits of Horizontal Stirred Mills HPGR / Stirred Mill Circuit Literature Summary Experimental Procedure Definition of Comminution Circuits Cone Crusher / Ball Mill Circuit HPGR / Ball Mill Circuit HPGR / Stirred Mill Circuit Sample Description Equipment High Pressure Grinding Roll Horizontal Stirred Mill Vibrating Screen Bond Test Ball Mill iv

5 4 Testing and Simulation Results Cone Crusher / Ball Mill Circuit Results Flowsheet Simulation Specific Energy Calculations HPGR / Ball Mill Circuit HPGR Pilot-Scale Testing Flowsheet Simulation Specific Energy Calculations HPGR / Stirred Mill Circuit The Stirred Mill Circuit The HPGR Circuit Circuit Energy Summary Discussion of Results Assessing Operating Parameters for Pilot-Scale Testing HPGR Operating Parameters Stirred Mill Operating Parameters Comparison of Comminution Circuits Preliminary HPGR / Stirred Mill Circuit Flowsheet Refined Procedure for Future Testing Conclusions and Recommendations References Appendix A JK SimMet Data Appendix B Bond Work Index Data Appendix C HPGR Data Appendix D Stirred Mill Data v

6 List of Tables Table 2-1 Summary of Energy Consumption for Comminution Table 2-2 Summary of Grinding Media Wear Rates Table 2-3 Normalized Effect of Decreasing Ball Size Table 2-4 Summary of Ball Mill Size Over the Years Table 2-5 Summary of Power Density for Grinding Mills Table 3-1 HPGR Machine Specifics Table 4-1 Equipment Selection for Cone Crusher / Ball Mill Circuit Table 4-2 Bond Work Indices for Cone Crusher / Ball Mill Circuit Table 4-3 Feed Conditions for Pressing Force Tests Table 4-4 Results for Cycle Four of Closed Circuit Testing Table 4-5 Equipment Selection for HPGR / Ball Mill Circuit Table 4-6 Summary Bond Ball Mill Work Indices for Cone Crusher and HPGR Product Table 4-7 Test Conditions for the 355µm Signature Plot Table 4-8 Test Conditions for the 710µm Signature Plot Table 4-9 Summary of Mill Operating Conditions for 1.2mm Testing Table 4-10 Revised Test Conditions for 710µm Signature Plots Table 4-11 Summary of HPGR Results for First Stage Open and Closed Circuit Testing Table 4-12 Comparison of Wet and Dry Screening Table 4-13 Summary of HPGR / Stirred Mill Energy Requirements Table 5-1 Statistics Summary of Circuit Energy Values Table C-1 HPGR Pilot-Scale Test Key Table C-2 HPGR Operating Data Phase One Table C-3 HPGR Operating Data Phase One (continued) Table C-4 HPGR Operating Data Phase One (continued) Table C-5 HPGR Operating Data Phase One (continued) Table C-6 HPGR Operating Data Phase Two Table C-7 HPGR Operating Data Phase Two (continued) Table C-8 HPGR Operating Data Phase Two (continued) Table C-9 HPGR Operating Data Phase Two (continued) Table C-10 HPGR Feed Size Distributions Phase One Table C-11 HPGR Feed Size Distributions Phase One (continued) vi

7 Table C-12 T1A01 Product Size Distributions Table C-13 T1A02 Product Size Distributions Table C-14 T1A03 Product Size Distributions Table C-15 T1A04 Product Size Distributions Table C-16 T1A05 Product Size Distributions Table C-17 T1A06 Product Size Distributions Table C-18 T1A07 Product Size Distributions Table C-19 T1A08 Product Size Distributions Table C-20 T1A09 Product Size Distributions Table C-21 T1A10 Product Size Distributions Table C-22 T1A11 Product Size Distributions Table C-23 HPGR Feed Size Distributions Phase Two Table C-24 HPGR Feed Size Distributions Phase Two (continued) Table C-25 HPGR Feed Size Distributions Phase Two (continued) Table C-26 T2A01 Product Size Distributions Table C-27 T2B01 Product Size Distributions Table C-28 T2B02 Product Size Distributions Table C-29 T2B03 Product Size Distributions Table C-30 T2B04 Product Size Distributions Table C-31 T2B05 Product Size Distributions vii

8 List of Figures Figure 2-1 Diagram of HPGR Comminution... 6 Figure 2-2 Flowsheet for Cerro Verde Figure 2-3 Flowsheet Comparison for Peňasquito Figure 2-4 Standard Photographic Evidence of Micro-cracking Figure 2-5 Example of Gravity-Induced Vertical Stirred Mill Technology (Vertimill ) Figure 2-6 Examples of Fluidized Vertical Stirred Mill Technologies Figure 2-7 IsaMill TM Layout Figure 2-8 IsaMill TM Grinding Mechanism Figure 2-9 Flowsheet for Mount Isa Mines Figure 2-10 Flowsheet for Regrind Circuit at Kumtor Mine Figure 2-11 Original Flowsheet for McArthur River Figure 2-12 Modified McArthur River Flowsheet with Coarse IsaMill TM Grinding Figure 2-13 HPGR Flowsheet for Fines Production in the Cement Industry Figure 2-14 HPGR Flowsheet the Sukhoy Gold Plant Figure 2-15 Example of an HPGR / Stirred Mill Circuit Figure 2-16 A Proposed HPGR / IsaMill TM Circuit Figure 2-17 Anglo Platinum s HPGR Test Circuit Figure 2-18 Anglo Platinum s HPGR / Stirred Mill Testing Flowsheets Figure 3-1 Cone Crusher / Ball Mill Flowsheet Figure 3-2 HPGR / Ball Mill Flowsheet Figure 3-3 HPGR / Stirred Mill Flowsheet (Open Circuit) Figure 3-4 HPGR / Stirred Mill Flowsheet (Closed Circuit) Figure 3-5 Geographic Location of Mesaba Figure 3-6 Mesaba Feed Size Distribution Figure 3-7 Pilot-Scale HPGR Installation Figure 3-8 M20 Stirred Mill Installation Figure 3-9 M20 Stirred Mill with Mixing Tanks Figure 3-10 ZS40 SWECO Vibrating Screen Figure 3-11 Bond Test Ball Mill Figure 4-1 Cone Crusher / Ball Mill JK SimMet Flowsheet Figure 4-2 Comparison of Specific Pressing Force and Product Size viii

9 Figure 4-3 Comparison of Specific Pressing Force and Specific Throughput Figure 4-4 Comparison of Specific Pressing Force and Specific Energy Consumption Figure 4-5 Comparison of Moisture Content and Product Size Figure 4-6 Comparison of Moisture Content and Specific Throughput Figure 4-7 Comparison of Moisture Content and Specific Energy Consumption Figure 4-8 Comparison of Roller Speed and Product Size Figure 4-9 Comparison of Roller Speed and Specific Throughput Figure 4-10 Comparison of Roller Speed and Specific Energy Consumption Figure 4-11 Product Size for Closed Circuit Testing Figure 4-12 Specific Throughput for Closed Circuit Testing Figure 4-13 Specific Energy Consumption for Closed Circuit Testing Figure 4-14 HPGR / Ball Mill JK SimMet Flowsheet Figure 4-15 Summary of Bond Work Indices Figure 4-16 Signature Plot for Top Size Testing of 355µm Figure 4-17 Signature Plot for Top Size Testing of 710µm Figure 4-18 Stirred Mill Dynamic Classifier Pegs Figure 4-19 Summary of Mill Parameters for 1.2mm Test # Figure 4-20 Summary of Mill Parameters for 1.2mm Test # Figure 4-21 Summary of Mill Parameters for 1.2mm Test # Figure 4-22 Replacement of Grinding Disc with Spacer Figure µm Signature Plot Results with Revised Operating Conditions Figure 4-24 Malvern and Screening Comparison for T Figure 4-25 Malvern and Screening Comparison for T Figure 4-26 Particle Size Distributions for Option A Figure 4-27 Particle Size Distributions for Option B Figure 4-28 Product Size for Second Stage Closed Circuit Testing Figure 4-29 Specific Throughput for Second Stage Closed Circuit Testing Figure 4-30 Specific Energy Consumption for Second Stage Closed Circuit Testing Figure 5-1 Summary Layout for Cone Crusher / Ball Mill Circuit Figure 5-2 Summary Layout of HPGR / Ball Mill Circuit Figure 5-3 Summary Layout of HPGR / Stirred Mill Circuit Figure 5-4 Summary of Specific Energy Consumption for Each Circuit Figure 5-5 Product Size Distributions for Each Comminution Circuit ix

10 Figure 5-6 Preliminary Layout for an HPGR / Stirred Mill Circuit Figure 5-7 Scoping Level Testing Procedure for HPGR / Stirred Mill Evaluation Figure B-1 Bond Work Index Data Crusher Product (150µm) Figure B-2 Bond Work Index Data Crusher Product (150µm) (continued) Figure B-3 Bond Work Index Data 3N/mm 2 HPGR Product (150µm) Figure B-4 Bond Work Index Data 3N/mm 2 HPGR Product (150µm) (continued) Figure B-5 Bond Work Index Data 4N/mm 2 HPGR Product (150µm) Figure B-6 Bond Work Index Data 4N/mm 2 HPGR Product (150µm) (continued) Figure B-7 Bond Work Index Data 5N/mm 2 HPGR Product (150µm) Figure B-8 Bond Work Index Data 5N/mm 2 HPGR Product (150µm) (continued) Figure B-9 Bond Work Index Data Crusher Product (106µm) Figure B-10 Bond Work Index Data Crusher Product (106µm) (continued) Figure B-11 Bond Work Index Data 3N/mm 2 HPGR Product (106µm) Figure B-12 Bond Work Index Data 3N/mm 2 HPGR Product (106µm) (continued) Figure B-13 Bond Work Index Data 4N/mm 2 HPGR Product (106µm) Figure B-14 Bond Work Index Data 4N/mm 2 HPGR Product (106µm) (continued) Figure B-15 Bond Work Index Data 5N/mm 2 HPGR Product (106µm) Figure B-16 Bond Work Index Data 5N/mm 2 HPGR Product (106µm) (continued) Figure C-1 T1A01 Particle Size Distributions Figure C-2 T1A02 Particle Size Distributions Figure C-3 T1A03 Particle Size Distributions Figure C-4 T1A04 Particle Size Distributions Figure C-5 T1A06 Particle Size Distributions Figure C-6 T1A06 Particle Size Distributions Figure C-7 T1A07 Particle Size Distributions Figure C-8 T1A08 Particle Size Distributions Figure C-9 T1A09 Particle Size Distributions Figure C-10 T1A10 Particle Size Distributions Figure C-11 T1A11 Particle Size Distributions Figure C-12 T2A01 Particle Size Distributions Figure C-13 T2B01 Particle Size Distributions Figure C-14 T2B02 Particle Size Distributions Figure C-15 T2B03 Particle Size Distributions x

11 Figure C-16 T2B04 Particle Size Distributions Figure C-17 T2B05 Particle Size Distributions Figure D-1 355µm Top Size Test Signature Plot Data Figure D-2 355µm Top Size Test Particle Size Distributions Figure D-3 710µm Top Size Test Signature Plot Data Figure D-4 710µm Top Size Test Particle Size Distributions Figure D-5 T2C02 Signature Plot Data (T1) Figure D-6 T2C02 Particle Size Distributions (T1) Figure D-7 T2C03 Signature Plot Data (T2) Figure D-8 T2C03 Particle Size Distributions (T2) xi

12 Acknowledgements I would like to thank Xstrata Technology and the National Science and Engineering Research Council (NSERC) for their generous financial support for my research. Special thanks are given to Josh Rubenstein and Mike Larson for their indispensable advice and technical knowledge, which has allowed me to better understand the fundamentals of stirred milling. Special thanks are also given to Dr. Andrew Bamber and BC Mining Research for their continued support of my research and their role in introducing me to the field of high-pressure grinding. The ability to complete a Masters degree while running a business unit was a challenge, but the skills I have taken out of it will help me immensely as I move forward. I would like to thank my research committee and especially my faculty advisor Dr. Bern Klein for helping me through the process and providing invaluable advice and guidance along the way. I would also like to thank Pius Lo and UBC for providing me the facilities to complete my research. I would like to acknowledge Teck Ltd. and especially Steve Wilson for supporting my research and allowing me access to samples. I hope the results of this thesis will help in making Mesaba an economically viable mine in the future. Special thanks go out to Stefan Nadolski, of Koeppern Machinery Australia, for his assistance with this research and for his support as both a friend and a colleague. The help in refurbishing the M20, provided by both himself and Darcy Houlahan, allowed this research to happen and for that I am truly grateful. Most of all, I would like to thank my family for their continuing support of my endeavours and for allowing me to be the best that I can be. xii

13 1 Introduction The mining industry will be faced with new challenges in the years ahead. The exponentiallyincreasing global population has resulted in an increased demand for raw resources. With the known rich, coarse-grained deposits depleted, attention has turned to development of low-grade deposits requiring increased tonnages to achieve adequate metal production. This increased tonnage has resulted in an increased energy demand associated with metal extraction. Coupled with this, society is becoming increasingly conscious of their footprint on the environment, and serious attempts have begun, to reduce carbon emissions and increase energy efficiency (Norgate and Haque, 2010). To adapt to this changing landscape, the mining industry must begin to accept and adapt new, more energy-efficient technologies and begin to focus on developing flowsheets capable of addressing the above issues. Comminution, the process of crushing and grinding ore to liberate valuable minerals, is the most energy-intensive part of the processing flowsheet, and accounts for upwards of 75-80% of the overall energy consumption of the processing plant (Abouzeid and Fuerstenau, 2009; Tromans, 2008). In addition, the main unit operations employed in this process, tumbling mills, are as low as 1% efficient (Fuerstenau and Abouzeid, 2002). Currently, the main comminution circuits employed in the mining industry to process hard-rock, low-grade deposits include some form of tumbling mill. This equipment utilizes steel balls (ball mills), competent ore (Autogenous Grinding (AG) mills) or the combination of the two (Semi- Autogenous Grinding (SAG) mills) to fracture rock using the breakage mechanisms of impact and abrasion. The rotation of these large, cylindrical mills requires a considerable amount of energy. Although their established circuit design and ability to process high tonnages is a huge benefit, the increased energy demand and inability to efficiently grind to liberation sizes below 45µm could slowly decrease their role in flowsheet designs of the future. In the past 20 years, new, more energy-efficient technologies have been developed and adapted for hard-rock mining comminution. High Pressure Grinding Rolls (HPGR), an innovative technology adapted from the cement and briquetting industries, have begun to be considered for more base metal projects now that roll surfaces have been developed to treat hard, abrasive ores (Dunne, 2006). Operating with two counter-rotating rolls, HPGRs create a compressive bed of particles between the rolls, utilizing the process of inter-particle breakage. 1

14 This form of breakage results in improved comminution performance with a decreased demand on energy (Klymowsky et al., 2006). Additionally, unlike tumbling mills, which require steel balls to act as an energy transfer medium, HPGRs transfer energy directly from the rolls to the bed of material, resulting in an increase in energy efficiency (Fuerstenau and Kapur, 1995). Another technology, known as a horizontal stirred mill or IsaMill TM, was adapted from the pharmaceutical industry in the early 1990s to help effectively process fine-grained ore bodies (Johnson et al., 1998). The IsaMill TM consists of a cylindrical tube with a centrally-rotating shaft, mounted with evenly-spaced grinding discs. Loaded with small ceramic grinding media (2-6mm) and operated at high speeds, the equipment utilizes high-intensity attrition breakage to reduce particles in size. The rotation of a central shaft, as opposed to the entire grinding chamber (tumbling mills), results in decreased energy requirements; while the combination of small grinding media and increased media velocity, has been shown to improve the energy efficiency of grinding in particle sizes below an f80 of 150µm (Burford and Clark, 2007). The goal of this research was to examine the possibility of incorporating the above-mentioned energy-efficient equipment into a single flowsheet and eliminating the need for a tumbling mill. The biggest obstacle surrounding this research was that the proposed circuit would be operating both pieces of equipment outside of their normal operating range. As HPGRs began being adapted to the hard-rock mining sector, they found the most functionality in a tertiary crushing role, preparing feed for the ball mill (Morley, 2006a). Therefore, the process envelope for an HPGR operating in hard-rock circuits typically has feed sizes of up to 70mm, and products normally no finer than 4mm (Gruendken et al., 2010). At the same time, horizontal stirred mill technologies such as the IsaMill TM have begun to be well-established in ultrafine grinding as a regrind mill, providing a more energy-efficient alternative for processing rougher concentrates with an f80 no larger than 100µm (Gao and Holmes, 2007). To design the proposed circuit, a suitable transfer size needed to be established, to utilize both pieces of equipment effectively. A review of the literature found that a suitable circuit layout would comprise of two stages of HPGR, followed by stirred milling (Daniel, 2007b). The literature provided very little operating data for this circuit layout and therefore, optimization of operating parameters was fundamental in making this circuit technically feasible. 2

15 The following were the proposed research objectives for the work summarized in this thesis: The organization and setup of appropriate pilot-scale research equipment, capable of testing the proposed circuit. Included in this objective was a complete refurbishing of a Netzsch M20 stirred mill, complete with an appropriate mixing system capable of handling the coarse particle sizes tested. Determination of a suitable transfer size between the second-stage HPGR and the stirred mill. Examination of possible circuit layouts for the two stages of HPGR comminution. This included assessing changes in operating parameters and their effects on comminution performance. Determination of the potential specific energy requirements necessary to operate the proposed circuit. Comparison of the determined energy requirements with two conventional circuits currently being used in the industry, a cone crusher / ball mill circuit and an HPGR / ball mill circuit. This analysis included a combination of testing and simulation to gain an appropriate baseline for comparison. Development of a preliminary HPGR / stirred mill circuit flowsheet. Formulation of a practical testing procedure, which could be applied for future examination of the proposed circuit with other ore types. 3

16 2 Literature Review The following chapter will review the current literature related to high pressure grinding rolls, stirred media mills, and the combination of the two. This review will include a summary of the fundamentals, an explanation for energy efficiency, the current flowsheet design for the equipment, and the potential advantages and disadvantages of each technology. 4

17 2.1 High Pressure Grinding Rolls Background The technology of high pressure grinding and its adaptation to comminution was first realized by Professor Klaus Schoenert in the late 1970s, with fundamental work on fracture physics in comminution (Schoenert, 1979). His results concluded that increased energy efficiency was possible with the use of compressive beds and the effects of inter-particle breakage. Using a similar machine design to those used for the briquetting of coal, he adapted this concept into the technology now known as High Pressure Grinding Rolls, or HPGRs for short (Schoenert, 1988). HPGRs initially found considerable success being used to grind soft clinker material in the cement industry. Converting closed circuit ball mills to semi-finish grinding circuits and incorporating HPGRs for feed preparation, the industry was able to reduce energy consumption by 15-30% (Patzelt, 1992). After its establishment in the cement industry, advocates of the technology began to look for new applications. In 1990, the Argyle diamond mine of Australia adopted the HPGR technology to help process the increased ore hardness of their deeper, competent, un-weathered lamproite (Lane et al., 2009). Due to its ability to select pressures strong enough to break the host rock, but weak enough to leave the valuable stones intact, the applications of the HPGR in the diamond industry have become well-established (Anguelov et al., 2008). Since then, HPGRs have also found a niche application crushing iron ore for pellet feed preparation (Pyke et al., 2006). The first attempt to apply the HPGR technology in hard-rock, base metal mining, occurred at the Cyprus Sierrita mining complex near Green Valley, AZ in July 1995 (Thompsen et al., 1996). The plant trials conducted on the copper-molybdenum ore resulted in important findings in the areas of circuit design and wear rates for continuous operation. Although this application of HPGRs was not successful in becoming a lasting operation, the trials presented a success in further understanding the challenges that lay ahead, before the HPGR technology could successfully be adapted to hard-rock mining (Morley, 2008). Over the next decade, continuous improvements in roller wear, especially in the area of studded lining, allowed HPGRs to become a more viable option for base metal mining. The improvements of roller wear life, inter alia, allowed the successful installation of the HPGR technology at the Cerro Verde copper mine in 2007, becoming the first large-scale HPGR 5

18 installation in hard-rock mining (Vanderbeek et al., 2006). Since then, installations at Boddington (gold), PT Freeport Indonesia (copper, gold) and Amplats Potgietersrust (platinum) have incorporated the HPGR technology into their flowsheets (Rosario and Hall, 2010) Technology Overview The technology of HPGRs is comprised of two counter-rotating rolls mounted on a sturdy frame (refer to Figure 2-1). One roll is fixed, while the other is allowed to float and move horizontally. Unlike the rolls crusher typically found in the coal industry, a force is exerted on the floating roller by a hydraulic oil cylinder system, exposing material to pressures as high as 300N/mm 2 (Daniel, 2002; Schoenert, 1988). Material is choke-fed between the two rolls, creating a compressive bed of particles and reducing the material in size through inter-particle breakage. To cushion the constant shock on the rollers, a set of nitrogen accumulators is installed behind the floating roller, providing a smooth operating gap. To maintain high pressures between the rolls, removable cheek plates are installed in the transverse direction. Due to this arrangement, a pressure profile is created along the rolls, resulting in a finer crush at the centre and a coarser crush along the edge. Figure 2-1 Diagram of HPGR Comminution (Daniel and Morrell, 2004; Napier-Munn et al., 1996) Due to the high compressive forces exerted on the material, the product tends to agglomerate, producing what is commonly referred to as flake. The competency of this flake varies depending on material type and although publications refer to flake competency testing (Morley, 2008); no quantitative method has been developed. Due to the presence of flake in the product, 6

19 some form of de-agglomerator, such as an impact crusher or scrubber, may need to be considered when designing HPGR circuits (Schoenert, 1988). Another feature unique to HPGR comminution is the concept of the floating operating gap. Unlike other crushing equipment, there is no closed side setting when operating an HPGR. Since the roller is floating, the operating gap varies during operation and the actual distance between the rolls is dependent on the material and the frictional forces between itself and the roll surface. Due to this unique operating gap, it is difficult to produce accurate throughput models for HPGRs without performing pilot-scale testing (Klymowsky et al., 2006). Normally, HPGRs are sized using the width and diameter of the rolls. Different aspect ratios between the diameter and width are used, depending on the supplier. The main supplier of HPGRs, Krupp-Polysius, uses a high aspect ratio, which is more expensive, but produces longer wear life, since a smaller roller width allows less contact with the material. The other two suppliers, KHD and Koeppern, utilize a smaller aspect ratio, which increases roller width, creating a relatively finer product, due to a decreased edge effect and an increased pressure near the centre of the roll (Morley, 2008). The roller size also dictates the allowable top size of the feed. Since HPGRs promote compressive bed breakage, too large a top size will result in single particle breakage, thus eliminating the benefits of HPGR comminution. According to Polysius website, the largest HPGR has a throughput upwards of 3,000tph, with a feed top size of 75mm (Polysius, 2011). Currently, the only reliable way to scale-up, and properly size and select an HPGR, is to perform pilot-scale testing. Unfortunately, this requires a large quantity of material and usually poses a challenge to greenfield operations, since normally; only expensive drill core is available. The lack of a reliable small-scale suite of tests that can be used for accurate size and selection is one of the biggest hindrances preventing HPGRs from being explored in more projects (Daniel, 2002; Morrell et al., 1997). HPGRs are mechanically very reliable, with typical availability as high as 98% (Morley, 2008). The main reason for downtime remains the wear-life of different parts of the equipment. The main wear components of HPGRs are typically the parts in direct contact with the material. These parts include the feed chute, the cheek plates, and the roller surface (Dunne, 2006). The roller surface is the most problematic, and a majority of downtime is associated with roller 7

20 change-out. Currently, the two main profiles used for roller surface are a smooth tyre and studded lining. The smooth surface roll profile was initially made of a Ni-hard, wear-resistant steel (Oberheuser, 1996), but over time, newer surfaces, such as Koeppern s Hexadur wear lining, have been developed. Hexadur consists of a hard, abrasive-resistant material set into a matrix of softer material. This softer material wears quicker and creates a grooved profile that promotes the formation of an autogenous layer. In the cement industry, Hexadur has been found to last for 30,000 hours of continuous operation, but currently the only example of Hexadur lining being used for processing hard ore is at the Bendigo gold mine in Australia and no wear rates are available (Pyke et al., 2006). Hexadur can operate at pressing forces up to 6N/mm 2 for industrial applications (Morley, 2008). The more applicable roll liner for hard-rock applications is the studded lining. Consisting of tungsten carbide studs mounted on to a tyre, studded lining improves wear life through formation of an autogenous layer developed between the studs. This autogenous layer reduces contact between the roller surface and the abrasive material, and allows a wear life of between 4,000 and 8,000 hours (Klymowsky et al., 2006). The literature has suggested that not only does studded lining promote longer wear life, it also improves throughput, although at the price of increased energy consumption and higher grinding forces (Lim and Weller, 1999). Studded lining is limited to a specific pressing force of 1-4.5N/mm 2, after which the pressing force will cause damage to the metal studs (Morley, 2008) HPGR Operating Parameters The two main operating parameters available when running an HPGR are the rotational speed of the rolls and the horizontal pressing force applied by the floating roll. These two parameters provide flexible control to an HPGR operator. A change in roller speed allows the operator to adjust the throughput of the machine, up to a certain point. Increasing the roller speed will subsequently increase throughput, decreasing operating gap and specific throughput, and resulting in a narrower product size distribution (Lim et al., 1997). Koeppern recommends that a nominal roller speed of 19.1 RPM or the equivalent of 1 roll diameter per second, be employed for optimal results. By selecting the appropriate pressing force to be exerted by the floating roller, operators have the ability to vary product size and control the p80 for downstream processing. A decrease in pressing force will result in a coarser product, while an increase will 8

21 result in a finer product. Unfortunately, past a certain point, an increase in force produces little size reduction, and results in an increase in wasted energy attributed to heat (Djordjevic, 2010; Schoenert, 1988). To ensure effective operation of an HPGR, too high a pressing force should not be used, as this would prevent the floating roller from floating. In this situation, an unknown amount of force is being applied directly to the frame, considerably reducing the effectiveness and efficiency of the machine. Therefore, determining the optimal pressing force is critical, since this value varies depending on the properties of the ore. To properly scale-up and size HPGR machines, a number of parameters have been developed over the years. The following is a list of the main terms used in HPGR sizing and selection. Process Specific Throughput Constant (m-dot) The specific throughput constant, or m-dot, of the HPGR provides a normalized value allowing for throughput comparison of different-sized roller presses. The m-dot is a function of roller width, roller diameter, roller speed, and press throughput. The m-dot corresponds to the throughput (t/h) of an HPGR fitted with roller dimensions of 1m width, 1m diameter and rotating at 1m/s. The m-dot is material specific and used to size the HPGR roller dimensions for a given press throughput (Klymowsky et al., 2006). m-dot = W D * L * v (1) Where: m-dot (ts/hm 3 ) = Specific throughput W (t/h) = Press throughput D (m) = Roller diameter L (m) = Roller width v (m/s) = Roller peripheral speed Specific Pressing Force (F sp ) The specific pressing force refers to the amount of force being applied by the floating roller to material located between the rolls, as a function of roller width and roller diameter. The specific pressing force is independent of roller size, allowing for comparison of process performance for HPGRs that vary in size (Klymowsky et al., 2006). 9

22 F sp = F D * L (2) Where: F sp (kn/mm 2 ) = Specific pressing force F (kn) = Total pressing force exerted Net Specific Energy Consumption (E sp ) The net specific energy consumption is the amount of energy transferred to the material running through the HPGR. This value can be calculated by recording the average power consumption of the machine before and after a test, and subtracting it from the power being consumed during a test. The net power consumption is then divided by the throughput (t/h) of the machine during stable operating conditions. Net specific energy consumption is used for motor sizing and was the value used for comminution energy requirements in HPGR testing. E sp = P t - P i W (3) Where: E sp (kwh/t) = Net specific energy consumption P t (kw) = Total main motor power draw P i (kw) = Idle main motor power draw Energy Efficient Comminution The process of comminution has been well-documented as a very energy-intensive process. Tromans (2008), using energy figures collected by national energy departments, summarized the actual figures related to this demand on energy. Tromans found that upwards of 39% of the overall energy consumption for mining activities is spent in the processing plant, and of that figure, 75% can be attributed to the process of comminution. Table 2-1 summarizes the author s findings for the energy consumption of mining between 2001 and 2002, for the USA, Canada, Australia, and South Africa. 10

23 Table 2-1 Summary of Energy Consumption for Comminution (Tromans, 2008) Country Annual Energy Consumption For Mining (PJ*) Percentage of Total National Energy Usage Consumed by Comminution United States x 10 3 ~0.39% Canada 7.89 x 10 3 ~1.86% Australia ~1.48% South Africa ~1.8% * 1 PJ = 1 x J Overall, the demand for energy in mining is considerable, and the ability to reduce this consumption by even a fraction would be very beneficial. To improve mining economics of the future, focus should be on improving the energy efficiency of circuit flowsheets. It follows from the above discussion, that comminution represents an area for potentially larger energy savings. The fracture of rock in comminution occurs when compressive forces cause pre-existing flaws in the rock to experience tensile stresses normal to the crack length (Hu et al., 2001; Tromans and Meech, 2002). Failure results once an increased propagation of these cracks produces new surface area, and a release of strain energy at the crack tip (Rumpf, 1973). The Merriam-Webster online dictionary defines efficiency as the ratio of the useful energy delivered by a dynamic system to the energy supplied by it (Merriam-Webster Dictionary, 2011). For comminution, and tumbling mills in particular, this could be defined as the energy required for the breakage and size reduction of rock, over the mechanical energy delivered to the system by rotating the mill. If this definition is used, efficiencies for comminution in the range of 0.1-2% have been well-documented in the literature (Fuerstenau and Abouzeid, 2002; Tromans and Meech, 2002; Tromans and Meech, 2004). To provide a more meaningful number for the efficiency of comminution, other definitions have been put forward by researchers. Some have argued that the output energy used in efficiency calculations is not reflective of the process itself. Fuerstenau and Kapur (1995) argued that, traditionally, only the surface energy required in the generation of new surfaces has been used, ignoring the strain energy required for the growth of crack length. With the inclusion of strain energy, the authors suggested that the baseline for comminution efficiency should be determined by the energy resulting from single-particle fracture experiments. This argument 11

24 stems from claims that the energy utilization of comminution for single particle breakage is the most efficient (Schoenert, 1979). If this definition were used for comminution efficiency, then ball milling has efficiencies lying mostly within the range of 7 and 12%, with some as high as 17%, while HPGR comminution can be as high as 45% (Fuerstenau et al., 1996; Gutsche and Fuerstenau, 1999). Tromans (2008) hypothesized that comminution efficiency must have some maximum ideal limiting efficiency. Using a theoretical analysis of fracture mechanics, the author found a value somewhere between 5 and 10%, depending on Poisson s ratio. Tromans summarized that using this limiting efficiency, relative efficiencies ratios for comminution are between 3 and 26%, depending on the material. Whittles et al. (2006) found that in terms of fracture mechanics, the most efficient forms of comminution are slow compression of single particles, followed by compression of a bed of particles. Although slow compression of a single particle, which minimizes energy loss due to heat and noise, is the most efficient form of comminution, this is not practical for large-scale applications. Instead, Schoenert (1988) determined that, for continuous operations, the application of a compressive force to a bed of particles was the most effective process. This concept allowed for the development of the HPGR technology, improving upon the inefficiencies inherent in ball milling. Although one of the main mechanisms in ball mill comminution is through particle bed breakage, the low probability of particle collisions makes ball milling inefficient (Gutsche and Fuerstenau, 1999). Ball milling involves a considerable waste of energy in the lifting and dropping of steel balls, resulting in imperfect collisions which may or may not actually produce enough impact force to result in particle breakage. Coupled with this, a large quantity of energy is put into the wasted generation of heat, and the adverse wear of liners and steel balls (Fuerstenau and Kapur, 1995). Compared to ball mills, the breakage in HPGR comminution is a much more direct consequence of the process. No medium is required in the transfer of energy to the material. The compression exerted by the rolls is transferred directly to the material, resulting in an improved utilization of energy (Fuerstenau and Kapur, 1995). Breakage results due to very high stresses generated at the contact points between particles in the confined compressive bed. Because of this inter-particle interaction, the pressure is amplified within the bed of particles, resulting in pressures high enough to exceed the Uniaxial Compressive Strength (UCS) of the rock. This 12

25 process results in improved energy efficiency over traditional tumbling mills (Fuerstenau et al., 1991; Fuerstenau et al., 1996). Although wasted heat is generated in HPGR processing, Djordjevic (2010) found this to be unavoidable in rock fragmentation, due to the activation of material friction and shear stresses acting along the fractured surface. Regardless of the definition used for comminution efficiency, HPGR comminution has been welldocumented as improving energy utilization in comparison to conventional grinding practices; however HPGR circuit configurations require an increased reliance on materials handling equipment and the overall energy benefits associated with HPGR comminution are reduced. Even with the increased energy demand of auxiliary equipment, several publications have documented the overall energy benefit of HPGR circuits over conventional hard-rock circuits such as the SAG Ball Mill Comminution Circuit (SABC). Oestreicher and Spollen (2006) conducted a comparative study between an SABC circuit and an HPGR / ball mill circuit using a combination of testing results, operational data, and simulation. The authors found an overall reduction in energy of 19.6%. Rosario and Hall (2010) presented a study comparing two case studies examining SABC and HPGR / ball mill circuits. The authors concluded that if only specific energy consumption of the comminution process was compared, then Case A had a reduction in energy of 25.1% and Case B had a reduction in energy of 30.2%; however if the overall circuit was examined, including all auxiliary equipment, then the energy reduction dropped by only 11.7% and 18.4%, respectively. Finally, Anguelov et al. (2008) summarized a number of trade-off studies performed by Wardrop Engineering for various mining projects. An average energy savings of 25% was determined when incorporating the HPGR into hard-rock flowsheets HPGR Flowsheet Considerations Several flowsheets options have been developed to capitalize on the energy benefits associated with the HPGR technology summarized in Section One of the most important aspects to consider when designing an HPGR circuit is that as pressure (energy) is increased in an HPGR, there is a limiting factor where the generation of fines begins to clog pores, preventing further breakage. At this threshold, an increased amount of energy is wasted in the form of heat (Djordjevic, 2010). Therefore, when designing an HPGR circuit, final preparation of product must incorporate a grinding mill to produce the adequate amount of fines for further processing. 13

26 When HPGRs were first introduced into the cement industry, their role in the flowsheet was to perform much of the work that was traditionally being done by a ball or tube mill (Aydogan et al., 2006). In most circumstances, the introduction of an HPGR before the ball mill, acting as a booster, helped improve throughput and reduce specific energy consumption (Fuerstenau et al., 1991). As HPGRs began to be adapted to hard-rock mining, the best use for the technology was found to be in replacing a cone crusher (tertiary crushing role) in a standard three-stage crushing circuit. This arrangement typically requires a feed top size of between 50 and 70mm, and produces a product size between 4 and 6mm. When operating an HPGR for this duty, the industry developed some guidelines after the Cyprus Sierrita plant trials of the mid 1990s, the most important of which was the careful preparation of feed for the circuit (Morley, 2006b). The presence of tramp metal in the feed can be detrimental to the roller lining, and a metal detector should be placed on the conveyor prior to the HPGR feed chute. Also, the feed top size should be no larger than the operating gap. This prevents the possibility of single particle breakage or the damaging of metal studs. To achieve this, the secondary crusher should be placed in closed circuit with a screen prior to the HPGR. Possible circuit configurations for HPGRs include: Open Circuit A possibility if employing two-stage grinding or designing a grinding circuit that could handle the coarse fraction. This configuration is the simplest, since no recycle feed is required, and is ideal for products with hard competent flake. The drawback of this circuit could be increased energy demand (Gruendken et al., 2010). Closed Circuit with Dry Screening A possibility if flake competency is weak and efficient screening can be achieved. This configuration is ideal over wet screening, since it prevents unwanted water from entering the circuit. The main drawback of this configuration is excessive dust generation (Morley, 2006a). Closed Circuit with Wet Screening The most likely circuit when dealing with more competent flake, or with less efficient screening (Morley, 2006a). Unfortunately, increasing moisture content in the feed leads to lower throughput and higher energy consumption (Fuerstenau and Abouzeid, 2007). Therefore proper circuit design must compensate for these limitations. Closed Circuit with Product Splitter This involves taking a split of the product, preferably the edge product, and recycling the coarser material without the need for classification. This configuration is ideal since no screening is required; however, since 14

27 HPGR product will still contain a fraction of coarse particles when entering the ball mill, the situation could arise where the ball mill is unable to effectively break these particles and a critical build-up could ensue (Gruendken et al., 2010). HPGRs have an advantage over conventional crushers when operating in closed-circuit operations. HPGRs can still produce an effective crush of oversize material, since unlike crushers, there is no closed side setting, limiting the probability of breakage (Gruendken et al., 2010). Morley and Daniel (2009) examined what future HPGR flowsheets should entail. The authors believe that the next generation of HPGR flowsheets should attempt to eliminate the need for auxiliary equipment, and allow the secondary crusher or the HPGR to operate in some form of open circuit. This would greatly decrease the capital cost and complexity of HPGR circuits. Although several ideas are put forward, each suggestion is met with substantial operating problems, making these flowsheet concepts unrealistic at the present time. The role of an HPGR as a tertiary crusher has thus placed it in direct competition with the SAG Ball Mill Comminution (SABC) circuit currently being used as the standard for ball mill feed preparation (Gruendken et al., 2010). Although an SABC circuit is currently the industry standard, an HPGR circuit could provide increased operating benefits for hard-rock, hightonnage operations. The advantages that an HPGR circuit has over an SABC circuit (refer to Figure 2-2) were first observed with the HPGR installation at Cerro Verde in Peru (Vanderbeek et al., 2006). Figure 2-2 Flowsheet for Cerro Verde (Rosario et al., 2011; Vanderbeek et al., 2006) 15

28 Another application currently being implemented at Gold Corps Peňasquito operation in Mexico, utilizes an HPGR to treat pebble crusher product in their SABC circuit (refer to Figure 2-3). The purpose of this arrangement is to improve the overall throughput of the comminution circuit, by further reducing the size of the pebble crusher product. The authors estimated that a 30% increase in total capacity can be achieved, with only an additional 4.8kWh/t of energy expenditure (Dixon et al., 2010). Figure 2-3 Flowsheet Comparison for Peňasquito (Dixon et al., 2010) Advantages and Disadvantages When applying HPGRs to a process flowsheet, there are several benefits that make this technology more appealing than the conventional SAG / ball mill and three-stage crushing circuits. A review of the literature has suggested that proper design of HPGR circuits can lead to the following advantages: Decreased Operating Costs Due to the reduction in energy consumption provided by HPGR comminution, lower energy costs will lead to lower overall operating costs. The elimination of steel grinding media can also lead to cost savings. Morley (2008) stated that, although roll tyre replacement is still required, the cost is approximately the same as SAG mill liners. Vanderbeek et al. (2006) estimated that a reduction in operating costs of $0.368/t was achievable by implementing an HPGR based circuit in place of an SABC circuit. 16

29 Inter-Particle Breakage The unique compressive bed and increased production of fines created in HPGR comminution provides an advantage over conventional crushers (cone) for preparation of ball mill feed. This benefit could lead to an increase in overall circuit throughput, providing an improvement in ball mill circuits (Danilkewich and Hunter, 2006). Lower Sensitivity to Ore Variability Unlike SAG mill operations, HPGRs can handle a large change in ore hardness, with little effect on throughput and comminution performance. Due to the ability to adjust the specific pressing force, an operator can directly account for increased ore hardness (Rosario, 2010). Small Machine Footprint With the size of an HPGR being considerably smaller than that of the SAG mills currently in operation, a reduced machine footprint can increase available space in the mill, increasing flexibility for operations (Danilkewich and Hunter, 2006). Short Equipment Lead Time HPGRs have a relatively short lead time in comparison to SAG mills, with differences between 6 and 14 months (Morley, 2008). The above advantages provided by HPGRs are hard to dispute, but the other main advantages resulting from the production of micro-cracks within the product, have yet to be fully proven. Due to the high stresses created within the compression bed, some of the product does not fracture, but does contain several micro-cracks. This can weaken the material and potentially improve downstream processes, such as flotation and leaching. After a review of the literature, a standard photograph (Figure 2-4) was presented when the topic of micro-cracking was discussed. No background on experimental procedure could be found. Daniel (2007a) provided the first detailed account that HPGR product exhibits microcracking. In Daniel s doctoral dissertation, photographic existence of micro-cracking was provided for a number of different ore types, using a Scanning Electron Microscope (SEM). 17

30 Figure 2-4 Standard Photographic Evidence of Micro-cracking (Morley, 2008) Several publications have reported that micro-cracks result in a reduction in the Bond ball mill work index of 10-25%, in comparison to conventional crusher product (Daniel, 2007a; Danilkewich and Hunter, 2006; Muranda, 2009; Norgate and Weller, 1994; Rule et al., 2008). Some have argued this is due to the higher presence of fines in HPGR product, providing the illusion that HPGR product is weaker than conventional crusher product. To prove this concept wrong, Rule et al. (2008) performed regular Bond work index tests, as well as a modified Bond work index test, where fines were removed from the HPGR product to create a comparable size distribution to the crusher product. A reduction of 12% was obtained for the unaltered HPGR product, while a reduction of 7% resulted from the modified product. According to the literature, increasing the specific pressing force results in an increased reduction in the Bond work index. Norgate and Weller (1994) performed Bond work index testing on zinc and gold ores at specific pressing forces, ranging between 1 and 12N/mm 2. The reduction was greatest between 4 and 8N/mm 2, with little difference resulting from higher pressures. The authors concluded from the results that the reduction in Bond work index should be used as additional criteria when selecting an optimum specific pressing force. In addition to size reduction and specific throughput, specific energy consumption of both HPGR, and ball mill grinding, should be evaluated. Tavares (2005) performed a comparative study between the HPGR and conventional crushing equipment (roll crusher and hammer mill), to evaluate the reduction of impact energy necessary to break particles. Copper and gold ores were tested at narrow size fractions using an impact load cell. Particle weakening occurred in HPGR product, but diminished with decreasing particle size. Tavares concluded that, below 1.5mm, particle weakening became insignificant 18

31 between the different pieces of equipment. Although this could be considered proof that particle weakening does not occur at finer size fractions, the impact load cell may not be the ideal way to test the strength of small particles. Daniel (2007a) found that extensive micro-cracking is still present in fractions finer than 850µm. To demonstrate that particle weakening leads to a reduction in the energy demand in ball mill grinding, HPGR and conventional crusher (cone) product could be processed through a pilotscale ball mill. Under continuous operation, energy requirements for grinding to a specified product size could be compared for each feed preparation method. So far, no literature was found that incorporates this approach. Unlike the presence of micro-cracking and the reduction of ball mill energy, the benefits microcracking provides to downstream processing are less clearly understood. Micro-cracks are believed to create a more porous material, leading to better leaching characteristics for gold and silver ores. In a report by Golden Queen Mining, for their Soledad Mountain Project in Southern California, bottle roll and column leach tests were performed to compare the HPGR with a Vertical Shaft Impact (VSI) crusher. Higher recoveries and shorter leach times were achieved with HPGR product, and subsequently, the choice of HPGR comminution was selected for their heap leach operation (Klingmann, 2005). In the case of flotation, no conclusive evidence has been put forth to confirm that HPGR product improves flotation kinetics and increases metal recovery (Palm et al., 2010). Hosten and Ozbay (1998) speculated that compressive bed breakage leads to material fracturing along grain boundaries, resulting in liberation of mineral grains at coarse size fractions. Daniel (2007a) attempted to answer the question of preferential liberation created by HPGR comminution. The author s work found no conclusive proof of this phenomenon. Although HPGRs provide a veritable array of benefits, there are a few drawbacks to this technology. The following are the main disadvantages associated with HPGR technology: Circuit Complexity Due to the stipulations placed on the feed requirements for HPGRs, circuit design typically includes a secondary crusher in closed circuit with a screen, a metal detector to prevent tramp metal from damaging the rolls, and a screening circuit to handle HPGR product prior to feeding the ball mill. These specifications require an 19

32 increased amount of materials handling equipment and increased capital costs associated with the circuit (Morley, 2008). Increased Capital Cost Due to the present limitation on bearing size, and thus roller size, an HPGR machine is limited to a maximum throughput of 2,500-3,000tph. For larger operations, multiple machines are required to perform the same duty as one SAG mill. Coupled with the increased auxiliary equipment mentioned above, HPGR circuits typically have higher capital costs. Vanderbeek et al. (2006) estimated that Cerro Verde capital costs for the complete HPGR comminution circuit (primary crushing through ball milling) were ~29% higher than a complete SABC circuit. Although higher in capital, the decrease in overall operating costs produced approximately 1.5% higher internal rate of return for the project. No Standard Energy and Throughput Model Although not a disadvantage of the technology, currently, there are no standard small-scale tests available to accurately predict energy and throughput for a given ore. Until small-scale tests are available that require a small quantity of material, HPGRs will not be considered in early stage circuit designs. Industry Acceptance Due to the mining industry s reluctance to embrace new technology, HPGRs are not considered as a primary option in circuit design. Although examples of HPGRs operating in hard-rock, high-tonnage operations are beginning to materialize, until this technology is successfully introduced in a Canadian mining operation, the status quo will remain in effect on this. Inability to Process Clayish Ore HPGRs are unable to process sticky clayish ores, due to slippage on the rolls, reduced throughput, and production of unreasonably large and competent flakes. Although this is the current situation, work has been done on a circuit design to handle such ores (Rosario et al., 2011). Poor Performance with Increasing Moisture Content When high moisture content is present in the feed, poor performance in terms of throughput and wear rates can be experienced. When processing wet material, the inability to produce a continuous autogenous layer on the roller surface can drastically decrease roller life (Fuerstenau and Abouzeid, 2007; Morley, 2008). Although there are limitations to the HPGR technology, proper circuit design and continuing research and development should lead to the mitigation of the associated risks. 20

33 2.2 Stirred Media Mills Background Stirred mill technology, or the concept of a centrally rotated shaft to agitate media, was first developed during the 1950s in Japan. This technology utilized a vertical orientation, and was used for grinding in the minerals industry. In 1979, the Metso grinding division (then known as Koppers) acquired the license, with the intention of applying it to base metal mining operations. Unfortunately, the design was not suitable for the rigors of hard and abrasive ore, and subsequent work was done to improve grinding efficiency, minimize maintenance and downtime, and improve wear rates. These developments led to the creation of Svedala s Vertimill, a tower mill with the ability to process material as coarse as 6mm, and produce product as fine as 20µm (Allen, 2009). In the early 1990s, Mt. Isa Mines Limited (now a part of Xstrata) was investigating technologies to economically process two of their fine-grained lead/zinc deposits: Mount Isa in Queensland and McArthur River in the Northern Territory. The McArthur River deposit was originally discovered in 1955, but no company had been able to economically process the fine-grained deposit with the grinding technology available. In the case of Mount Isa, the gradual decrease in metallurgical performance in the mid 1980s, due to finer liberation size, resulted in recoveries dropping to 50% by the early 1990s (Anderson and Burford, 2006; Burford and Niva, 2008). High media costs, impractical energy requirements, and poor flotation performance from steel media contamination, led to the decision that the available grinding technologies, ball mills and tower mills, were unsuitable for ultrafine grinding to sub-10µm, and new technology was needed to address this challenge. After researching other industries that require ultrafine grinding, the process team at Mount Isa settled on the horizontal stirred mill technology, manufactured by Netzsch of Germany. The technology was being used to process high value manufactured products, such as printer inks, pharmaceuticals, paint pigments, and chocolate. These applications required small mills, run in batch operation, using high cost, sanitary grinding media. To adapt the mill to the metals industry, work was done to increase the mill capacity, allow for continuous operation, and apply low-cost grinding media (Pease, 2007). This development work resulted in the creation of the M3,000 IsaMill TM, leading to its installation at Mount Isa in 1994 and McArthur River in 1995 (Burford and Clark, 2007). 21

34 Since the mid 1990s, the development of inert ceramic grinding media and increased mill size to the M10,000 have allowed the IsaMill TM to move away from ultrafine grinding (<10µm) and establish itself as a regrind mill, producing product with a p80 between 20 and 40µm. The ability to further liberate rougher concentrate, without contaminating mineral surfaces, has led to successful installations at Kumtor (gold), Western Limb (platinum tailings), Prominent Hill (copper/gold), and Potgietersrust (platinum) (Anderson and Burford, 2006; Burford and Clark, 2007; Curry et al., 2005) Vertical Stirred Mill Technology There are two main orientations for stirred mill technology: vertical and horizontal. Each orientation has its own advantages and disadvantages, but both use attrition as the main breakage mechanism for size reduction. Vertical stirred mill technologies can be classified into two sub-categories, gravity-induced and fluidized, depending on how the grinding media is circulated within the mill and the speed at which the shaft operates. In the case of gravity-induced circulation, a centrally-mounted, double-helical screw is suspended into the cylindrical grinding chamber, and rotated at a low speed in the range of 100 RPM (Sinnott et al., 2006). The chamber is filled with grinding media, typically steel for coarse applications, and ceramic for fine applications. As the screw rotates, media is drawn up the centre of the mill, and eventually cascades off the edge of the screw, creating a gravity-induced, downward flow of media along the mill perimeter (Allen, 2009; Sinnott et al., 2006). Material, fed as slurry, enters at the top of the mill and circulates down along the perimeter, being drawn back upwards with the aid of the rotating screw (Cleary et al., 2006). This action creates continuous contact with grinding media, initiating size reduction through attrition. As material is ground finer, it overflows the mill and is sent to a classifier, where coarse material is re-circulated back to the mill, and finer material is sent on as final product. Figure 2-5 shows an example of particle flow in the Vertimill. Examples of gravityinduced stirred mills include the tower mill and the Vertimill. Lichter and Davey (2006) stated that tower mills are more efficient at a coarser feed size; however most installations operate in regrind circuits at fine particles sizes (Allen, 2009). 22

35 Figure 2-5 Example of Gravity-Induced Vertical Stirred Mill Technology (Vertimill ) (Metso, 2010) For fluidized stirred mill technology, a centrally-rotating shaft is suspended within a cylindrical grinding chamber, but unlike gravity-induced circulation, the shaft is equipped with either pins or grinding discs and operates at a high speed in the range of 250 RPM (Sinnott et al., 2006). The rotating shaft agitates the grinding media, creating a fluidized bed. Slurry is fed in at the bottom (Deswik) or top (Stirred Media Detritor) of the mill, and passes through the fluidized bed of media, resulting in high intensity media particle interactions. Product then passes through the media retention screens and overflows the mill as product. Typically, these mills operate with ceramic or sand grinding media, and are best suited for ultrafine grinding applications (Metso, 2010; Rule et al., 2008). Examples of fluidized vertical stirred mill technologies include Imerys Stirred Media Detritor (SMD) (pin configuration) and the Deswik mill (disc configuration). Both of these technologies are shown in Figure 2-6. Vertical stirred mill technology is a more efficient comminution technology compared with conventional tumbling mills, and Metso has claimed that a 30-50% reduction in energy can be achieved, depending on how fine a grind is required (Metso, 2010). Since a ball mill relies on rotation of the entire grinding chamber to create slurry media interactions, a greater expenditure of energy is required. Size reduction is usually achieved by attrition and impact, but impact is not as effective, due to the probability that media will collide with other media or the mill lining, resulting in wasted energy. Allen (2009) proposed that the most efficient zone in the 23

36 ball mill is what is referred to as the kidney, in reference to its shape. In this zone, media and particles are in constant contact with each other, resulting in an increased rate of attrition. A similar zone of intense attrition is consistent throughout the stirred mill chamber, resulting in improved grinding efficiency. Although this restricts the top size fed to the mill because this zone reduces impact breakage, limiting fracture of coarse particles. Figure 2-6 Examples of Fluidized Vertical Stirred Mill Technologies Stirred Media Detritor (Left) and Deswik Mill (Right) (Capstick, 2010; Metso, 2010) Horizontal Stirred Mill Technology In the metal mining industry, the main example of a horizontal stirred mill technology is the IsaMill TM. This technology is comprised of a centrally-rotating shaft, enclosed by a fixed cylindrical grinding chamber (refer to Figure 2-7). The shaft is installed with 7-8 evenly-spaced polyurethane grinding discs, and operates at very high speeds, between 1,200 and 2,000 RPM (Larson et al., 2008). The rotation of the grinding discs creates tip speeds of 19-23m/s, while a tower mill and the SMD operate at 3m/s and 8m/s, respectively (Anderson and Burford, 2006; Parry, 2006). Material is fed as slurry at one end of the mill and passes through the fluidized media zone, where high-intensity attrition reduces the particles in size (Arburo and Smith, 2009). High-intensity attrition allows the IsaMill TM to process fine particles at a high throughput. Attached to the end of the rotating shaft is a dynamic classifier that utilizes centrifugal forces to retain grinding media and coarse particles, while allowing fine particles to exit the mill. A 24

37 diagram detailing this process is shown in Figure 2-8. Currently, the largest IsaMill TM available is the M10,000, which is equipped with a 3MW motor. Figure 2-7 IsaMill TM Layout (Burford and Clark, 2007) Figure 2-8 IsaMill TM Grinding Mechanism (Burford and Clark, 2007) Originally, IsaMill TM operations used close-proximity grinding media, including slag, ore gravel, and sand. This grinding media was beneficial because it was cheap and in constant supply. Unfortunately, this type of media is not hard enough to produce efficient grinding, and suffers 25

38 from high wear rates. Kwade and Schwedes (2002) stated that the stress intensity exerted by grinding media, adheres to the following relationship: SI = d 3 * ρ * v 2 (4) Where: SI (N*m) = Stress intensity per media particle d (m) ρ (kg/m 3 ) v (m/s) = Media diameter = Media density = Media velocity Since an IsaMill TM already operates at high speeds, to improve the effectiveness of grinding media, an increase in diameter or density is required. The original grinding media had a low Specific Gravity (SG) (2.4) and small diameter (<1mm), leading to milling inefficiencies and limitation of feed size. With the introduction of ceramic grinding media, exhibiting higher SG (3.7) and larger diameter (3.5mm), the IsaMill TM can operate at a coarser feed size (<150µm) while providing a lower media wear rate (Burford and Niva, 2008). Table 2-2 shows typical wear rates for different grinding media, including MT1, a ceramic grinding media manufactured by Magotteaux International. Table 2-2 Summary of Grinding Media Wear Rates (Curry and Clermont, 2005) Media Type Consumption Rate (g/kwh) Relative Consumption MT1 (-4 +3 mm) Alumina 1 (-4 +3 mm) Alumina 2 (-4 +3 mm) Australian River Pebble (-4 +3 mm) Australian Silica Sand (-6 +3 mm) Ni Slag (-4 +1 mm) The energy requirements for a full-scale IsaMill TM can be determined using a laboratory mill. Gao et al. (1999) determined that a 1:1 energy scale-up exists between a lab-scale M4 mill and industrial-scale M4,000 mill. This ratio is attributed to the grinding mechanism shown in Figure 2-8, which prevents short-circuiting in the mill, allowing for uniform grinding. Curry et al. (2005) reported that results obtained in an M4 can accurately scale-up to an M10,

39 2.2.4 Horizontal Stirred Mill Operating Parameters When running a horizontal stirred mill, several operating parameters pertain to performance in the mill. Some parameters are related to the feed conditions entering the mill (feed density and volumetric flow rate), while others relate to the operating conditions of the mill (mill speed, media volume, and media size). The following section will discuss each of these parameters and their effect on mill performance. Feed Density The feed density is related to the density of the solids component and the percentage by weight occupied in the slurry. Several publications reported that operating at low percent solids, below 30-40%, results in lower energy efficiency (Gao et al., 1999; Lichter and Davey, 2006); however this limit may be material-dependent, depending on SG of the ore (Larson et al., 2008). Larson et al. (2008) suggested that operating at 50% solids results in optimal energy efficiency and greater than 65% may result in poor efficiency due to viscosity issues. Feed density is a main parameter used to control retention time in the mill. An increase in solids content allows the mill to operate at the same throughput, but decreases the amount of energy transmitted to the material. Volumetric Flow Rate The volumetric flow rate refers to the amount of slurry passing through the mill in a given time interval. Larson et al. (2008) found that the effect of flow rate has little influence on energy efficiency and only affects residence time in the mill. This residence time will affect the size reduction of the product, but the energy usage will adhere to the same curve. Mill Speed The mill speed refers to the rotational speed of the agitator. Typical values for operation vary depending on mill size (disc diameter), but result in tip speeds of 19-23m/s. Larson et al. (2008), using a lab-scale M4 mill, determined that mill speed has very little effect on energy efficiency, and a linear relationship exists between speed and mill power draw. Parry (2006) suggested that varying mill speed can control the stress intensity exerted by the grinding media and could be used to process soft and hard minerals. 27

40 Media Volume The media volume refers to the percentage of bulk media occupying the grinding chamber when the shaft and disc volume are removed. The generally accepted operating range for media volume is between 60 and 80%. When operating below this range, insufficient media is available for grinding and the possibility of unbroken solids can lead to clogging of the mill (Larson et al., 2008). The adjustment of media volume is one of the options available to operators to prevent over-grinding when a decrease in throughput is experienced. Termed turn down, operators can decrease the media volume in the mill to reduce the energy input into the material and operate at a lower throughput with the same grind (Curry et al., 2005). Media Size The media size, measured by the diameter, is the most critical parameter available to optimize energy efficiency. The selection of media size is crucial, since coarse media is required to break the larger particles, but must be small enough to provide efficient grinding for finer particles. Jankovic (2003) found a difference in energy efficiency of 40% when poorly selected media size was tested. The impact of media size is shown in Table 2-3, where a decrease in media size leads to an increase in media surface area, resulting in an increase in media particle collisions. Table 2-3 Normalized Effect of Decreasing Ball Size (Lichter and Davey, 2006) Ball Size (mm) Surface Area (m 2 /t) Number of Balls (per tonne) Number of Balls Normalized , , , ,144, ,648, ,314,560 1,000 Mankosa et al. (1986) suggested a selection ratio for media size to mean (80% passing) particle size of 20:1 for fine grinding. Using this ratio, a feed f80 of 300µm would require a media top size of 6mm. This ratio provides a balance between an increased probability of media particle 28

41 interaction and the capability for media to catch and break particles. Yue and Klein (2006) confirmed these assumptions using a geometric analysis and suggested that this ratio allowed for the capture of 4 to 5 particles within bead voids. So far no literature was found related to media selection for coarse stirred milling and this ratio may be lower for coarser applications Energy Efficiency for Stirred Media Mills As discussed in Section 2.1.4, operation of conventional tumbling mills requires a substantial amount of energy to rotate large cylindrical mills filled with steel media and slurry. This rotating action creates the lift for steel balls to tumble, thereby reducing coarse particles in size through impact breakage, while providing the motion necessary to grind particles between steel balls for attrition breakage. The combination of these mechanisms allows ball mills to be applicable to a wide range of sizes from an f80 of 4mm down to a p80 of 45µm, below which these mechanisms lose an increasing amount of energy to ball ball and ball liner collisions. These two mechanisms of breakages are effective; however the low probability of particle ball collisions leads to low energy efficiency (Fuerstenau and Abouzeid, 2002). Section referred to the kidney zone in a ball mill. In this zone, constant ball particle collisions results in high fines generation through attrition. Unfortunately, this zone is not consistent throughout the mill, because an open volume is still required for media to tumble and to create impact breakage. Due to these shortcomings, over the years, compensation for low efficiencies has resulted in the installation of larger ball mills, increasing the power requirements for grinding (refer to Table 2-4). Stirred mill technology has evolved over the years to help improve upon this increased energy requirement. The energy benefits associated with increased media particle interactions, resembling the kidney zone in a ball mill, accompanied with lower power draw necessary to rotate a central shaft, has led stirred milling to become a viable option for regrind mills (Lichter and Davey, 2006). 29

42 Table 2-4 Summary of Ball Mill Size Over the Years (Daniel, 2007a; Lynch and Rowland, 2005) Year Diameter (m) Length (m) Power (kw) , , , ,440 Kwade and Schwedes (2002) stated that the stress intensity exerted by media is proportional to the velocity squared. The speed of media in tumbling mills is limited by the speed at which centrifuging of mill contents begins and effective breakage from cascading media ceases, referred to as critical speed (Kapur et al., 1992). This limits the size of the grinding media in a ball mill because a smaller media size cannot subject particles to the high-stress intensities required for breakage (Wang and Forssberg, 2007). Breakage characteristics in stirred mills are dependent on the stress intensity exerted on the particle and the number of stress events experienced by feed particles and their resulting daughter fragments (Kwade and Schwedes, 2002). For grinding of coarse particles, the stress frequency is high because larger particles have a higher probability of making contact with media. In addition, coarse particles also exhibit a higher degree of flaws, resulting in lower required stress intensity for breakage (Wang and Forssberg, 2007). If an appropriate-sized media were selected, then it should be possible to efficiently grind coarse particles in a stirred media mill. As particles are reduced in size, they require a higher number of collisions and increased stress intensity to cause further breakage. To achieve this, stirred mill technology, such as the IsaMill TM operates at high impeller speeds and small media size, resulting in increased energy intensity and a higher frequency of collisions. Table 2-5 compares the power density in an IsaMill TM to other grinding mills. The power density in a stirred mill is considerably higher than in 30

43 a tower mill. If this extra energy is utilized effectively through the optimization of operating parameters, then increased energy efficiency should be achievable. Table 2-5 Summary of Power Density for Grinding Mills (Burford and Niva, 2008) Mill Type Installed Power (kw) Mill Volume (m 3 ) Power Density (kw/m 3 ) Autogenous Mill 6, Ball Mill 2, Regrind Mill Tower Mill 1, IsaMill TM M10,000 3, Several publications found that for fine size ranges up to 150µm, the mechanisms mentioned above result in stirred mill technology being a more energy-efficient option over ball mills, independent of orientation (Allen, 2009; Anderson and Burford, 2006; Anyimadu et al., 2007; Lichter and Davey, 2006; Pease, 2007). As the feed increases in particle size, the effectiveness of the ball mill becomes more apparent. Shi et al. (2009) conducted a study to determine whether stirred mills could achieve higher energy efficiency than ball mills at coarse particle sizes. The first test examined the energy comparison between a vertical stirred mill and a Bond ball mil for processing material with a feed top size of 3.35mm down to a p80 of 75µm. The stirred mill achieved energy reductions of 25%, 37%, and 27% for the three ore types tested. These results are similar to other researchers suggestions that vertical stirred mills are more energy-efficient than ball mills for primary grinding applications (Lichter and Davey, 2006). Allen (2009) noted that although Vertimills have been applied to regrind applications, they are best suited for primary grinding. The second test by Shi et al. (2009) investigated the energy requirements of an M4 IsaMill TM and a batch ball mill (300mm x 300mm) for processing a sample with a feed top size of 1mm down to varying product sizes. The ball mill was able to achieve slightly lower energy requirements for coarser grinds (p80 > 40µm), but became less efficient with finer grinds. Although the ball mill was more efficient at coarser particle sizes, it is evident that operating conditions for testing were not optimized. Testing conditions consisted of operating at 30% solids, using 3.5mm ceramic grinding media. Section mentioned that media size plays an 31

44 important role in the effectiveness of breakage at coarse sizes. The recommended media size ratio of 20:1 was not used, which may explain the ineffectiveness of grinding at coarse sizes. More energy was required by the smaller media to break the coarse particles, resulting in a lower efficiency. Based on the 20:1 ratio, a media top size of 6-8mm should have been used. Other publications have examined the energy efficiency between tower mills and the IsaMill TM and found that the IsaMill TM operates more effectively at finer sizes while a tower mill becomes more efficient at coarser sizes (Burford and Niva, 2008; Harbort et al., 2010). Parry (2006) performed a comparative study between tower mills installed at the Red Dog mine in Alaska and lab-scale stirred mills. Using an IsaMill TM and a Stirred Media Detritor, Parry found a 50% reduction in energy for grinding from an f80 of 29µm to a p80 of 22µm. Although the tower mill may function more effectively at coarse sizes, its orientation poses considerable problems in scale-up. When designing vertical stirred mills, the motor selection is dependent upon the required torque necessary to rotate the media from rest. With larger units, this torque begins to dominate the mechanical design, requiring a substantially larger motor and support frame. Due to the orientation of the stirrer in a horizontal configuration, scaling up does not result in this problem and the design of larger mills is more feasible (Pease, 2007). Overall, there are benefits and limitations to both stirred mill orientations, but the use of stirred mill technology has the potential to improve the energy efficiency of grinding and with proper flowsheet design, could be a viable option over conventional ball milling Horizontal Stirred Mill Flowsheet Options Originally the IsaMill TM was developed for ultrafine grinding applications, to efficiently process finely-disseminated ores. Typical flowsheets place the IsaMill TM in the regrind circuit, accepting rougher concentrate product from the regrind ball mill, at an f80 of 25µm, and further liberating minerals by grinding to below a p80 of 10µm (Gao et al., 2002). An example of this is shown in Figure 2-9 where the IsaMill TM is installed at Mount Isa Mines to treat the lead and zinc concentrate for further liberation. 32

45 Figure 2-9 Flowsheet for Mount Isa Mines (Gao et al., 2002) After establishing itself as an energy-efficient alternative for ultrafine grinding to sub 10µm, the IsaMill TM is now being considered for coarser applications, accepting feed sizes up to an f80 of 150µm. Combining the increased grinding effectiveness of harder, higher SG ceramic media with an overall increase in mill size, design of regrind circuits are beginning to incorporate IsaMiIls TM as the mill accepting feed directly from rougher flotation (Burford and Clark, 2007). An IsaMill TM installed at the Kumtor gold mine in Kyrgyzstan was originally designed to process product from a regrind ball mill, but during ball mill maintenance, the M10,000 was used as its replacement (refer to Figure 2-10). During this time, the IsaMill TM was required to accept an f80 of µm and produce a product p80 of 60-65µm. Although there was insufficient time to optimize media top size and other operating conditions, the circuit was able to operate effectively. Had these conditions been optimized, plant operators believed that a reduction in power draw and a decrease in product size could have been achieved as well (Anderson and Burford, 2006). 33

46 Figure 2-10 Flowsheet for Regrind Circuit at Kumtor Mine (Burford and Clark, 2007) The Phu Kham copper/gold mine in Laos, currently in development, will utilize an M10,000 IsaMill TM for regrinding flotation concentrate. The circuit will operate at a throughput of 168tph, processing an f80 of 106µm and producing a product p80 of 38µm (Burford and Clark, 2007). IsaMill TM technology is beginning to be applied to coarse grinding applications. Pease (2007) presented a summary of an IsaMill TM at McArthur River operating as a secondary grinding mill, treating SAG mill product. Originally the IsaMill TM was installed to grind 50µm concentrate down to a p80 of 7µm, while a SAG mill was used to prepare feed for flotation (p80 of 45µm). To increase tonnage a tower mill was installed to process a portion of the SAG underflow stream (refer to Figure 2-11). Lab and pilot-scale tests were carried out on SAG underflow product to determine the potential grinding limits for coarser IsaMill TM applications. With the plus 1mm fraction screened out, the IsaMill TM achieved a finer product for the same energy input as the installed tower mill. Unfortunately a build-up of steel scats and coarse particles provided problems for continuous operation. The IsaMill TM was subsequently tested using minus 1mm cyclone overflow with a reduced f80 of 300µm. Plans are now underway to increase throughput at McArthur River by installing two M10,000 mills to treat SAG screened product (f80 of 300µm) and grind to a flotation feed p80 of 45µm (refer to Figure 2-12). 34

47 Figure 2-11 Original Flowsheet for McArthur River (Pease, 2007) Figure 2-12 Modified McArthur River Flowsheet with Coarse IsaMill TM Grinding (Pease, 2007) Process Benefits of Horizontal Stirred Mills There are several process benefits available for horizontal stirred mills which can help improve the economics of the circuit. The following are the main advantages offered: Small Machine Footprint Although not a process benefit, the compact size and energy intensity offered by stirred mills allows for a small footprint, resulting in an increase in available space, in case plant expansion and an increased throughput is desired. Inert Grinding Media The use of ceramic grinding media eliminates the potential for steel contamination of mineral surfaces, a problem that may hinder flotation kinetics. Several publications have noted, especially in the precious metals sector, that the use of 35

48 inert ceramic grinding media can help improve flotation response for fines flotation (Rule et al., 2008). The use of steel media can lead to deposits of iron hydroxide on the surface of sulphide minerals, resulting in poor flotation selectivity and increased reagent dosage. The use of ceramic grinding media has eliminated this problem and can allow for cleaner liberated mineral surfaces at finer sizes, improving the economic benefits of fine-grained mineral deposits (Arburo and Smith, 2009; Pease et al., 2006). Internal Classification The dynamic classifier installed at the end of the mill agitator shaft, allows the mill to operate in open circuit, producing a sharper product size distribution. This configuration leads to the elimination of a recycle stream, reducing maintenance costs, and increasing throughput capacity. The uniform grinding action experienced throughout the mill, leads to a reduction of over-grinding and the prevention of ultrafines, which may prove problematic in the flotation circuit (Burford and Clark, 2007). 36

49 2.3 HPGR / Stirred Mill Circuit To take advantage of the two energy efficient technologies summarized in Sections and 2.2.5, an appropriate circuit flowsheet must be developed. Section stated that an HPGR is limited in the size reduction achievable from one pass through the rolls. To create a product fine enough to process through a stirred mill, a second stage of HPGR comminution is necessary. Several publications have documented the effects of processing material through multiple passes of an HPGR. Norgate and Weller (1994) performed tests on a gold ore to determine whether operating several stages of HPGR at a lower specific pressing force could be more energy-efficient than single-stage operation at a higher specific pressing force. The results showed that with subsequent passes through the mill, the specific energy consumption trended downward. By pass four, specific energy consumption began to flatten out. The largest decrease, at 31%, was experienced between pass one and two. In terms of size reduction, a positive linear relationship was found between the total specific energy consumed for all passes and the reduction ratio. It should be noted that this test was also repeated with a deagglomeration step performed between each pass, resulting in no significant change in results. The data presented by Norgate and Weller (1994) shows that the operation of multiple HPGRs in series can lead to an increased size reduction, producing the feed range acceptable for stirred milling. Daniel (2007b) conducted tests on a pilot-scale HPGR to assess whether several HPGRs in series could produce a similar grind size to a ball mill, at a lower specific energy consumption. The first two passes through the HPGR produced the highest size reduction ratios and subsequent passes became less efficient. Daniel concluded that two passes through the mill is the limit for efficient crushing of hard ores. Rule et al. (2008) conducted testwork on a Labwal HPGR to determine the effect of passing material through two open-circuit stages of HPGR comminution, while varying the specific pressing force in the second stage. Good size reduction was achieved between the first and second pass, but no significant difference in size reduction resulted from increasing the specific pressing force in stage two. To optimize the specific energy consumption for operation of a two stage HPGR circuit, an increase in specific pressing force in stage two is unnecessary. Fuerstenau et al. (1999) summarized work showing the benefits that an HPGR / ball mill circuit can achieve by optimizing the ball size. The characteristics of HPGR product, including fines 37

50 content and micro-cracking, allows for the ability to improve the efficiency in the ball mill by reducing ball size. The authors predicted that optimization may be reached when 50-60% of the total energy load is handled by the HPGR. Pease (2007) found that the IsaMill TM can process particles sizes as coarse as 1mm. If the main attributes of HPGR product, including the high proportion of fines and the presence of micro-cracking, are also considered, a successful transition between the two pieces of equipment should be feasible. To prevent coarse particles from entering the stirred mill and causing critical build-up, either a screen or air classifier, is required. Since HPGR product tends to form a cake, depending on the competency, either an impact style de-agglomerator or wet screening is required. HPGRs operating in closed circuit to achieve a fine product cut size can be found in the cement industry. An example of this circuit design is shown in Figure An impact style de-agglomerator is employed to break up flake and an air classifier is used to prepare a fine product for ball milling (Aydogan et al., 2006). Figure 2-13 HPGR Flowsheet for Fines Production in the Cement Industry (Aydogan et al., 2006) The Sukhoy gold plant in Russia, shown in Figure 2-14, operates an HPGR in closed circuit with a 1.4mm screen and utilizes a scrubber for product de-agglomeration. The main challenge experienced in operating this circuit, was proper control of the moisture content in the HPGR feed (Gruendken et al., 2010). 38

51 Figure 2-14 HPGR Flowsheet the Sukhoy Gold Plant (Gruendken et al., 2010) Wang et al. (1998) and Wang et al. (2006) performed testing on an HPGR / stirred mill combination using lab-scale equipment. Calcium carbonate and limestone wet filter cake (<150µm) were subjected to multiple passes through a lab-scale HPGR and processed through a horizontal stirred mill or attrition mill. Increasing the number of passes through the HPGR resulted in improved throughput and size reduction in the stirred mill. The authors concluded that the breakage mechanisms in HPGR comminution can lead to a subsequent increase in the breakage rates in stirred milling. Valery and Jankovic (2002) proposed the first concept of a combination HPGR / stirred mill circuit in a study examining the need for a reduction in the energy requirements of comminution. Simulating results for a more energy efficient circuit, a high-intensity blasting, two stage HPGR / Vertimill circuit (refer to Figure 2-15) was compared to a conventional blasting, SAG / ball mill circuit. The simulation results predicted an energy savings of 45%, but no actual testwork was conducted. 39

52 Figure 2-15 Example of an HPGR / Stirred Mill Circuit (Valery and Jankovic, 2002) Figure 2-16 A Proposed HPGR / IsaMill TM Circuit (Pease, 2007) Pease (2007) presented the concept of an HPGR / IsaMill TM circuit in his discussion of coarse stirred milling at McArthur River (refer to Figure 2-16). No testing was carried out but Pease predicted that this circuit could be an example of comminution flowsheet design of the future. 40

53 Ayers et al. (2008) described the first operation of an HPGR / IsaMill TM circuit using pilot-scale equipment. The authors documented Anglo Platinum s research into applying the IsaMill TM to coarser feed applications. A test rig was set up, incorporating two 5tph HPGRs in series with screens (refer to Figure 2-17). Figure 2-17 Anglo Platinum s HPGR Test Circuit (Ayers et al., 2008) A continuously operating circuit was establishing using the coarse HPGR in closed circuit with a dry screen, followed by wet screening of the undersize, at a cut size of 850µm (refer to Figure 2-18). The screen product was fed to an M250 IsaMill TM operating with 3.5mm MT1 ceramic grinding media. With an f80 of 300µm and a product p80 of 45µm, the IsaMill TM circuit achieved 1.3tph, with a specific energy consumption of 75kWh/t and a total circuit energy consumption of 80kWh/t. Another circuit tested in this research project incorporated a ball mill before the IsaMill TM (refer to Figure 2-18) and resulted in improved IsaMill TM circuit performance, but at a higher total circuit specific energy consumption of 85kWh/t. In both cases the optimal media size was not used and improved circuit energy requirements could have been achieved with an increased media size. 41

54 Figure 2-18 Anglo Platinum s HPGR / Stirred Mill Testing Flowsheets (Ayers et al., 2008) The authors concludes with examples of future circuit arrangements to be tested by Anglo Platinum, but as of the writing of this thesis, no published literature summarizing this work can be found in the public domain. 42

55 2.4 Literature Summary The HPGR can achieve improved energy efficiency over a SAG mill due to the application of inter-particle breakage, and the ability to transfer input energy directly to the material via the grinding rolls (Section 2.1.4). A stirred mill can achieve improved energy efficiency over a ball mill, at fine particle sizes, due to reduced energy requirements associated with utilizing a centrally-rotating shaft and the ability to grind efficiently with high speeds and small grinding media (Section 2.2.5). The incorporation of these two energy-efficient comminution devices could result in an overall reduction in the specific energy requirements for comminution. HPGR product, with a high percentage of fines and micro-cracks, could successfully be transferred to a stirred mill circuit with two successive passes through the HPGR. Since very few operating examples were found (Section 2.3), pilot-scale testing would be required to successfully determine appropriate design criteria. Overall, combining an HPGR and a stirred mill to produce a more energy-efficient circuit has the potential for a wide variety of processing and operating advantages. With the benefits of lower operating costs related to reduced energy consumption and operation of an open-circuit grinding configuration, coupled with the flexibility available to grind efficiently to finer particle sizes with inert grinding media, this combination has the potential to be the future for energyefficient comminution. 43

56 3 Experimental Procedure This chapter describes the methodology and the equipment used to address the objectives of this research. The main objective was to examine the technical feasibility of combining an HPGR and a stirred mill into a novel flowsheet. To achieve this objective, a pilot-scale testing program was carried out on a copper-nickel sulphide ore from Teck Limited s Mesaba deposit in Minnesota. Evaluation of the potential energy benefits for the proposed circuit design required a basis for comparison. Lab and simulation work was carried out on two alternate comminution circuits, a cone crusher / ball mill circuit and an HPGR / ball mill circuit. Results from this study were used to draw conclusions on which of the three circuits required the lowest specific energy consumption for comminution. 44

57 3.1 Definition of Comminution Circuits The three circuits examined for the energy comparison study were: a cone crusher / ball mill circuit, an HPGR / ball mill circuit and the novel HPGR / stirred mill circuit. The feed size to each circuit was fixed at an f80 of 21mm, and a product p80 of 75µm was chosen as a suitable feed size for flotation. The circuits were evaluated solely on the power consumed per tonne of material in order to achieve an equivalently sized product from an equivalently sized feed. Energy requirements of material handling equipment such as conveyors, pumps and screens were not taken into account. The approach in all cases was to determine an appropriate set of design criteria for each flowsheet and to calculate the specific work index for each stage of comminution based on the work index determined and the transfer sizes selected Cone Crusher / Ball Mill Circuit The first circuit examined was a cone crusher / ball mill circuit, typically found in a three-stage crushing flowsheet. This circuit was the industry standard for hard-rock comminution prior to the establishment of SAG mill technology. The circuit comprised of a cone crusher in closed circuit with a screen followed by a ball mill in closed circuit with a cyclone. The flowsheet of this circuit is shown in Figure 3-1. Data for the circuit was generated from a combination of Bond grindability testing and simulation using JK SimMet software. Feed f80 = 21mm Cone Crusher Ball Mill Product p80 = 75µm 4mm Screen Cyclone Figure 3-1 Cone Crusher / Ball Mill Flowsheet 45

58 3.1.2 HPGR / Ball Mill Circuit The second circuit examined was an HPGR / ball mill circuit. This circuit mimics the standard HPGR comminution flowsheet currently being used in the hard-rock mining sector (refer to Section 2.1.5). The circuit comprised of a high pressure grinding roll in closed circuit with a screen followed by a ball mill in closed circuit with a cyclone (refer to Figure 3-2). Data for this circuit was generated using a combination of HPGR pilot-scale testing, Bond grindability testing and simulation using JK SimMet software. Feed f80 = 21mm HPGR Ball Mill Product p80 = 75µm 4mm Screen Figure 3-2 HPGR / Ball Mill Flowsheet Cyclone. For HPGR pilot-scale evaluation, tests were carried out to assess the influence of different process parameters on comminution performance. These tests included the variation of specific pressing force, roller speed and feed moisture content, as well as closed-circuit testing with a 4mm screen. Data from this study was entered into JK SimMet to model fit an appropriate HPGR model for Mesaba ore. The T10H and HPGR power coefficient model parameters were fitted using the procedure outlined by Daniel and Morrell (2004). After calibration of the HPGR model, simulation was carried out for the HPGR / ball mill circuit HPGR / Stirred Mill Circuit The final circuit tested comprised of two stages of HPGR followed by a horizontal stirred mill. Since no operating examples were found in the literature, pilot-scale testing on both pieces of equipment was performed to determine appropriate transfer sizes between each stage of 46

59 comminution. From these tests, an appropriate circuit design and layout was generated for comparison to the above mentioned circuits. The appropriate transfer size between each stage of HPGR comminution was evaluated using two separate flowsheet options. Option A examined the first-stage HPGR in open circuit (Figure 3-3), while Option B examined the first-stage HPGR in closed circuit with a 4mm screen (Figure 3-4). Each option was evaluated to assess how a change in transfer size between HPGRs affected specific throughput, specific energy consumption and size reduction. f80 = 21mm Feed HPGR HPGR Stirred Mill Product p80 = 75µm Fine Screen Figure 3-3 HPGR / Stirred Mill Flowsheet (Open Circuit) f80 = 21mm Feed HPGR HPGR Stirred Mill Product p80 = 75µm 4mm Screen Fine Screen Figure 3-4 HPGR / Stirred Mill Flowsheet (Closed Circuit) For determination of transfer size between the second-stage HPGR and the stirred mill, three different top sizes (355µm, 710µm and 1.2mm) were evaluated. These particle sizes tested the limits for stirred mill grinding and an evaluation was based on the specific energy requirements 47

60 for comminution and whether the stirred mill could operate and grind effectively. Once a transfer size of 710µm was selected, pilot-scale testing was performed to generate data for the circuit. 48

61 3.2 Sample Description The Mesaba copper-nickel deposit is located in the Mesabi Range of the Duluth intrusive complex located in North-eastern Minnesota (refer to Figure 3-5). This complex is comprised of mafic volcanics (tholeiitic basalt) with layered intrusions of primarily a gabbro-troctolite composite (Minnesota Geological Survey, 2010). Mineralogy of the Mesaba deposit comprises mainly of massive and disseminated sulphides with the main minerals of interest being chalcopyrite (copper), cubanite (copper) and pentlandite (nickel). The inferred resource stands at 700Mt, with a grade of 0.46% Cu and 0.12% Ni (Infomine, 2001). Figure 3-5 Geographic Location of Mesaba (Mayhew et al., 2009) The samples used for this study were originally excavated as part of a bulk sample taken in 2001 after Teck Limited s acquisition of the property. The majority of the sample was used to create a bulk flotation concentrate for hydrometallurgical testing, while the remaining material was kept stored on site. As part of hydrometallurgical testing, the head grade of the bulk sample was determined to be 0.32% Cu and 0.12% Ni (Teck, 2010). In December 2009, approximately 5 tonnes of sample, at nominally 100% minus 100mm, was shipped to UBC. The material was screened and crushed in a laboratory jaw crusher to 100% minus 32mm and homogenized and split into sixteen 45 gallon drums using a rotary sample splitter. A representative sample was taken for size distribution, bulk density and moisture 49

62 content determination. A moisture content of 1% and a bulk density of 2.16t/m 3 were established for the ore. The particle size distribution of the sample is shown in Figure 3-6. Figure 3-6 Mesaba Feed Size Distribution 50

63 3.3 Equipment The following section describes the main pieces of test equipment used and the methodology used for calculating specific energy consumption High Pressure Grinding Roll HPGR testing was conducting using a pilot-scale unit manufactured by Koeppern. The pilot unit is custom made for obtaining design information for sizing and selection of industrial-scale units. Table 3-1 summarizes the technical data provided by Koeppern for the machine. Table 3-1 HPGR Machine Specifics Roller Diameter 750 mm Roller Width 220 mm Press Drive Dual Output Shaft Gear Reducer Feed System Gravity Wear Surface Hexadur WTII Installed Power 200 kw Maximum Pressing Force 1800 kn Maximum Specific Pressing Force 8.5 N/mm 2 Variable Speed Drive up to 40 RPM (1.55 m/s) Experimental data was recorded every 200ms through the programmable logic controller (PLC) data logger and downloaded to a laptop. The computer system measures: time, roller gap (left and right), pressing force (left and right) and power draw. A picture of the HPGR pilot unit is shown in Figure

64 Figure 3-7 Pilot-Scale HPGR Installation A pilot test with the HPGR comprises the crushing of one 45 gallon drum of material (~375kg). The material is loaded into a feed hopper with the use of an overhead crane and drum tipper. Once the machine conditions are stabilized, the slide gate of the feed hopper is opened and the test begins. The material flows with the aid of gravity through the HPGR rollers and drops on to the product conveyor located below the rolls. Using the equations presented in Section 2.1.3, specific throughput and specific energy consumption are then determined for the test. Since the HPGR does not grind uniformly across the roller width, a splitter gate is installed on the end of the product conveyor to separate the product into centre, edge and waste streams. The centre portion is finer than the edge portion and during testing a particle size distribution is performed on each to accurately predict size distributions for full-scale operations. For square rollers, where roll diameter is equal to roll width, the proportion of centre and edge product is observed to be approximately 85% centre and 15% edge for industrial units. All of the HPGR product size distributions presented in Chapter 4 account for this through scaling of centre and edge size distributions at a ratio of 85:15. Material collected during unstable operation, initial response and material run-out periods, was designated as waste material and only material which has been crushed during stable press operation was collected for analysis. 52

65 3.3.2 Horizontal Stirred Mill Stirred mill testing was carried out using Netzsch s M20 horizontal stirred mill. The mill has a capacity of 20 litres and installed with an 18.6kW motor. The mill was upgraded to include a new mechanical seal, updated grinding disc configuration and the replacement of a hand-crank, variable-speed pulley system with a fixed pulley system and Variable Frequency Drive (VFD). The installation of the VFD allowed for direct readings of mill power and mill speed. To monitor the mill, sensors were installed for feed pressure, and both feed and product temperature. A PLC interface and data logger was also installed to control the mill settings and record all important mill parameters during testing. A picture of the upgraded mill is shown in Figure 3-8. The mill configuration, including grinding disc design, was based on recommendations from Xstrata Technology and allowed for the ability to scale-up results to what would be expected for industrial units. Figure 3-8 M20 Stirred Mill Installation A Watson-Marlow and Bredel SPX 25 hose pump and corresponding VFD were used to feed the mill. The pump has a capacity of 25L/min and was designed to handle viscous slurries. The installation of a VFD for the 1.5kW pump motor allowed for accurate monitoring and control of mill flow rate. 53

66 The mixing system was comprised of two 180L-capacity mix tanks with corresponding 250W variable speed agitators and it was designed to mix slurries at upwards of 60% solids with a particle top size as coarse as 1.2mm. The piping system for the circuit was setup so that each mix tank could easily be switched from product to feed with little hesitation. The final setup is shown in Figure 3-9. Figure 3-9 M20 Stirred Mill with Mixing Tanks For testing of the stirred mill energy requirements, a graph of specific energy consumption and p80 grind size was generated. This graph, known as a signature plot, is the common method used in industry for accurate sizing of full-scale IsaMills TM and has a scale-up ratio of 1:1 (Gao et al., 1999). The procedure entails running the material through the mill a select number of times and recording the energy requirements and product size after each pass. The passes are carried out consecutively in order to observe the energy consumption as the size of the product decreases. The results provide a series of points plotted on a log log graph that shows the relationship between energy input and product size (p80). 54

67 Particle sizing for this work was done using a Malvern Mastersizer This laser sizing equipment utilizes the principle that grains of different sizes diffract light at different angles; a decrease in size produces an increase in diffracted angle. This equipment has become the standard for analyzing size ranges unrealistic for screening (Larson et al., 2008) Vibrating Screen All screening work carried out for HPGR closed-circuit testing was performed using a SWECO Vibro-Energy Separator. This vibrating screen, model ZS40, is equipped with a 373W motor and a counterweight system to produce both vertical and horizontal vibrating motion. The screener is equipped with 1m diameter wire mesh screens. A picture of the equipment is shown in Figure Figure 3-10 ZS40 SWECO Vibrating Screen Bond Test Ball Mill Energy requirements for ball mill grinding were determined using Bond ball mill work indices for cone crusher and HPGR product. Representative samples were screened at minus 3.35mm and processed through a standard Bond ball mill measuring 305mm in length and 305mm in 55

68 diameter, with a 285 ball charge weighing 20,125 g (refer to Figure 3-11). Testing was carried out using the standard Bond Ball Mill Grindability Test procedure developed by Bond (1961). For the crushing work index, insufficiently sized material was available to perform impact testing; therefore a traditional approach was taken and the Bond work index was used. The resulting indices were then used with the Bond equation to calculate specific energy consumption for both crushing and grinding. Figure 3-11 Bond Test Ball Mill 56

69 4 Testing and Simulation Results The following chapter summarizes the results obtained for each circuit described in Section 3.1. Simulation and lab results from each circuit are presented and the specific energy consumption for comminution is calculated. 57

70 4.1 Cone Crusher / Ball Mill Circuit Results Flowsheet Simulation The specific energy consumption of comminution for the circuit was determined with a flowsheet developed using JK SimMet software. The circuit was designed for 250tph capacity and equipment was sized based on a product p80 of 75µm. Table 4-1 summarizes the equipment sized for the circuit. Table 4-1 Equipment Selection for Cone Crusher / Ball Mill Circuit Cone Crusher Closed Side Setting (mm) 2.8 Re-circulating Load (%) ~30% Product Screen Aperture Size (mm) 4 Ball Mill Diameter (m) 5 Length (m) 10 Critical Speed (%) 70% Media Charge (%) 40% Media Top Size (mm) 35 Re-circulating Load (%) ~250% Hydrocyclones Quantity 6 Cyclone Diameter (mm) 420 Inlet Diameter (mm) 175 Vortex Finder Diameter (mm) 150 Apex (Spigot) Diameter (mm) 113 Length (mm) 500 Cone Angle (degree) 20 A flowsheet was simulated to determine the appropriate transfer size between the cone crusher and the ball mill. The flowsheet generated in JK SimMet is shown in Figure 4-1. Simulation results determined that the appropriate transfer size between the crushing circuit and the ball mill circuit would be 80% passing 2.12mm. For a summary of JK SimMet results refer to Appendix A. 58

71 Figure 4-1 Cone Crusher / Ball Mill JK SimMet Flowsheet 59

72 4.1.2 Specific Energy Calculations To calculate the overall specific energy consumption for the cone crusher and ball mill, work indices were determined for the material. In the case of the cone crusher, no material was available for the size requirements, 50-75mm, necessary to perform impact testing (Tavares and Carvalho, 2007); therefore a traditional approach was taken and the Bond ball mill work index was used. Locked-cycle testing was performed using two sieve sizes (106µm and 150µm), to allow for the comparison of different product sizes. The results for the work indices of the circuit are shown in Table 4-2. For a complete breakdown of results refer to Appendix B. Table 4-2 Bond Work Indices for Cone Crusher / Ball Mill Circuit Ball Mill Work Index (kwh/t) 150 µm µm 16.5 Using the Bond work indices and the transfer size determined in Section 4.1.1, the theoretical energy requirements were calculated for the proposed circuit using the Bond equation (Bond, 1961). W = 10 BWi 1 p80 1 (5) f80 Where: W (kwh/t) = Specific Energy Consumption BWi (kwh/t) p80 (µm) f80 (µm) = Bond Work Index = 80% Passing Product Size = 80% Passing Feed Size Calculation of the cone crusher energy requirements used the Bond work index at a sieve size of 150µm. This coarser screen size provided a lower estimate for the energy requirements of a cone crusher and provides a best-case scenario for the crushing circuit. Calculation of the ball mill energy requirements used the Bond work index at a sieve size of 106µm. The final product for the circuit was set at a p80 of 75µm and the Bond work index, at a sieve size of 106µm, better reflects the grinding energy requirements to grind to this finer particle size. 60

73 Cone Crusher Circuit W = kwh t 1 1 2,120µm 21,000µm Specific Energy Consumption for Crushing = 2.36kWh/t Ball Mill Circuit W = kwh t µm 2,120µm Specific Energy Consumption for Grinding = 15.47kWh/t TOTAL ENERGY CONSUMPTION = 17.83kWh/t 61

74 4.2 HPGR / Ball Mill Circuit This section summarizes the testing and simulation work carried out to determine the specific energy consumption required for an HPGR / ball mill circuit HPGR Pilot-Scale Testing Section stated that the only reliable method for assessing a material s response to HPGR comminution, and determining scalable operating parameters, is to perform pilot-scale testing. The first parameter evaluated was the appropriate specific pressing force that achieved the optimum balance of specific throughput, specific energy consumption and size reduction. Once identified, testing was done using this pressing force to assess the effect of different operating conditions on HPGR performance. A complete summary of HPGR operating data, including particle size distributions, can be found in Appendix C. Identifying Specific Pressing Force Four initial tests were done to determine the effect of specific pressing force on the material. Pressures of 2N/mm 2, 3N/mm 2, 4N/mm 2 and 5N/mm 2 were chosen and comparisons were made with respect to product size, net specific energy consumption and m-dot. The feed conditions for each test are shown in Table 4-3. All tests were performed at a roller speed of 0.75m/s. Table 4-3 Feed Conditions for Pressing Force Tests Moisture Content 2.5% Bulk Density 2.16 t/m 3 f mm f mm The comparison of product particle size at different specific pressing forces is shown in Figure 4.2. As the pressing force increased, both the p50 and p80 decreased, although the effect on p80 was more pronounced than the effect on p50. This result is due to an increased force being exerted on the particles as they flow through the rolls. An increased force would promote increased breakage and the effect would be more pronounced on larger sized particles, hence the steeper trend for p80. 62

75 Figure 4-2 Comparison of Specific Pressing Force and Product Size The comparison of specific throughput (m-dot) at different specific pressing forces is shown in Figure 4-3. As the pressing force increased, the specific throughput decreased. This trend is due to the gap between the rollers decreasing slightly with increasing pressing forces, resulting in the reduction of throughput in the machine. Figure 4-3 Comparison of Specific Pressing Force and Specific Throughput 63

76 The comparison of specific energy consumption at different specific pressing forces is summarized in Figure 4-4. As the pressing force increased, the energy consumption also increased. This is typical of the process because more energy is being transmitted into the material at higher pressures. Figure 4-4 Comparison of Specific Pressing Force and Specific Energy Consumption A specific pressing force of 4N/mm 2 was selected for the remainder of pilot-scale testing. The results indicated that a pressing force of 4N/mm 2 provided a fine balance between energy consumption and size reduction without a significant change in specific throughput. Variation of Moisture Content Section described that moisture can have an adverse effect on HPGR comminution. Testing was performed to assess the effect of moisture on size reduction, specific throughput and energy consumption. Both drier (1%) and wetter (5%) tests were performed at the selected specific pressing force (4N/mm 2 ) with a roller speed of 0.75m/s. The effect on size reduction is shown in Figure 4-5. An increase in moisture tends to increase the product size, although not to a significant degree. This trend may result from the presence of moisture causing slippage on the rolls and a slight decrease in the effectiveness of the compression bed. 64

77 Figure 4-5 Comparison of Moisture Content and Product Size The comparison of specific throughput (m-dot) at different feed moisture contents is shown in Figure 4-6. An increase in moisture content had a drastic effect on the specific throughput, dropping it to ts/hm 3. This may result from moisture acting as a lubricant, decreasing the frictional forces between the material and the roll surface, and decreasing the operating gap. Figure 4-6 Comparison of Moisture Content and Specific Throughput 65

78 The comparison of specific energy consumption at different feed moisture contents is shown in Figure 4-7. An increase in moisture content caused an increase in energy consumption. This may be caused by the wet material requiring more power to draw it through the rolls, coupled with the decreased operating gap, reducing flow through the rolls. Figure 4-7 Comparison of Moisture Content and Specific Energy Consumption The results indicate that increased moisture content had a negative effect on HPGR comminution performance and drier material produced better results. Unfortunately in operations, dry material requires a complex dust suppression system, and therefore, a compromise on moisture content must be made. The remainder of pilot-scale tests were conducted at 2.5% moisture as a compromise. Varying Roller Speed The effect of roller speed on comminution performance was assessed using two tests at higher (0.9m/s) and lower (0.6m/s) speeds. The effect on size reduction is shown in Figure 4-8. The graph shows that roller speed has very little effect on size reduction. 66

79 Figure 4-8 Comparison of Roller Speed and Product Size The comparison of roller speed with specific throughput is shown in Figure 4-9. An increase in roller speed caused an increase in throughput. This should be expected, since a faster rotation of the rolls causes an increase in the amount of material that can be processed in a given time interval. Figure 4-9 Comparison of Roller Speed and Specific Throughput 67

80 The effect on specific energy consumption at different roller speeds is shown in Figure The change in roller speed had no effect on the specific energy consumption. Figure 4-10 Comparison of Roller Speed and Specific Energy Consumption Closed Circuit Testing To test the effect of closed-circuit operation, locked-cycle testing was conducted using a 4mm screen. Material was processed through the HPGR at 4N/mm 2 and the product screened at 4mm using a SWECO 1m vibrating screen. Using the product size distributions from testing, the percentage of minus 4mm was calculated (at 90% screening efficiency) and then combined with fresh feed and re-run through the HPGR. This process was repeated two more times to simulate closed-circuit operation. Results were generated to determine size reduction, specific throughput and specific energy consumption for each cycle. The resulting product size for each cycle is shown in Figure The chart shows that the introduction of a re-circulating load decreased the product size and began to stabilize by cycle four. 68

81 Figure 4-11 Product Size for Closed Circuit Testing The effect on specific throughput for closed-circuit testing is shown in Figure Closedcircuit operation had little effect on specific throughput. The variation between each cycle can probably be attributed to testing error. Figure 4-12 Specific Throughput for Closed Circuit Testing 69

82 The results for specific energy consumption are displayed in Figure The closed-circuit testing had little to no effect on specific energy consumption. Figure 4-13 Specific Energy Consumption for Closed Circuit Testing Results from the last cycle of testing are summarized in Table 4-4. These results will be used for energy calculations, as well as the experimental data required for model fitting with JK SimMet. Table 4-4 Results for Cycle Four of Closed Circuit Testing F mm F mm p mm p mm Percentage Passing -4 mm 67.4% Net Specific Energy Consumption 1.45 kwh/t (-4 mm) Net Specific Energy Consumption 2.15 kwh/t Specific Throughput 304 ts/hm 3 70

83 4.2.2 Flowsheet Simulation Using the results from the last cycle of HPGR closed-circuit testing, model fitting of an HPGR circuit was performed using JK SimMet. The T10H and HPGR power coefficient model parameters were fitted using the model fit tool in JK SimMet. This tool uses an iterative function to fit experimental data to simulated data by adjusting model parameters until a correlation can be achieved. The T10h and HPGR power coefficient parameters relate to the breakage mechanisms in the compression zone of the HPGR and the product size for closedcircuit testing was used as the experimental data. The procedure used for calibrating the HPGR model was outlined by Daniel and Morrell (2004). The resulting model fit was able to simulate a product size distribution similar to the one generated experimentally. Once an HPGR model was calibrated for use with Mesaba ore, a flowsheet was designed for 250tph capacity with a product p80 of 75µm. The equipment sized for the flowsheet is summarized in Table 4-5. Table 4-5 Equipment Selection for HPGR / Ball Mill Circuit High Pressure Grinding Roll Roller Diameter (mm) 1,200 Roller Width (mm) 1,000 Re-circulating Load (%) ~45% Product Screen Aperture Size (mm) 4 Ball Mill Diameter (m) 5 Length (m) 9.1 Critical Speed (%) 70% Media Charge (%) 40% Media Top Size (mm) 27.5 Re-circulating Load (%) ~250% Hydrocyclones Quantity 7 Cyclone Diameter (mm) 350 Inlet Diameter (mm) 175 Vortex Finder Diameter (mm) 150 Apex (Spigot) Diameter (mm) 113 Length (mm) 450 Cone Angle (degrees) 20 71

84 A flowsheet was simulated to determine the appropriate transfer size between the HPGR and the ball mill. The flowsheet generated in JK SimMet is shown in Figure Simulation results determined a transfer size of 80% passing 1,6mm between the HPGR circuit and the ball mill circuit. For a summary of JK SimMet results, refer to Appendix A. 72

85 Figure 4-14 HPGR / Ball Mill JK SimMet Flowsheet 73

86 4.2.3 Specific Energy Calculations Section described that one of the benefits of HPGR comminution is particle weakening and the subsequent reduction in energy requirements for ball mill grinding. To confirm this advantage, Bond ball mill work indices were determined for HPGR product at different specific pressing forces. Samples were taken from HPGR centre product and screened at 3.35mm with no additional crushing. As with Bond work indices for cone crusher product, two separate screens sizes (106µm and 150µm) were tested to allow for the comparison of different product sizes. The results, including cone crusher product for comparison, are summarized in Table 4-6 and shown in Figure For a complete breakdown of results, refer to Appendix B. Table 4-6 Summary Bond Ball Mill Work Indices for Cone Crusher and HPGR Product Bond Ball Mill Work Indices (kwh/t) Locked Cycle Cone HPGR Product Screen Size Crusher (µm) Product 3 N/mm 2 4 N/mm 2 5 N/mm Figure 4-15 Summary of Bond Work Indices A reduction in Bond work index was achieved between cone crusher and HPGR product; however the reduction went from 8.8% to 4.8% with a decrease in screen size. Results also showed a reduction in Bond work index with increasing specific pressing force, although this 74

87 effect is reduced with a decrease in screen size. The reduction in screen size may have caused a decrease in the effectiveness of product micro-cracking and a relatively higher amount of energy was required to produce the finer product. Using the results in Table 4-6, coupled with HPGR tests data, specific energy consumption can be calculated for the circuit. HPGR Circuit Results from closed-circuit HPGR testing (Section 4.2.1) produced an HPGR Specific Energy Consumption equal to 2.15kWh/t Ball Mill Circuit Using the Bond work index resulting from a specific pressing force of 4N/mm 2 and a screen size of 106µm, coupled with the transfer size determined in Section 4.2.2, the following energy calculation can be made: W = kwh t µm 1,600µm Specific Energy Consumption for Grinding = 14.2kWh/t TOTAL ENERGY CONSUMPTION = 16.35kWh/t 75

88 4.3 HPGR / Stirred Mill Circuit The HPGR / stirred mill circuit required considerably more pilot-scale testing than the previous two circuits, since very few operating examples could be found in the literature. The results of pilot-scale testing determined the appropriate transfer size between each step of comminution and provided the corresponding specific energy consumption for circuit energy summation The Stirred Mill Circuit Section described that the typical f80 for an IsaMill TM ranges from µm, usually operating with 3.5mm or finer ceramic grinding media. Only a few examples were found of the IsaMill TM grinding a coarser feed size; however these examples did not employ a large enough media size (Ayers et al., 2008; Shi et al., 2009). Design of an HPGR / stirred mill flowsheet requires selection of an effective feed top size for stirred mill grinding. Using properly sized grinding media, transfer sizes were tested to achieve a balance between energy consumption and practicality. Once an appropriate transfer size was selected, additional pilot-scale testing was performed with improved operating conditions in order to assess the specific energy requirements for grinding to a p80 of 75µm. Three feed top sizes were chosen to cover a range from fine to coarse and included 355µm, 710µm and 1.2mm. Each feed size was run through an IsaMill TM to produce a signature plot to compare the energy versus size reduction relationships. Traditionally signature plots are performed using an M4 IsaMill TM, due to a small sample requirement of only 15kg. The results from these tests have been found to accurately scale-up 1:1 to industrial sized units (Curry et al., 2005). Tests are typically conducted with a media top size no larger than 3.5mm, but when running coarse grinding tests, a larger media size is required. Unfortunately size constraints in the mill chamber prevent an M4 from operating effectively with a larger media size. Therefore, when running tests at coarse particle sizes above a p80 of 200µm, signature plots can no longer be run with an M4 unit and testing requires a larger mill to operate effectively with media size above 3.5mm. For analysis of the 355µm sample, an M4 could still be used, but for the 710µm and 1.2mm samples, an M20 was employed. Testing of the 355µm sample was performed using a graded charge (57.5% 3mm, 30% 2mm and 12.5% 3.5mm) of MT1 ceramic grinding media manufactured by Magotteaux International. 76

89 The operating conditions for the test followed the standards laid out in Section and are summarized in Table 4-7. An f80 of 204µm was determined for the feed. This corresponds to a media to mean (80% passing) particle size ratio of 17.5:1. Had the 20:1 ratio suggested by Mankosa et al. (1986) been applied, the media top size would have been 4mm. For a complete breakdown of test results, refer to Appendix D. Table 4-7 Test Conditions for the 355µm Signature Plot f µm Feed Wt. 15 kg Solids Content 43% Flow Rate 2.5 L/min Media Volume 70% Mill Speed 1,215 RPM A graph of the resulting signature plot is shown in Figure The graph shows that grinding to a product p80 of 75µm required 14.17kWh/t. Figure 4-16 Signature Plot for Top Size Testing of 355µm Testing of the 710µm sample was performed using a composite of two types of media: 70% by volume 8mm diameter ceramic media manufactured by Rojan Advanced Ceramics and 30% by volume 3mm diameter MT1 ceramic media manufactured by Magotteaux International. In retrospect, this mixture may not have been ideal for testing. The media top size appears to 77

90 have been oversized for the application, as well as a combination of two different types of ceramic media, with different sizes and SGs, may not have been as efficient as a graded charge comprising of a single type of media. The test conditions used for the signature plot are shown in Table 4-8. An f80 of 321µm corresponds to a media to particle size ratio of 25:1. Had the 20:1 ratio suggested by Mankosa et al. (1986) been applied, the media top size would have been 6.4mm. For a further breakdown of results, refer to Appendix D. Table 4-8 Test Conditions for the 710µm Signature Plot f µm Feed Wt. 75 kg Solids Content 40% Flow Rate L/min Media Volume 70% Mill Speed 1,169 RPM The resulting signature plot is shown in Figure The graph shows that the first pass produced a very fine product (p80 of 38µm) and that extrapolation was necessary for calculating the specific energy consumption to grind to a p80 of 75µm. With extrapolation, an estimated specific energy consumption of 14.72kWh/t was required. Net SPecific Energy Consumption (kwh/t) y = x R² = p80 p98 y = 10469x R² = Particle Size (µm) Figure 4-17 Signature Plot for Top Size Testing of 710µm 78

91 For testing the 1.2mm sample, the same ratio of 8mm Rojan Ceramic media and 3mm Magotteaux media was used. This testing was expected to approach the limits of feasibility for coarse grinding in a stirred mill, and as a result, numerous tests were conducted at a variety of test conditions. Since coarse particles may have trouble passing through the dynamic classifier and exiting the mill, a test was conducted to assess whether removal of dynamic classifying pegs (refer to Figure 4-18) helped improve coarse grinding. Table 4-9 summarizes the operating conditions for each test. Figure 4-18 Stirred Mill Dynamic Classifier Pegs Six Classifying Pegs (Left) and Three Classifying Pegs (Right) Table 4-9 Summary of Mill Operating Conditions for 1.2mm Testing Test No. f80 (µm) Solids Content (%) Media Volume (%) Mill Speed (RPM) Mill Flow Rate (L/min) Number of Dynamic Classifier Pegs , , , The first test was unable to grind coarse particle sizes with the operating conditions selected. The first pass through the mill recorded a power draw 56% higher than the allowable power rating for the motor. The VFD safety featured shut down the mill at the 75 second mark to prevent damage to the motor. The stirred mill was attempting to grind material from a feed top size of 1.2mm to a product p80 of 50µm in one pass, requiring more power than was available. A summary of key mill parameters is shown in Figure

92 Mill Shaft Speed (rpm) Motor Power (kw), Flowrate (l/m), Product Temp (degrees C) Time (s) Shaft Speed (shaft sensor) Product Temp Flowrate (pump spd) Motor Power Figure 4-19 Summary of Mill Parameters for 1.2mm Test #1 The second test was also unable to successfully grind coarse particle sizes, with similar results to test one. The VFD safety feature shut down the mill at the 2 minute mark because the mill was drawing 26% more power than the motor could handle. A summary of key mill parameters is shown in Figure Mill Shaft Speed (rpm) Motor Power (kw), Flowrate (l/m), Product Temp (degrees C) Time (s) Shaft Speed (shaft sensor) Product Temp Flowrate (pump spd) Motor Power Figure 4-20 Summary of Mill Parameters for 1.2mm Test #2 80

93 The third test was carried out to assess whether removal of the dynamic classifier pegs helped improve coarse grinding. The mill recorded a dramatic drop in power draw during the first pass; however an increase in feed pressure at the 380 second mark resulted in the mill shutting down. The main mill parameters plotted in Figure 4-21 show that during testing, the motor power slowly increased and the shaft speed slowly decreased. This is consistent with a hypothesis that, due to a decrease in energy input, critical sized particles were being retained in the mill; slowly building up until the mill chamber was full, triggering an increase in pressure at the feed inlet Mill Shaft Speed (rpm) Motor Power (kw), Flowrate (l/m), Product Temp (degrees C) Time (s) Shaft Speed (shaft sensor) Product Temp Flowrate (pump spd) Motor Power Figure 4-21 Summary of Mill Parameters for 1.2mm Test #3 For the range of operating conditions tested, grinding a feed size of 1.2mm was not possible; however if a higher rated motor was installed, the possibility of operating at these conditions may have been successful. With the results presented above, a transfer size of 710µm was selected as the most feasible option for an HPGR / stirred mill circuit. Although a transfer size of 355µm produced the lowest energy requirement for grinding to a p80 of 75µm, the savings of 0.5kWh/t was not sufficient to overcome the increased energy requirements for operating an HPGR in closed circuit with a 355µm screen. 81

94 After designating 710µm as the transfer size between the second-stage HPGR and the stirred mill, two additional pilot-scale tests were performed to attempt to reduce the specific energy requirements for the circuit. Operating conditions were chosen to target a specific energy input of 7-9kWh/t per pass through the mill. This would create a coarse first pass for the signature plot, eliminating the problem of extrapolation experienced with the original test. The first grinding disc on the mill shaft was replaced with a spacer (refer to Figure 4-22) to reduce the energy input of the mill. A summary of the revised operating conditions for testing are shown in Table For a complete summary of results, refer to Appendix D. Figure 4-22 Replacement of Grinding Disc with Spacer Table 4-10 Revised Test Conditions for 710µm Signature Plots f µm Feed Wt. 100 kg Percent Solids 57% Flow Rate 20.4 L/min Mill Speed 1,169 RPM Improvements were also made in the selection of grinding media for the tests. A graded charge (40% 4-6mm, 30% 2-4mm, 20% mm and 10% 2-2.2mm) of Cenotec ceramic grinding media was used at a media charge volume of 70%. The ratio of media top size (6mm) to f80 (340µm) was adjusted, representing a ratio of 17.6:1. The results for both tests are shown in Figure The graph shows that substantially reduced energy consumption was achieved with properly sized media and revised operating conditions. 82

95 Figure µm Signature Plot Results with Revised Operating Conditions Testing showed an average specific energy consumption of 9.73kWh/t was required to grind to a p80 of 75µm, an improvement of 34% over the original test. This value was selected to be used for energy calculations of the HPGR / stirred mill circuit. The size measurements used to generate the signature plots in Figure 4-23 were performed using a Malvern Mastersizer For a comparison, pass one product was also sized using screens. Since Malvern sizing is based on volume, while screening is based on weight, results will not be identical. All other testwork performed for this flowsheet relied on size results from screening; therefore a comparison should be made. Malvern and screening comparisons for T1 and T2 are shown in Figure 4-24 and Figure 4-25, respectively. The screening results indicated a finer product than the Malvern results. These results show that the signature plots generated using Malvern sizing, can be considered a conservative estimate for energy consumption, since size results may have been finer using screens. Unfortunately screening is impractical below 38µm (the product size after pass two), so Malvern sizing was used for all stirred mill products in order to remain consistent. 83

96 Figure 4-24 Malvern and Screening Comparison for T1 Figure 4-25 Malvern and Screening Comparison for T The HPGR Circuit Pilot-scale testing was conducted to produce suitable data for the HPGR section of the HPGR / stirred mill circuit. Since size reduction is limited with one stage of HPGR comminution (refer to Section 2.3), design of an HPGR / stirred mill circuit required at least two consecutive stages of 84

97 HPGR comminution to produce a particle size acceptable for stirred milling. With the transfer size between the second-stage HPGR and the stirred mill determined in Section 4.3.1, work was done to determine the appropriate transfer size between each stage of HPGR crushing. Two options were examined to find the appropriate circuit layout. In Option A, the first-stage HPGR was placed in closed circuit with a 4mm screen, while in Option B, the first stage remained open circuit and the second stage accepted product directly from stage one. For Option A, product from closed-circuit testing in Section was processed again through the HPGR at the same roller speed (0.75m/s) and specific pressing force (4N/mm 2 ). The use of the same specific pressing force for second-stage HPGR crushing stems from work performed by Rule et al. (2008), in which they found that no difference was observed when changing the specific pressing force in the second stage of two-stage HPGR crushing. For Option B, fresh feed was processed through two consecutive stages of HPGR comminution using the same roller speed and specific pressing force as Option A. The results for both options are summarized in Table The size distributions for Options A and B are presented in Figure 4-26 and Figure 4-27, respectively. For a complete breakdown of results, refer to Appendix C. 85

98 Table 4-11 Summary of HPGR Results for First Stage Open and Closed Circuit Testing HPGR Stage 1 HPGR Stage 2 OPTION A OPTION B f mm mm f mm 13.7 mm HPGR p mm 7.68 mm HPGR p mm 1.88 mm Circuit p mm 7.68 mm Circuit p µm 1.88 mm Circuit Reduction Ratio Net Specific Energy Consumption 1.45 kwh/t 1.54 kwh/t Percentage Passing 4 mm 67.4% (-4 mm) Net Specific Energy Consumption 2.15 kwh/t Specific Throughput 304 ts/hm ts/hm 3 f mm 7.68 mm f µm 1.88 mm HPGR p mm 2.79 mm HPGR p µm 462 µm Circuit p µm 339 µm Circuit p µm 142 µm Circuit Reduction Ratio Net Specific Energy Consumption 1.2 kwh/t 1.23 kwh/t Percentage Passing 0.71 mm 71.3% 56.5% (-0.71 mm) Net Specific Energy Consumption 1.68 kwh/t 2.18 kwh/t Specific Throughput 236 ts/hm ts/hm 3 TOTAL SPECIFIC ENERGY CONSUMPTION 3.83 kwh/t 3.72 kwh/t 86

99 Figure 4-26 Particle Size Distributions for Option A Figure 4-27 Particle Size Distributions for Option B 87

100 Operating the first stage of HPGR crushing in open circuit required less energy compared with operating in closed circuit with a screen. If looked at strictly from an energy perspective, very little difference is gained choosing one circuit over the other, but if design and operating factors are considered, the choice of operating the first stage in open circuit becomes the better option. The ability to operate the circuit without a screen allows for the elimination of extra auxiliary equipment such as screens and conveyors, while the absence of an additional stage of wet screening would help to reduce the adverse effects that increased moisture content would have on HPGR performance. Although the increased re-circulating load resulting in the second stage would require an increase in tonnage and machine size, this would be countered by the decreased machine size required for stage one. Overall, the reduced complexity offered by open circuit configuration led to selecting this configuration for further testing. Once the open circuit configuration was selected for stage one, additional pilot-scale testing was performed to evaluate how comminution performance would be affected by operating the second sage in closed circuit with a 710µm screen. Testing was conducted in a similar manner to Section Product from Option B was screened at 710µm and a calculated split of oversize was mixed with fresh product from stage one and processed through the HPGR. This procedure was repeated two more times in order to simulate closed-circuit operation. The resulting product size for each cycle is shown in Figure The product size increased slightly with the introduction of a re-circulating load. This is in contrast to the results in Section 4.2.1, where the introduction of a re-circulating load caused a decrease in product size. This increase may have been the result of a finer re-circulating load reducing the breakage within the compressive bed. For a complete breakdown of results, refer to Appendix C. 88

101 Figure 4-28 Product Size for Second Stage Closed Circuit Testing The results for the effect of closed-circuit operation on specific throughput are displayed in Figure The introduction of a re-circulating load had no effect on specific throughput. Figure 4-29 Specific Throughput for Second Stage Closed Circuit Testing 89

102 The results for the effect of closed-circuit operation on specific energy consumption are summarized in Figure As with specific throughput, the introduction of a re-circulating load had no effect on specific energy consumption. Figure 4-30 Specific Energy Consumption for Second Stage Closed Circuit Testing To achieve efficient screening at 710µm for an industrial operation, the practice of wet screening is necessary. Section showed that the introduction of moisture to an HPGR circuit leads to adverse effects on throughput and energy consumption. The effect of moisture on secondstage HPGR crushing was tested using product from the final closed-circuit cycle. The sample was wet screened over a 710µm screen to determine the potential moisture content for oversize in a closed-circuit operation. The saturated oversize, with a measured moisture content of 10.5%, was then used to run an additional closed-circuit cycle. A summary of the results are presented in Table 4-12 To allow for a direct comparison of the effects of wet screening, the results from cycle four (dry) are presented as well. For a complete breakdown of results, refer to Appendix C. As expected, the results show an adverse effect on throughput and energy consumption, although the product size became considerably finer. The data generated for the wet screening cycle represents the worst-case scenario, and thus will be used for the energy calculations for the circuit. 90

103 Table 4-12 Comparison of Wet and Dry Screening Dry Cycle Wet Cycle Feed Moisture Content 2.4% 5.8% F mm 6.41 mm F mm 1.95 mm p mm 2.88 mm p µm 523 µm Percentage Passing -710µm 49.8% 54.8% Net Specific Energy Consumption 1.45 kwh/t 1.96 kwh/t (-710µm) Net Specific Energy Consumption 2.91 kwh/t 3.58 kwh/t Specific Throughput 304 ts/hm ts/hm Circuit Energy Summary The results obtained from pilot-scale testing and the resulting specific energy consumption for the circuit are summarized in Table For a comparison, results from dry and wet screening for second-stage HPGR are included. Results show that the implementation of wet screening would result in a 4.7% increase in specific energy consumption for the circuit. Table 4-13 Summary of HPGR / Stirred Mill Energy Requirements Comminution Circuit Section Reference Specific Energy Consumption with Dry Screening (kwh/t) Specific Energy Consumption with Wet Screening (kwh/t) First Stage HPGR Second Stage HPGR Stirred Mill Total Specific Energy Consumption

104 5 Discussion of Results The following chapter describes research outcomes based on results presented in Chapter 4. With these results, conclusions are reached on the choice of operating parameters for pilotscale testing and what changes could be made for future testing. This chapter will also discuss which circuit required the lowest specific energy consumption for comminution; while preliminary design work is presented of a potential flowsheet for an HPGR / stirred mill circuit. The chapter concludes with a refined testing procedure for future HPGR / stirred mill studies. 92

105 5.1 Assessing Operating Parameters for Pilot-Scale Testing Operating parameters used to conduct HPGR and stirred mill testing were chosen based on a review of the literature and recommendations made from industry. This section will review the main parameters used for testing and comment on future variations to be studied HPGR Operating Parameters The main operating parameters identified in Section as having the most influence on comminution performance were specific pressing force and feed moisture content. Specific Pressing Force The specific pressing force selected for testing of Mesaba ore was 4N/mm 2. This force provided a balance of size reduction and specific energy consumption without drastically changing the specific throughput. This force also approaches the limits for safe operation with studded lining. Morley (2008) indicated that the safe operating range for studding lining is 1-4.5N/mm 2 and anything higher risks damaging the metal studs. Therefore selection of 4N/mm 2 can be safely operated in industrial units and provides good size reduction at low specific energy consumption. The choice to keep specific pressing force constant for second-stage HGPR crushing was based on results presented by Rule et al. (2008) and summarized in Section 2.3, in which varying specific pressing force had little effect on size reduction for second-stage HPGR crushing in a Labwal. Further testing could be done to confirm this conclusion with pilot-scale results. Feed Moisture Content The feed moisture content used during testing was 2.5% by weight. The moisture content was selected taking into account results obtained in Section Lower moisture content resulted in improved HPGR comminution performance; however dust suppression must be accounted for. The optimum results occurred at 1%, when specific throughput was high, specific energy consumption was low and product size was finer. With increasing moisture content, a decrease in comminution performance was observed. Unfortunately operating at 1% moisture is unfeasible due to excessive dust generation. In both pilot-scale testing and industrial 93

106 operations, dust suppression is an issue and higher moisture content mitigates these risks. The selection of 2.5% provided a good balance in comminution performance and a substantial reduction in dust generation. For industrial applications, moisture content is rarely as low as 1% and values above 2-3% are more realistic. Moisture content testing for second-stage HPGR crushing (Section 4.3.2) showed a considerable increase in specific energy consumption when addition of saturated oversize was used for closed-circuit testing. The effect of wet screening resulted in an increase in moisture content to 5.8%. The data generated for this test provide an indication of what is expected when an HPGR operates in closed circuit with wet screening. Further testing could be done to examine a full range of moisture contents for second-stage HPGR crushing Stirred Mill Operating Parameters The main operating parameters discussed in Section for stirred milling were feed density, flow rate, mill speed and media size. Section found that mill geometry also has an effect on grinding. Feed Density The average solids content used for 710µm stirred mill testing was ~57% by weight and the SG of the ore was measured to be 3.0. This resulted in a solids content by volume of 30.6% and a slurry SG of This solids content was chosen to reduce grinding, by decreasing the amount of energy transmitted per pass through the mill. Normally a solids content of 50% is recommended to achieve optimum results. Larson et al. (2008) claims that operating with higher solids content reduces efficiency due to increased viscosity. For testing Mesaba ore, viscosity issues were negligible because power draw dropped with subsequent passes through the mill. High viscosity would have caused power draw to remain constant or increase with finer size. The sharp product size distribution produced during testing resulted in a lower ultrafines content (<15µm), which allowed solids content to remain high without clogging the mill. Further testing could be performed to determine whether a decrease in solids content has an improved effect on grinding efficiency. 94

107 Mill Flow Rate The flow rate chosen for feeding slurry to the mill was 20.4L/min. This flow rate was chosen to reduce the grinding residence time in the mill. An increased flow rate resulted in a coarser product, a desirable characteristic when generating results for grinding to a p80 of 75µm. Mill flow rate does not have an effect on grinding efficiency because feed and mill properties are not changing. The only change is the speed at which particles travel through the mill. The only circumstances where mill flow rate would affect stirred mill operations are at very high and very low flows. If too high a flow rate was selected, material may have a hard time exiting the mill, resulting in clogging the mill. If too low a flow rate was selected, solids would have time to settle in the feed line, resulting in blockage. The flow rate chosen for testing exhibited neither situation and no further testing is required to assess different flow rates. Mill Speed The speed chosen for testing was set at 1,200 RPM with 1,169 RPM achieved, the maximum attainable speed for the mill with the currently installed motor and drive system. This corresponds to a tip speed of ~11m/s. Larson et al. (2008) found that mill speed has little influence on grinding efficiency at high speeds. The authors claim that mill speed affects the power draw of the mill and the energy transmitted to the material but does not affect the grinding efficiency. Further testing could be done to validate this claim for coarse stirred milling by running tests at higher and lower speeds. Media Size The size of grinding media used in testing was a graded charge with a top size of 6mm. The recommended ratio for fine grinding is 20:1 media size to mean particle size (80% passing). The f80 for 710µm testing was 340µm, corresponding to a ratio of 17.6:1. Had a 20:1 ratio been used, the media would have had a top size of 6.4mm. Since only one media size was used, further testing could be done to confirm an optimum bead size. Different media top sizes could be tested, ranging from 3-8mm, improving the media selection guidelines for coarse stirred milling. 95

108 Mill Geometry The mill geometry was varied during stirred mill testing by removing the first grinding disc and reducing the number of classifying pegs. The former resulted in a decrease in grinding energy, while the latter resulted in a drop in grinding efficiency. Further work could be done to examine the spacing of grinding discs and the effect on grinding efficiency. Feed Top Size The maximum feed top size tested for coarse stirred milling was 1.2mm. The operating conditions chosen for these tests did not result in effective grinding. Further work could be done to evaluate this top size with a larger-sized motor. With the use of a larger motor size, increased power draw could be achieved and some of the operating conditions selected could be fully explored. Further testing could also be done on an intermediate top size of 1mm. 96

109 5.2 Comparison of Comminution Circuits From the conclusions reached in Chapter 2, it was predicted that the incorporation of both an HPGR and a stirred mill into a novel circuit design, could lead to a reduction in the specific energy requirements for comminution. In Chapter 4, lab and simulation work was done to investigate whether these conclusions were valid. The resulting transfer sizes, operating work indices and circuit layout are summarized in Figures 5-1 to 5-3. Feed f80 = 21mm WI = 2.36 kwh/t t80 = 2.12mm WI = kwh/t Product p80 = 75µm Figure 5-1 Summary Layout for Cone Crusher / Ball Mill Circuit Feed f80 = 21mm WI = 2.15 kwh/t t80 = 1.6mm WI = 14.2 kwh/t Product p80 = 75µm Figure 5-2 Summary Layout of HPGR / Ball Mill Circuit 97

110 f80 = 21mm WI = 1.54 kwh/t Feed t80 = 6.4mm WI = 3.58 kwh/t t80 = 0.34mm WI = 9.73 kwh/t p80 = 75µm Product Figure 5-3 Summary Layout of HPGR / Stirred Mill Circuit Using the results presented above, a bar graph was generated to summarize the specific energy consumption for each stage of comminution (refer to Figure 5-4). The graph shows that the HPGR / stirred mill circuit required the lowest specific energy consumption and achieved a reduction of 9.2% and 16.7% over the cone crusher / ball mill and HPGR / ball mill circuits, respectively. The results presented in this thesis were obtained from pilot-scale testing on a single test for each operating variable. Since pilot-scale testing required a significant quantity of material per test, 350kg for HPGR and 100kg for stirred mill, the reproducibility and standard deviation could not be determined for each changing variable; however some repeatability testing was performed on pilot-scale HPGR testing using 5 homogenized drums. Results showed that specific energy consumption had a standard deviation of and specific throughput a standard deviation of For stirred mill testing, since only two signature plots were generated at similar conditions, the standard deviation could not be calculated and instead the median of 0.23 was considered. The energy figures associated with Bond grindability testing were found to have a standard deviation of , when comparing the three results of HPGR product at a screen size of 106µm. With these results, testing errors were calculated for each circuit at a 95% confidence interval. Table 5-1 summarizes the statistics related to each circuit energy result. 98

111 Figure 5-4 Summary of Specific Energy Consumption for Each Circuit Table 5-1 Statistics Summary of Circuit Energy Values Standard Deviation of the Mean 95% Confidence Interval Sample Set Mean (kwh/t) Standard Deiation Upper Limit Lower Limit Cone Crusher Specific Energy Value Ball Mill Specific Energy Value HPGR Energy Value Ball Mill Energy Value Stage 1 HPGR Energy Value Stage 2 HPGR Energy Value Stirred Mill Energy Value* *Median used intead of standard deviation Total Specific Energy Consumption With 95% Confidence Interval Cone Crusher / Ball Mill Circuit HPGR / Ball Mill Circuit HPGR / Stirred Mill Circuit / / / The error values displayed in Table 5-1 show that the HPGR / stirred mill circuit contained the most potential for a variation in reported results. With the inclusion of testing error, the HPGR / stirred mill circuit still required the lowest specific energy consumption for comminution. The testing error presented here can be considered only an approximation because the results are based on only a few tests and the accuracy of JK SimMet modelling is not accounted for. Although JK SimMet is a sophisticated tool for process simulation; the outputs are still dictated by a mathematical model and are not a result of actual testing. 99