Advances in Comminution - Kowatra

ADVANCES IN COMMINUTION

Published by the Society for Mining, Metallurgy, and Exploration, Inc.

Society for Mining, Metallurgy, and Exploration, Inc. (SME) 8307 Shaffer Parkway Littleton, Colorado, USA 80127 (303) 973-9550 / (800) 763-3132 www.smenet.org SME advances the worldwide mining and minerals community through information exchange and professional development. SME is the world’s largest association of mining and minerals professionals. Copyright ” 2006 Society for Mining, Metallurgy, and Exploration, Inc. All Rights Reserved. Printed in the United States of America. Information contained in this work has been obtained by SME, Inc., from sources believed to be reliable. However, neither SME nor its authors guarantee the accuracy or completeness of any information published herein, and neither SME nor its authors shall be responsible for any errors, omissions, or damages arising out of use of this information. This work is published with the understanding that SME and its authors are supplying information but are not attempting to render engineering or other professional services. If such services are required, the assistance of an appropriate professional should be sought. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Any statement or views presented here are those of the author and are not necessarily those of SME. The mention of trade names for commercial products does not imply the approval or endorsement of SME. ISBN-13: 978-0-87335-246-8 ISBN-10: 0-87335-246-7

Library of Congress Cataloging-in-Publication Data Advances in comminution / edited by S. Komar Kawatra. p. cm. Includes bibliographical references and index. ISBN-13: *978-0-87335-246-8 ISBN-10: 0-87335-246-7 1. Stone and ore breakers--Technological innovations. 2. Crushing machinery--Technological innovations. 3. Mining engineering--Technological innovations. I. Kawatra, S. K. TN510.A38 2006 622'.73--dc22 2005057533

Preface This third international symposium and proceedings, Advances in Comminution, have come at a critical time. Because of rapidly rising energy prices, it is important that the latest information be made available for improving the efficiency of highly energy-intensive comminution processes. The contributors and topics for this third international symposium have been carefully selected to provide a balance between academic and industrial practice so that the reader can readily find information on current best practices and evaluate future industry trends. Two previous symposiums, also organized by the Society for Mining, Metallurgy, and Exploration, were great successes. The first conference was held in 1992, at a time when there was much discussion about switching from traditional rod mill and ball mill circuits to autogenous grinding. The second comminution symposium, held in 1997, focused on initial installations of high pressure grinding rolls (HPGRs). Now, in 2006, the HPGRs are becoming part of hard-rock grinding circuits. They have proven to be a very economical addition to many comminution processes because of lower energy consumption and easy integration into existing conventional systems. The 2006 conference focuses on the dilemma of needing to grind materials to everfiner sizes while maintaining reasonable energy costs. The selection and sizing of stirred mills for regrinding and ultrafine grinding applications do not lend themselves to conventional methodologies; therefore, new approaches are being developed. There is also a great deal of activity directed toward improving ore characterization to predict AG/ SAG mill energy requirements, as well as developing improved models and instrumentation for optimization and control of comminution circuits. Instrumentation, modeling, and control functions in particular have benefited from rapidly advancing computer technology, with calculations that were formerly extremely time-consuming becoming rapid and routine. These advances will keep energy waste to a minimum and will provide the increased energy efficiency needed to maintain ongoing industry success. It is hoped that the symposium and these proceedings will be useful to those who are working toward major advances in industrial practice. Appreciation is extended to members of the organizing committee, who were instrumental in acquiring high-quality papers and reviewing them on very short notice, and to the SME staff, particularly Ms. Tara Davis and Ms. Jane Olivier, for their assistance in organizing the third international symposium and publishing these proceedings.

vii

Contents EDITORIAL BOARD PREFACE PART 1

v

vii

ADVANCED COMMINUTION TECHNOLOGIES

1

High-Pressure Grinding Rolls—Characterising and Defining Process Performance for Engineers

PART 2

PART 3

3

High-Pressure Grinding Rolls—A Technology Review

15

Some Basics on High-Pressure Grinding Rolls

41

High-Pressure Grinding Rolls for Gold/Copper Applications

51

Selection and Sizing of Ultrafine and Stirred Grinding Mills

69

Effects of Bead Size on Ultrafine Grinding in a Stirred Bead Mill

87

Specific Energy Consumption, Stress Energy, and Power Draw of Stirred Media Mills and Their Effect on the Production Rate

99

AG/SAG Mill Circuit Grinding Energy Requirement—How to Predict It from Small-Diameter Drill Core Samples Using the SMC Test

115

COMMINUTION PRACTICES

129

Causes and Significance of Inflections in Hydrocyclone Efficiency Curves

131

Simulation-Based Performance Improvements in the Ispat Inland Minorca Plant Grinding Circuit

149

Determining Relevant Inputs for SAG Mill Power Draw Modeling

161

Cement Clinker Grinding Practice and Technology

169

Extended Semiautogenous Milling: Smooth Operations and Extended Availability at C.M. Doña Ines de Collahuasi SCM, Chile

181

LIBERATION AND BREAKAGE

191

Shell and Pulp Lifter Study at the Cortez Gold Mines SAG Mill

193

Breakage and Damage of Particles by Impact

205

The Rationale behind the Development of One Model Describing the Size Reduction/Liberation of Ores

225

Influence of Slurry Rheology on Stirred Media Milling of Limestone

243

iii

Experimental Evaluation of a Mineral Exposure Model for Crushed Copper Ores

261

Linking Discrete Element Modeling to Breakage in a Pilot-Scale AG/SAG Mill 269

PART 4

Significance of the Particle-Size Distribution in the Quality of Cements with Fly Ash Additive

285

Modeling Attrition in Stirred Mills Applying Statistical Physics

293

MILL DESIGN

307

Design of Iron Ore Comminution Circuits to Minimize Overgrinding

309

Evaluation of Larger-Diameter Hydrocyclone Performance in a Desliming Application

321

Selection and Design of Mill Liners

331

The Importance of Liner Geometry and Wear in Crushing

377

Bond’s Method for Selection of Ball Mills

385

Developments in SAG Mill Liner Design

399

The Gearless Mill Drive—The Workhorse for SAG and Ball Mills 413 Optimizing Hydrocyclone Separation in Closed-Circuit Grinding PART 5

INSTRUMENTATION, MODELING, AND SIMULATION

435

445

Use of Multiphysics Models for the Optimization of Comminution Operations

447

Batu Hijau Model for Throughput Forecast, Mining and Milling Optimization, and Expansion Studies

461

The Use of Process Simulation Methodology in Process Design Where Time and Performance Are Critical

481

Modeling and Simulation of Comminution Circuits with USIM PAC

495

Remote and Distributed Expert Control in Grinding Plants

513

Developments in Sensor Technology for Tumbling Mills

527

Ball Mill Circuit Models for Improving Plant Performance

539

INDEX

547

iv

PART 1

Advanced Comminution Technologies

1

High-Pressure Grinding Rolls— Characterising and Defining Process Performance for Engineers Richard Bearman*

ABSTRACT

High-pressure grinding rolls (HPGRs) are increasingly becoming a part of the hard-rock processing picture through their energy efficiency, the ability to induce microcracks and preferential liberation, coupled with high throughput and high reduction ratio. Given that the machine is still not regarded by many as an off-the-shelf piece of process equipment, there is work required to define guidelines for its use and to provide engineers with tools they can use. This paper examines the current knowledge around the HPGR process performance and explores key relationships available to engineers, whilst considering current approaches to simulation. INTRODUCTION

High-pressure grinding rolls (HPGRs) have struggled for acceptance into the hard-rock mining sector. Many of the issues that restricted their widespread use have now been conquered, but it is still regarded as an “immature” technology. Why is this the case? In contemplating an answer to the issue of the “immaturity,” the status of other accepted technologies must be examined. As an example, the traditional compressionstyle cone-gyratory crushers can be considered. When a plant design is being assembled, every well-equipped engineer will be able to turn to numerous rules of thumb associated with these crushers—even without reference to textbooks or suppliers. The types of rules referenced above include ƒ Product-size distribution will be approximately 80% passing the closed-side setting—

with poor applications dropping to 50%. ƒ Centralized and circumferentially distributed feed is required to extract the best

performance. ƒ Profile and condition of the crushing liners is critical to deliver the best distribu-

tion of energy into the crushing chamber. ƒ Low-bulk-density feeds reduce throughput. ƒ Maximum product bulk density is 1.9 to 2.1 t/m3 for average limestone feedstock. ƒ Secondary applications are power driven, whilst tertiary duties are pressure driven. * Rio Tinto Technical Services, Perth, Western Australia 3

4

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

ƒ Mostly 5%–10% of the feed-size distribution is the maximum less than the closed-

side setting—except with modern cones that are trying to generate interparticle crushing. ƒ Maximum feed size should not exceed 80% of the open-side feed opening. ƒ Feed moistures >4% should be avoided.

Given this type of knowledge, it is easy for the designer to determine the position within the flowsheet and to then calculate the feed rates, type of feed arrangement, and the pre- and postclassification required. Why do these rules of thumb, or guidelines, not exist for HPGRs? There are several reasons for this lack of clarity, namely: ƒ Number and type of applications ƒ Genesis of the HPGR concept ƒ Industry position on technology ƒ Existence of process models

First, there are very few actual, or operating, applications in hard-rock duties. The only hard-rock applications that have been in existence for any length of time are restricted to the diamond and iron ore (pellet-feed) sectors. Another consideration is that the HPGR is a very rare breed of machine, in that its development stemmed from fundamental research. Given the types and focus of early publications, much was made of the nature of the interparticle breakage at the heart of the technology. Obviously, given the ground-breaking nature of the invention, this focus was fully justified, but it led—unfairly—to the HPGR being regarded as an academic device searching for an industry application. The language used about the HPGR, and unfamiliar terms such as “m-dot” (denoting specific throughput), further led to an air of mystique around the HPGR. Was it a crusher or a mill? Its place in the world was unclear. Another element restricting the rate of application was the lack of process models. Simulation is a large part of the flowsheet design exercise and this inevitably requires process models to exist for each piece of equipment. In the case of the HPGR, much of the effort was placed in scale-up procedures. Several organisations did produce process models of HPGRs, but they have been fragmented in their acceptance. Currently, the most complete model approach is that reported by Daniel and Morrell (2004), who have developed an approach from the earlier model of Tondo (1997). It is interesting to note that the Tondo model came out of the first major process study of HPGRs, namely the AMIRA P428 that was completed in 1997. If these points above are added to the naturally conservative stance of the mining industry, this provides a view of why, even after mechanical/wear issues have been overcome, there is still a slow rate of acceptance. As of today, the situation has changed. The features and benefits have become clear to many practitioners, including ƒ Energy efficiency ƒ Preferential liberation at natural grain boundaries ƒ Microcracking and enhanced extraction ƒ Small footprint in terms of throughput and size reduction ƒ Minimal vibration from machine into drive mechanisms and support structure

Of increasing importance is the energy-efficiency issue. It was not too long ago that the mining industry regarded energy consumption as somewhat of a side issue. The Kyoto Protocol and the greenhouse debate changed this view forever (Ruben 2002).

HPGRS—CHARACTERISING AND DEFINING PROCESS PERFORMANCE

5

CRITICAL HPGR PARAMETER S

HPGR roll diameters typically range from 0.5 m to 2.8 m, depending on the supplies, and roll widths vary from 0.2 m to 1.8 m. The aspect ratio of the rolls also varies as a function of manufacturer. Typical HPGR throughput rates range from 20 to 3,000 tph, with installed motor power as high as 3,000 kW per roll. The roll surface is protected with wear-resistant materials, and it has been these that have traditionally stymied HPGR acceptance, but solutions are now in place (Maxton, Morley, and Bearman 2004). When operating an HPGR, the two most important operating parameters are ƒ Operating pressure ƒ Roll speed

The two key operating parameters are inherently linked to the following: ƒ Specific throughput ƒ Specific pressing force ƒ Maximum pressure between the rolls ƒ Specific energy input

Detailed descriptions of the derivation and formulation of the parameters are given in numerous texts, and as such, the following section provides only a précis of the critical formulas, with some examples of actual relationships from testwork. Specific Throughput

The specific throughput, m-dot, is regarded by many as the key parameter for sizing the rolls. Specific throughput is defined as the throughput (tph), divided by the roll diameter (m), roll width (m), and the peripheral roll speed (m/s). For the purposes of brevity, only the equations for this parameter are reported here. Further details are provided in earlier works. (Schönert 1991). Part of its importance is that the equation allows comparison between any size of rolls providing that the surfaces are the same. m• = M/(D u L u u) where M D L u m•

= = = = =

(EQ 1)

throughput rate (tph) roll diameter (m) roll width (m) roll speed (m/s) specific throughput (ts/hm3)

The throughput can also be calculated from the continuity equation as follows: M = L u s u u u Uc u 3.6

(EQ 2)

where s = operating gap (mm) Uc = density of the product cake (t/m3) Combining equations (1) and (2), one obtains: m• = (s/D) u Uc u 3.6

(EQ 3)

For a given material and operating conditions, the gap scales linearly with the diameter of the rolls, and hence the specific throughput can be assumed to be constant.

6

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

250 30 Bar 38 Bar 52 Bar

m-dot, ts/hm 3

230

210

190

170

150 1.40

1.45

1.50

1.55

1.60

1.65

1.70

1.75

Bulk Density, t/m 3

FIGURE 1 Variation in specific throughput as a function of feed-bulk density for various operating pressures using a pilot-scale HPGR

It should be noted that recent work by Daniel (2005) has examined the determination of an equivalent diameter for piston press tests. Daniel proposes D = (Ucf u xc u xd) / ((Ucf u xc ) – (xg u Ug )) where Ucf xc xd xg Ug

= = = = =

(EQ 4)

feed bulk density, lightly compacted initial bed height in piston press displacement of piston final bed height (i.e., operating gap) density of product flake

This relationship has potential to assist in translating piston press results to engineering parameters. Variation in Throughput with Key Variables

Figure 1 shows the variation in the specific throughput as a function of the feed bulk density. The relationship appears to be linear over the range of feeds tested. Given that the specific gravity of the feed material is 2.85 t/m3, it would be unlikely that the loose feed bulk density would exceed 1.8 t/m3; therefore, this graph suggests that the relationship is relevant over a vast majority of cases. It should be noted that throughput is highest at the lowest pressure, with larger changes associated with the all-in (high bulk density) feed types. Figure 2 shows the type of linear increase in specific throughput associated with increasing operating gap. Figure 3 shows a plot of all tests versus the specific energy (power) consumed. It is interesting to note that the data appear in two distinct clusters. The right-hand cluster consists purely of the all-in feed types with no truncation of the feed-size distribution at the lower end, whilst the left-hand cluster is formed from feeds with fines truncation.

7

HPGRS—CHARACTERISING AND DEFINING PROCESS PERFORMANCE

250 240 230

m-dot, ts/hm 3

220 210 200 190 180 170 160 150 15

16

17

18

19

20

21

Operating Gap, mm

FIGURE 2 Variation in specific throughput as a function of operating gap using a pilot-scale HPGR at an operating pressure of 38 bar

60 58 56

m-dot, ts/hm 3

54 52 50 48 46 44 42 40 40

90

140

190

240

290

Power, kW

Variation in specific throughput as a function of operating gap using a pilot-scale HPGR

FIGURE 3

Specific Pressing Force

The specific pressing force is defined as the grinding force applied to the rolls (kN), divided by the diameter (m) and width (m) of the rolls (Schönert 1988). The specific pressing force has the unit of N/mm2. Fsp = F/(1,000 u D u L) where Fsp F D L

= = = =

specific pressing force (N/mm2) applied grinding force (kN) roll diameter (m) roll width (m)

(EQ 5)

8

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Ranges for specific pressing force vary considerably in the range 1–9 N/mm2, with studded machines normally restricted to 5 N/mm2 maximum pressure. Specific pressing force is a key parameter used in scale-up and for comparison purposes between different machine sizes. Maximum Pressure between Rolls

The maximum pressure applied to the material between the rolls has been estimated by several workers, and it is generally assumed to be in the range of 40 to 60 times the specific pressing force. It is generally accepted that the following equation (Schönert 1988) holds true: Pmax = F/(1,000 u D u L u k u D ip ) where Pmax F D L k D ip

= = = = = =

(EQ 6)

maximum pressure (MPa) applied grinding force (kN) roll diameter (m) roll width (m) material constant (0.18–0.23) compression angle (6–10 degrees)

The parameter D ip can be calculated from the operating gap, with a detailed description being given by Schönert and Lubjuhn (1990). Specific Energy Input

The specific energy consumption of an HPGR is a familiar quantity to process engineers. As with all other instances of the parameter, it is calculated from the net power input to the rolls divided by the ore throughput rate. It is important to note that specific energy input (kWh/t) is proportional to the specific pressure applied to the rolls. Typical specific energy values for studded rolls range from 1 to 3 kWh/t. As with all direct comminution devices, harder material will absorb more energy compared to a softer material, for a given size reduction. A rule of thumb is that the ratio of specific pressing force to specific energy input is 1.8–3:1, with this ratio decreasing towards 1.0 for finer comminution. Figure 4 shows the type of response mentioned. In this case, the slope of the graph indicates a ratio of 1.5:1. Specific energy consumption is markedly impacted by the feed-size distribution, as illustrated in Figure 5. As the feed distribution lengthens (i.e., the bulk density increases), the specific energy consumption drops. The major impact of specific energy input is the product fineness. As with all comminution equipment, a point of diminishing returns will occur where extra energy does not generate a commensurate increase in fineness. Figure 6 shows a range of energies and fines generation. At the levels displayed in Figure 6, the point of diminishing returns has not been reached. SIMULATION OF HPGR PERFOR MANCE

As with all modeling and simulation of process equipment, there is a sliding scale from the simplest spreadsheet-based feed-product transfer function at one end, through empirical representations, to mechanistic models, and finally to detailed fundamental descriptions. The key process issues that need to be estimated, or predicted, during the design phase of a process plant are

9

HPGRS—CHARACTERISING AND DEFINING PROCESS PERFORMANCE

Specific Energy Consumption kWh/t

2.1 1.9 1.7 1.5 1.3 1.1 0.9 0.7 0.5 1.0

1.5

2.0

2.5

3.0

3.5

4.0

Specific Pressing Force, MPa

FIGURE 4 Relation between specific energy consumption and specific pressing force using a pilot-scale HPGR

Specific Energy Consumption kWh/t

2.1 30 Bar 38 Bar 52 Bar

1.9 1.7 1.5 1.3 1.1 0.9 0.7 0.5 1.40

1.45

1.50

1.55

1.60

Feed Bulk Density, t/m

1.65

1.70

1.75

3

FIGURE 5 Relation between specific energy consumption and feed bulk density using a pilot-scale HPGR, at various operating pressures

ƒ Throughput ƒ Size reduction (product and oversize) ƒ Power consumption (energy efficiency) ƒ Required hydraulic stiffness ƒ Target gap and operating pressure

Using these parameters, it is then possible to insert the HPGR into a flowsheet and make sensible comparisons against other types of equipment and flowsheet configurations. The additional benefits of preferential liberation and enhanced extraction must be assessed via laboratory tests and incorporated with the full analysis.

10

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

35

Net –1 18 mm Generation

33 31 29 27 25 23 21 19 17 15 0.5

1.0

1.5

2.0

Specific Energy Consumption, kWh/t

FIGURE 6 HPGR

Relation between specific energy consumption and fines generation using a pilot-scale

Piston Press Testing and Ore Characterisation

The main ore characterisation tests for HPGR modeling are the piston-press and dropweight procedures. The drop-weight test is the Julius Kruttschnitt Mineral Research Centre (JKMRC) -developed, single-particle test and is used to examine areas in the HPGR where the breakage is of a single-particle nature. The piston-press test is for characterisation of the packed-bed breakage zone in the HPGR. The purpose of the piston-press test is to generate an appearance function as per the drop-weight test, but for packed-bed breakage. Hence, the piston-press appearance function is used to characterise the predominant breakage action in the HPGR. The piston press can be used in an analogous manner to the traditional drop-weight test (i.e., breakage parameters and an appearance function can be determined). In terms of the breakage characteristics, Table 1 provides an example of the comparison of the “b” parameters from the drop-weight and piston-press tests for material from Argyle Diamonds. The immediate observation regarding the data in Table 1 is that the piston press “b” parameters are higher than the single-particle test, with the inference being that the material appears softer in a packed-bed environment. Given the mode of compression (i.e., slow interparticle versus transient compression), Table 1 could represent an efficiency factor relating the two forms of breakage. Of more practical importance is that the use of the packed-bed, piston-style test is critical to the formation of a representative model of HPGR performance. Application of Piston Press to Provide Conceptual-Level HPGR Performance Estimates

A variety of workers are now using piston-press tests to research the action of HPGRs. The press arrangement at Freiberg University has recently been used to test a copper ore supplied by Rio Tinto. The aim of the tests is to determine the amenability of the ore to HPGR treatment and to examine the use of the piston press for conceptual-level evaluations. A series of tests at pressures from 80 to 320 MPa were undertaken with the results presented in Table 2. The maximum pressures reported in Table 2 were chosen to mimic those seen in the HPGR pilot tests, and the results appear to be good approximations to those obtained

HPGRS—CHARACTERISING AND DEFINING PROCESS PERFORMANCE

TABLE 1

Single-particle breakage parameters

Sample Unweathered lamproite Siliceous waste

TABLE 2

11

Single-Particle Test

Packed-Bed Test

b 0.44 0.40

b 0.940 0.703

Flake density results from piston-press tests Maximum Pressure, MPa

Flake Density, t/m3

77.24 157.29 230.53 310.98

2.14 2.32 2.32 2.38

from pilot-scale HPGR work. Given this agreement, it is suggested that the piston press be used to provide a conceptual-level envelope of performance. The suggested sequence is 1. Estimate m-dot value from Equation (3), by substitution of the product flake

density, operating gap (final bed depth from piston press), and use of Equation (4) to determine D. 2. Estimate throughput from the rearranged Equation (1), with assumed values for

roll diameter (D), roll width (L), and roll speed (u) relating to the desired scale of equipment. These values can be determined in association with manufacturers. It should be noted that the scale independence of m-dot, due to the linearity of operating gap versus roll diameter, is a major assumption in this step. 3. Calculate the specific pressing force (Fsp) from Equation (5) using the applied

grinding force from the piston press and the D and L values used above. With these key parameters, it is possible to ensure that the size of rolls and the bearing selection is correct. To estimate comminution performance: ƒ Determine the specific energy consumption from assumed relationship with spe-

cific pressing force. Values for the ratio Fsp:Wsp can be assumed to vary from 1:1 for very fine comminution through to 3:1 for very coarse duties. A value of 1.5:1, as shown in Figure 5, is a good general value for moderate comminution of hard ores. Care should be taken—although particle-size distribution is a major part of the bulk properties that dictate the relationship between Fsp and Wsp, other factors also influence the bulk behaviour including ore hardness, friction, and moisture (M.J. Daniel, personal communication, 2005). ƒ Specific energy consumption is inherently linked to product-size distribution via the

traditional breakage and appearance type mapping employed in single-particle dropweight tests. Using the A and b parameters from the piston-press test, these along with the specific energy consumption can be substituted into the following equation: t10 = A(1 – e–b. Ecs) where t10 = percentage passing one tenth of the feed size A and b = breakage characteristics from piston-press tests Ecs = specific energy consumption (kWh/t)

(EQ 7)

12

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Using the standard single-particle relationships between t10 and the other size distribution markers (i.e., t2, t4, t25, t50, t75), the entire size distribution of the product can be generated. Theoretically, this is a combination of packed- and singlebed approaches, but, as Tondo (1997) showed, the packed-bed t10 versus tn relationship underestimates size reduction in coarse sizes, compared to single-particle tests. Given that the variable edge effect generates coarser products, it is likely that any underestimation from the packed-bed parameters is simply an approximation to the coarser edge comminution. This approach is backed up by the fact that various workers have chosen to deal with this in different ways, whilst still obtaining satisfactory results. Tondo (1997) used both single-particle and packedbed A and b parameters with separate appearance functions in his work, whilst Daniel (2002) assumes a 10% split to edge and uses the single-particle function for all breakage with a t10 of 30. This conceptual-level approach, although not rigorous, helps engineers to obtain a “feel” for HPGR performance and at least obtain a quick, first-pass estimate of the operational envelope. It should be noted that no account is taken of precrush or edge effects. Analysis of this technique suggests that both throughput and product fineness are overstated, but as the scale of machine increases, the discrepancy lessens. This reduction in error with scale can probably be assigned to the decreasing proportion of machine performance impacted by edge effects. Detailed HPGR Modeling

For a more complete treatment of performance estimation in a modeling sense, true models are required. The work of Daniel and Morrell (2004) represents the most complete current description. The basis for their work is shown schematically in Figure 7. Daniel and Morrell outline information required for modeling, as shown in Table 3. To undertake the simulation, there are a variety of parameters relating to the breakage and classification of material in the three different zones as defined in Figure 7. The main parameters are listed in Table 4. This extremely comprehensive treatment is then used in a verification and scale-up scheme procedure; full details can be found in works by Daniel and Morrell (2004). CONCLUSIONS

There is an increasing body of knowledge around the application of HPGRs in hard-rock duties. In terms of selection and sizing, much has already been written, particularly by the suppliers. For process performance, the increasing application is allowing the development of some rules and shortcuts that can allow a first-pass evaluation of HPGRs for flowsheet purposes—a critical element on the pathway to engineering acceptance. In many ways, this paper seeks to provide a pragmatic engineering basis for the assessment of HPGR performance. This message was also the theme expressed by Klymowsky and Liu (1996), where they sought a Bond work-index analogy for HPGRs. There is no doubt that a standardized, accepted HPGR “work index” would be a great boost to HPGR acceptance. Beyond these engineering views of HPGRs, the detailed modeling and simulation of HPGR process performance is finding common ground, and workers have developed comprehensive approaches that provide the required accuracy and resolution. Assimilation of this understanding within the industry, along with simpler measures and guidelines, will accelerate HPGR implementation, particularly now that mechanical issues are predominantly of historical interest only.

HPGRS—CHARACTERISING AND DEFINING PROCESS PERFORMANCE

13

Feed to HPGR

Entry Zone Single-Particle Breakage

Centre Zone Packed-Bed Breakage

Edge Effect SingleParticle Breakage

Product from HPGR

After Tondo 1997.

FIGURE 7

TABLE 3

Schematic representation of Daniel and Morrell model

Model inputs and outputs Measured Input

Sample mass Roll diameter (D) Roll width (L) Roll speed (U) Bulk “compacted” density (qc) Feed-size distribution

Measured Output

Calculated Output

Working gap (xg) Flake thickness (xgf ) Flake density (qg) Product-size distribution (measured) Batch process time Working pressure (pw), power (kW)

Measured throughput (Qm) Calculated throughput (Qcalc) Specific energy (Ecs) Specific force (Fsp) Critical gap (xc) Product-size distribution

Source: Daniel and Morrell 2004.

TABLE 4

Model parameters Fixed Default Parameters

t10p, t10e—breakage for edge and precrusher K1p, K2p, K3p—precrusher model parameter K1e, K2, K3—edge-crusher model parameter K1h, K2h, K3h—compression zone parameter Split factor (c) Kp(edge)—power coefficient (edge)

Critical Model Parameters

Kp(HPGR)—power coefficient (compression zone) t10h—breakage for compression zone crusher

Source: Daniel and Morrell 2004.

ACKNOWLEDGMENTS

The author gratefully acknowledges all practitioners in the field of HPGR technology that have contributed to this paper through discussions. In particular, the discussions and advice from Mike Daniels, JKMRC, showed that a considerable amount of effort is still being applied to the issue of HPGR application.

14

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

REFERENCES

Daniel, M.J. 2002. HPGR model verification and scale-up. Master’s thesis. Brisbane, Australia: Julius Kruttschnitt Mineral Research Centre, Department of Mining and Metallurgical Engineering, University of Queensland. ———. 2005. Paper submitted to Randol Pacific Gold Forum, Perth, Australia. Daniel, M.J., and S. Morrell. 2004. HPGR model verification and scale-up. Minerals Engineering 17:1149–1161. Klymowsky, I.B., and J. Liu. 1996. Towards the development of a work index for the roller press. In Comminution Practices, SME Symposium 1996. S.99/105. Maxton, D., C. Morley, and R. Bearman. 2004. A quantification of the benefits of high pressure rolls crushing in an operating environment. Minerals Engineering 16:827–838. Ruben, E.S. 2002. Learning our way to zero emissions technologies. IEA Zero Emission Technologies Strategies Workshop, Washington, DC, March 19. Schönert, K. 1988. A first survey of grinding with high-compression roller mills. International Journal of Mineral Processing 22:401–412. ———. 1991. Advances in comminution fundamental, and impacts on technology. Pages 1–21 in Proceedings of the XVII International Mineral Processing Congress. Volume 1. K. Schöenert, ed. Ljubijana, Yugoslavia. Schönert, K., and U. Lubjuhn. 1990. Throughput of high compression roller mills with plain and corrugated rollers. Pages 213–217 in 7th European Symposium on Comminution. Tondo, L.A. 1997. Phenomenological modelling of a high pressure grinding roll mill. Master’s thesis. Brisbane, Australia: Julius Kruttschnitt Mineral Research Centre, Department of Mining and Metallurgical Engineering, University of Queensland.

High-Pressure Grinding Rolls— A Technology Review* Chris Morley†

ABSTRACT

The development of high-pressure grinding rolls (HPGRs) technology is reviewed, with an emphasis on aspects relevant to hard-rock comminution. Case histories are investigated and lessons learned are discussed in the particular context of the application of the device as a supplement to, or replacement for, conventional crushing and semiautogenous milling circuits. The potential for the more widespread use of this technology as a comminution method in hard-rock processing is examined. The use of the technology as a metallurgical tool is addressed, and future flowsheet concepts are introduced that make progressively greater use of the energy efficiency of HPGRs. INTRODUCTION

High-pressure grinding roll (HPGR) technology has its genesis in coal briquetting in the early twentieth century, but it was not until the mid-1980s that it was adopted for comminution applications, when it was applied in the cement industry to treat relatively easily crushed materials. Since then, it has been applied to progressively harder, tougher, and more abrasive materials, generally successfully, but as would be expected, not without some developmental problems. Machines are now also in use in the following applications: ƒ Kimberlites in secondary, tertiary, and recrush roles ƒ Iron ores for coarse crushing, autogenous mill pebble crushing, regrinding, pre-

pelletising, and briquetting ƒ Limestone crushing ƒ Concentrates fine grinding ƒ Gold ore crushing

Other prospective applications include phosphates, gypsum, titanium slag, copper and tin ores, mill scale, and coal. Hard-rock operations that use HPGRs as an alternative or supplement to conventional comminution devices include Argyle, Diavik, Premier, Kimberley, Jwaneng, Venetia and

* Updated from the original paper, “HPGR in Hard Rock Applications,” published in Mining Magazine, September 2003, www.miningmagazine.com † Fluor, Australia 15

16

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Courtesy of Köppern.

FIGURE 1

Coal briquetting press—early twentieth century

Ekati (diamonds), CMH-Los Colorados, CVRD, Empire and Kudremukh (iron ore), and Suchoj Log (gold ore). Hard-rock operations to have considered using HPGR and conducted pilot testing include Mt. Todd, Boddington, and KCGM, all in Australia. A full plant trial of an HPGR was conducted on a particularly arduous duty at Cyprus Sierrita between 1995 and 1996; and, more recently, HPGR has been piloted at Lone Tree, Nevada, in the United States, and Amplats Potgietersrus in South Africa. Currently, HPGR-based comminution plants are under construction at Bendigo, Australia (gold), and Cerro Verde, Peru (copper), and at final feasibility study stage for the Soledad Mountain, California (heap leach gold/silver), and Boddington, Australia (gold/copper), projects. There are currently three recognised manufacturers of HPGR machines, namely Polysius (a Thyssen Krupp company), KHD Humboldt Wedag AG, and Köppern, all based in Germany. T H E TE C H N O L O G Y

Machine Design

The HPGR machine comprises a pair of counterrotating rolls mounted in a sturdy frame. One roll is fixed in the frame, while the other is allowed to float on rails and is positioned using pneumohydraulic springs. The feed is introduced to the gap between the rolls and is crushed by the mechanism of interparticle breakage. The pressure exerted by the hydraulic system on the floating roll largely determines comminution performance. Typically, operating pressures are in the range of 5–10 MPa, but can be as high as 18 MPa. For the largest machines, this translates to forces of up to 25,000 kN. The rolls are protected with wear-resistant surfaces, and the ore is contained at the roll edges by cheek plates. Technology Motivators

Generally, the primary motivation for the use of the HPGR as a comminution alternative is its energy efficiency when compared to conventional crushers and mills. This improved

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

17

Courtesy of KHD Humboldt Wedag AG.

FIGURE 2

HPGR machine

efficiency is due to the determinate and relatively uniform loading of the material in the HPGR compression zone, whereas the loading in conventional crushers and (particularly) tumbling mills is random and highly variable, and therefore inefficient. The most energy-efficient method of breakage is the slow application of pressure to individual particles to cause structural failure, such that the energy lost as heat and noise is minimised. However, until a device is invented that can perform this task on a commercial scale, the HPGR remains the most energy-efficient comminution technology available. A major operating cost in conventional semiautogenous-based comminution circuits treating hard and abrasive ores is that of grinding media. One effect of the use of HPGRbased circuits is that semiautogenous mill grinding media is eliminated, and while ballmill media costs typically are slightly greater (due to the increased transfer size from HPGRs), the overall media savings are typically of the same order of magnitude as the energy savings. In addition to its energy and media benefits, the HPGR may be regarded as a metallurgical tool offering improved gravity, flotation and leach recoveries, and enhanced thickening, filtration, and residue deposition performance. These effects can be attributed to the phenomenon of microcracking of individual progeny particles due to the very high stresses present in the HPGR compression zone. Microcracking occurs predominantly at grain boundaries and so increases liberation and lixiviant penetration, while the effective reduction in milling work index caused by microcracking reduces overgrinding and slimes generation. In addition to being ore dependent, the extent of microcracking is a direct function of the operating pressure—and therefore energy input—of the HPGR, and in any given operation, the benefits of microcracking must be weighed against the incremental power required to achieve those benefits. The HPGR’s mechanism of interparticle breakage is particularly beneficial in the processing of diamond-bearing kimberlites, which undergo a form of differential comminution

18

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Courtesy of Polysius AG.

FIGURE 3

Cone crusher product particle (conventional crushing)

Courtesy of Polysius AG.

FIGURE 4

HPGR product particle (internal microfractures after Polycom treatment)

whereby the host rock is shattered while the diamonds are liberated undamaged—provided, of course, that the diamonds are smaller than the operating gap of the HPGR. This effect is also of benefit in the treatment of gold ores containing coarse gravity-recoverable gold grains, which would be flattened in conventional tumbling mills and rendered more difficult to recover. Technology Status

The HPGR, considered a mature technology in the cement industry, is now the norm rather than the exception in modern diamond plant design and is becoming common in iron ore processing, particularly in the field of pellet feed preparation. However, although some of the current diamond and iron ore applications can be regarded as hard-rock duties, HPGR is regarded by many as unproven in true hard-rock mining, and this perception is reinforced by the experience at Cyprus Sierrita in 1995– 1996. This application is widely considered to have been unsuccessful because it did not lead to a commercial sale; however, the fact that the comminution performance of the machine was impressive is not in dispute. The difficulties experienced related to the behaviour of the wear surfaces, and many valuable lessons were learned from this operation regarding the precautions necessary in circuit design and unit operation for the protection of the studded roll surfaces and the successful application of HPGR technology.

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

FIGURE 5

19

Cyprus Sierrita installation

The following is a summary of the more important issues arising from observations of the HPGR operation at Cyprus Sierrita and elsewhere: ƒ The technology is approaching a level of maturity allowing it to be seriously con-

sidered for hard-rock applications. ƒ HPGRs are sensitive to segregation and tramp metal in the feed. ƒ Mechanical availability of HPGRs is relatively high, and loss of machine utilisation

in hard-rock applications is predominantly wear related. ƒ The smooth and profiled hard-metal roll surfaces commonly used in the cement

sector are unsuitable for hard abrasive ores. Instead, the more recently introduced autogenous wear layer concept should be used, in which crushed ore is captured in the interstices between metal carbide studs or tiles. ƒ On hard-rock applications in particular, HPGRs are sensitive to feed top size,

which ideally should not exceed the roll operating gap. Oversize material in the feed can lead to stud breakage. ƒ Roll wear surfaces may be formed as segments or as cylindrical sleeves or tyres.

Segments may be used for softer ores and lower operating pressures, while tyres are recommended for hard-rock duties and higher pressures as they present a uniform, uninterrupted wear surface to the ore and thereby avoid the preferential wear that occurs at segment boundaries. In addition, tyres are easier to fabricate than segments and so are less expensive. ƒ Tyres involve long change-out times due to the need to remove the roll assemblies

from the mainframe, while segments can be changed in situ. Some machine designs aim to minimise change-out times for tyres by allowing roll assembly removal without the need for dismantling of the feed system and superstructure. ƒ Wear of the roll edges and cheek plates (the static wear plates used to contain the

ore at the roll edges) remains an issue, and development in this area is ongoing. A

20

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

few operations use rock boxes (chutes at the edges of the rolls) instead of cheek plates, allowing part of the feed material to flow around the rolls and so relieve the pressure on, and wear of, the roll edges. This does, however, introduce the disadvantage of passing uncrushed feed to product. Technology Hindrances

Hindrances to the adoption of HPGRs in hard-rock processing include ƒ The generally conservative nature of the mining industry ƒ A perception of high cost, particularly of the replacement wear parts in abrasive

applications ƒ Uncertainties regarding the reliability of modeling and scale-up from laboratory

or pilot operations to commercial installations ƒ A lack of definition of the requirements for robust flowsheet design of an HPGR-

based comminution circuit. Of these, it is generally acknowledged that high wear rates constitute the major obstacle to the ready acceptance of the technology in hard-rock applications. However, the HPGR can prove a cost-effective comminution device, even when the high cost and frequency of replacement of wear surfaces in highly abrasive duties are considered. Scale-up procedures have been the subject of many technical publications and should now be considered reliable. They are mentioned here only briefly for the sake of completeness. The characteristics of HPGRs that have a significant impact on flowsheet design will be considered as the main emphasis of this analysis. SCALE OF OPERATION

A common perception is that a project must be of relatively large scale before the use of HPGRs can be justified. However, HPGR units of almost any size can be produced (up to the current practical unit capacity limit of about 2,200 t/h), and this technology deserves serious consideration over a much wider range of plant capacities than might initially be imagined. Ultimately, HPGRs can be justified if they offer benefits to metallurgical performance and/or project economics, and the potential for such benefits can usually be assessed at the prefeasibility study phase by conducting preliminary tests. The manufacturers have test facilities in Germany, and small-scale laboratory facilities are available at various locations globally. Pilot-scale machines are available at several research facilities in Perth, Western Australia, and a Polysius mobile pilot unit used for trials at an operation in North America in 2003 was subsequently relocated to South Africa for evaluation on a hard-rock mining operation. THE MANUFACTURERS AND THEIR DESIGNS

Polysius, KHD, and Köppern are widely represented globally, but the machines are manufactured exclusively at their respective facilities in Germany. Polysius favours a high-aspect-ratio design—large diameter, small width—while KHD and Köppern prefer a low-aspect ratio. The high-aspect-ratio design is inherently more expensive but also offers an intrinsically longer wear life for a given application, as the operating gap is larger and the roll surfaces are exposed to a correspondingly smaller proportion of the material processed. The high-aspect-ratio design also produces a coarser product due to the greater influence of the edge effect; however, this difference is relatively slight, particularly with larger units. Nevertheless, for closed-circuit applications,

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

21

ATWAL

REGRO LABWAL Data of Test Units: Diameter of Rolls: Width of Rolls: Speed of Rolls: Top Feed Size:

0.71 m 0.21 m 0.29–1.10 m/s 16–35 mm

Diameter of Rolls: Width of Rolls: Speed of Rolls: Top Feed Size:

0.30 m 0.07 m 0.2–0.9 m/s 8–12 mm

Courtesy of Polysius AG.

FIGURE 6

Polysius test facility

this additional coarseness does increase the circulating load and tends to offset the wear life benefits, as a higher total throughput is required for the same net product. The use of tungsten carbide studs to create an autogenous wear layer on the roll surface is covered by a patent held by KHD, from whom this technology is available under license. Both Polysius and KHD have experience with minerals applications and studded roll technology, and are able to supply machines with capacities of up to about 2,200 t/h. Although Köppern has limited minerals experience, their HPGRs are successfully operating in the cement industry. For highly abrasive materials, Köppern recommends HPGRs fitted with their Hexadur wear protection. The Hexadur surface comprises hexagonal tiles of a proprietary abrasion-resistant material set into a softer matrix, which wears preferentially in operation, allowing the formation of an autogenous wear protection layer at the tile joints. The tiles and matrix material are fully bonded together and to the substrate in a high-temperature, high-pressure furnace. By contrast, KHD’s studs are inserted into drilled holes. As a result, the tiles are inherently stronger and more resistant to breakage due to oversize ore or tramp metal. Köppern supplies patterned and profiled surfaces in both segment and tyre format, whereas Hexadur is generally available only in tyre format due to the dimensional control difficulties inherent in the fabrication and furnace treatment of segments. However, research into the commercial production of Hexadur segments is ongoing. Meanwhile, the maximum Hexadur roll diameter available currently (and for the foreseeable future) is 1.5 m, constrained by furnace dimensions. This constraint limits Köppern’s unit capacity to about 1,000 t/h for hard-rock comminution applications using Hexadur. However, Köppern also offers machines with studded roll surfaces supplied by KHD, effectively lifting this capacity constraint.

22

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Courtesy of KHD Humboldt Wedag AG.

FIGURE 7

Studded roll wear surface

Courtesy of Köppern.

FIGURE 8

Hexadur wear surface for hard-ore comminution

Köppern has an established design in which the ends of the mainframe hinge outwards to allow the roll assemblies to be removed without disturbing the feed system and superstructure. This allows roll change-out times for tyre replacement of about the same duration as for in-situ segment change-out. Polysius also offers a design that allows rapid roll assembly removal, but without the need for a hinged frame design. In more recent developments, KHD has unveiled a rapid change-out concept to be offered on new

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

23

Courtesy of Köppern.

FIGURE 9

Köppern HPGR

Courtesy of Köppern.

FIGURE 10

Köppern hinged frame

machines and which can be retrofitted to existing units, and Köppern has introduced their “C-frame” design that allows the removal of both roll assemblies from one end of the frame, so offering a maintenance advantage over their earlier design. KHD uses cylindrical roller bearings that allow the choice of grease or circulating oil lubrication systems, as there is no relative movement between the bearings and seals. Polysius and Köppern use grease-lubricated, self-aligning spherical roller bearings. OPERATING CHARACTERISTICS

There are many factors to be considered when specifying an HPGR and selecting an appropriate flowsheet for a given application. The following subsections summarize the more important issues.

24

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Ore Characteristics

The compressive strength of the material to be crushed determines the amount of useful energy that can be absorbed by the material, which in turn dictates the bearing and motor sizes required for a given duty. With studded roll wear surfaces, the compressive strength of the ore, in combination with the feed particle top size and operating pressure, will largely determine the probability of stud damage—the higher the values of each of these variables, particularly when they occur together, the higher the likelihood of incurring stud damage. Ongoing development of stud technology is aimed at reducing the sensitivity of the studs to these variables. The abrasion index of the material being crushed will determine the wear rate (as distinct from the breakage rate) of the studs, as well as that of the substrate metal. For example, the wear life at the iron ore operations at Los Colorados and Empire are about 14,000 and 10,000 hours, respectively, while those at the Argyle and Ekati diamond mines were about 4,000 hours initially, but increased to 6,000–8,000 hours and beyond with ongoing development of stud and edge protection configurations. HPGRs are not generally suitable for the treatment of highly weathered ores or feeds containing a large proportion of fines. (This of course does not apply to applications where all the feed material is fine, such as fine grinding of concentrates.) Fine and weathered material tends to cushion the action of the rolls and so reduces the efficiency of comminution of the larger feed particles. For example, Argyle bypasses its primary HPGRs when very fine ore is being mined. On these ore types, the fine or weathered material should be removed by prescreening if HPGR treatment of the coarser component is required. HPGRs are not generally suitable for comminution of feeds containing excessive moisture, which tends to cause washout of the autogenous layer on studded rolls and increases slippage on smooth rolls. In both cases, accelerated wear is the result. For example, Ekati bypasses the –4+1 mm feed fraction around the HPGR when the prevailing ore type results in inherently high moistures. Specific Pressure

The specific pressure (specific press force) is the force (Newtons) divided by the apparent (or projected) area of the roll—that is, the product of roll diameter and length: specific pressure (N/mm2) = force (N)/(D (mm) u L (mm)) Typical practical operating values are in the range of 1–4.5 N/mm2 for studded roll surfaces and up to 6 N/mm2 for Hexadur. The required specific pressure determined in tests is used for scale-up of the required operating hydraulic pressure for the commercial unit. Specific Energy Input

The specific energy input (SEI) is the net power draw per unit of throughput: specific energy input (kWh/t) = net power (kW)/throughput (dry t/h) Typical operating values are in the range of 1–3 kWh/t. In general, a given ore will absorb energy up to a point beyond which little additional useful work (i.e., size reduction) is achieved—a zone of diminishing returns is approached. For equivalent size reduction, a hard, competent ore of high compressive strength will result in a higher SEI than a softer ore of low compressive strength. The energy input is governed by the hydraulic pressure, of which it is a roughly linear function. Generally, specific energy input in coarse crushing applications is numerically

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

25

about one half to one third of the specific pressure, so that a specific pressure in the typical operating range of 3–4.5 N/mm2 can be expected to correspond to a specific energy input of 1–2.5 kWh/t. In fine-grinding duties, this ratio is typically higher—for example, a ratio of 1.05 applies at the Kudremukh pellet feed operation. The best method of determining the optimum specific energy is to conduct tests to derive a graph of product fineness against specific energy. The graph generally displays an initial steep slope that flattens out to approach the horizontal at high SEI values (e.g., 3.5–4.5 kWh/t). The optimum SEI can then be selected. Microcracking

Although the size reduction graph frequently enters an area of diminishing returns with increasing specific energy, it has been demonstrated on some ores that the reduction in effective work index due to microcracking (also known as microfracturing or microfissuring) does not always display the same tendency. As a result, it may be beneficial from an overall comminution energy perspective to operate at a higher specific energy than corresponds to the optimum for size reduction in the HPGR stage, to maximise the benefits of microcracking. In this regard, the final grind size must also be taken into account, as the effects of microcracking are felt more in the coarser fractions, so that an application with a coarse grind will benefit more than one with a fine grind. It is important to conduct sufficient tests to quantify the optimum point of increased fines generation and reduced product work index, to ensure an HPGR is specified that is capable of transmitting the necessary power. Feed Top Size

For hard-rock applications, the feed top size is a critical variable in the successful operation of an HPGR crusher. For smooth rolls, too large a top size results in reduced nip efficiency, slippage, and accelerated wear; for studded rolls, tangential forces at the roll surface due to early nipping—effectively causing single-particle breakage by direct contact with the roll surfaces—can cause stud breakage. Constraints on feed top size have been related in the literature both to roll diameter and to operating gap. Figures of up to 7% of roll diameter and three times the gap have been quoted as appropriate limits on feed top size, even though the latter ratio implies some direct contact of the larger particles with the surfaces of both rolls, leading to singleparticle breakage. These figures are now considered much too optimistic in hard-rock applications, and it is generally accepted that, to minimise the likelihood of stud breakage, feed top size should not exceed the expected operating gap. This will normally demand a closedcircuit crushing operation upstream to ensure this top size is positively controlled. For softer materials, this rule can be relaxed—for example, some kimberlite operations successfully treat open-circuit secondary crushed products with top size–gap ratios of about 1.8–2.0 using studded rolls. By interpolation, ratios of around 1.3–1.5:1 are tolerable when treating ores of moderate hardness. Where uncertainty exists regarding ore hardness categorisation, it is considered prudent to adopt a ratio of close to 1:1 initially, and then relax this incrementally if and when it is established that stud breakage is not an issue. As a guide, the direct-contact nip angle (for single-particle breakage and possible stud damage) is normally in the range of 10˚ to 13˚ while interparticle breakage commences at angles of 5˚ to 7˚. By using a scale diagram of an HPGR unit of a given roll diameter, and showing these angles and an appropriate operating gap, estimates can be made of the particle size above which single-particle breakage is likely to occur and below which interparticle breakage commences.

26

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Unit Capacity

The capacity of an HPGR is fundamentally a function of the ore characteristics. Capacity is generally expressed in terms of specific throughput mx (m-dot), which is a function of the roll diameter, length, and peripheral speed: mx (t·s/m3·h) = throughput (t/h)/(diameter (m) u length (m) u speed (m/s)) The value of mx is determined in pilot tests and used in scale-up to the commercial unit, taking into account the change in the relative proportions of product from the centre of the rolls and from the edges where poorer comminution occurs (the “edge effect”), and also whether the commercial unit is to be operated with cheek plates or rock boxes for roll edge protection. In addition to its fundamental relationship to the ore characteristics, the value of mx is a function of many variables. The following should be regarded as general trends for the majority of ores, rather than as statements of universal fact—there will always be the exception that proves the rule: ƒ Ore hardness—mx increases with ore hardness. ƒ Specific pressure—mx decreases slightly with increasing pressure. ƒ Roll surface—mx increases with increasing “texture” of the roll surface, due to the

reduced slip (increased kinetic friction) and improved nip between the rolls. Thus, smooth rolls give the lowest values, with profiled surfaces in the mid-range, and studded surfaces the highest (typically about 50% higher than for smooth rolls). ƒ Roll speed—for smooth rolls, mx decreases with roll peripheral speed, so that

actual throughput increases with increasing speed but at a progressively diminishing rate due to increased slippage. The effect is much reduced with profiled or studded rolls due to the inherently higher kinetic friction of these surfaces. ƒ Feed top size—the available evidence is not conclusive, but it appears that mx

might increase slightly with an increase in feed top size. ƒ Feed bottom size—mx decreases significantly as feed bottom size is increased.

Thus, the highest value of mx occurs with a full-fines feed, and this value decreases progressively as the fines cut-off or truncation size is increased. This is due to the increased voidage in the truncated feeds, which results in a lower back pressure on the rolls and a consequent reduction in the operating gap. ƒ Feed moisture—for moisture levels greater than about 1%, mx decreases with

increasing moisture due to the replacement of solids with water in the compacted product flake; higher moisture levels can result in excessive slippage and ultimately to washout of the autogenous layer on studded rolls. Below 1% moisture, there is some evidence of reduced m• values with studded rolls due to the difficulty in generating and maintaining a competent autogenous wear layer with very dry feeds, as the crushed product is too friable to form a compacted layer between the studs. Operating Gap

The operating gap is directly related to the unit capacity, all else being equal, so “gap” can be interchanged with mx in the above analysis. Depending on the application, the ratio of operating gap to roll diameter will normally lie in the range of 0.010 to 0.028. Circuit Capacity

The capacity of an HPGR circuit, as distinct from the unit capacity discussed above, is obviously a function of the circuit design. Of the above variables, the feed bottom size

HIGH-PRESSURE GRINDING ROLLS—A TECHNOLOGY REVIEW

27

is particularly relevant in this regard, as a truncated feed necessarily implies the presence of a screen or other classification device upstream of the HPGR. It has been noted that capacity decreases with truncated feeds; however, the capacity of the circuit would increase if the amount of fines removed from the HPGR feed exceeded the reduction in HPGR unit capacity. Whether this occurs in practice remains the subject of some debate (and in any event is probably ore specific), but recent modeling of pilot test data for two prospective applications indicates that this is the case, and this is supported by the limited evidence available in the literature. However, an increase in circuit throughput achieved in this way may be offset by a decrease in product fineness and/or reduced microcracking such that, depending on the downstream processing route, a full-fines HPGR feed may be preferable to a truncated feed. For any given application, the more efficient flowsheet can be determined only by comprehensive tests and modeling, but where doubt exists, the circuit should, if possible, be designed with the flexibility to operate with full fines or truncated feed to allow circuit performance to be optimised. This flexibility normally comprises the prescreening of the feed and a facility to recycle to HPGR feed a portion of either the HPGR product or, where the HPGR operates in closed circuit with a screen, the screen undersize. Product Sizing

As noted earlier, product fineness increases with operating pressure (and therefore power), generally up to a point of diminishing returns. It has been observed elsewhere that it is more energy efficient to operate an HPGR at low pressures and in closed circuit with a screen, so that less energy is wasted on compacting the product. However, this generally would require more or larger HPGRs to handle the increased circulating load. Also, it is not clear whether the analysis included the cost of conveying the increased circulating load of screen oversize. Product fineness generally decreases with increasing “texture” of the roll surface; so smooth rolls give the finest product, with profiled surfaces in the mid-range and studded surfaces the coarsest. This is due to the reduced slip between the rolls and the ore, giving a higher throughput for a given power draw. For the same product fineness, therefore, a studded or profiled roll machine would have to be operated at higher pressures than a smooth roll unit. However, the effect is relatively small, and the benefits of profiled or studded rolls usually outweigh the reduced product fineness. Furthermore, the effect appears to be ore specific, and some operations (e.g., Jwaneng) have recorded an increase in fineness with studded rolls compared to smooth rolls. Increasing roll speed leads to a reduced product top size and improved F50/P50 reduction ratio, without significantly changing the fine end of the sizing spectrum. A slight mismatch or differential in roll speeds has been found to enhance grinding performance, and though this could be considered intuitively plausible, it might also be expected that adopting this as a deliberate control strategy could lead to increased roll surface wear rates due to this imposed speed differential. This effect is therefore regarded as being of academic interest rather than practical significance. Product sizing is largely independent of feed moisture. Product sizing is a function of roll aspect ratio. A high aspect ratio gives an inherently coarser product for the following reasons: ƒ The proportion of edge material in the product is greater. ƒ The pressure peak in the compression zone is lower (for a given specific pressure).

However, the overall effect is generally fairly modest. The shape of the HPGR product sizing curve is dissimilar to that of conventional crushers, so that for products with nominally the same P80, the HPGR product contains

28

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

considerably more fines below this size than from a conventional crusher. The implications of this are that, where the product is delivered to, for example, a ball milling operation, mill capacity will be greater when treating HPGR product than predicted by the standard Bond equation. Milling power requirements are thus reduced by both the sizing of the HPGR product and the microcracking of the product particles, and are therefore best determined by pilot testing. Roll Surface Wear

Increasing roll speed increases turbulence in the feed material and slip of feed against the roll surfaces, leading to elevated wear rates. This should generally be a concern only at the top end of the practical speed range. In this respect, Polysius traditionally uses a rule of thumb to the effect that the peripheral speed of the rolls (in meters per second) should not exceed roll diameter (in meters), although Köppern does not support this view and regularly nominates speed–diameter ratios of up to 1.3. KHD also uses these higher ratios for their smaller-diameter machines but generally uses 1. When the HPGR is added to an existing mill circuit the bonus becomes Wspec millbefore – Wspec millafter B = ------------------------------------------------------------------------Wspec HPGR

(EQ 7)

Wspecmill is the kilowatt-hours per ton for the existing mill before and after adding the HPGR. The energetic advantage of the HPGR is that the bonus is greater than 1, which means that 1 HPGR kW or kWh/t will do more grinding work than 1 mill kW or kWh/t. In other words, adding an HPGR to an existing mill circuit will reduce specific energy

SOME BASICS ON HIGH-PRESSURE GRINDING ROLLS

45

consumption and increase production. The savings in specific power consumption at the mill main drives is 'P = Wspec millbefore –  Wspec millafter + Wspec HPGR

(EQ 8)

The overall system savings are somewhat lower because of added equipment, such as conveying or screening. Bonus values depend mainly upon feed material, circuit configuration, and reference mill. Some typical values (from Köppern operating and test data) are ƒ Cement clinker

1.8–2.5

ƒ Blast furnace slag

2.5–3.8

ƒ Limestone

1.7–2.0

ƒ Kimberlite

1.6–2.0

The Machine Design

Figure 4 views an HPGR from the side where the hydraulic system is located. The two grinding rolls are suspended with self-aligning roller bearings in bearing blocks, which are mounted in the machine frame. Each roll has its own drive train with planetary gear reducers. Torque arms are provided to neutralize the countertorques generated by the drives. This particular machine design features a hinged frame that swings open for easy roll exchange. The machine has the following characteristics: Roll diameter Roll length Circumferential speed Installed power Installed grinding force Throughput Bonus achieved

2,140 mm 1,300 mm variable, max. 1.58 m/sec 2 u 1,300 kW 19,500 kN 850 tph cement clinker 2.1

The energy density in the nip zone is quite high, about 400 times compared to a ball mill. Correspondingly high are the loads and stresses on the rolls, especially on the roll surfaces. Figure 5 shows three basic roll designs. The roll surfaces are of particular importance not only from the wear aspect but also for their capability to draw in the material. Figure 6 shows a studded roll surface (according to sources at KHD Humboldt Wedag, Cologne, Germany) where material builds up between the studs, thereby forming an autogenous wear protection and providing a rough surface for good friction. Figure 7a shows a worn, welded hard surface, and Figure 7b shows metallurgical powder-based wear elements applied by a hot isostatic pressure process. These are just three examples; there are several others that have been developed over the years. HPGR Applications in Mining

Figures 8 through 11 show some typical applications for HPGRs to increase throughput and lower specific energy consumption of a grinding circuit. Figure 8 shows the HPGR after a semiautogenous grinding (SAG) mill for pebble grinding. The HPGR product is returned either to the SAG mill or to the screen. In Figure 9, the HPGR is located after the secondary crusher. The HPGR product is screened, and the oversize is returned to the HPGR. Figure 10 has the HPGR as single-pass pregrinder in front of the milling plant.

46

ADVANCES IN COMMINUTION

Hinged Frame

FIGURE 4

1

2

Roll

Hydraulic System

Torque Arms

Gear Reducers

HPGR assembly at the workshop (view from the hydraulic side)

Grinding Surface

3

Bearing Journal

(a) Solid Roller

FIGURE 5

ADVANCED COMMINUTION TECHNOLOGIES

1

Roller Core

4

1

Roller Core

5

Segments

2 (b) Tire-Shaft Roller

(c) Segments

Press tools

The HPGR can apply about 2.3 kWh/t to the feed material. At a bonus of, for example, 1.8, it would add 2.3 u 1.8 = 4.14 kWh/t ball-mill equivalent to the grinding circuit. The HPGR can also be located after heavy-media separation (HMS) with or without a crusher grinding the wet oversize (Figure 11). The best HPGR location for a given mine needs to be decided for each case; bottleneck identification, space availability, conveying distances, power grid, and other conditions must be considered. CONCLUSIONS

Versatility and efficient energy utilization have made high-pressure grinding an established comminution technology in the minerals processing industries. Relatively simple formulae can be used to describe and understand the underlying mechanics of HPGRs. Careful consideration must be given to the overall grinding process in order to take full advantage of the special features offered by high-pressure comminution.

SOME BASICS ON HIGH-PRESSURE GRINDING ROLLS

Tungsten Carbide Studs

FIGURE 6

Studded roll surface

Circumferential Wear Grooves

(a) Worn Surface Welding

FIGURE 7

Autogenous Wear Protection

Roll surface examples

Hexadur Tiles

Softer Interstice Material

(b) Hexadur Roll Surface

47

48

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Run-of-Mine Ore

Primary Crusher

SAG Mill

Screen

Crusher

Grinding Circuit

HPGR

Sorting Section

FIGURE 8

HPGR after SAG mill for pebbles grinding

Run-of-Mine Ore

Primary Crusher

Intermediate Stockpile

Double-Deck Screen Secondary Crusher

HPGR

Screen

Concentration Section

FIGURE 9

HPGR after secondary crusher with screen

SOME BASICS ON HIGH-PRESSURE GRINDING ROLLS

Concentrate Stockpile

Storage Bin

HPGR

Grinding Circuit

Pelletizing Plant

FIGURE 10

HPGR in single-pass grinding before milling plant

Ore Storage

Screen

HMS Float Sink HPGR Final Concentration

Screen

HMS Float Sink Crusher Final Concentration HPGR

Final Concentration

Screen Fines

Tailings

FIGURE 11

HPGR grinding wet oversize

49

High-Pressure Grinding Rolls for Gold/Copper Applications Norbert Patzelt,* Rene I.B. Klymowsky,* Johann Knecht,* and Egbert Burchardt*

ABSTRACT

Successful pilot-plant demonstrations carried out in 2003 and 2004 have proven the operational reliability of high-pressure grinding rolls (HPGRs) in hard-rock applications. As a result of these breakthroughs, six HPGRs will be commissioned in two copper concentrators in 2006. INTRODUCTION

High-pressure grinding rolls (HPGRs) are well established in the diamond and iron ore industries. Process advantages of HPGRs had been recognised by the minerals industry for many years. However, unresolved issues pertaining to wear have made the industry reluctant to adopt this technology. Starting in 2003, a successful pilot-plant demonstration on an exceptionally hard and abrasive gold ore proved that the wear issues could be resolved by the design of an appropriate wear-protection system, and availabilities in excess of 90% could be achieved. The pilot-plant results built up confidence in the minerals industry and a second pilot-plant trial was conducted on another extremely hard ore with the aim of determining if anything could break the machine. The machine demonstrated even higher availabilities than in the previous case. A commercial breakthrough then came when one of the world’s leading copper producers decided to build a new concentrator in South America based on HPGR technology. Four Polycoms, 24/16 in size, each equipped with two 2,500-kW motors, will be used in tertiary crushing duty in closed circuit with wet screens. Shortly thereafter, a second major copper producer ordered two large Polycom 20/15 units for an existing copper concentrator in Indonesia. In both cases, it was the energy savings and low operating costs of the HPGRs that attracted the producers. This paper examines the conditions (such as the press force necessary) that lead to energy savings, lower operating costs, and the optimum performance of the HPGRs. Wide variations occur in ores, even within one deposit. These variations, insofar as they affect the performance of an HPGR, need to be quantified with meaningful HPGR indices. Two such indices are the ATWAL Wear Index (ATWI) for wear due to abrasion and the Polycom Grinding Index (PGI) for quantifying the fines production.

* Polysius AG, Neubeckum, Germany 51

52

ADVANCES IN COMMINUTION

FIGURE 1

ADVANCED COMMINUTION TECHNOLOGIES

Installation of an HPGR in a copper concentrator in the United States in 1994

A laboratory ball mill test, the Labmill test, is described to overcome the uncertainty about the energy required for ball milling after an HPGR. The test is aimed to deal specifically with the special features of an HPGR product (i.e., microcracking and the high amount of fines in the product). EXPERIENCES WITH HPGRS IN HARD-ROCK APPLICATIONS

Cyprus Sierrita

The first serious application of HPGRs in the hard-rock mining industry was the installation of an HPGR in a copper concentrator in the United States in 1994 (Figure 1). The expected performance in terms of throughput, fines production, and energy consumption was met. However, the hardness and abrasiveness of the ore was by far higher than that of ores treated in HPGRs previously. It soon became apparent that the wear protection of HPGRs, in particular, the stud technology, was not advanced enough at that time to allow for a smooth and easy transition into continuous operation. Different stud qualities had to be tested and changed in order to suit the requirements of the ore. The change-outs were facilitated by having the rolls equipped with segments; however, these also contributed to wear problems. Finally, stud qualities were found that provided a reasonable lifetime at low cost (~0.10 to 0.15 US$/t) even under these difficult circumstances. In the end, the unit was decommissioned after treating more than 7,000,000 t of ore when the initial investment plans for the mine were abandoned. Despite the positive operating results, this installation was widely viewed by the industry as a failure of HPGR technology, and its acceptance was set back for years. Newmont Gold, Lone Tree (Nevada)

Following the Cyprus Sierrita demonstration, there was little progress made towards improving the technology or improving the wear protection for hard-rock applications. The next milestone in HPGR development came in April 2003, when Newmont Mining Corporation began a 3-month trial of a pilot-sized HPGR. Polysius designed a new roll surface specifically for the grinding of hard and abrasive copper and gold ores (Figure 2). The new roll surface consisted of a replaceable shrink-fitted tyre, armed with a new design of tungsten carbide studs, and a new edge-protection system intended to eliminate

HIGH-PRESSURE GRINDING ROLLS FOR GOLD/COPPER APPLICATIONS

FIGURE 2

TABLE 1

53

A pilot-sized HPGR equipped with a new roll surface designed by Polysius

Material data of Lone Tree ore Ball Mill Work Index, Wi (BM) Unconfined compressive strength Silica content Bond Abrasion Index, Ai ATWAL Wear Index, ATWI

20 kWh/t 200 MPa 78%–84% 0.64 >40 g/t

repair welding of the roll edges. The rolls were also provided with cheek plates to contain the material within the gap. The roll dimensions were 950 mm diameter u 350 mm width. Each roll was driven by a 160-kW motor and was operated at a speed of 21 rpm. The capacity of the unit was about 80 tph. The unit was run in closed circuit with an 8-mm square-mesh screen, and was protected from tramp metal by an overhead magnet and a metal detector on the main conveyor belt. The machine was operated 24 hours a day, 7 days a week, for a period of 87 days with no mechanical downtimes. The initial operational availability of the unit was 89% due to a programming glitch, which occurred after startup, after which a 92% availability was achieved. The properties of the feed are specified in Table 1. The Lone Tree trial was a true milestone, as no other pilot unit previously had been operated continuously on a 24/7 basis, and no mechanical or welding repairs had been required. The operators and maintenance staff were encouraged by the operation of the HPGR. No stud failures occurred during the more than 1,600 operating hours of the demonstration trial. In a commercial-scale application on a similar hard, abrasive ore, an HPGR would have achieved more than 3,000 hours of service and run considerably longer on a less competent and less abrasive ore. Relationships between wear and particle-size distribution were obtained that will improve future understanding of wear and wear life, benefiting the industry as a whole. The Polysius ATWAL laboratory abrasion test accurately predicted the wear rate in the trial. Individual tests, conducted over the course of 1 year on several representative samples obtained prior, during, and after the trial, were found to be reproducible within 10% of each other, validating the method used for determining and predicting wear in larger units.

54

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

Anglo American Platinum, Potgietersrust Platinum Mine

In October 2004, a pilot-plant unit was commissioned in the Potgietersrust platinum mine. This was a further step forward and the first approach of HPGR technology to the platinum industry. The HPGR was again operated in closed circuit with 8-mm screens. The feed material was prepared by two-stage crushing. Feed size was initially –25 mm but was increased to –35 mm for extended operating periods. The HPGR treated a total of 188,000 tph, resulting in 115,000 tph of final product until it was decommissioned in April 2005. The operating time was in excess of 3,000 hours. The ore was even tougher than that used in the previous application, which is reflected in the Bond Work Index (BWI) of 22 kWh/t, and in the operational problems experienced at the crushers. The HPGR was equipped with the latest wear protection. Results from abrasion testing indicated low abrasion, which was confirmed in the field. Availabilities as high as 97% were achieved. The test installation was declared a success by the operator who, in his words, “had tried very hard to break the machine.” IMPLEMENTATION OF HPGRS IN GRINDING CIRCUITS

In greenfield installations, HPGRs will have their place as the tertiary crushing stage in front of ball mills. Even this commitment still leads to many different flowsheet configurations. One possible flowsheet is shown in Figure 3. The secondary crusher and the HPGR, which replaces conventional tertiary crushers, are operated in closed circuit with dry screens. The product of the crushing circuit is stockpiled. In this configuration, the crushing circuit and the ball mill circuit are decoupled, allowing both circuits to be operated at a different utilisation. However, this decoupling has two implications. First, an additional stockpile is required. Stockpiling of the HPGR product, which contains a lot of fines, remains a challenge and requires extensive dust suppression. Second, screening of the HPGR product has to be done dry because a wetscreen undersize cannot be stockpiled. This entails a coarser product. Wet screening of the HPGR discharge may provide significant improvements. It is advantageous from the point of energy efficiency to shift as much grinding work as possible to the HPGR and feed the ball mills with a finer product. This approach requires a finer mesh size for the screen, 4 to 6 mm. Fine screening usually has a lower efficiency, especially if the discharge from the HPGR is in the form of highly compacted flakes. Wet screening will address the disagglomeration of the HPGR discharge and will definitely improve screening efficiency. It also will facilitate wetting of the material. A flowsheet illustrating a wet-screen arrangement for an HPGR is shown in Figure 4. The HPGR and ball mill circuits are combined, whereas the secondary crusher is decoupled. Alternatively, if the circuit consists of multiple crushing and grinding units, all three can be combined, eliminating the stockpile by oversizing the equipment. This arrangement allows for the lower availability of the crushers, whereas the HPGR availability is expected to be high enough for in-line operation with the ball mills. OPTIMUM HPGR PERFOR MANCE IN CLOSED-CIRCUIT OPERATION

In tertiary applications, HPGRs have to be operated in closed circuit. Consequently, the ball mill feed is not the “discharge” of the HPGR but is the product of the size distribution of the HPGR discharge and the mesh size of the closing screen. This raises two questions: first, what influence do HPGR operating parameters have on the feed-size distribution to the ball mill; and secondly, what is the most efficient way to operate an HPGR? In open circuit, the operating parameter that manifests the most influence on the particle-size distribution is the press force applied to the rolls. The energy absorbed by the material has been shown to be proportional to the applied press force.

HIGH-PRESSURE GRINDING ROLLS FOR GOLD/COPPER APPLICATIONS

FIGURE 3

55

HPGR in closed circuit with dry screens

Optional

FIGURE 4

HPGR in closed circuit with wet screens

In a closed circuit, however, the influence of the press force on the product size of the circuit is lost. This is demonstrated with two examples, one taken from tests on a semi-industrial scale unit and the other from tests on a laboratory-scale HPGR (Figure 5). Results were taken from single-pass tests on these units, in order to prove that the findings were independent of the machine size. The press forces applied were in the range of 2.7 to 4.3 N/mm2 on the semi-industrial unit, and from 2.3 N/mm2 to a higher value of 8.4 N/mm2 on the laboratory-scale unit. The impact of the press force on the throughput and energy consumption of the circuit are also shown. A screen undersize, representing the circuit product, was calculated on the basis of 100% screen efficiency from the discharge. Cut sizes were 4 mm for the semi-industrial circuit and 1 mm for the laboratory-scale circuit. It was assumed that the recirculation of screen oversize did not affect the size reduction in the HPGR significantly. On this basis,

56

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

2.5 Feed 1 Feed 2

Specific Energy, kWh/t

2.0

1.5

1.0

0.5

0.0 0

1

2

3

4

5

Specific Press Force, N/mm 2

FIGURE 5

Absorbed specific energy versus press force

a projection of the circuit performance in terms of throughput and energy consumption was also made. This approach may be considered simplistic but is adequate to explain some of the principles. Figure 6 shows discharge size distributions of a copper ore treated in a semi-industrial test unit. The grinding force was steadily increased from test R4 to R6, resulting in a finer discharge. The circuit product of an HPGR with a 4-mm classifying screen was calculated on the basis of 100% screening efficiency. Figure 7 shows discharge size distributions of a platinum ore treated in a lab-scale test unit. The grinding force was steadily increased from test L1 to L4, resulting in a finer discharge. The circuit product of an HPGR with a 1-mm classifying screen was calculated on the basis of 100% screening efficiency. The conclusions drawn from Figures 6 and 7 were that the size distribution of the final circuit product did not vary much, no matter if the HPGR discharge was finer or not. The fineness of the circuit product was largely determined by the mesh size of the closing screen. However, the applied press force had a strong influence on the circulating load and the circuit throughput with an HPGR of given size, as well as on the energy consumption, as shown in Figures 8 and 9. In Figure 8, the specific energy input was increased by 30% while the throughput of the closed circuit increased by only 10%. In Figure 9, the specific grinding energy was increased even further by 100% to 8.2 N/mm2, whereas the throughput of the closed circuit only increased by 40%. These examples show that operation at high specific press forces, 8 N/mm2, reduces energy efficiency drastically. The following general conclusions were drawn with regard to optimum operation of HPGRs in closed circuit: 1. The product-size distribution of an HPGR in closed circuit with screens is not

influenced by the applied press force. 2. The applied press force determines the circulating load and the energy consump-

tion of the HPGR circuit.

HIGH-PRESSURE GRINDING ROLLS FOR GOLD/COPPER APPLICATIONS

100 Feed R4 Discharge R4 S/U R5 Discharge R5 S/U R6 Discharge R6 S/U

Fineness Cumulative Passing, %

80

60

40

20

0 0.01

0.10

1.00

10.00

100.00

Particle Size, mm NOTES: Discharge denotes the HPGR discharge from respective tests. S/U denotes the screen undersize from respective tests.

FIGURE 6

Semi-industrial HPGR test with copper ore

100 Feed L1 – D L1 – S/U L3 – D L3 – S/U L4 – D L4 – S/U

Fineness Cumulative Passing, %

80

60

40

20

0 0.01

0.10

1.00

10.00

Particle Size, mm NOTES: D denotes the HPGR discharge from respective tests. S/U denotes the screen undersize from respective tests.

FIGURE 7

Lab-scale HPGR tests with platinum ore

100.00

57

58

ADVANCES IN COMMINUTION

ADVANCED COMMINUTION TECHNOLOGIES

4

80

60

3

40

2

20

1

0

Specific Energy Input w(sp), kWh/t

Throughput M, tph

M w(sp)

0 3

2

4

5

6

2

Specific Grinding Force ϕ, N/mm

FIGURE 8

Semi-industrial circuit projection (copper ore 1,000 N—and plotted in Figures 12 to 14. These figures show that

LINKING DISCRETE ELEMENT MODELING TO BREAKAGE IN A PILOT-SCALE AG/SAG MILL

279

4.5 4.0

Translational Velocity, m/s

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.0

0.2

0.4

0.6

0.8

1.0

Time, sec , 101

FIGURE 8 PFC3D estimates of the time history of the translational velocity of a small particle (black—higher maxima) and a large particle (gray)

8.0 7.5 7.0 6.5

Rotational Velocity, rad/s, 101

6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.0

0.2

0.4

0.6

0.8

1.0

Time, sec , 101

FIGURE 9 PFC3D estimates of the time history of the rotational velocity of a small particle (black) and large particle (gray) showing minimal difference

280

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

6.0 5.0 4.0

Vertical Coordinate, m, ×10–1

3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0 –6.0

–4.0

–2.0

0.0

2.0

4.0

6.0

Horizontal Coordinate, m, ×10–1

FIGURE 10 Spatial pattern of motion in the vertical plane of a randomly selected single small particle in a mixed size charge in the PFC3D model

6.0 5.0

Vertical Coordinate, , m, ×10–1

4.0 3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0 –6.0

–4.0

–2.0

0.0

2.0

4.0

6.0

Horizontal Coordinate, m, ×10–1

FIGURE 11 Spatial pattern of motion in the vertical plane of a randomly selected single large particle in a mixed size charge in the PFC3D model

LINKING DISCRETE ELEMENT MODELING TO BREAKAGE IN A PILOT-SCALE AG/SAG MILL

281

6.0 5.0 4.0

Vertical Coordinate, m, ×10–1

3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0 –6.0

–4.0

–2.0

0.0

2.0

4.0

6.0

Horizontal Coordinate, m, ×10–1

FIGURE 12

Spatial pattern in the vertical plane of shear contact forces of 10 to 100 N

6.0 5.0 4.0

Vertical Coordinate, m, ×10–1

3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0 –6.0

–4.0

–2.0

0.0

2.0

4.0

6.0

Horizontal Coordinate, m, ×10–1

FIGURE 13 1,000 N

Spatial pattern in the vertical plane of shear contact forces with intensities of 100 to

282

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

6.0 5.0 4.0

Vertical Coordinate, m, ×10–1

3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0 –6.0

–4.0

–2.0

0.0

2.0

4.0

6.0

Horizontal Coordinate, m, ×10–1

FIGURE 14

Spatial pattern in the vertical plane of shear contact forces >1,000 N

low- and medium-intensity shear forces occur predominantly within the charge, whereas strong shear forces occur during freefall and at the resultant impact. Significantly higher intensity shear forces also occur in the base of the mill charge at the interface between the mill liners and the particles. CONCLUSIONS

Based on the results presented, 3D DEM can accurately predict net power draw of the pilot experimental AG mill. This success does not necessarily imply similar accuracy in prediction of power draw in operating mills, because only a small quantity of ore fines and water are present in this pilot mill. As the no-load power of the larger AG mills can be predicted with sufficient accuracy using established empirical models (Napier-Munn et al. 1996), the DEM model may well be capable of accurate prediction of total mill power draw if slurry behaviour can be included. A strong correlation exists between the observed power utilisation and modeled net frictional power consumed within the mill charge. Therefore, subject to further validation of DEM, the modeled net-frictional power may be able to be used in a similar manner to the empirical power utilisation parameter for optimisation of performance of industrialscale AG mills. The insight offered into the interactions between lifter height and particle size goes well beyond liner damage caused by excessive liner height and offers the possibility of liner design for maximum production. The DEM results provide a rationale for the approximately constant rock-wear rate reported in the literature—even though there are significant differences due to the interaction with the mill lifters by different sized particles. Hence, there is also a strong possibility that a single, measurable characteristic may be used to model this type of rock wear when combined with a DEM model.

LINKING DISCRETE ELEMENT MODELING TO BREAKAGE IN A PILOT-SCALE AG/SAG MILL

283

ACKNOWLEDGMENTS

The DEM modeling work reported in this paper has been partially funded by the Centre for Sustainable Resource Processing. REFERENCES

Cleary, P.W. 2001. Recent Advances in DEM Modelling of tumbling mills. Minerals Engineering 14:1295–1319. Cundall, P.A., and O.D.L. Strack. 1979. A discrete numerical model for granular assemblies. Geotechnique 29(1):47–65. Djordjevic, N. 2003. Discrete element modelling of the influence of lifters on power draw of tumbling mills. Minerals Engineering 16(4):331–336. Djordjevic, N., F.N. Shi, and R.D. Morrison. 2004. Determination of lifter design, speed and filling effects in AG mills by 3D DEM. Minerals Engineering 17(11–12):1135–1142. Itasca Consulting Group. 1999. PFC3D (Particle Flow Code in 3 Dimensions). Minneapolis, MN: Itasca Consulting Group. Loveday, B.K. 2004. The use of FAG and SAG batch tests for measurement of abrasion rates of full-size rocks. Minerals Engineering 17(11–12):1093–1098. Loveday, B.K., and D. Naidoo. 1997. Rock abrasion in autogenous milling. Minerals Engineering 10(6):603–612. Loveday, B.K., and W.J. Whiten. 2002. Application of a rock abrasion model to pilot-plant and plant data for fully and semi-autogenous grinding. Transactions of the Institution of Mining and Metallurgy 111:C39–C43. Luckan, P.I., and K. Pillay. 2004. The development of an autogenous model for quartzite by using semi-batch laboratory-scale milling. Laboratory Project 2004 DNC4IP1. Durban, South Africa: University of KwaZulu-Natal. Unpublished. Mishra, B.K., and R.K. Rajamani. 1994. Simulation of charge motion in ball mills. Part 1: Experimental verifications. International Journal of Mineral Processing 40(3–4): 171–186. Morrison, R., and P.W. Cleary. 2004. Using DEM to model ore breakage within a pilot scale SAG mill. Mineral Engineering 17:1117–1124. Morrison, R., P.W. Cleary, and W. Valery. 2001. Comparing power and performance trends from DEM and JK modelling. Pages 284–300 in SAG 2001. Volume IV. Vancouver, BC: University of British Columbia, Department of Mining and Mineral Process Engineering. Napier-Munn, T.J., S. Morrell, R.D. Morrison, and T. Kojovic. 1996. Mineral Comminution Circuits: Their Operation and Optimisation. Brisbane, Australia: Julius Kruttschnitt Mineral Research Centre.

Significance of the Particle-Size Distribution in the Quality of Cements with Fly Ash Additive Viktória Gável* and Ludmilla Opoczky*

ABSTRACT

The fineness of fly ash used as a cement additive influences the quality of the cement. The cement industry characterizes fineness based on the Blaine surface area. However, according to our investigations, the Blaine surface does not reflect exactly the actual fineness and particlesize distribution of fly ash or cements with fly ash added. Cements with fly ash added are characterized by “coarser” and “narrower” particle-size distributions than cements without fly ash added, but both types of cement typically have approximately the same Blaine surface area. The value of specific surface area calculated from the particle-size distribution of fly ash, using the exponential approximation method, gives a better estimate of its real fineness. Given this knowledge, the fineness and, therefore, the quality of cements with fly ash additive can be influenced favorably during cement production. INTRODUCTION

Fly ash—the by-product of coal powder–fired thermal power stations—has been utilized as an additive to cement for several decades. Recently, several problems concerning the production of composite cements that contain fly ash were raised regarding the technology of grinding and the analysis of particle size. The most important issues were (1) the role of the fineness (specific surface and particle-size distribution) of fly ash in the development of the cement quality; and (2) the revision and improvement of methods used for characterization and testing of the fineness (Opoczky and Juhász 1990). This paper presents the principal results of our investigations that were carried out in this field. EXPERIMENTAL MATERIALS AND METHODS

To conduct the experiments, we used various types of cements produced in Hungarian cement plants, as well as fly ash from two Hungarian power stations. The quality of the materials investigated (strength, water demand, pozzolanic activity, etc.) was tested according to the related European and Hungarian standards. Particle composition of the ground products was determined by a Cilas (Marcoussis, France) Model 715 laser granulometer. For the characterization of the particle-size * Research & Development Ltd. for the Cement Industry (CEMKUT Ltd.), Budapest, Hungary 285

286

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

distribution, we used two parameters from the Rosin-Rammler-Sperling-Bennett (RRSB) equation: the fineness number ( x ) and the uniformity coefficient (n). Uniformity coefficient n characterizes the dispersion (width) of the particle-size distribution curve (i.e., the lower the value of n, the “wider” [more disperse] the particle-size distribution). Fineness number x characterizes the fineness of the ground product (i.e., the smaller the value of x , the finer the ground product; Beke 1981).

Rx = e

x – § --· © x¹

(EQ 1)

where R(x) = oversize x = particle size x = fineness number Fineness and degree of dispersion of the initial materials and the ground products were characterized by their specific surface, as determined by the widely used Blaine apparatus (based on measuring the permeability of a packed bed of powder). This is commonly referred to as the “Blaine surface” (S), as it is generally taken to be related to the specific surface area of the particles. Fineness was also calculated from the particlesize distribution data using an exponential approximation method referred to as “calculated surface.” The essence of this calculation is that the individual particles are assumed to be spherical, which is similar to the laser granulometric analysis where the apparatus expresses the size of the particles through the diameter values of the equivalent spheres. In this way, the specific surface area, Smg, for the assembly consisting of different continuous spheres of various sizes can be expressed as x max

S mg

6 = ----- ˜ US

³

1 --- ˜ f  x dx x

(EQ 2)

x min

where x = particle size (i.e., the diameter of the equivalent sphere) f(x) = function describing the particle-size distribution (frequency curve) A single function seldom describes the size distribution exactly; therefore, approximations were made for the individual sections of the curve using either the same type of functions with various parameters or using different types of functions. Additionally, the definite integrals were summarized by sections. Knowing that the smallest particles play a decisive role in constituting the specific surface, and that the individual particle classes are usually not sufficiently narrow as compared to the size of the particles, a more exact result was achieved when using an exponential approximation to the distribution function. The specific surface by particle classes could be calculated using the following formula: mi F F i–1 · - ˜ § ----i – -------'S mgi = -------------m i – 1 © x i x i–1 ¹

(EQ 3)

The exponent of the power function by particle classes also could be calculated as follows:

PARTICLE-SIZE DISTRIBUTION IN THE QUALITY OF CEMENTS WITH FLY ASH ADDITIVE

lg  F i e F i–1 m i = --------------------------lg  x i e x i–1

287

(EQ 4)

where mi = exponent of the power function for the given particle class Fi = cumulative distribution function for the given particle class RESULTS AND DISCUSSION

The degree of dispersion, or fineness, of fly ash used as an additive to cement has a significant influence on the quality of the fly ash and that of the composite cements with fly ash additive (Opoczky 1996; Opoczky and Tamás 2002). The fineness of fly ash is characterized in the cement industry by the Blaine surface, and the fly ash is also qualified by this value. According to our investigations, assessment by the Blaine surface value does not provide adequate information on the actual particle composition or fineness of the fly ash (Opoczky 2001; Gável 2003). This is shown in Figure 1 and Table 1 where the particle-size distribution of two products of about the same Blaine surface, ground clinker and fly ash, are illustrated in the RRSB system of coordinates (DIN [German Institute for Standardization] 66145), accepted in both the European and Hungarian practice. Particle-size distribution of the fly ash that had about the same Blaine surface (~3,500 cm2/g) when characterized by the fineness number ( x = 60 Pm) proved to be much more coarse than that of the ground clinker ( x = 18 Pm). In order to approach the particle-size distribution of the clinker, fly ash had to be ground to a Blaine surface of 2 ~6,000 cm /g. A similar conclusion also resulted in the case of fly ash and cement(s) when the Blaine surface values were compared with those calculated from particle-size distribution measurements data using the exponential approximation method (calculated surfaces) (Table 2). In the case of cement(s) with no additive, the calculated surface values (arrived at from the particle-size distribution measurements data using the exponential approximation method) do not differ significantly from the Blaine surface determined by the permeability method. However, in the cases of fly ash and of cements with fly ash additive, the difference between the Blaine surface value and the calculated surface value is significant. The difference between the Blaine surface values of cement and fly ash can be explained, on the one hand, by a significant difference in their particle-size distribution and, on the other hand, by the fact that the fly ash always contains—in addition to relatively coarse particles—very fine particles of elementary carbon, the presence of which significantly increase the Blaine surface value. The particle-size distribution of fly ash plays an essential role in the quality of composite cements containing fly ash additive. According to our investigations, there is a definite relationship between the uniformity coefficient, n, of composite cements and their water demand. Cements of higher uniformity coefficient (n)—that is, of more “narrow” particle-size distribution—usually require more water. The more narrow the particle-size distribution, the less tightly the particles can pack together, and so more water is required to fill the pores and gaps (Opoczky and Tamás 2002). As the fly ash has more narrow particle-size distribution than the cements, it increases the uniformity coefficient (n) of the cements and simultaneously increases its water demand when being added to the cement (Figure 2). This adversely affects the

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

1 10

99 90 75 50

Oversize, %

36.8 75

25

90

10

95

5

99

Clinker, 3,500 cm2/g Fly ash, 6,600 cm2/g Fly ash, 3,500 cm2/g 1

10

100

Undersize, %

288

1

1,000

Particle Size, µm

FIGURE 1

TABLE 1

Particle-size distribution

Particle-size distribution and Blaine surface Parameters of the RRSB Equation

Denomination of Materials Investigated

Blaine Surface, cm2/g

Ground clinker Original fly ash Ground fly ash

~3,500 ~3,500 ~6,000

TABLE 2

Fraction Composition, %

Fineness Number, μm

n Uniformity Coefficient

0–3 μm

3–32 μm

32–192 μm

~18 ~60 ~19

0.9013 1.0870 1.0835

15.90 3.70 13.20

64.00 34.70 67.70

20.10 61.60 19.10

x

Fineness characteristics of fly ash and cements with fly ash additive Parameters of the RRSB Equation

Denomination of Materials Investigated

Original fly ash* Ground fly ash CEM I 42.5N* CEM II/A-V 42.5N* CEM II/A-V 32.5R* CEM II/B-V 32.5N*

Fly Ash Content, m/m %

Fineness Number, μm

n Uniformity Coefficient

Blaine Surface, cm2/g

Calculated Surface, cm2/g

100 100 0 20 20 35

95 48 19 21 25 24

1.0237 1.0342 0.9953 0.9555 0.9575 1.0249

3,460 3,840 3,570 3,750 3,590 3,340

1,940 3,230 3,620 3,640 3,420 3,190

x

* European Standard: EN 197-1.

strength, workability, and other application properties of the cements and mortars or concretes made of such cements. By applying an adequate fine grinding, the particle-size distribution of the fly ash can be influenced favorably. Interrelations between the fineness number ( x ) and Blaine surface value of fly ash are shown in Figures 3 and 4. From Figures 3 and 4 one can determine what Blaine surface value the fly ash should be ground to in order for its fineness and particle-size distribution ( x – fineness number, calculated surface) to approach the fineness characteristics of the cement without fly ash additive. For example, in order to achieve a fineness number x , 2 ~25 Pm the fly ash should be ground to a Blaine surface value of at least 5,500 cm /g.

289

PARTICLE-SIZE DISTRIBUTION IN THE QUALITY OF CEMENTS WITH FLY ASH ADDITIVE

36 34

1.00

32 30

0.95

28 n Wd

26

0.90

24 0 100

10 90

Fly Ash Clinker

20 80

m/ % m

35 65

FIGURE 2 Change of the uniformity coefficient (n) and water demand (Wd) of the cement, depending on the fly ash content

Fineness Number (x), µm

80 70 60 50 40 30 20 10 0 2,000

3,000

4,000

5,000

6,000

7,000

2

Blaine Surface (S), cm /g

FIGURE 3

Connection between the fineness number and Blaine surface of ground fly ash

4,500 Calculated Surface, cm 2/g

4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 2,000

3,000

4,000

5,000

6,000

7,000

2

Blaine Surface (S), cm /g

FIGURE 4

Connection between the calculated surface and Blaine surface of ground fly ash

Water Demand (Wd), m/m %

Uniformity Coefficient (n)

1.05

290

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

Blaine surface (S) of cement with fly ash content in cm2/g

60

3,000 3,500 4,000

Compressive Strength, N/mm 2

50

40

30

20

10

0 0 10 20 35

3

28 7 Days

m

Fly Ash Content, /m %

FIGURE 5 Change of compressive strength of cement, depending on fly ash content and fineness of cement (co-grinding)

Other examples of this connection are ƒ It is possible to reduce the adverse effect of the fly ash additive on the strength—

particularly the initial strength—of the cement, as shown in Figure 5. ƒ It is possible to reduce the adverse effect of the fly ash additive on the water-

retaining capability of the cement (Figure 6). ƒ It is possible to reduce sulfate-caused expansion of the cement or to produce

cements of increased sulfate resistance (Figure 7). CONCLUSIONS

The fineness of fly ash used as a cement additive has a significant influence on the quality of composite cements containing fly ash. In the cement industry, the fineness of fly ash is characterized through the Blaine surface, which is determined by a permeability method. According to our investigations, the Blaine surface value does not provide adequate information on the actual particle composition of the fly ash. Namely, in the case of having approximately the same Blaine-specific surface value, fly ash is usually characterized by coarser and narrower particle-size distribution ( x = fineness number, n = uniformity coefficient) than cements without additives. We arrived at a similar conclusion when comparing the Blaine-surface and calculatedsurface values of fly ash and cements. Knowing the correlations between particle-size distribution, calculated surface, and Blaine surface of the fly ash, one can establish to what Blaine surface value the fly ash should be ground in order for its fineness and particle-size distribution to approach the fineness characteristics of the cement.

PARTICLE-SIZE DISTRIBUTION IN THE QUALITY OF CEMENTS WITH FLY ASH ADDITIVE

291

Water-Retaining Capability, m/m %

100

90

80

70

60

50 3,200 Cement

2,200 Fly Ash

3,000 Fly Ash

4,000 Fly Ash

Blaine Surface (S), cm 2/g

FIGURE 6

Water-retaining capability of cements with 20% different fineness of fly ash

Fly Ash SBlaine, ~6,000 cm2/g

Fly Ash SBlaine, cm 2/g 1.2

0.5 ~4,000

Expansion, mm/m

Expansion, mm/m

~3,000

0.8

0.4

0

0.4

0.3

0.2 CEM I 42.5

10 Fly Ash Content, m/m %

FIGURE 7

15

S-54 Cement

10

15

Fly Ash Content, m/m %

Effect of different fineness of fly ash on the sulfate resistance of cements in 28 days

Through the proper adjustment of the particle-size distribution of the fly ash, the properties (water demand, water retaining capability, strength, sulfate resistance, etc.) of fly ash containing composite cements can be influenced favorably. REFERENCES

Beke, B. 1981. The Process of Fine Grinding. Volume 1. Developments in Mineral Science and Engineering series. Budapest, Hungary: Akadémiai Kiadó and Martinus Nijhoff/ Dr. W. Junk Publishers. Gável, V. 2003. Description of grinding fineness of fly-ash and cements with fly-ash (in Hungarian). Paper presented at the 20th Cementipari Konferencia, HortobágyMáta, Hungary, October 13–15.

292

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

Opoczky, L. 1996. Grinding technical questions of producing composite cement. International Journal of Mineral Processing 44–45:395–404. ———. 2001. Silicate-chemical properties of fly-ashes. Pages 255–262 in Oilfield Chemistry. Volume 3. Edited by I. Lakatos. Budapest, Hungary: Akadémiai Kiadó. Opoczky, L., and A.Z. Juhász. 1990. Mechanical Activation of Minerals by Grinding: Pulverizing and Morphology of Particles. Budapest, Hungary: Akadémiai Kiadó and Ellis Horwood Publishers. Opoczky, L. and F. Tamás. 2002. Multicomponent composite cements. Pages 559–594 in Advances in Cement Technology: Chemistry, Manufacture and Testing. 2nd edition. Edited by S.N. Gosh. New Delhi, India: Technical Books International.

Modeling Attrition in Stirred Mills Applying Statistical Physics Thomas Neesse,* Johann Dück,* and Friedrich Schaaff*

ABSTRACT

Grinding attrition is a new purification process that can be used for mineral residues. It can be applied to efficiently remove surface-adsorbed contaminants in the particle size range tc, the following is obtained: 2

 1 – Dm f0 t ln § ---- · m f = m f0 + ----------------------------© tc ¹ DE

(EQ 9)

With this equation, the parameters D, E, and tc can be determined using experimental data. The total quantity of the produced fines 'mf during attrition can be determined on the basis of Equations 5 and 8, according to which the increase in the fine particles diminishes over time. The production rate can be calculated according to the following relationship: dm dm f --------- = ---------f dt dt

t=0

1 ------------------1 + t e tc

(EQ 10)

Here a process duration tp is to be defined, at which the production rate of the fines decreases on 0.01. This corresponds to a time interval of tp = 99 tc and is at the same time the interpretation of the parameter tc.

MODELING ATTRITION IN STIRRED MILLS APPLYING STATISTICAL PHYSICS

299

For the total fines production 'mf, the following results from Equation 6: 2

 1 – Dm f0 'm f = 4.6 ----------------------------DE

(EQ 11)

Figure 6 shows the measured fines fraction over ln t for three experimental data sets. The linear progression of this dependence confirms the validity of Equation 9. The parameters of the process equation for the three experiments are listed in Table 2. Equation 5 is thus the result of a physically based model of the attrition process, proceeding from a statistic distribution of the stress intensity in the milling chamber. With this model, it is possible to present the fines production in the form of a dimensionless equation with three parameters. These parameters D, E, and tc must be experimentally determined. MACROPROCESS OF THE GRINDING ATTRITION

In the previous section, grinding attrition was studied as a microprocess. Figure 7 illustrates the model of the macroprocess referring to the entire milling chamber and considering the main process parameters. These are ƒ Size and density of the grinding media dGM, UGM, and the particles of the attrition

material dAM, UAM ƒ Volume ratio of grinding media/attrition material VGM/VAM ƒ Volume fraction of water cv in the milling chamber ƒ Specific mechanical energy input Em referred to as the mass of the attrition material

The modeling of the macroprocess is again based on the assumption of a stochastic movement of the grinding media and the particles. Blecher (1993) showed that energy dissipates in two main portions, one portion at the tip of the stirrer and another one at the inner surface of the mill. Contrary to real comminution in stirred mills, for grinding attrition, there are not any zones with a high local energy dissipation. Due to a high solids content and a moderate energy input, the energy dissipation of the grinding media occurs in an equidistributed manner in the mill chamber. The Reynolds number of the movement in the mill is very high and indicates a turbulent regime. Turbulence in the stirred mill is also the basis of the investigations of Theuerkauf and Schwedes (1999) modeling the movement of the filling of stirred media mills. The turbulent movement is statistically recordable as a random walk of the turbulence elements in all three directions in space. Thus, turbulence is responsible for the impacts between particles and grinding media. In the following discussion, grinding attrition is characterized by a homogeneous turbulence with the mean velocity v' 2 = v . To model the macroprocess, the relations of the mean free path length O with respect to the mean impact time W in the milling chamber must be found. STRESS NUMBER

The macromodel is based on determining the stress numbers for one particle and one grinding body per unit of time. According to Figure 8, a cylindrical cell with the cross-sectional area (rGM + rAM)2 and the length O = v ˜ 't is observed, where a particle moves through with the velocity v during 't. The mean impact number in this volume results from the contact number of the particle and the grinding media, whose central point is located in the cell observed. Here, it should be noted that the impact partners (particle and grinding media) have different sizes.

300

ADVANCES IN COMMINUTION

LIBERATION AND BREAKAGE

Case 1 Case 2 Case 3 Regression

0.28

Δmf

0.24

0.20

0.16 Case 1: Δmf = 0.0399 In t + 0.1498 Case 2: Δmf = 0.0542 In t + 0.0591 Case 3: Δmf = 0.0438 In t + 0.1153 0.12 1

2

3

4

5

In t

FIGURE 6 TABLE 2

Comparison of the nonlinear model with the experimental data Values of the parameters according to Equation 9 for the experimental data Case 1

Case 2

Case 3

D = 1.57 E = 14.8 tc = 0.024 min

D = 2.02 E = 8.22 tc = 0.34 min

D = 1.67 E = 12.6 tc = 0.072 min

The time W between two impacts is W = O --v

(EQ 12)

Z = 1 --W

(EQ 13)

and the impact frequency Z is

The mean free path length of one particle can now be determined in analogy to gas molecules, considering different sizes of particle and grinding media: 1 O = --------------------------------------------2 S  r AM + r GM c GM where rAM = radius of attrited material rGM = radius of grinding media cGM = volume concentration of grinding media

(EQ 14)

MODELING ATTRITION IN STIRRED MILLS APPLYING STATISTICAL PHYSICS

Diameter of the Ring-Type Stirrer Solids Concentration (attrition material + grinding media + water)

Revolution Number

Filling Volume

Size and Density of Particles and Grinding Media

Volume Ratio of Particles and Grinding Media

FIGURE 7

Volume Element for Determining the Stress Number

Scheme of the macroprocess, including the main process parameters

Particle Does Not Hit Grinding Medium

r = rGM + rAM

>r 30% solids were required to produce the target U/F density of >55% solids. Pressure variation had little influence on U/F density. Table 2 shows the typical U/F densities produced by different hydrocyclones at the same feed density. Note that both of the Type 1 hydrocyclones produced higher U/F densities than the Type 2s at equivalent conditions. Mass Recovery to Underflow

Table 2 also shows that mass recoveries to underflow were somewhat different for the different units, indicating the influence of hydrocyclone geometry (long cylinder versus long cone). Across the pressure range tested, only the standard-diameter Type 1 hydrocyclone failed to achieve the target recovery. Also notable was the performance of the standarddiameter Type 2 unit, which exhibited mass recoveries to underflow of around 90%.

LARGER-DIAMETER HYDROCYCLONE PERFORMANCE IN A DESLIMING APPLICATION

327

Typical separation performance

TABLE 1

Type (diameter)

VF,* mm

Sp,* mm

Press, kPa

Density, %

I

d50c

1 (100 mm) 2 (100 mm) 1 (250 mm) 2 (250 mm)

40 41 60 61

25 26 40 38

180 180 180 180

30 30 30 30

0.50 0.38 0.28 0.41

12.57 9.74 10.61 10.73

* VF = vortex finder; Sp = spigot diameter.

70 35% Solids Primary 30% Solids Primary 25% Solids Primary 65

U/F Density, % Solids

60

55

50

45

40 100

150

200

250

Pressure, kPa

FIGURE 6

TABLE 2

Hydrocyclone pressure versus U/F density

U/F density and mass recoveries to underflow for each make (feed density 30% w/w)

Type (Diameter)

Pressure, kPa

VF, mm

Sp, mm

U/F density, % w/w

Mass Recovery to Underflow, %

1 (100 mm)

120 180 240 120 180 240 120 180 240 120 180 240

40 40 40 41 41 41 60 60 60 61 61 61

25 25 25 26 26 26 40 40 40 38 38 38

56.59 60.62 60.99 55.08 53.74 52.38 60.42 60.48 60.38 56.10 54.77 56.89

83.46 79.74 82.91 90.33 90.83 91.04 85.55 84.05 84.32 84.77 82.61 85.09

2 (100 mm)

1 (250 mm)

2 (250 mm)

328

ADVANCES IN COMMINUTION

TABLE 3

MILL DESIGN

Change in cut-point with pressure (30 wt % solids)

Type (Diameter)

Pressure, kPa

VF, mm

Sp, mm

Area Ratio

d50c

1 (100 mm)

120 180 240 120 180 240 120 180 240 120 180 240

40 40 40 41 41 41 60 60 60 61 61 61

25 25 25 26 26 26 40 40 40 38 38 38

2.56 2.56 2.56 2.49 2.49 2.49 2.25 2.25 2.25 2.58 2.58 2.58

12.90 12.57 10.83 9.98 9.74 9.26 11.74 10.61 9.52 11.23 10.73 9.46

2 (100 mm)

1 (250 mm)

2 (250 mm)

Cut-point d50c

As with tests manipulating the feed density, the vortex finder–spigot area ratio was held as constant as possible to highlight the effects of different pressures on the measured cut-point. In all cases, the corrected cut-point decreased as pressure increased. However, the impact of the shorter cone length for the 100-mm-diameter Type 1 was evident in that it failed to achieve the target cut-point of 10 Pm at any of the pressure settings tested. Both makes of 250-mm-diameter hydrocyclones achieved the target within the pressure range tested. Table 3 shows typical results for each unit. Hydrocyclone Throughput/Flow Rate

It was expected that the hydrocyclone flow rate and, hence, throughput would increase with increased hydrocyclone diameter. Table 4 shows that this was true. However, there are higher reported flow rates for the Type 2 units compared with the Type 1 units of the same diameter. This result quantifies the extra capacity of the long-cone units. As expected, Figure 7 shows that the differences in flow rate translated into similar differences in dry throughput as a function of pressure. The results presented in Table 4 and Figure 7 highlight some interesting differences among the hydrocyclones types. As noted earlier, the 100-mm Type 1 hydrocyclone was a long-cylinder unit, whereas the 100-mm Type 2 unit was a long-cone unit. Note that while the intention with the additional cylinder length in the Type 1 unit was to provide similar capacity to a long-cone unit, the results show that the long-cone unit achieved consistently higher flow rate and throughputs at equivalent conditions. The contrast is even more dramatic for the larger-diameter units. It is important to note that the 250-mm Type 1 unit was not fitted with the extra cylinder, so it was of a conventional short-cone design. Clearly, the results on the 250-mm-diameter units show the substantially higher capacity for the long-cone unit. Secondary hydrocycloning tests were also conducted for both types of hydrocyclones using primary hydrocyclone underflow as feed, and the trends in the results were consistent with those described above. A final comparison was conducted at optimum conditions for secondary hydrocycloning. In terms of the cone length, given that only the Type 2 hydrocyclones were directly comparable, it is this comparison that is shown in Table 5. Clearly, both units satisfy all criteria for this assessment; however, in all measures, it is the larger-diameter unit that provided superior performance.

LARGER-DIAMETER HYDROCYCLONE PERFORMANCE IN A DESLIMING APPLICATION

329

Flow rates for different operating pressures (30 wt % solids)

TABLE 4

Type (Diameter)

VF, mm

Sp, mm

Pressure, kPa

Flow Rate, L/sec

1 (100 mm)

40 40 40 41 41 41 60 60 60 61 61 61

25 25 25 26 26 26 40 40 40 38 38 38

120 180 240 120 180 240 120 180 240 120 180 240

3.94 5.04 6.19 4.34 6.56 7.79 11.30 15.50 19.12 14.40 18.20 22.74

2 (100 mm)

1 (250 mm)

2 (250 mm)

40

35

Throughput, tph

30

250-mm Type 2 250-mm Type 1 100-mm Type 2 100-mm Type 1

25 20 15

10 5 0 100

120

140

160

180

200

220

240

260

Pressure, kPa

FIGURE 7

Throughput performance for the units tested

TABLE 5 Comparison of the performance of the Type 2 hydrocyclones (secondary hydrocyclone optimum) Performance Characteristics

Pressure, kPa Feed density, % I Dry throughput, tph U/F density, % Recovery to underflow, % d50c , mm

Type 2 (100 mm) VF 41 mm, Sp 26 mm

Type 2 (250 mm) VF 61 mm, Sp 38 mm

180 33 0.37 10.2 56 88 10.1

180 33 0.34 32.4 65 93 8.1

330

ADVANCES IN COMMINUTION

MILL DESIGN

CONCLUSIONS

The results achieved in this assessment demonstrate that larger-diameter hydrocyclones may be used in fine desliming applications provided that careful attention is paid to setting appropriate operating conditions. In this work, the performance of larger-diameter (250 mm) units at least equaled—and in the optimum secondary hydrocycloning case, exceeded—that of the smaller-diameter (100 mm) standard unit in terms of the performance criteria, namely corrected cut-point of d10 Pm; high-quality separation, as evidenced by a low partition imperfection (55% solids by weight; high mass recovery to underflow (>85%); and high throughput/capacity. These results illustrate the opportunity to realize significant potential gains in circuit capacity, lower circuit footprint, and simpler operating configuration, and hence, lower costs with no loss of metallurgical performance by the use of larger and fewer hydrocyclones in this application. REFERENCES

Chu, L-Y., M-W. Chen, and X-L. Lee. 2000. Effect of structural modification on hydrocyclone performance. Separation and Purification Technology 21:71–86. ———. 2001. Effects of geometric and operating parameters and feed characters on the motion of solid particles in hydrocyclones. Separation and Purification Technology 26:237–246. Kawatra, S.K., A.K. Bakshi, and M.T. Rusesky. 1996. Effect of viscosity on the cut (d50) size of hydrocyclone classifiers. Minerals Engineering 9(8):881–891. Kraipech, W., W. Chen, F.J. Parma, and T. Dyakowski. 2002. Modelling the fish-hook effect of the flow within hydrocyclones. International Journal of Mineral Processing 66:49–65. Lynch, A.J., and T.C. Rao. 1975. Modelling and scale-up of hydrocyclone classifiers. Pages 9–25 in Proceedings 11th International Mineral Processing Congress, Cagliari, Italy. Plitt, L.R. 1976. A mathematical model of the hydrocyclone classifier. CIM Bulletin 69:114.

Selection and Design of Mill Liners Malcolm Powell,* Ian Smit,† Peter Radziszewski,‡ Paul Cleary,§ Bruce Rattray,** Klas-Goran Eriksson,†† and Leon Schaeffer‡‡

ABSTRACT

Dramatic shortcomings of mill liner designs, especially of large semiautogenous grinding (SAG) mills—such as rapid failure, mill shell damage arising from the charge impacting directly on the liner, and unsuitable spacing of lifter bars yielding unfavourable compromises between lifter bar height and liner life — highlight the significance of correct mill liner selection. Liners protect the mill shell from wear and transfer energy to the grinding charge. A careful balance is required to optimise these conflicting requirements. This review serves to highlight these problems and addresses logical and often inexpensive resolutions by considering charge trajectories and liner spacing criteria, in conjunction with liner wear monitoring. An overview of the principal types and materials of liner construction is given, with a focus on liner design based on the best technology available, combined with experience and logical engineering practice. Methods of monitoring the progressive wear of liners and their relation to the performance of the mill are presented. The value of wear monitoring in ongoing liner optimisation and cost saving, through balancing the longevity of the lifters and shell plates and providing reliable comparative data for testing different liner materials and designs, is explained. Wear-testing techniques and their drawbacks and limitations are discussed, along with new tests that are under development. The contribution of advanced computation techniques, such as the Discrete Element Method (DEM), to predict the wear profiles of liners and integrate this information into optimising the overall performance of the mill from a production and cost perspective are considered in some detail. This takes into account the change of the charge trajectories, energy transfer, and milling efficiency as the mill liner wears and the profile changes. It is hoped that this review will enable mill operators to select suitable mill liners, with a view toward decreasing production costs while maintaining mill performance near optimal levels. INTRODUCTION

Poor liner design has a detrimental effect on milling performance and on liner life (Powell 1991a), which can result in a loss of revenue and increased operational costs. Reduced * Mineral Processing Research Unit, University of Cape Town, Cape Town, South Africa † Anglo Research, Johannesburg, South Africa ‡ Department of Mechanical Engineering, McGill University, Montreal, Canada § CSIRO Mathematics and Information Sciences, Clayton, Victoria, Australia ** Castech Solutions, South Fremantle, Perth, Australia †† Mill Linings Systems/Technical Support, Metso Minerals, Ersmark, Sweden ‡‡ Mill Linings, Weir Rubber Engineering, Salt Lake City, Utah 331

332

ADVANCES IN COMMINUTION

MILL DESIGN

milling efficiency can cause excess power usage and decreased recovery of valuable minerals. Excess liner wear results in exorbitant liner materials costs and excessive downtime, which reduces mill availability and impacts on plant throughput. For a plant with a number of mills, this also entails the employment of extra mill relining staff and the risks and costs associated with frequent relining. Optimised liner design can be used to strike the best economic balance between liner life and mill grinding performance, thus enhancing the profitability of a mining operation. Protection of the mill shell from the aggressive impacting and abrasive environment inside a mill is the primary purpose of mill liners. Generally, the care of liners came under the maintenance and engineering department, where the objective was to utilise a liner that lasted as long as possible, or was as inexpensive as possible, or, preferably both. Liners were treated merely as a cost overhead and a cause of downtime, and the maintenance approach has been to reduce the cost while remaining within acceptable downtime constraints. Cost savings led to the development of profile liners and lifter bars, as these dramatically increase the life of the liner. The downtime constraints and high stresses in large SAG mills helped to drive the development of greatly improved liner materials. However, this cost-engineering approach ignored the mill performance and overlooked the other key function of mill liners. The second critical function of a liner is to transfer rotary motion of the mill to the grinding media and charge. After all, the liner is the interface between the mill and the grinding charge. Although literature on the grinding action in mills has been published for 100 years (White 1905; Davis 1919), the first publication on the influence of liner design on the charge motion appeared about 70 years later (McIvor 1983). With the advent in the 1980s of larger SAG mills running in single-stream circuits, it became apparent to the operating staff that the liner was having a significant influence on mill performance. This had been hidden previously by the regular changing of liners over a number of mills in the older plants that had many mills in parallel function. In fact, this is generally still the case in the multistream plants, where mill liner design and selection is only tackled on a cost-consumables basis. However, the gains to be had through good liner design and selection are just as great as on the large SAG mills. This paper discusses recognising problems in liner design and selection in existing operations and reviews liner selection for new applications. TY P E S O F L I N E R S

Design and Structure

The design of a liner is driven by the material of construction and the application, and is limited by casting, moulding, and handling constraints. For large mills with wide inlet trunnions in excess of 1.5 m, liner handling machines are now in common use, and this has allowed the evolution of large integral liner blocks, each weighing up to 1.5 t (Figure 1). This holds great advantage for minimising relining time, as there are fewer blocks to handle. For example, at the Kalgoorlie Consolidated Gold Mines (Western Australia), the number of outer head liners in a 36-ft SAG mill was reduced from 36 to 18 pieces, and in doing so, the time to replace them was reduced by 9 hours at a cost-downtime savings of about US$25,000/hr. In this case, the liners sections are 3.5 t each. For smaller mills, the liners have to be handled and installed manually, so smaller blocks with removable lifter bars are generally favoured (Figure 2). Following is a list of the primary types of liners, including comments on their application, advantages, and disadvantages:

SELECTION AND DESIGN OF MILL LINERS

FIGURE 1

Solid steel liners with integral lifter bars

FIGURE 2

Removable lifter bars

333

1. Solid liners: These types of liners have an integral lifter and liner, as shown in

Figure 1. They have fewer pieces and are easier to install, but they tend to have a high scrap weight, as once the lifter section is worn down, liner performance drops and necessitates change-out. 2. Removable lifter: In a liner with a removable lifter, the lifter can be changed

rather than the complete liner (Figure 2), thereby maximising liner life and assisting in manually relined mills. The drawback is that there are more pieces to be installed, and the liners can move during relining. If they are not well secured against the backing liner, the lifter can shift and work loose; this is especially a problem if the bolts begin to stretch. 3. Grid liners: Pocketed grid liners is a system (that appears to be unique to South-

ern Africa) where the grinding media packs in the grid structure and forms an integral part of the liner (Figure 3). Often the liners have a flat profile, suited to the high speeds (85% to 90% of critical) at which most of the older mills operate. These liners have been demonstrated to be economically unbeatable for highly abrasive ores in small- to medium-size mills (Powell 1991a). They are lightweight and make use of the grinding media hardness to provide an effective wear material. They must be manufactured in manganese steel to wedge the steel balls, but the manganese steel spreads on impact and can make removal difficult. Safety

334

ADVANCES IN COMMINUTION

FIGURE 3

MILL DESIGN

Austenitic manganese steel grid liners

FIGURE 4 Single-direction top-hat liners: an integral liner (left), and bidirectional liners with removable lifters (right)

aspects should be considered because of the risk of balls dislodging when the mill is entered for inspection or relining. The liners require a thorough hosing down to prevent this. 4. Wedged liners: Wedged liners were common in the first half of the last century

but are dangerous to install and no longer used. Liner blocks are now wedged in by bolted lifter bars, which allows simple castings of the liner blocks. 5. Integral wave blocks: These are commonly used in ball mills, and the profile of

the liners has become sufficiently sophisticated to enable the liner profile to be maintained as the liner wears. 6. Uni-direction profiled liners: The lifter has different leading and trailing profiles

(Figure 4). The profile can be better customised to suit mill speed and filling and therefore optimise performance. It allows more material in the lifter for a given base width, but the mill must only run in one direction. 7. High–low double-wave ball mill liners: These liners are a refinement of the wave

liner (Figure 5). This was applied to Cadia Hill gold mine through evaluation of their existing wear profile and wear rate, and it provided a more consistent wear profile through the liner’s working life. The correct wave face angle needs to be calculated and applied because an incorrect angle can lead to ball segregation and loss of grind. An indication of our limited ability to accurately “design” liner profiles is that few liners are optimal at original installation or in the post-commissioning set, and it is imperative as a user to vigorously pursue improvement of the design to get the most out of the liners.

SELECTION AND DESIGN OF MILL LINERS

FIGURE 5

335

High–low wave ball mill liner

Materials

The selection of the construction material is a function of the application, abrasivity of ore, size of mill, corrosion environment, size of balls, mill speed, and so forth. Liner design and material of construction are integral and cannot be chosen out of context. Following is a list of the primary materials of construction, including particular uses and strengths. 1. Austenitic manganese steel (AMS): AMS is used for grid liners generally, in

smaller mills. Its great advantage is that it hardens under stress, yet the substrate remains tough and can withstand extreme impacting without fracture. Its primary disadvantage is that it spreads with impact, so solid liners begin to squeeze together and become extremely difficult to remove, and can damage a mill shell if the stress is allowed to build to an extreme level. 2. Low-carbon chrome moly steel (300 to 370 BHN [Brinell hardness number]):

Generally used for mill liners (autogenous grinding [AG], SAG, and ball) prior to the movement to higher-carbon-content steels. It has excellent wear characteristics with some impact resistance and is generally used for discharge grates where slightly better impact resistance is required, compared to the higher carbon chrome moly steels or for thinner section liners. 3. High carbon chrome moly steel (325 to 380 BHN): This steel is now considered

the main material used for SAG mill liners. There are a number of variations with either different carbon or chrome contents. The variations tend to have a bearing on the size of the liner and its section thickness. There is ongoing development within this area as the size of the liners are outstripping the properties provided by the standard high chrome moly steels. 4. Nihard iron (550 BHN): The use of this type of material generally began with rod

mills and ball mills, where impacts were considered low enough for this brittle yet highly abrasive-resistant wear material to perform well. However, it is considered obsolete in light of the use of high chrome irons and chrome moly white iron.

336

ADVANCES IN COMMINUTION

FIGURE 6

MILL DESIGN

Rubber lining in a ball mill and feed head metal-capped lifter

5. High chrome irons (+600 BHN) chromium iron: This iron is considered to have

superior wear-abrasion characteristics, and it is generally used in rod and ball mills. It is more cost-competitive and more brittle than chrome moly white irons. 6. Chrome moly white irons (600 to 700 BHN): This cast material is considered to

be the ultimate and was developed and used to date for abrasion resistance in milling. It is commonly used in cement mills and in some of the largest ball mills in the world, and where performance has not been improved to date. Rubber Liners

The interplay of a material and its configuration are especially significant in rubber liners. During the last half century, rubber mill linings have been used successfully in secondary and regrind milling applications and are specified today for these new applications (Figure 6). However, now with improved materials and computer-aideddesign programs, they are being used more and more in primary grinding applications. In addition to their abrasion resistance, they also are resistant to most chemicals (Schnarr, Schaeffer, and Weinand 2002). The more technical term for rubber is “elastomer.” A good elastomer for a mill liner would have an elongation of 500% to 600%, which means that it can be stretched five to six times its length without damaging it. In addition, the tensile strength should be around 20.68 MPa (3,000 psi). The third important physical characteristic is hardness, and this should be between 55 and 70 durometer on the A scale. The material used for a rubber mill liner usually consists of a blend of a natural and synthetic rubber. In some applications, the material may be all synthetic. The mixture of the rubber and synthetic materials plus various chemicals and fillers is called a “compound.” Each rubber mill-lining manufacturer has their own recipes for their compounds, as well as their own designation. In designing a rubber lining, the same computer tools as described elsewhere in this paper are used. Whether the lining material is metal or rubber, the same type of comminution is required in the charge; therefore, the same simulation tools can be used with some adjustments for the lining material. For maximum life, rubber performs best with a 90˚ impact, so this is taken into consideration when designing. Many improvements in

SELECTION AND DESIGN OF MILL LINERS

FIGURE 7

337

Configurations of metal-capped rubber liners

rubber compounds have been made over the years, but current research and development is providing nanotechnology, which should further improve wear life in the future. Rubber and Steel Composites. Rubber and steel have been used successfully in many applications. In some cases, the rubber and metal can be separate components (i.e., metal lifter bars and rubber plates). During the last two decades more emphasis has been placed on metal-capped rubber lifter bars (Figures 6 and 7). The material used for a metal cap is similar to that used for a metal lining, but a hardened steel plate can also be used. The joining of the metal and rubber has to be with a chemical bond plus a mechanical type of attachment to ensure a positive fastening of the two materials for the life of the component. With the use of computer simulations and careful inspection of the existing wear profile, a greatly improved liner design can be generated. Thus, all high, or high–low, or lower-sloped lifters are recommended for different applications, as illustrated in Figure 7. Rubber is one-seventh the weight of metal, and in many countries the cost is less, so it is very beneficial to utilise rubber wherever possible. By strategically placing the metal cap material with the minimal amount of metal, the best economy can be obtained. Some applications require a metal leading face only. Others require a metal face and top protection (Figure 7). An important feature of combination linings for ball mills is the configuration of the lining, where the lifting action is transferred to the charge, and therefore will remain constant throughout the life of the lining; whereas solid linings will wear more on the lifting portion and become smoother with less lifting action as the lining wears down. The different wear characteristics of the two materials in Skega Poly-Met (Mill Linings Systems, Metso Minerals, Ersmark, Sweden) make it possible to design a lining that will maintain its profile throughout its life, as illustrated in Figure 8. Magnetic Liners

The lining system in magnetic liners consists of permanent magnets embedded in a rubber moulding. The powerful magnets keep the lining in place without liner bolts and ensure that the lining attracts magnetically susceptible material available in the mill (Figure 9). The particles attracted to the surface of the magnetic lining form a thin, continuous layer in a wave profile. The total thickness of the magnetic lining, including the wear layer, is much less than that of a conventional lining. The mill will thus have a larger effective diameter. The lining configuration is ideal for fine grinding, giving an efficient grinding performance in

ADVANCES IN COMMINUTION

MILL DESIGN

90

60

338

FIGURE 8

Designing steel-capped liners for even wear

Homogeneous Bed of Fine Magnetic Material Coarser Bed of Small Pieces of Magnetic Material Fluid Bed of Fine and Coarse Magnetic Material

S

N

N

S Rubber

FIGURE 9

Steel Shell

Permanent Magnet

Magnetic liner (Metso OreBed)

these applications. A combination of the previously described features has resulted in higher throughput (or lower energy consumption) and, in several cases, a lowering of media consumption by at least 10%. Because of the complicated manufacturing process, the magnetic liner elements are much more expensive than conventional rubber lining, but in an ideal application, the wear on the lining is almost negligible and therefore can give years of trouble-free operation. The limitation for this lining concept is that the magnets are not very resistant to impact because they are brittle. They are suitable in mills of t12 ft diameter using maximum 1-in. balls, and in mills d12 ft in diameter using maximum 112-in. balls. Magnetic liners have been utilised successfully in vertical stirred mills. An example of the application of the Metso OreBed lining is at the LKAB Kiruna iron ore operation in Sweden (K. Tano, personal communication, 2005). A test on the primary ball mill showed that the conventional Poly-Met lining slightly outperformed the OreBed lining in terms of a 0.5-tph higher throughput, so the plant remained with PolyMet. However, they installed the OreBed lining in all their pebble mills, where they have successfully operated for more than 10 years without any maintenance or replacement. Figure 10 shows the liner with the coating of magnetite ore and slurry. It was concluded at the site that the magnetic lining works well for secondary grinding, where abrasion is more important than impact.

SELECTION AND DESIGN OF MILL LINERS

FIGURE 10

339

OreBed liner in secondary ball mill, and a single panel

HOW GOOD IS CURRENT DESIGN?

Liner installations have resulted in variable performance, from outstanding to disastrous. This range indicates the potential of good design and application, as well as the potential for poor installations. In this section, a few case studies will highlight this range, what was identified as the cause, and the lessons learned. Good Examples

Two examples of successful liner design applications that highlight key design aspects are as follows: ƒ Cadia Hill gold mine 40-ft SAG mill—reduced rows: The primary objective was to

reduce the packing of material between the existing high–low liner design. A twothirds row design was installed, which allowed for increased spacing between lifters and thus the use of a larger release angle. Packing was eliminated and the mill was able to run at a higher speed, thereby increasing grinding performance and reducing ball breakage. ƒ Codelco Andina 20-ft ball mill—high–low ball mill liners (white iron): The objec-

tive was to increase liner life, thereby increasing plant availability and reducing running costs. The high–low wave profile liner developed at Cadia was installed in chrome moly white iron. The mill operating performance was not hampered, yet liner life increased by more than 50% over the previous sets of double-wave liners that had been in use. Bad Examples

Some disasters with the emphasis on cause, and the lessons learned, are presented: ƒ Large-diameter SAG mill in South America—pulp lifter design: The outer pulp

lifter had not been designed correctly to allow for ease of removal without the need to remove the shell liners adjacent to it. This arose from a lack of knowledge of how relines are carried out and the importance of timely removal during shutdowns. This highlights the need to discuss liner design changes or new concepts with the maintenance crews and reliners so as to detect design retailing flaws.

340

ADVANCES IN COMMINUTION

FIGURE 11

Bolt-hole cracking

FIGURE 12

Heavily dimpled, peened, and cracked liners

MILL DESIGN

ƒ Large ball mill—white iron liners: The existing liner design had always been con-

structed from high carbon chrome moly steel; however, to increase liner life, the move was made to use chrome moly white iron. However, the liners cracked severely during installation. Bolt-hole details had remained unchanged, but on close examination, the bolt-hole profile was found to not be a tight fit with the liner bolt. This allowed pinpoint loading to occur, which acted like a guillotine, cracking the liners down the bolt-hole centre line, as illustrated in Figure 11. This demonstrates the need for a full reevaluation of the liner design, with close attention to fit faces (curves, angles, etc.), clamping face, lifting lugs, and bolt-hole shape when changing liner material. ƒ Dimpled, peened, and cracked SAG liner: Examples of dimpled, peened, and

cracked SAG liners in a 24-ft SAG mill are shown in Figure 12. The mill was found to have the correct lifter profile and filling, but the feed was excessively diluted (to assist in flushing feed through a poorly designed feed chute). From listening carefully to the mill (in the absence of a proper microphone system), it was concluded that the angle of repose of the charge was abnormally steep, resulting in the toe being very low in the mill. The dilute charge also significantly removed the padding influence of the charge contents. The feed chute had to be reconstructed before this problem could be resolved.

SELECTION AND DESIGN OF MILL LINERS

341

It is difficult to comprehend, given the number of SAG mills in operation globally, that the design of SAG mill liners is still largely troublesome. A number of papers from the SAG 2001 conference (CIM 2001) refer to the following problems: ƒ Candelaria: “Liner design progressed from 72 lifter rail type design, with aggres-

sive face angle, to 36 lifter design with 35 degree face angle. Mill throughput increased by 15%.” ƒ Alumbrera: “Three years worth of trials have been conducted to optimise lifter

geometry. Liner progression from 72 to 48 to 36 rows of lifter bars. Could originally not operate SAG mill at speeds in excess of 70% of critical because of impact on the shell. Operation at reduced speed resulted in low power draw and reduced throughput.” ƒ Los Pelambres: “SAG mill liner progression from 72 rows with 8 degree face angle

to 36 rows with less aggressive 30 degree face angle. Changes allowed for the mills to be operated safely at higher steel loads without increased risk of liner damage. Increased power draw resulted in increased primary mill throughput.” ƒ Collahuasi: “The original SAG mill liner design, Hi-Hi with a 6 degree contact

angle, was changed to a 17 degree angle and later to 30 degree angle. An 11% increase in mill throughput was achieved.” It is common practice to rely on the mill vendor to supply the liner design. This is quite often a contractual requirement with respect to the vendor guaranteeing the performance of the mill. When the mill does not achieve the required throughput rate— because it has to be operated differently from that originally intended in order to prevent liner damage—most vendors appear to adopt a trial-and-error approach, leading to the iterations such as those referred to previously. The net result is a lengthy ramp-up time and loss of production. SYMPTOMS OF POOR LINER DESIGN

When liners are performing to expectations, they are usually left as is. It is usually only when they suffer from premature failure, or come under the cost spotlight, that they are assessed and their performance scrutinised. This section of the paper provides basic guidelines to assessing whether the liners may require reevaluation and identifies possible causes of problems in liners that are known to be problematic. Noisy Mill

A distinct impact rattling indicates that the grinding media is impacting directly on the liner, rather than on the toe of the grinding charge. The consequences of balls impacting on the liner are 1. Greatly accelerated liner wear due to high-energy impacts on the liner. For a 5-m-

diameter mill, these will be in excess of 8 m·s–1; for a 10-m-diameter mill, these exceed 12 m·s–1. This causes high chrome liners to spall and crack, and they may even fracture. This can reduce liner life from a year to less than a few months. Rubber liners can split and tear under these excessive stresses, especially when worn to partial thickness. 2. Reduced milling efficiency from the highest impact collisions occurring against

the liner instead of on the toe of the charge—where effective work can take place 3. Lowered power draw from dilation of the mill charge and the balls returning

energy to the mill shell rather than to the mill charge

342

ADVANCES IN COMMINUTION

MILL DESIGN

4. Ball fracture from the high-velocity impacts directly onto steel—an impulse force

that many balls cannot withstand. This results in a loss of top-size grinding media and reduced milling efficiency, plus increased ball addition costs. 5. Loosening of liner bolts arising from high stress on the liners stretching the bolts.

The damage consequences of rocks impacting the liner are essentially the same as for ball impact, but the impact forces are lower and less detrimental. Autogenous milling takes place when the larger rocks land on smaller rocks and transfer their large energy to the smaller rocks, which can then be sufficient to fracture them (Napier-Munn et al. 1996). As the high-energy collisions are occurring on the liner rather than on the rocks in the toe of the charge, the milling efficiency is reduced. Mill Listening Devices. It is of great advantage to monitor the sound of impacts on the mill shell to warn of direct impacting. This has recently taken a step beyond simple decibel monitoring to full Fourier analysis of the frequency spectrum in research work conducted at the Julius Kruttschnitt Mineral Research Centre (JKMRC; Pax 2001). Broken Liners

Broken liners can result from media impacting directly on the liner, and this is particularly severe for large AG/SAG mills. A high incidence of fracture of lifter bars without corresponding evidence of porosity or casting faults generally indicates impact breakage. Excessive Liner Wear

If the liners have a low or flat profile, this generally indicates excessive slip of the grinding media on the liner. Consequences of excessive slip are as follows: ƒ Liner wear increases substantially and can show evidence of circumferential

grooving of the lining. ƒ There is a substantial loss of energy transfer to the mill. Although a small fraction

of grinding may take place at the liner–ball interface, a 10% loss in energy due to slip results in about a 10% loss of energy transfer to the grinding media. ƒ Reduced mill throughput results from reduced milling efficiency.

Often a lining wears through a favourable profile regime. This can occur in a number of different manners: ƒ Mill throughput drops markedly when new liners are installed. This can indicate

an unfavourable new profile, or an excessively thick new lining—to counteract excessive wear from poor liner design. ƒ Mill throughput peaks during the liner life, usually at the end of the life of the

liner. This is often a symptom of oversized lifter bars, or lifters with too vertical a face angle. ƒ Mill throughput decreases towards the end of the liner life, and the liner has a

flattened profile—a sign of excessive slip, and the liner should be replaced sooner. An incorrect mill product size can result from the incorrect tumbling action within the mill: ƒ Primary mills require a vigorous action with high-energy impacts to fracture the

ore. If the action is primarily gentle cascading, then a fine product and low throughput would result.

SELECTION AND DESIGN OF MILL LINERS

343

Rate of Production of Fines for Different Liner Configurations at Two Mill Speeds 0.80 80% 90%

0.78

0.75 0.70 0.71

0.64

0.64

0.55

0.57 0.55

0.60

0.60

70 mm, 90˚

0.60

70 mm, 50˚

Rate, kg/min

0.65

0.56

0.50 0.45

FIGURE 13

40 mm, 50˚

Grids

Smooth

70 mm, 70˚

40 mm, 70˚

Grids

Smooth

0.40

Influence of liner profile on mill performance

ƒ Regrind mills require a cascading action to maximise the frequency of abrasion

interactions. High-energy impacting wastes energy, reduces the rate of abrasion interactions, and reduces the grinding pressure, thus reducing the ability of the mill to produce fines. This can result in a high recirculating load of oversized particles and a reduced mill throughput, as limited by the required product size. P I L O T TE S T S O F T H E I N F L U E N C E O F L I N E R D E S I G N

It can be difficult to assess the influence of liner design upon mill performance in a production environment, as it tends to be a small influence superimposed upon a number of operational variations, especially in semiautogenous milling. Standard pilot tests do not account for the liner design; they use a standard liner profile for all testwork, generally designed to give adequate lift to the charge at standard mill operating conditions—75% of critical speed and 25% filling for autogenous/semiautogenous milling. In this section, a test procedure is presented that can be applied to batch pilot milling. To assess the direct influence of liner design on mill efficiency, a 1.8-m-diameter pilot mill was utilised (Powell and Vermeulen 1994). The use of batch milling meant that reasonable size samples (6 PLI 0–3 PLI 3–6 PLI >6 PLI 0–3 PLI 3–6 PLI >6

INSTRUMENTATION, MODELING, AND SIMULATION

RQD 0–30 V-FS V-FH D-FS D-FM

RQD 30–60 RQD >60 V-CS V-MM V-CM V-MH V-CH D-CS D-CM D-CH

IT-MM

IT-CH IT-MH YT-CH

Sixteen ore-hardness domains

High Grade Medium Grade Low Grade Young Tonalite Intermediate Tonalite Diorite Ultimate Pit V-FS V-FH V-CS V-MM V-MH V-CM V-CH D-FS D-FM D-CS D-CM D-CH IT-MM IT-MH IT-CH YT-CH

FIGURE 7

Batu Hijau ore body description in 16 ore hardness domains

Historical Models

Several mill throughput models were developed from 2001 to 2004, using the techniques outlined in the following subsections. Simple Regression Modeling (2002). Basic multiple linear regression models were generated by fitting several months of historical mill throughput data to daily proportions of each ore domain. Initial models were based on simple lithology separation and later subdivided as RQD was used to further delineate lithology. The models generated provided reasonable throughput indicators; however, they did not account for throughput variations resulting from differences in mill feed size distribution. Bond Work Index Modeling (2002). Bond Work Index–based modeling was used to determine mill power as a function of work index:

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION

469

mill power = fn(WiC) + fn(Wirm) + fn(Wibm) where WiC = Bond Crusher Work Index WiRM = Bond Rod Mill Index WiBM = Bond Ball Mill Index This model places the greatest emphasis on ball mill power and assumes fixed feed size to the SAG mill for all ore types. Throughput predictions produced were significantly different to actual plant performance. This was particularly true for tonalite ores, which have the lowest actual mill throughput rates and also the lowest Bond rod and ball mill work indices. The tonalite ores are generally much coarser than the other lithologies, as demonstrated by their inherently higher RQD. This modeling method was similar to models used to design the Batu Hijau grinding circuit. Figure 8 shows that the Bond Work Index modeling approach is not valid for modeling SAG mill throughput rates as it does not describe SAG mill breakage rates or account for variations in feed size to the mill. BOCCOST Empirical Modeling (early 2003). BOCCOST, or blasting optimization crushing conveying optimizing SAG throughput, is a database that captures approximately 200 variables from blasting, mining, and milling operations and assigns them to a common point in space in the geological model in real time (Pontin and Setiawan 2002). The intention was that the relationships between these variables should be determined with a view to optimize the overall blasting, blending, and crushing process to maximize mill throughput. Interpretative analysis of about fifteen variables from the BOCCOST database was used to model the following: ƒ The relationship between size of material delivered to the primary crusher and pit

geology and blasting practices ƒ The relationship between SAG mill feed rate, pit geology, and blasting practices

Again, essentially basic multiple linear regression analysis was used over three defined ore-body zones. The zones were differentiated broadly based on major differences in faulting and lithology that corresponded with perceived mill throughput zones. Initial data screening was used to isolate the variables that had the most impact on mill throughput and blasted rock size. Model output was used as a tool to assist short- and medium-term mine planning. Prediction of blasted rock size from BOCCOST was viewed to be more accurate than the throughput predictions. JKSimMet Modeling (early 2003). JKSimMet modeling, based on mill surveys and nine drop-weight results across RQD and lithology-defined ore domains, used the JKSimMet ta estimation of mill feed size P80. This model produced similar results to the simple linear regression model. JKSimMet Modeling (late 2003). JKSimMet modeling used mill feed size estimation from the BOCCOST empirical blast fragmentation model. JKSimMet modeling from early 2003 was used to generate throughput versus mill feed F80 relationships for “hard,” “medium,” and “soft” hardness ores of the following form: throughput = M u SAG F80b

470

ADVANCES IN COMMINUTION

INSTRUMENTATION, MODELING, AND SIMULATION

Empirical Ore Delivery Method JKSimMet Model (early 2003) Bond Work Index Model

8,000 7,000

Mill Throughput, tph

6,000 5,000 4,000 3,000 2,000 1,000

FIGURE 8

Volcanic RQD >50

Volcanic

Diorite RQD >50

Diorite

Young Tonalite RQD >50

Young Tonalite

Tonalite RQD >50

Tonalite

0

Mill throughput predictions

The constants M and b were determined based on functions of RQD for each ore hardness. SAG feed F80 was estimated from a simple log fit of historical measured SAG F80 and BOCCOST blasted rock P80. This relationship is very robust, as demonstrated in Figure 9. Throughput prediction from this method was an improvement over previous modeling attempts in that it allowed for variable mill feed size distribution; however, it still lacked adequate accuracy for incorporation into the ore block model revenue calculation. INTEGRATED MILL THROUGHPUT MODELING

In early 2004, Batu Hijau contracted MMPT-AP to assist with ore characterization and produce domain-specific mill throughput models for use in long-term production forecasting and evaluation of mill expansion options. Revised models were required due to pebble circuit flowsheet changes in August 2003, which produced a step change improvement in mill performance, such that previous throughput models were no longer valid. MMPT-AP conducted modeling of the full comminution process using blasting, crushing, and grinding models to calculate SAG mill feed size, throughput rate, and flotation feed grind size (Jankovic et al. 2004). The simplified model is presented in Figure 10. This model consists of the following combined mechanistic models: ƒ Blast fragmentation model ƒ Primary crusher model ƒ Milling circuit model (SAG mill, ball mill, and cyclone models, pebble crusher in

closed circuit) The results of the simulations combining the blast fragmentation model, the primary crusher model, and the complete SAG/ball/pebble/screen models were divided into coarse/hard, mean, and fine/soft for each lithology. The inputs to these models were ƒ The rock mass data consisting of RQD, which was used to estimate the volumetric

joint distribution and then converted into a mean in-situ block size ƒ UCS inferred from PLI data ƒ Drop-weight test data and their correlations to PLI

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION

471

85

SAG Feed F80, mm

80 75 70 65 60 55 50 45 17-Jun-05

8-May-05

29-Mar-05

17-Feb-05

8-Jan-05

29-Nov-04

20-Oct-04

10-Sep-04

1-Aug-04

22-Jun-04

13-May-04

3-Apr-04

23-Feb-04

14-Jan-04

40

Date Average SAG F80 mm Predicted SAG F80 Based on BOCCOST Shovel P80 7 per Moving Average (Average SAG F80 mm) 7 per Moving Average (Predicted SAG F80 Based on BOCCOST Shovel P80)

FIGURE 9

Actual and predicted SAG F80 and BOCCOST model P80 comparison

Ore Characterization

Blast Design

MMPT Blast Fragmentation Model

Lithology Zones Rock Strength - PLI - DWi, A×b, ta - WiC, WiBM, WiRM, Ai Rock Structure - RQD, Mapping

ROM Ore Size Distribution

Primary Crusher Model (JKSimMet/MMPT)

SAG Feed Size Distribution

Grinding Circuit Model (JKSimMet)

tph

FIGURE 10

P80

Schematic of the modeling approach

The blast fragmentation model assumed a blast design with bench heights of 15 m, using blast-hole diameters of 311 mm, and a burden and spacing of 7 and 6 m, respectively. The polygon and shovel P80s were used to validate the model predictions. The envelopes of ROM size distributions (soft/fine, mean, hard/coarse) were obtained using Monte Carlo sampling followed by model simulation. Monte Carlo sampling refers to the traditional technique for using random or pseudo-random numbers to sample from a probability distribution. These techniques are applied to a wide variety of complex problems involving random behavior. Each Monte Carlo iteration involves obtaining an estimate of each of the input variables to the model based on their mean value and standard deviation.

472

ADVANCES IN COMMINUTION

INSTRUMENTATION, MODELING, AND SIMULATION

The ROM size distribution is calculated using the model and these randomly created input variable parameter estimates. This process is repeated 1,000 times to obtain an envelope of possible ROM size distributions based on the variation in the input data. The coarse, mean, and fine ROM size distributions are the upper, mean, and lower size distributions based on the 5% and 95% confidence intervals. The ROM fragmentation is used as input to the primary crusher model to predict the feed size for SAG mills. Simulations were conducted with a crusher closed side setting of 125 mm. The primary crusher product size distributions were used as feed to the SAG mill model, and simulations of the entire grinding circuit were conducted. The resulting simulated throughputs for each ore type were “blended” according to the reported oreblending information and compared to the actual average daily throughput. Preliminary Models for Ore Characterization

Metso modeling was conducted on three different sets of ore domains to determine the level of detail required to produce a representative and accurate mill throughput model. The models were validated and compared to daily production information: ƒ Eight ore domains used historically based on four lithologies and two ranges of

RQD (40%) ƒ Nine ore-hardness domains (matrix) based on three ranges of RQD and PLI but

not lithology ƒ Twenty-four domains based on four lithologies, two ranges of RQD (40%), and three ranges of PLI (6) Model prediction compared to actual production is shown in Figure 11. The 24domain model produced the best fit, indicating that a combination of lithology, RQD, and PLI was required to accurately model mill throughput rates. The 9-domain model results show poor sensitivity to high and low throughput rates. Overall, the results indicated the need to retain lithology as a parameter in ore domain definition. The results of this work led to the application of three RQD and three PLI ranges to each of the four lithologies to produce 36 domains. Spatial and quantitative ore characterization reduced the 36 domains to 16 domains, as shown in Figure 6. Predicting Mill Throughput for Varied Blast Designs

The Metso modeling approach was applied to the 16 ore domains shown in Figure 6, using two different drill and blast regimes: ƒ Historical powder factor, little variation between ore types (0.25–0.35 kg/t) ƒ Drill and blast cookbook, high powder factors in harder ores (up to 0.54 kg/t)

Blast-fragmentation simulations indicate a large variation in ROM fragmentation for each lithology as a consequence of different rock strength (PLI), RQD, and blast design. Figure 12 shows the simulated ROM for volcanic lithology. Intense blast designs were implemented in mine operations in 2004. The resultant improvement in ROM fragmentation size in harder ores is shown in Figure 13. The points plotted are annual averages for each of the 16 ore domains. The reduction in variability of primary crusher feed size improved crusher operation and increased average power draw for harder ores as the crusher can be operated consistently at the target closed side setting of 120 mm. The primary crusher product (SAG mill feed) was obtained by “passing” the ROM ore through the primary crusher model. This approach produced a SAG mill feed size distribution for each of the 16 ore domains (and two blast designs).

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION

473

7,000

6,500

Model, tph

6,000

5,500

5,000 8-Domain R2 = 0.16 24-Domain R2 = 0.33 9-Domain R2 = 0.35 4,500 4,500

5,000

5,500

6,000

6,500

7,000

Actual, tph

FIGURE 11 Preliminary Metso models versus actual November 2003–March 2004 mill production (4-day moving average composites)

100 90

Cumulative % Passing

80 70

V-CHc V-CMc V-MHc V-MMc V-FHc V-FSc V-FS hb V-CSc V-CS hb V-MM hb V-MH hb V-CM hb V-CH hb

60 50 40 30 20 10 0 1

10

100

1,000

Size, mm

FIGURE 12

ROM volcanic (hb = blast designs from the cookbook; c = old blast design)

10,000

474

ADVANCES IN COMMINUTION

INSTRUMENTATION, MODELING, AND SIMULATION

400 Year 2003, Low Powder Factor Blast Designs Year 2004, Intense Blast Designs

ROM Fragmentation F80, mm

350

300

250

200

150

100 0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.55

Powder Factor, kt/t

FIGURE 13

Actual ROM fragmentation improvement from intense blast designs

Domain-specific ore breakage parameters for input to a SAG mill model were obtained using the JKTech drop-weight test Aub result versus point load index (Is50) relationships from Figure 2. This approach allowed evaluation of the variability of breakage rates and indicated a “possible throughput range” for each domain and blast design (powder factor) simulated. Additionally, the “average throughput” for each domain and blasting powder factor was determined based on “best fit” to actual daily throughput. The average-throughput results presented in Figure 14 show that a 10%–15% mill throughput rate increase could be expected in the harder ore types with more intense blasting. Model Validation

The 16-domain model mill throughput rates were applied to the daily ore delivery blends for year 2004 and compare well on a monthly/annual basis, as shown in Figure 15. Model prediction errors were estimated using a sum-of-squared-error method, weighted for total tons milled in each time interval. The results shown in Table 3 and Figure 16 indicate that estimation of mill throughput for weekly ore delivery could be utilized to optimize short-term mine plans. The daily mill throughput prediction is inherently inaccurate as the daily mine ore delivery will not match average RQD, PLI, and powder factor used to predict the average throughput rate for each domain. Additional external effects, such as the particle size segregation on mill feed stockpiles and the nonlinear effect of ore blending on mill throughput, cannot be accounted for in the model, thus, further increasing the error of estimate for shorter time periods. APPLICATION OF THE MODEL

The 16-domain model was integrated with the ore control block model to assign a mill throughput rate factor as “mill run time required to process each ore block” (tons per

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION 8,000

475

1.2

1.0 Mill tph (Metso Model)

6,000 0.8

5,000

4,000

0.6

3,000 0.4

2,000 0.2

Drill and Blast Powder Factor, kg/t

7,000

1,000 Volcanic

Diorite

Tonalite

0

Historical Mill tph Historical Powder Factor

FIGURE 14

YT-CH

IT-CH

IT-MH

IT-MM

D-CH

D-CM

D-CS

D-FM

D-FS

V-CH

V-CM

V-MH

V-MM

V-CS

V-FH

V-FS

0.0

Intense Blast Mill tph Intense Blast Powder Factor

Effect of blast powder factor on mill throughput

8,000

Total Mill Throughput Rate, tph

7,000 6,000 5,000 4,000 3,000 2,000 Actual Daily Production 16-Domain Model Prediction

1,000 0 1-Jan-04

FIGURE 15

1-Apr-04

1-Jul-04

30-Sep-04

31-Dec-04

Actual mill throughput for 16-domain model predictions for 2004

block/mill tons per hour). This mine scheduling was simplified to selection of sufficient ore blocks to match forecast mill run-hours for a time period. This approach also enables optimization of ore delivery on the basis of both payable metal content (dollar index) and production rates. The 16-domain model was also used to evaluate mill throughput rates during a recent mill expansion study based on the properties of the incremental ore added to the ore delivery plans to feed the expanded plant.

476

ADVANCES IN COMMINUTION

TABLE 3

INSTRUMENTATION, MODELING, AND SIMULATION

Sixteen-domain model predictions for 2004 Average, tph

Actual production Annual model prediction Monthly Weekly Daily

Average Error, tph

Average Error, %

9 258 401 578

0 4 7 10

6,062 6,070

8,000

+10%

7,000

Model, tph

–10%

6,000

5,000

Daily Weekly Monthly Annual 4,000 4,000

5,000

6,000

7,000

8,000

Actual, tph

FIGURE 16

Model predictive performance

CONTINUOUS THROUGHPUT MODEL

The 16-domain mill throughput model is useful for monthly and annual throughput forecasting as the average properties of the ore delivered over these longer time periods is generally close to the average RQD and PLI within each domain. The model output of an average throughput rate per domain does not account for throughput variability within the domains; hence, significant step changes in throughput prediction occur at the domain boundaries. To be able to improve the accuracy of the mill throughput predictions on a shorter time scale, a continuous model was required that could accept any combination of RQD, PLI, and blast powder factor per domain to reflect ore blocks delivered on a daily basis.

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION

FIGURE 17

Continuous model user interface

16-Domain Model

Continuous Model

8,000

8,000

7,000

7,000

6,000

6,000

5,000

5,000

Mill tph

Mill tph

477

4,000 3,000

4,000 3,000 2,000

2,000 1,000

0–30 30–60 RQD

0 0–3

3–6

>60 >6

PLI

FIGURE 18

1,000 0

1.5

3

4.5

6

7.5

75

45 60

15 30 RQD

PLI

Sixteen-domain model and continuous model for volcanic ore lithology

Metso conducted more than 40 complete model runs to generate multiple fragmentation and mill throughput predictions per domain and converted the results into a userfriendly software package, as shown in Figure 17. The user can input any combination of RQD, PLI, and powder factor per domain to reflect specific ore parcels. Figure 18 illustrates the difference in the prediction resolution between the 16-domain model using a single throughput value per domain and the continuous model. Batu Hijau engineers have found the continuous model to be informative but not entirely practical for monitoring of long time periods (monthly/annual), as the model is coded into the software developed by Metso and hence cannot be integrated into a spreadsheet calculation. FUTURE MODEL REFINEMENTS

The path forward for refinement of ore domain characterization and mill throughput modeling includes additional comminution testwork on drill cores, updates to the geological model from in-fill drilling, and revisions to the mill throughput models based on the revised geology. Future refinement possibilities are as follows: ƒ Development of a continuous function equation for each ore domain that can be

used in a spreadsheet rather than using the continuous model program. Some work has started in this area to allow throughput prediction based on variation in RQD within each ore domain.

478

ADVANCES IN COMMINUTION

INSTRUMENTATION, MODELING, AND SIMULATION

ƒ Refinement of the model for periods where peaks and dips are not picked up.

More than 80 additional modified drop-weight tests on 2004 and 2005 in-fill drill samples will be conducted in 2005 to boost model integrity. ƒ Modeling of ball mill and pebble crusher limiting ore—requires additional ball

mill work index tests to produce a spatially representative ball mill work index model. Some lower-grade ores appear to have a significantly higher ball mill work index based on plant performance and testing of grabs samples. The geological model is not complete in low-grade ores. ƒ Development of a blend model to describe the nonlinear impact of blending on

the throughput rate estimates from each ore source. ƒ Determination of the impact of mill throughput on flotation feed size distribution

and consequently on flotation recovery. Increasing grind size with throughput has variable impact for Batu Hijau ore, and significant losses can be incurred in some of the softer ore types. This will enable an improved revenue cutoff relationship to be determined to truly optimize blending and ore delivery strategy for maximizing revenue. CONCLUSIONS

An ore-characterization and mine-to-mill throughput modeling approach has been successfully applied at Batu Hijau. This has produced significantly better mill throughput rate predictions than previous modeling has allowed. The ore domain definition and modeling approach relies on definition of ore domains based on ore lithology, rock structure (RQD), and rock strength (PLI) and takes the in-situ ore through blasting, crushing, and milling models to produce an accurate prediction of mill throughput for medium- to long-term production forecasting. Short-term predictions are influenced by external factors, such as particle-size segregation and residence time on process surge piles, and the nonlinear impact of blending. Improved ore definition and understanding of the Batu Hijau resource has resulted in the following benefits: ƒ Enormous improvement in communications between the mine and the mill at

Batu Hijau. Mine engineering, operations, and geology personnel understand the variables that drive mill throughput. The mine and the mill work with a common purpose. ƒ Development of a blasting “cookbook” for generation of an optimum product size

for feed to the mill ƒ Accurate prediction of blast fragmentation size, which is the key to accurate pre-

diction of SAG mill feed size (F80) ƒ Generation of a throughput model that has superior accuracy for annual and

monthly throughput budgeting purposes and evaluation of mill expansion options ACKNOWLEDGMENTS

The authors would like to thank the PT Newmont Nusa Tenggara engineering, operations, and management team at Batu Hijau for their efforts and continued support of this program, and for their permission to publish this paper. BIBLIOGRAPHY

Butcher, A., and W. Valery Jr. 2005. Establishing the links between ore characteristics and crushing and grinding performance. Paper presented at SME 2005 Conference, Salt Lake City, UT, 28 February–2 March.

BATU HIJAU MODEL FOR FORECAST, OPTIMIZATION, AND EXPANSION

479

Flores, L. 2005. Hardness model and reconciliation of throughput models to plant results at Minera Escondida LTDA, Chile. Paper 16 in Proceedings of Canadian Minerals Processors, 2005. The Canadian Institute of Mining, Metallurgy and Petroleum, ISBN 1-894475-47-X. Grundstrom, C., S. Kanchibotla, A. Jankovic, and D. Thornton. 2001. Blast fragmentation for maximising the grinding circuit throughput at Porgera gold mine. Pages 383– 399 in Proceedings of the 27th Annual Conference on Explosives and Blasting Technique. Orlando, FL: International Society of Explosives Engineers. Hart, S., W. Valery Jr., B. Clements, M. Reed, S. Ming, and R. Dunne. 2001. Optimisation of the Cadia Hill SAG mill circuit. SAG2001: International Conference on Autogenous and Semiautogenous Grinding Technology, Vancouver, BC. Jankovic, A., D. La Rosa, W. Valery Jr., and Y. Tai. 2004. Throughput Model for Forecasting Production at Batu Hijau. Final report to PT Newmont Nusa Tenggara. Asia Pacific, Australia: Metso Minerals Process Technology. Kanchibotla, S.S., W. Valery Jr., and S. Morrell. 1999. Modelling fines in blast fragmentation and its impact on crushing and grinding. Explo’99 Conference, Kalgoorlie, Western Australia. Lam, M., A. Jankovic, W. Valery Jr., and S. Kanchibotla. 2001. Increasing SAG mill circuit throughput at Porgera gold mine by optimising blast fragmentation. SAG2001: International Conference on Autogenous and Semiautogenous Grinding Technology, Vancouver, BC. Morrell, S. 2000. Point load/drop-weight test correlation. Progress Report JKMRC/ AMIRA. Project D438 A, confidential to sponsors. Australian Mining Industry Research Association, Melbourne. Morrell, S., and W. Valery Jr. 2001. Influence of feed size on AG/SAG mill performance. SAG2001: International Conference on Autogenous and Semiautogenous Grinding Technology, Vancouver, BC. Morrell, S., W. Valery Jr., G. Banini, and S. Latchireddi. 2001. Developments in AG/SAG mill modelling. SAG2001: International Conference on Autogenous and Semiautogenous Grinding Technology, Vancouver, BC. Pontin, D., and L.E. Setiawan. 2002. BOCCOST modelling—tracking and relating mine and mill performance indicators in real time to increase SAG mill throughput. Value Tracking Symposium AusIMM (October). Valery, W., Jr. 1997. A model for dynamic and steady-state simulation of autogenous and semi-autogenous mills. Ph.D. dissertation. Brisbane, Australia: JKMRC, University of Queensland. ———. 2004. Process integration and optimisation in aggregates production. Paper presented at the 2nd International Seminar on Construction Aggregates. Campinas, Brazil, October 25–28. Valery, W., Jr., et al. 2001. Mine to mill optimisation and case studies. Paper presented at VI Southern Hemisphere Conference on Minerals Technology, Rio de Janeiro, Brazil, May 27–30. Valery, W., Jr., D. La Rosa, and A. Jankovic. 2004. Mining and milling process integration and optimisation. Paper presented at SME 2004 Conference, Denver, CO, February 23–25.

The Use of Process Simulation Methodology in Process Design Where Time and Performance Are Critical Kent T. Tano,*† Bertil I. Pålsson,† Johanna Alatalo,* Lars Lindqvist‡

ABSTRACT

In 2004, LKAB started a basic engineering study to expand its existing production lines at the Malmberget mine site. Current limitations in the underground mine capacity require the use of ore from other mine sites, which results in varying ore properties regarding grindability and chemical composition. It was necessary to determine if the required particle size from an agglomeration point of view could be obtained with extreme ore types by design in a robust process and proper control strategy. In addition, project time constraints forced a decision on final design of the ore beneficiation process to be based on a combination of pilot-scale campaigns and process simulations. Steady-state simulations using ModSim software are used extensively to study process performance for different ore-type feed and different design parameters, such as mill dimensions and varying feed particle size. Population balance models of the ball mills are combined with simple flowsheet models of magnetic separation. Model parameters are derived from pilot-plant data and, when possible, from existing circuits. Dynamic process simulation (Dymola) based on ModSim data is used to evaluate necessary instrumentation as this called for early decisions in the ongoing project. The dynamic simulator will be used to parameterize the progressive, integration and derivative (PID) function controllers and also as a tool in the education of production personnel. The final goal is to facilitate and minimize startup of the plant and to reach design capacity rapidly. INTRODUCTION

The iron ore market is expected to grow in the coming years, mostly due to the tremendous growth in China. Most steel producers in the blast furnace industry, as well as those in the direct reduction route of iron making, have decided to invest in capacity expansions or plan to do so. To maintain market share, LKAB must increase its pelletizing capacity. Existing plants are run at their maximums, and investment in new pellet capacity is a necessity. As a first step, a decision was made in November 2004 to invest in a new pellet plant at Malmberget. The application for an environmental permit at this site has already been approved by the authorities, thus, actual startup production near the * LKAB, Research and Development, Malmberget, Sweden † Luleå University of Technology, Division of Mineral Processing, Luleå, Sweden ‡ Optimation AB, Luleå, Sweden 481

482

ADVANCES IN COMMINUTION

INSTRUMENTATION, MODELING, AND SIMULATION

end of 2006 is promising. The total production capacity of pellet and sinter fines at the Malmberget operation is today limited by the production system for run-of-mine ore. An immediate increase in pellet production without dropping the production of sinter fines will therefore need ore transported from LKAB’s other mine site (Kiruna), approximately 100 km away. This situation will most likely last for only a few years until present bottlenecks at the Malmberget mine are eliminated. This must be taken into account in designing the beneficiation process. The differences in ore mineralogy and chemical composition for the Malmberget and Kiruna ore will put extra control demands on the grinding and mixing of pellet feed concentrates. The ore from Kiruna has to be ground to a finer size for liberation, approximately 80%