Energy Efficient Mineral Liberation Using HPGR Technology: Valid Solutions Space

Energy efficient mineral liberation
using HPGR technology
Valid
solutions
space
Search Space
Constraints
PhD
Path
by
Michael John Daniel

B.Sc.(Eng), M.Sc. Eng (Minerals Processing)
A thesis submitted at the University of Queensland in fulfillment of the requirements for

the degree of Doctor of Philosophy.
Julius Kruttschnitt Mineral Research Centre,

Department of Mining, Materials and Minerals Engineering,
University of Queensland.
June, 2007
STATEMENT OF ORIGINALITY
I declare that the work presented in this thesis is to the best of my knowledge, original,
except as acknowledged in the text, and that the material has not been submitted, either
in whole or in part, for another degree at this or any other university.
The subject matters that comprise original contributions to this field of knowledge are:
• To quantify the energy efficiency of comminution in circuits employing HPGR by

comparing these circuits to conventional circuits which produce the same product.
• To critically evaluate the Bond methodology of determining energy requirements

for milling, using the locked cycle tests and third theory equation in comparison to
direct energy measurements.
• To test the hypothesis whether HPGR promotes or enhances mineral liberation and
micro-cracking when compared to conventional comminution processes producing
the same product.
• To develop flowsheets that demonstrate step change reductions in energy

requirements whilst at the same time achieving improved mineral recovery that
could be realised utilizing solely HPGR technology for crushing and grinding.
• To better understand comminution energy and sustainability, by reviewing energy

efficiency drivers or “eco-efficiency” within the context of economics “dollar
cost” and energy balances “energy cost”.
• To quantify global comminution energy consumption in respect to the production

of various metaliferous and non-metaliferous commodities so as to identify which
areas within the minerals industry are most likely to be effected by matters
concerning eco-efficiency and increasing “dollar cost” of energy.
M. J. Daniel
June, 2007
Energy efficient mineral liberation using HPGR technology

STATEMENT OF CONTRIBUTION TO PUBLISHED WORK
Listed below are references of papers published by the candidate that report the results of
this thesis. Daniel (2006a) was presented at the 2nd international JKMRC student
conference, and Daniel (2006b) was presented at the XXVIII IMPC, Istanbul Turkey. A
co-authored paper by Shi, Lambert and Daniel (Shi et al 2006) was presented at SAG
2006 in Vancouver, Canada.
PUBLICATIONS BY THE CANDIDATE RELEVANT TO THIS

THESIS
Daniel, M.J., 2006a. The researcher who rummaged through Bond’s toolbox., Second
international JKMRC student conference, Brisbane, 7-8 March 2006.
Daniel, M.J., 2006b. Measurement of electrical energy consumed during a Bond ball mill locked
cycle test., XXVIII IMPC, 3-8 September 2006, Istanbul, Turkey.
JOINTLY-PUBLISHED WORK BY THE CANDIDATE RELEVANT

TO THIS THESIS
In this paper, the candidate completed all of the experimental work and data analysis for
the comminution component of this work that included novel energy efficiency
measurements and reference to the reduction in so called Bond work indices of HPGR
products.
Shi, F., Lambert, S., Daniel, M.J., 2006, Measurement of the effect of HPGR treating platinum
ores. SAG 2006, Vancouver, 23-27 September 2006.
RELEVANT PUBLICATIONS NOT FORMING PART OF THIS

THESIS
Daniel, M.J., 2002. High Pressure Grinding Rolls Model Verification and Scale-up. 11th Annual
JKMRC Conference, Keeping afloat in a changing world, pp 67-88.
Daniel, M.J., 2003. HPGR Model verification and scale-up. Masters thesis, University of
Queensland, JKMRC, Australia
Daniel, M. J. and S. Morrell. 2004. HPGR model verification and scale-up. Minerals Engineering
17(11-12): pp1149-1161.
Daniel, M.J., 2005. Particle bed compression comminution using a piston-die to predict the
performance of a HPGR. Randol Perth Forum, Perth August 21-24, 2005.
M. J. Daniel
June, 2007
ii

ACKNOWLEDGEMENTS
Dr. Brian Smith of BHP-Billiton is gratefully acknowledged for the donation of the
HPGR laboratory unit to the JKMRC. This has been a critical element to the success of
this work. Rod Nicholson of BHP-Billiton, Bert Knopjes of Lonmin and Dr. David Way
of Xstrata are gratefully acknowledged for the supply of ore samples and AMIRA P9N
sponsorship. Without the committed involvement of these companies and its people, this
study would not have been possible at this level of detail.
The assistance and guidance of Dr. Steve Morrell, Professor Emmy Manlapig and
Professor J.P. Franzidis are sincerely acknowledged. Thanks are due to the staff and
management of the JKMRC, JKTech, and JKMineralogy and in particular the staff of the
pilot plant who assisted whenever assistance was needed.
This thesis is dedicated to my beautiful wife Cheryl, and loving children, Michael,
Travers, Cresley, Kyran, Olivia and Jefferson. All of whom have gotten to understand
a little bit about ball mills. For their generation and future generations, who someday
might read, and silently absorb this small contribution to scientific literature. And to those
that do, analyse thoughtfully, thoroughly and most importantly critically these results and
data which are presented.
This thesis is based wholly on the research that I have conducted over the past four years
and on my current understanding of comminution processes. This has developed from
critically reviewing existing literature and conducting hundreds of HPGR and Bond ball
mill tests. This thesis represents “my truth” of a technical issue relevant to understanding
what drives comminution behavior in the minerals industry.
It is a truth based on the assumption that others (like Fred Bond) have provided a
“truthful” baseline from which I have continued to build upon. I would like to thank Fred
Bond for clearly documenting the Bond locked cycle test, and energy per mill revolution
(Excerpt form Bond’s 1949 paper is given below), as this was the most important piece of
literature information that has enabled me to bond (excuse the pun) this thesis together.
Finally I’d like to thank Jack Holmes (former technical. director of Anglo American) for
his paper published in 1957. It’s the finest comminution paper I have read (see appendix).
iii

ABSTRACT
This study evaluates the effect of using high pressure grinding rolls (HPGR) in new
comminution circuit designs in comparison to conventional comminution devices such as
SAG and ball mills, in terms of energy efficiency. The total energy to achieve similar
product sizes was measured directly in the process by using a new digital energy meter.
The difference between the measured energy as a percentage of the conventional ball mill
energy constituted the energy efficiency or energy saving of the circuit configuration.
The analysis is based on laboratory data that compares the performance of hybrid
HPGR/ball mill circuits against that of conventional grinding mill circuits. The circuits
are evaluated in terms of the total comminution energy, resultant mineral liberation and
the eco-efficient effect of significantly reducing grinding media consumption. Three
different ore types are studied, viz. a lead/zinc ore, a bauxite/aluminum ore and a
platinum/chrome ore. Total circuit energy is compared using both Bond’s third theory and
work index values, and a newly developed methodology of direct energy measurement.
In order to validate the results of the digital energy meter, measured energy results made
in the Bond ball mill were compared to several other energy evaluation techniques that
included mill power models, DEM and Bond energy “back-calculation” methods.
Comparable results have confirmed that the digital energy meter can measure specific
comminution energy directly.
Bond’s empirical method is also critically reviewed. The Bond method which uses an
empirical equation and the well known Bond work index to predict comminution energy
requirements appears to be based on 60 J/rev (Bond, 1952a 1952 b). This so-called mill
energy defines an equivalent net energy in the Bond ball mill test to realise the same for a
2.4 meter wet grinding mill. Bond’s empirical equation results can thus be reproduced
using 60J/rev and the mill test data.
Bond’s original paper published in 1949 stated that the net energy input to the laboratory
scale ball mill is 93 J/rev (Bond, 1949). This is comparable to the digital energy meter
measurement of 91.4 J/rev. Bond’s empirical relationship thus suggests that there is a
built in scale factor of 0.645 which accounts for the differences between dry laboratory
milling tests and a full scale (2.4m) wet grinding mill and is the reason for the stated 60
J/rev.
The digital energy meter was subsequently used to measure the energy consumption in the
ball mill and compares this with the energy required to produce the same product size
distribution in several hybrid HPGR/ball mill circuits. It was observed that on average a
measured energy saving of 25-40% was achieved through the hybrid HPGR/ ball milling
circuit when compared with the conventional ball mill circuit.
Results obtained from measuring energy directly are compared with results obtained by
using Bond’s work index and third theory equation. When Bond’s method is used, a
method frequently used in research, negative energy savings sometimes result. The Bond
third theory and work index should not be used under these circumstances.
iv

The characteristics of the liberated mineral products were measured using the JKMRC
mineral liberation analyser (MLA). For each of the experimental circuit products,
liberation characteristics concluded that the enhancement of liberated mineral through the
use of HPGR technology was not founded. However, extensive particle micro-cracking
was visible in the HPGR produces that were visible in the MLA backscatter images. This
was not observed in the conventionally crushed products. The HPGR micro cracks are
believed to be responsible for particle weakening and the resultant reduced milling energy
requirements of ball mill processes that follow with the total circuit energy requirements
significantly reduced.
Greater eco-efficiency can also be realised by reducing the consumption of mill liners and
grinding media. Though the “dollar cost” of comminution is normally accounted for as a
direct electricity expense in the process and is rarely considered for its overall energy cost
or “embodied energy” of manufacturing the steel which amounts to up to 4-6 kWh/t. Eco-
efficient and sustainable development initiatives are linked to “energy cost” and not
always “dollar cost” savings. Rather the direct and indirect energy cost savings and the
impact on the environment should be targeted.
Direct energy use in comminution processes is reviewed. It is shown that 0.56% (87
TWh) of the global net electrical energy consumption of 15,500 TWh per annum is used
to crush and grind non-ferrous ores. Of this, 33% and 53% of the energy is required to
process gold and copper ores respectively. This suggests that the HPGR should be
targeted at gold and copper mining operations in the future to be effective in reducing
carbon emissions. As such new eco-efficient flowsheets that use multiple HPGR’s in
series could be considered. HPGR would be employed as primary comminution devices
followed by small ball mills.
All of the data presented in this thesis is provided in the accompanying CD.

TABLE OF CONTENTS
STATEMENT OF ORIGINALITY........................................................................................................ I
STATEMENT OF CONTRIBUTION TO PUBLISHED WORK ......................................................II
PUBLICATIONS BY THE CANDIDATE RELEVANT TO THIS THESIS ....................................II
JOINTLY-PUBLISHED WORK BY THE CANDIDATE RELEVANT TO THIS THESIS...........II
RELEVANT PUBLICATIONS NOT FORMING PART OF THIS THESIS ...................................II
ACKNOWLEDGEMENTS .................................................................................................................. III
ABSTRACT ........................................................................................................................................... IV
TABLE OF CONTENTS...................................................................................................................... VI
LIST OF FIGURES............................................................................................................................... IX
LIST OF TABLES ..............................................................................................................................XVI
LIST OF SYMBOLS........................................................................................................................XVIII
LIST OF SYMBOLS........................................................................................................................XVIII
CHAPTER 1 ........................................................................................................................................ - 1 -
INTRODUCTION ............................................................................................................................... - 1 -
Introduction ................................................................................................................................. - 1 -
Energy efficiency and new comminution circuit designs.........................................................................- 3 -
Enhanced mineral liberation and micro-cracking ....................................................................................- 3 -
Direct comminution energy measurement versus Bond’s methodology ..................................................- 4 -
Thesis Objectives ......................................................................................................................... - 4 -
CHAPTER 2 ........................................................................................................................................ - 5 -
LITERATURE REVIEW..................................................................................................................... - 5 -
Introduction ................................................................................................................................. - 5 -
Focusing topics............................................................................................................................ - 6 -
Topic A: Direct ball mill energy measurement in comparison to Bond’s empirical method. ..................- 6 -
Topic B: Bond’s third theory and direct energy measurements to assess circuit efficiency.....................- 6 -
Topic C: HPGR micro-crack and liberation characteristics.....................................................................- 6 -
Topic D: Comminution energy within a global sustainable development context...................................- 6 -
The energy efficiency debate ....................................................................................................... - 8 -
Theories of comminution ........................................................................................................... - 10 -
Bond, Von Rittinger and Kick...............................................................................................................- 10 -
Controversies’ over size energy descriptions (Bond, Hukki, Rittinger, Holmes, Kick and Charles).....- 14 -
The Morrell and Holmes equations .......................................................................................................- 18 -
The equilibrium contradiction of the Bond “locked cycle” test ................................................ - 23 -
Test description .....................................................................................................................................- 24 -
Equilibrium reached at 250% recycle. ...................................................................................................- 24 -
Equilibrium reached when grindability changes direction.....................................................................- 24 -
Ball mill work index ..............................................................................................................................- 25 -
The potential role of HPGR....................................................................................................... - 27 -
HPGR Overview....................................................................................................................................- 29 -
HPGR breakage mechanism ..................................................................................................................- 30 -
HPGR used in fine grinding ..................................................................................................................- 31 -
Comparative methods for comminution energy efficiency and recovery................................... - 33 -
Contribution by Klingmann...................................................................................................................- 33 -
Contribution by Appling and Bwalya ....................................................................................................- 34 -
Contribution by Evelin et al ..................................................................................................................- 37 -
vi

HPGR in hybrid grinding systems .........................................................................................................- 38 -
HPGR micro-cracking and mineral liberation .......................................................................... - 40 -
Perceived fracture mechanisms taking place in HPGR inter-particle comminution. .............................- 40 -
Types and mechanisms of mineral liberation.........................................................................................- 46 -
Stephenson’s contribution .....................................................................................................................- 48 -
Literature review summary........................................................................................................ - 53 -
CHAPTER 3 ...................................................................................................................................... - 57 -
EXPERIMENTAL PROGRAM ......................................................................................................... - 57 -
Introduction ............................................................................................................................... - 57 -
Background to experimental design.......................................................................................................- 57 -
Methodology .............................................................................................................................. - 59 -
Direct energy measurement in the Bond ball mill..................................................................................- 59 -
Circuit energy efficiency........................................................................................................................- 59 -
Enhanced preferential liberation and micro-cracking ............................................................................- 59 -
Summary................................................................................................................................................- 60 -
Experimental samples................................................................................................................ - 60 -
Identified samples..................................................................................................................................- 61 -
Feed size distribution.............................................................................................................................- 62 -
Flowsheets- Circuit scenarios ................................................................................................... - 62 -
Study limiting particle size range ..........................................................................................................- 64 -
Sample requirements and test design for circuit scenario evaluation ....................................... - 70 -
Brief description of Laboratory comminution tests ................................................................... - 72 -
Laboratory HPGR tests..........................................................................................................................- 72 -
Bond rod mill.........................................................................................................................................- 72 -
Bond ball mill........................................................................................................................................- 72 -
Characteristics of the Bond “locked cycle” test and third theory equation. ............................. - 72 -
Error analysis of the Bond rod and ball mill tests..................................................................................- 79 -
Description of the Laboratory scale HPGR and test procedure................................................ - 82 -
HPGR test measurements and calculated outputs..................................................................................- 84 -
HPGR Pre-test settings – Initial gap and the hydraulic pressure system ..........................................- 84 -
Calculated HPGR test throughput as a check for measured throughput...........................................- 85 -
HPGR measured throughput ............................................................................................................- 85 -
Measured specific comminution energy ...........................................................................................- 87 -
Bulk “compacted” density of feed. ...................................................................................................- 89 -
HPGR “Flake” density determination ..............................................................................................- 89 -
Product size distribution measurement procedure............................................................................- 90 -
Summary .................................................................................................................................... - 91 -
CHAPTER 4 ...................................................................................................................................... - 93 -
MEASUREMENT OF ELECTRICAL ENERGY CONSUMED DURING A BOND BALL MILL LOCKED
CYCLE TEST................................................................................................................................... - 93 -
Introduction ............................................................................................................................... - 93 -
Energy measurements and definitions ....................................................................................... - 94 -
The new digital electrical energy metering device .................................................................... - 95 -
Active power calculation .......................................................................................................................- 97 -
Energy calculation .................................................................................................................................- 98 -
The Bond test methodology and ball mill power models ........................................................... - 99 -
Mill power and energy consumed per mill revolution .........................................................................- 102 -
Levin’s methodology...........................................................................................................................- 107 -
Morrell’s mill power model.................................................................................................................- 110 -
Measured Bond ball mill energy results .................................................................................. - 111 -
Digital energy meter results .................................................................................................................- 111 -
Summary of results .................................................................................................................. - 119 -
CHAPTER 5 .................................................................................................................................... - 121 -
ENERGY EFFICIENCY OF HYBRID HPGR CIRCUITS COMPARED WITH CONVENTIONAL ROD
AND BALL MILLS ........................................................................................................................ - 121 -
Introduction ............................................................................................................................. - 121 -
Bond mill energy test evaluation methodology........................................................................ - 122 -
Comminution energy measurement in HPGR.......................................................................... - 126 -
Bond rod mill circuit comparison test results.......................................................................... - 127 -
Comminution circuit scenarios comparison results................................................................. - 129 -
vii

Evaluation of circuit scenarios 1 and 4 of direct unit process comparisons........................................- 129 -
Evaluation of circuit scenarios 2 and 3, HPGR and rod milling..........................................................- 133 -
Evaluation of circuit scenario 5 hybrid HPGR/ball mill vs. conventional ball milling........................- 134 -
Description of how the results are presented. ......................................................................................- 136 -
Results of the Mt Isa ore......................................................................................................................- 137 -
Results of the Worsley Bauxite ore .....................................................................................................- 143 -
Results of the platinum/chrome ore .....................................................................................................- 147 -
Specific energy test results consolidated, compared and summarised .................................................- 152 -
Holmes theoretical approach to comminution and the relation to Bond energy ..................................- 154 -
Summary .................................................................................................................................. - 157 -
CHAPTER 6 .................................................................................................................................... - 159 -
RESULTS- ENHANCED PREFERENTIAL MINERAL LIBERATION .............................................. - 159 -
Introduction ............................................................................................................................. - 159 -
The JKMRC MLA system......................................................................................................... - 159 -
Existence of particle micro-cracking ....................................................................................... - 160 -
Worsley Bauxite ore MLA micro-crack results ...................................................................................- 161 -
Mt Isa Results- Lead zinc ore MLA micro-crack results .....................................................................- 163 -
Lonmin platinum ore MLA micro-crack results ..................................................................................- 169 -
Enhanced preferential mineral liberation ............................................................................... - 177 -
Summary .................................................................................................................................. - 182 -
CHAPTER 7 .................................................................................................................................... - 183 -
THE IMPORTANCE OF COMMINUTION ENERGY AND ENERGY EFFICIENCY ON THE DEVELOPMENT
OF SUSTAINABLE MINERAL PROCESSING ................................................................................ - 183 -
Introduction ............................................................................................................................. - 183 -
Future sustainable development – “Energy efficient mineral liberation” .............................. - 183 -
Real issues facing the mining industry ................................................................................................- 184 -
“Dollar cost” and “Energy cost” of mineral extraction........................................................................- 185 -
Comminution energy within a mining operation as a whole. .................................................. - 190 -
Comminution energy within a global context. ......................................................................... - 193 -
Global comminution energy based on grinding media consumption...................................................- 194 -
Eco-efficiency and the “energy cost”(Emergy) of steel grinding media ..............................................- 196 -
Share of comminution energy in crushing and grinding.....................................................................- 197 -
Summary .................................................................................................................................. - 200 -
CHAPTER 8 .................................................................................................................................... - 201 -
CONCLUSIONS & RECOMMENDATIONS ................................................................................... - 201 -
Introduction ............................................................................................................................. - 201 -
Direct Bond ball mill energy measurements............................................................................ - 201 -
Energy efficiency – Hybrid triple pass HPGR/ball mill circuit ............................................... - 202 -
HPGR micro-cracking and mineral liberation ........................................................................ - 203 -
Main thesis conclusions........................................................................................................... - 203 -
Recommendations.................................................................................................................... - 204 -
REFERENCES ................................................................................................................................ - 205 -
APPENDICES ................................................................................................................................. - 215 -

All data accessed and used in this thesis is located on the CD provided on the back page .... - 215 -
APPENDIX A........................................................................................................................... - 215 -
viii

LIST OF FIGURES
Figure 1 shows stamp mill batteries used circa 1897 - 14 -

Figure 2 shows tube mills at the Randfontein Estates Gold Mines, South Africa in 1911. The
caption in the book said that the tube mills were driven by electricity (Hocking,
1986). - 14 -
Figure 3 graphically shows the size energy relationship in comminution from Kelly and
Spottiswood (1997) and Hukki (1962). - 15 -
Figure 4 shows the Morrell function f(x) used to describe the variation of the integrated
exponent “n” in the Walker equation (Morrell, 2004). - 20 -
Figure 5 shows the variation in size energy relationship as determined by (Bond, 1959) and
(Morrell, 2004). - 21 -
Figure 6 shows the comparison of the Morrell (2004) equation and the well known Bond third
theory equation for an ore with a work index of 14. - 22 -
Figure 7 compares the equations of Bond(1952) and Holmes(1957). - 23 -
Figure 8 shows two different process configurations that detail the Bond ball mill locked cycle
test. - 25 -
Figure 9 shows the main components of a 650 tph Polysius HPGR unit that is being used in an
iron ore peletising plant in Brazil. Web reference at
http://www.goldenqueen.com/gall07.htm. - 29 -
Figure 10 shows a schematic representation of the basic principles of the HPGR comminution
process. - 30 -
Figure 11 the relationship between bed pressure, energy and fraction finer than 80 microns
from Schönert’s first survey of grinding with high-compression roller mills. - 31 -
Figure 12 shows on the left had side a HPGR treating coarse particles, where the feed particle
size is greater than the working gap, where as the HPGR on the right hand side, is
a condition where the feed size distribution is less than or equal to the working
gap. - 32 -
Figure 13 shows the product size distribution of the samples treated in a Barmac VSI crusher
and an HPGR (Klingmann, 2005). - 33 -
Figure 14 shows the bottle roll test gold recovery for the 1990 and 1997 samples treated in a
Barmac VSI crusher and an HPGR (Klingmann, 2005). - 34 -
Figure 15 shows energy consumption, recoveries for different conditions (Apling and Bwalya,
1997) - 35 -
Figure 16 shows energy consumption against mineral recovery for different grinding
mechanisms and conditions (Apling and Bwalya, 1997) - 36 -
Figure 17 shows product particle size variation against recovery for different minerals and
HPGR conditions (Apling and Bwalya, 1997). - 36 -
Figure 18 shows how particle size distributions differ depending on the type of comminution
devices used. No indication of the energy expended in these devices is given. (Evelin
et al, 1997). - 37 -
Figure 19 shows the increased gold recovery with the varied crushing devices. - 38 -
Figure 20 shows a comparison of the baseline ball mill circuit and the single pass hybrid
HPGR- ball mill circuit. - 39 -
Figure 21 shows the direct comparison between a ball mill in closed circuit, against a single
pass HPGR in closed circuit. - 39 -
Figure 22 illustrates typical fracture processes that may take place within the HPGR. - 40 -
Figure 23 shows the two most famous and widely published pictures that claim that HPGR
produces microcracks in the products (Kirsch 2006, ,Knecht 1994, Knecht and
Baum 1994, Klymowsky et al 2002). - 42 -
Figure 24 claims to show inter-granular cracks of two pyretic gold ore particles - 850 Hm
1180 Hm (Stephenson 1997). - 43 -
Figure 25 on the LHS shows a siliceous gold ore – 850 Hm 1180 Hm and on the RHS a pyretic
gold ore - 75Hm 106Hm (Stephenson 1997). - 43 -
Figure 26 on the LHS shows a pyretic gold ore – 425 Hm 300 Hm and on the RHS a pyretic
gold ore - 850 Hm 1180 Hm (Stephenson 1997). - 44 -
Figure 27 shows the product size distributions of the conventionally crushed material and
corresponding calculated gold leach extractions (McNab 2006). - 45 -
Figure 28 shows the product size distributions of the conventionally crushed material and
corresponding calculated gold leach extractions (McNab 2006). - 45 -
ix

Figure 29 shows how gold extraction is claimed to be improved through the use of HPGR in a
Gold heap leaching application. In this example by McNab (2006), The products
are described as a jaw crushed product with a P80 of 8.2 mm and an HPGR
product with a P80 of 7.8 mm. The microcracks in these images appear to show
trans-granular cracking, but the severity of the cracks appears to be more
prominent in the bright shaded minerals. - 46 -
Figure 30 shows how the product size distribution of a conventional cone crusher is compared
with the products of an HPGR. Both product size distributions display a P80 of
about 8 mm which highlight the effect of possible errors should the p80 be used to
describe the entire size distribution as required in the Bond equations (after Kirsch
2006). - 46 -
Figure 31 shows particle orientation in a piston and die, before bed compression(Hilden,2005). - 47 -
Figure 32 shows force/energy dispersion routes as the force is applied and energy imparted
by the contact forces into the bed of particles(Hilden,2005). - 48 -
Figure 33 shows localized areas within the bed which due to particle orientation appear not to
be exposed to the energy imparted by the contact forces during bed
compression(Hilden,2005). - 48 -
Figure 34 Leach Results of Semi-Refractory Gold Ore (Nevada) in Size Fraction (Patzelt et al,
1995; Patzelt et al, 1997). - 52 -
Figure 35 shows the interaction between the survey or experimental methodology. - 58 -
Figure 36 depicts the origin and quantity of sample required of each of the three ores tested in
this study. - 60 -
Figure 37 shows the large rotary splitter on the left, and Gilson screens on the right used to
generate representative homogeneous test feed samples and to de-agglomerate
HPGR products. - 61 -
Figure 38 shows how comminution devices may be selected relating to size related classes
(courtesy of Metso handbook, Sandgren et al, 2002). - 63 -
Figure 39 shows the studies conceptual HPGR flowsheet comparison structure. - 63 -
Figure 40 shows a graphical representation of the well known size energy relationship for
tumbling mill devices after (Hukki 1962). - 65 -
Figure 41 shows the path of energy size relationship that could be realised should HPGR be
incorporated into a hybrid comminution circuit. - 65 -
Figure 42 shows the high level structure of the thesis study experimental plan. - 66 -
Figure 43 shows how circuit scenarios 1 was designed to compare a rod mill and a circuit
comprising of a HPGR in closed circuit. - 67 -
Figure 44 shows how circuit scenarios 2 is designed to compare a rod mill and a hybrid
HPGR/rod mill circuit comprising of three HPGR’s in series in open circuit
followed by a rod mill in closed circuit. The circuit is designed to produce identical
products at different grind sizes. - 67 -
Figure 45 shows how circuit scenarios 3 is designed to compare a rod mill and a circuit
comprising of a three HPGR’s in series in open circuit. The circuits is not
expected to produce identical products, but the rather a measure of the extent of
liberation at this stage of the process. - 68 -
Figure 46 shows how circuit scenarios 4 was designed to compare a ball mill and a circuit
comprising of a HPGR’s in closed circuit. - 68 -
Figure 47 shows how circuit scenarios 5 is designed to compare a ball mill and a hybrid
HPGR/ball mill circuit comprising of a three HPGR’s in series in open circuit
followed by a ball mill in closed circuit. The circuit is designed to produce identical
products at different grind sizes . - 69 -
Figure 48 shows circuit scenarios 6 that is designed to compare a ball mil circuit to an
HPGR circuit comprising of three HPGR’s in series. - 69 -
Figure 49 shows the breakdown of the three proposed ores and their origin. - 70 -
Figure 50 shows the breakdown of the conventional rod milling versus HPGR test series. - 71 -
Figure 51 shows the detailed breakdown of the conventional ball milling versus HPGR test
series. - 71 -
Figure 52 shows how a digital energy meter is hooked up with the rod and ball mill tests. - 73 -
Figure 53 shows the ball charge in the standard “bico” Bond mill at the JKMRC. - 74 -
Figure 54 shows the Bond work index in relation to the Bond grindability at a closing screen
size of 150 Hm. - 76 -
Figure 55 shows the Bond work index in relation to the Bond grindability at a closing screen
size of 106 Hm. - 77 -

Figure 56 shows the same data as Figure 55 plotted using normal axes. - 78 -
Figure 57 shows the relationship between the Bond work index and the Bond third theory
equation energy at various closing screen settings. - 79 -
Figure 58 shows the newly refurbished laboratory scale HPGR at the JKMRC installed with a
new digital energy meter. - 82 -
Figure 59 shows the condition of the rolls and cheek plates of the JKMRC 300 mm by 70 mm
HPGR. - 82 -
Figure 60 shows the movable roll working off the stops during HPGR processing. - 83 -
Figure 61 shows the JKMRC HPGR feed hopper and “profiled” rolls surface. - 86 -
Figure 62 shows the compacted HPGR “flake sections” used to determine the equivalent gap
thickness. - 87 -
Figure 63 shows the labelled and coated flakes that are used to determine flake density. - 90 -
Figure 64 shows the Mt Isa ore passing through the HPGR - 91 -
Figure 65 depicts the waveforms of electrical energy used in the active power calculation. - 98 -
Figure 66 depicts the digital energy integration mechanism in the new energy metering device. - 99 -
Figure 67 represents a statistical histogram of the Bond test work indices that were based on
more than 1000 test results. (Mosher and Tague, 2001). - 100 -
Figure 68 shows the full range of theoretical particle grind sizes versus specific comminution
energy when Bond equation represents all particles sizes and comminution
devices. - 101 -
Figure 69 shows the limited range of particle grind size (10 mm – 100 Hm) versus specific
comminution energy in ball mills for which the Bond equation was originally
designed to represent. - 101 -
Figure 70 shows how a direct relationship between ball mill mass and gross power exists
(Morrell, 1993) - 102 -
Figure 71 shows the predictive performance of various power models as developed by Arbiter
& Harris, Rose & Evans, Austin and Bond (after Morrell, 1993). - 102 -
Figure 72 depicts new feed at steady state conditions where Bond estimated the mill energy to
be 60 J/rev(See also Figure 8’s left hand side flowsheet, the same conditions where
Figure 72 represents when steady state condition during a locked cycle test. - 103 -
Figure 73 shows the relationship between the Bond specific comminution energy based on 60
J/rev and Bond’s specific comminution energy using the empirical equation and
work index value. - 105 -
Figure 74 shows the characteristics of the raw Bond data as assembled from a 250 tests
database. - 105 -
Figure 75 shows how the Bond third theory equation energy is linked to a 60 J/rev net energy
consumption in the Bond locked cycle test. - 106 -
Figure 76 shows the relationship between the Bond specific comminution energy, BWI and Gbp
(Grindability) – Equation 19. - 107 -
Figure 77 shows the log- log representation of the data in Figure 86. - 107 -
Figure 78 shows the variation in mill energy per revolution as determined by Levin where the
percent product in the feed is varied from 5% -25%, and the closing size is fixed at
106 Hm. - 109 -
Figure 79 the variation in mill energy per revolution as determined by Levin where the closing
size is varied from 75 Hm to 300 Hm, assuming 10 % of the product in the feed. - 109 -
Figure 80 DEM representation of the Bond Ball Mill with the specified charge of 285 ball of
20,116 g. - 110 -
Figure 81 shows the DEM output power profile of the rotating Bond ball mill, which averages
out at 73.7 W for a mill rotating at 70 rpm. - 110 -
Figure 82 shows the two digital energy meters that were used to test the repeatability and
accuracy of the totalises Wh measurements made by the device. - 112 -
Figure 83 shows the energy consumption against mill revolutions using digital energy meter
number 1. - 113 -
Figure 84 shows the energy consumption against mill revolutions using digital energy meter
number 2. - 113 -
Figure 85 shows the comparison between the two digital energy meter measurements and the
error that is expected should different devices be used to measure the energy. - 114 -
Figure 86 show the Bond mill test data represented as instantaneous power for different mill
load conditions. - 114 -
Figure 87 shows a comparison of the measured digital energy (kWh/t) plotted against the Bond
equation's net comminution energy and the calculated energy based on 60 J/rev. - 115 -
xi

Figure 88 show the measured digital energy (kWh/t) – left axis and Bond’s (1949) mill energy
of 93 J/rev – right axis, plotted against the Bond equation net comminution energy
for the same data set. This confirms that the digital energy meter achieves the same
result as Bond stated in 1949 more than 55 years ago. - 117 -
Figure 89 shows the specific comminution energy using Bond’s (1949) mill energy versus the
measured specific comminution energy using the digital energy meter (kWh/t) - 117 -
Figure 90 shows the equivalent of the Bond locked cycle test at steady state. - 123 -
Figure 91 shows an equivalent continuous ball mill at steady state, provided the feed rate is
controlled to provide steady state conditions. - 123 -
Figure 92 shows the equivalent of Figure 91, but represented as the Bond ball mill locked cycle
test. - 124 -
Figure 93 shows the locked cycle test at steady state 500 ml of conditioned ore plus 200 ml of
fresh feed totalising 700 ml of ore as a standard, with the energy input represented
as the number of mill revolutions. - 124 -
Figure 94 shows the composition of the mill load(approximately 500 ml of conditioned ore,
where at steady state the new feed approximately 200 ml representing a circulating
load of 250% is added. - 125 -
Figure 95 shows a DEM representation of the mill charge at steady state, which consists of
balls (20125 g) and 500 ml of conditioned ore. The ore constitutes grinding media
and plays a part in the generation of new surface area in the milled products. - 125 -
Figure 96 shows similar and parallel product size distributions of the duplicate Bond rod mill
tests. - 128 -
Figure 97 shows the change in feed size distribution and products when circuit scenario 1 was
evaluated in comparison to the products after two and three passes (Mt Isa Ore). - 131 -
evaluated in comparison to the products after two and three passes (Worsley
Bauxite ore). - 131 -
evaluated in comparison to the unique product size distribution shape after two
and three passes with the Lonmin ore. - 132 -
Figure 100 shows the similar products of the three pass tests for both the HPGR(coarse feed)
and HPGR(fine feed) tests that represent circuit scenarios 2 and 5 (Lonmin ore). - 132 -
Figure 101 shows the similar products of the three pass tests for both the HPGR(coarse feed)
and HPGR(fine feed) tests that represent circuit scenarios 2 and 5 (Xstrata Mt Isa
ore). - 133 -
Figure 102 shows how the products of the three pass HPGR were finer than the products of
the rod mill, resulting in circuit scenario 2 to not be evaluated as planned. - 134 -
Figure 103 shows the flowsheets of the two circuits that were compared as part of this study. - 135 -
Figure 104 shows the product size distributions of single, double and triple pass HPGR
processing for the lead/zinc ore, and the confidence at which these products were
obtained through triplicate tests. - 136 -
Figure 105 show the summarised results of the lead/zinc ore, and includes the feed size
distributions to both circuits, with the intermediate product of the triple pass
HPGR circuit making up the feed to the ball mill in the hybrid HPGR-ball mill
circuit. - 138 -
Figure 106 shows a log-log plot of the feed and product size distributions of the lead zinc ore,
as observed during the Bond ball mill tests. - 138 -
Figure 107 shows the measured mill energy under conditions of no-load, balls, and balls and
ore. The linear coefficients are used to determine the net energy per mill revolution
for the lead zinc ore. - 139 -
Figure 108 shows the summarised results of the bauxite ore, and includes the feed size
distributions to both circuits, with the intermediate product of the triple pass
circuit. - 143 -
Figure 109 shows a log-log plot of the feed and product size distributions of the bauxite ore, as
observed during the Bond ball mill tests. - 144 -
for the bauxite ore. - 145 -
Figure 111 show the summarised results of the platinum/chrome ore, and includes the feed
size distributions to both circuits, with the intermediate product of the triple pass
xii

circuit. - 147 -
Figure 112 shows a log-log plot of the feed and product size distributions of the
platinum/chrome ore and the non-parallel lines as observed during the Bond ball
mill tests. - 148 -
for the platinum/chrome ore. - 150 -
Figure 114 shows the correlation of the Bond energy determination methodology which is
based on a constant mill energy consumption of 60 J/rev. - 152 -
Figure 115 shows the measured energy size relationship for three different ores that were
subjected to a conventional milling circuit and a hybrid HPGR grinding circuit.
Suspect platinum/chrome ore data points have been omitted . - 153 -
Figure 116 shows the measured energy size relationship for three different ores that were
subjected to a conventional milling circuit and a hybrid HPGR grinding circuit,
and that show the fitted functions to the data. - 154 -
Figure 117 shows a schematic of how a one meter cubic block of ore of a known density could
be subjected to a size reduction process whereby each block is cleaved along the
planes which would produce eight equally sized cubes. - 155 -
Figure 118 shows the number of particles and particle surface area increase as the particle size
decreases. - 156 -
Figure 119 shows the co-incidence of the bond test surface area and comminution energy on a
log -log scale. - 157 -
Figure 120 shows a picture of the JKMRC MLA unit along with the mineral identification
image on the right and the black and white back scatter image on the left. The
insert picture shows the sample mount shuttle capable to analysing fourteen
samples. - 160 -
Figure 121 shows MLA identified mineral particles within the bauxite sample on the left hand
side, and the corresponding back scatter image on the right hand side. These
images represent the ball mill circuit showing the -425+300 Hm size fraction. No
microcracks are seen. - 162 -
Figure 122 shows MLA identified mineral particles within the bauxite sample on the left hand
side, and the corresponding back scatter image on the right hand side. These
images represent the hybrid HPGR/ball mill circuit showing fully liberated gibbsite
particles within the -425+300 Hm size fraction. The liberated gibbsite particle
shows evidence of microcracks as seen in the back scatter image. - 162 -
Figure 123 shows MLA identified mineral particles within the Mt Isa Lead Zinc ore sample on
the left hand side, and the corresponding back scatter image on the right hand side.
These images represent the ball mill circuit showing liberated sphalerite and
galena particles within the -150+106 Hm size fraction. The liberated particles do
not show any evidence of microcracks as seen in the back scatter image. - 163 -
Figure124 shows an MLA identified galena mineral particle within the Mt Isa Lead Zinc ore
sample on the left hand side, and the corresponding back scatter image on the right
hand side. These images represent the ball mill circuit showing liberated particles
within the -150+106 Hm size fraction. The liberated particles do not show any
evidence of microcracks as seen in the back scatter image. - 164 -
Figure 125 shows MLA identified mineral particles that are individually examined for
liberated mineral content as well as the extent of the microcracks. The figure
illustrated that the MLA block mounts contain thousand of particles, each of which
either display features of microcracks if they have been subjected to HPGR and
no-microcracks when subjected to conventional comminution technologies. - 165 -
Figure 126 shows MLA identified mineral particles within the Mt Isa Lead Zinc ore sample
on the left hand side, and the corresponding back scatter image on the right hand
side. These images represent the HPGR products (ie after the triple pass HPGR)
and show how complex the mineralogy is within the -425+300 Hm size fraction.
The un-liberated particles show evidence of visible trans-granular and inter-
granular microcracks as seen in the back scatter image. - 165 -
Figure 127 shows large particle back scatter images of the Mt Isa Lead zinc ore. These images
represent feed particles to both comminution circuits. Complex unliberated
particles within the -425+300 Hm size fraction are show in on the LHS, whilst
particles within the -3350+850 Hm size fraction are show in on the RHS. These un-
xiii

liberated particles do not show any evidence of microcracks whatsoever as seen in
the back scatter image - 166 -
Figure 128 shows a single large particle back scatter image of the Mt Isa Lead zinc ore. These
image represents products of the triple pass HPGR. Complex unliberated minerals
within the+850 Hm particles are partially liberated by the presence of
microcracks. The microcracks appear to be random and are described as trans-
granular. - 167 -
Figure 129 again shows a single large particle back scatter image of the Mt Isa Lead zinc ore.
These images are similar to those in Figure 128 and represent products of the triple
pass HPGR . - 167 -
Figure 130 shows multiple particles of the Mt Isa Lead zinc ore particles within the -600+425
Hm size fraction. These images are similar to those in Figure 128 and 129, finer in
size and represent products of the triple pass HPGR. - 168 -
Figure 131 shows multiple particles of the Mt Isa Lead zinc ore particles within the -425+300
Hm size fraction and represent products of the triple pass HPGR . Valuable
unliberated binary mineral particles appear to be selectively chosen for micro-
cracking. - 168 -
Figure 132 shows multiple particles of the Lonmin platinum ore, particles are within the -
300+212 Hm size fraction and represent products of the triple pass HPGR . - 170 -
Figure 133 shows multiple particles of the Lonmin platinum ore which have undergone ball
mill processing. The particles represent the -150+106 Hm sub-size fraction. Semi-
valuable liberated chalcopyrite mineral particles appear to have no micro-
cracking. - 170 -
Figure 134 shows multiple particles of the Lonmin platinum ore which have undergone
hybrid HPGR /ball mill processing. The particles represent the -300+212 Hm sub-
size fraction. Semi-valuable liberated chalcopyrite mineral particles have severe
micro-cracking. - 171 -
Figure 135 shows multiple particles of the Lonmin platinum ore which have undergone
HPGR processing. The particles represent mainly edge effect HPGR material in
the +850 Hm size fraction. A very large semi-valuable liberated sphalerite mineral
particle is identified within the sample mount in and amongst the rounded un-
liberated chromite mineral phases. - 173 -
Figure 136 shows thousands of Lonmin platinum ore particles which have undergone HPGR
processing. The particles represent material in the -425 +300 Hm size fraction.
Most of the chromites are perfectly liberated at this size range providing a reason
for the kink in product size distribution at this size. Grinding the chromites to sizes
below this size results in shard/needle shaped particles. - 173 -
Figure 137 shows a single particle of the Lonmin platinum ore which have undergone HPGR
processing. The particle is thought represent mainly edge effect HPGR material in
the +850 Hm size fraction and has no micro-cracking visible. - 174 -
Figure 138 shows two images of single particles of the Lonmin platinum ore which have
undergone HPGR processing. The particles are thought to represent a partial bed
compression zone within HPGR processing. The particles are of the +850 Hm size
fraction and show evidence of trans-granular and inter-granular particle micro-
cracking. - 175 -
Figure 139 shows four single particles of the Lonmin platinum ore which have undergone
HPGR processing. The particles are thought to represent the centre bed
compression zone region of HPGR processing. The particles are from the +850 Hm
size fraction and show evidence of severe trans-granular and inter-granular
particle micro-cracking. - 176 -
Figure 140 shows the 2D particle size distribution generated by the MLA data for the Mt Isa
lead zinc ore. The comparisons represent the ball mill products against the hybrid
HPGR/ball mill products. - 178 -
Figure 141 shows the grade versus recovery composition for Aluminium as generated by the
MLA mineralogical analysis. The analysis represents three different particle sub
size fractions - 179 -
Figure 142 shows the grade versus recovery composition for lead (Pb) as generated by the
MLA mineralogical analysis. The analysis represents a single +38 Hm particle sub
size fraction. - 180 -
xiv

Figure 143 shows the grade versus recovery composition for lead (Pb) as generated by the
MLA mineralogical analysis. The analysis represents a single +38 Hm particle sub
size fraction s - 181 -
Figure 144 shows how key people representing the mining industry have prioritised the main
issues facing the industry. - 185 -
Figure 145 shows the relationship between the ore grade of several commodities and the
“dollar cost” of producing them. - 186 -
Figure 146 shows how global copper cash operating costs for the year 2004 are ranked for
each operating mine. Costs range from between 8 - 98 c/lb, producing a total of
15,7 million tonnes of copper. - 187 -
Figure 147 shows the evolution of increased tonnes per day being processed in several large
copper mines. - 188 -
Figure 148 shows in the figure on left hand side how Copper resource grades have declined
over the past 100 years after Ruth, (Ruth, 1995) and where the abundance of the
various commodities as they exist in the earths crust is found relative to estimated
reserves in the figure on the right hand side (Kesler and Macmillan, 1994) - 189 -
Figure 149 shows the how the size of a copper ore body increases with decreasing copper head
grades (Kesler and Macmillan, 1994) - 189 -
Figure 150 shows the energy costs of mineral extraction with respect to ore grade composition
in a study by (Chapman and Roberts, 1983) for copper and nickel ores. - 189 -
Figure 151 shows the energy intensity of various industries, of which mining is one of the most
intensive. - 190 -
Figure 152 depicts the annual energy use in the US mining industry (US Department of
Energy, 2001). - 191 -
Figure 153 shows an example of the energy footprint being used nowadays by mining groups
to manage energy consumption and equivalent green house gas emissions
(Angloplatinum, 2005). - 192 -
Figure 154 shows how the total energy consumption per tonne of Aluminium produced is
steadily decreasing as a result of improved efficiency and technologies (Marks,
2006). - 193 -
Figure 155 show how world “net” electrical energy consumption has doubled since 1980(EIA,
2004) - 194 -
Figure 156 shows the cumulative production of world steel ball grinding media, and the
companies that represent the global production market (Smorgon-Steel-Group,
2005). - 195 -
Figure 157 quantifies global steel production and shows the steady increase in production. - 196 -
Figure 158 depicts an energy balance over the production of each tonne of the steel grinding
medial - 197 -
xv

LIST OF TABLES
Table 1 shows how ball mill length and diameters have increased in size over the past 100
years(Lynch and Rowland, 2005). - 17 -
Table 2 shows increase in SAG mill size since 1948 (Lynch and Rowland, 2005) - 18 -
Table 3 lists the data contents of a typical Bond Work index test after each cycle - 26 -
Table 4 lists the output data from the Bond locked cycle test that is required to determine the
Bond work index. - 26 -
Table 5 column leach test % of gold extraction vs. energy input (Evelin et al, 1997) - 38 -
Table 6 lists the +0.85 mm Leaching Results (Knecht, 1994) - 51 -
Table 7 shows the error and variation of standard Bond rod mill locked cycle tests. - 80 -
Table 8 shows the error and variation of standard Bond ball mill locked cycle tests. - 81 -
Table 9 List of the HPGR measured experimental test data. - 84 -
Table 10 List of the critical HPGR process calculated outputs. - 85 -
Table 11 lists the ranges over which the Bond work index and UCS are characterised. - 100 -
Table 12 summarises the results of the Bond mill conditions calculated by the use of the Hogg
and Fuerstenau model (1972). - 116 -
Table 13 Summary of the Bond mill power as determined by various methodologies - 118 -
Table 14 give a summary of the HPGR specific comminution energy measurements (HPRC
represents coarse feed (-12.5 mm) and HPGR represents fine feed (-3.35 mm). - 126 -
Table 15 gives the results of duplicate Bond rod mill work index test results for the Worsley
bauxite ore. - 127 -
Table 16 shows the variation in Bond work index test results when the closing screen is
changed while treating a platinum ore. (Shi, Lambert and Daniel, 2006). - 128 -
Table 17 give the energy comparisons of using the Bond equation and Rod mill work index in
comparison to direct energy measurements made on the Rod mill. - 134 -
Table 18 give the energy comparisons of using HPGR in a locked cycle arrangement against
the multiple pass scenarios. - 134 -
Table 19 gives the standard BWI and third theory equation energy for the lead /zinc ore. - 139 -
Table 20 gives the equivalent third equation energy of the lead/ zinc ore when calculated
using a mill energy input of 60 J/rev. - 140 -
Table 21 gives the calculated energy for the lead/zinc ore based on the measured energy per
mill revolution and the effective coarse fraction of the new feed that is subjected to
milling. - 141 -
Table 22 gives the calculated energy required for the lead/zinc ore based on the measured
energy per mill revolution and the effective new feed as determined at steady state
when the fraction of new feed equals the 250% recycle fraction - 141 -
Table 23 gives the calculated energy comparison for the lead zinc ore between the two
circuits, determined through the use of the Bond third theory equation and the
Bond work index as measured during the Bond locked cycle test. - 142 -
Table 24 gives the measured energy comparison for the conventional and hybrid HPGR/ball
mill circuits for the lead/zinc ore, based on the measured energy per mill revolution
and the effective new feed as determined at steady state when the fraction of new
feed equals the 250% recycle fraction. - 142 -
Table 25 gives the standard BWI and third theory equation energy for the bauxite ore. - 144 -
Table 26 gives the equivalent third equation energy of the bauxite ore calculated using a mill
energy input of 60 J/rev. - 144 -
Table 27 gives the calculated energy required for the bauxite ore based on the measured
energy per mill revolution and the effective coarse fraction of the new feed that is
subjected to milling. - 145 -
energy per mill revolution and the effective new feed as determined at steady state
when the fraction of new feed equals the 250% recycle fraction. - 145 -
Table 29 gives the calculated energy comparisons for the bauxite ore between the two circuits,
determined through the use of the Bond third theory equation and the Bond work
index that was determined using the data from the Bond locked cycle test - 146 -
Table 30 gives the measured energy comparison for two bauxite comminution circuits, based
on the measured energy per mill revolution and the effective new feed as
determined at steady state when the fraction of new feed equals the 250% recycle
fraction. - 146 -
xvi

Table 31 gives the BWI and third theory equation energy for the platinum/chrome ore. - 148 -
Table 32 gives the equivalent third equation energy of the platinum/chrome ore when
calculated using a mill energy input of 60 J/rev. - 148 -
Table 33 gives the calculated energy required for the platinum/chrome ore based on the
measured energy per mill revolution and the effective coarse fraction of the new
feed that is subjected to milling. - 149 -
measured energy per mill revolution and the effective new feed as determined at
steady state when the fraction of new feed equals the 250% recycle fraction. - 149 -
Table 35 gives the calculated energy comparison for the platinum/chrome ore between the
two circuits as determined through the use of the Bond third theory equation and
the Bond work index measured during the Bond locked cycle test. - 150 -
Table 36 gives the measured energy comparison for two platinum/chrome comminution
circuits, based on the measured energy per mill revolution and the effective new
feed determined at steady state when the fraction of new feed equals the 250%
recycle fraction. - 151 -
Table 37 shows how the new surface area is generated from the systematic theoretical
comminution of a 1 cubic meter block into small cubes as described in Figure 119 - 156 -
Table 38 shows the extent of microcracks for the bauxite ore of a hybrid HPGR/ball mill
circuit against a ball mill circuit producing three different closing screen size
distributions and multiple MLA sub-size fractions. - 161 -
Table 39 shows the extent of microcracks for the bauxite ore for the HPGR circuit against the
feed sample to both circuits at multiple MLA sub-size fractions. - 163 -
Table 40 shows the extent of microcracks for the lead-zinc ore with the hybrid HPGR/ball
mill circuit compared against the ball mill circuit producing three different closing
screen size distributions and multiple MLA sub-size fractions. - 166 -
Table 41 shows the extent of microcracks for the lead-zinc ore for the HPGR circuit against
the feed sample to both circuits at multiple MLA sub-size fractions. - 169 -
Table 42 shows the extent of microcracks for the Lonmin platinum ore with the hybrid
HPGR/ball mill circuit compared against the ball mill circuit producing three
different closing screen size distributions and multiple MLA sub-size fractions. - 172 -
Table 43 Economic grades, reserves, production rates and years of supply (Norgate and
Rankin, 2002). - 186 -
Table 44 show where energy is being consumed in the global materials and minerals industry,
and the status of where HPGR technology is being applied. - 199 -
Table 45 show where energy is being consumed in the global bulk materials industry, and the
status of where HPGR technology is being applied. - 199 -
xvii

LIST OF SYMBOLS
D HPGR Roll diameter (m)
L HPGR Length of the rolls (mm)
xg HPGR rolls working gap (mm)- measured distance between the two rolls surfaces
xgf HPGR roll working gap (mm)- measured flake thickness measurement
g Bulk density of the ore within the working gap (g/cm3)
c Bulk density of the ore at the boundary between the pre-crushing zone and the
compressed bed zone (g/cm3)- also called the "bulk compacted density"
f HPGR bulk density of the flake produced (g/cm3)
Qg HPGR throughput at the working gap (t/hr)
Qcalc HPGR model calculated throughput (t/hr)
Ecs Specific comminution energy (kWh/t)
Pno-load No-load power to the HPGR (kW)
Pshaft Shaft power of the HPGR (kW)
Qm Experimentally measured throughput HPGR (t/hr)
u HPGR roll peripheral speed (m/s)
M HPGR rolls shaft torque (Nm)
Pgross Gross power consumed by the HPGR (kW)
Fsp HPGR specific grinding force (N/mm2)
Pw Working oil pressure (MPa)
E Walker equation specific comminution energy (kWh/t)
n Walker equation exponent constant
k Walker equation constant
K Morrell equation partial constant
Cg(x) Morrell forms of the Walker/Hukki constant
f(x) Morrell form of the Walker equation exponent constant
a,b Morrell equation constant
x Walker equation particle size parameter
Er Specific energy using Rittinger (kWh/t)
Ek Specific energy using Kick (kWh/t)
Eb Specific energy using Bond (kWh/t)
W Specific energy (general form of the Bond third theory equation) (kWh/t)
kr Rittinger constant
Kk Kick constant
kb Bond constant or Bond work index
Wi Bond work index common form (kWh/t)
Mi Morrell’s work index common form (kWh/t)
P,P80. 80% passing size for the product
F,F80 80% passing size for the feed.
x2 P80=80% passing size for the product
x1 F80=80% passing size for the feed.
Gbp Bond locked cycle test “grindability”. Net grams per revolution (g/rev)
B Net energy per mill revolution (Levin’s methodology)
U Percentage of undersize in the feed (Levin’s methodology)
Eno-load Measured no load energy (digital energy meter)
Egross Measured gross energy (digital energy meter)
xviii

CHAPTER 1
INTRODUCTION
Introduction
Comminution is broadly defined as a group of mineral processing techniques used in
extractive metallurgy to manipulate the particle sizes of rocks. Comminution processes
are used to break rocks into powder as subsequent processing generally requires finer
particle sizes for the minerals of interest to be recovered. A particle becomes liberated
when the valuable mineral is physically freed from other minerals or gangue minerals
through the application of mechanical energy by stressing the particles. The extent to
which the particles are liberated has a direct influence on the amount of mineral of interest
that may be recovered.
The devices used for comminution are generally divided into broad categories based on
the size of the fragments they produce. Devices producing relatively coarse chunks (500
mm – 10 mm) are called crushers. Devices that produce finer particles are called grinders
10 mm – 10 Om). This managed process of particle size reduction remains today as the
only physical method by which the valuable minerals in massive rocks are liberated for
further treatment. The HPGR is a technology that is reported to be more energy efficient
as it simultaneously crushes and grinds ore particles within a compressed bed of particles.
HPGR is limited to a certain amount of energy that is applied to the ore to affect the inter-
particle breakage as the material passes between the rolls.
Previous research in high pressure grinding rolls (HPGR) energy efficiency often states
energy saving percentages out of context and uses those percentages to support
concluding arguments and statements. Comparative measurement techniques and analysis
may be flawed and often are not directly comparable. Energy consumption in
comminution devices still rely on using Bond’s empirical work index and third theory
equation to calculate the comminution energy requirements rather than measuring the
energy. This is common practice due to the difficulty in actually measuring energy
consumption reliably in laboratory devices when small quantities of rock are crushed or
ground. In many instances energy measurements of existing circuits are used as a
benchmark. When new technologies are proposed, there is no benchmark from which to
work, and the risk of an expensive process failing is high.
Existing literature relating to general topic of energy efficiency in comminution circuits

that employ high pressure grinding rolls (HPGR) and their effect on mineral liberation is
scarce. The literature that is available is reviewed in the next chapter. HPGR are
beginning to be more widely accepted in the minerals industry, however the HPGR
comminution device when analysed as more energy efficient device in comparisons to
conventional circuits lacks real numbers. The opportunity to improve mineral liberation
and the connection to energy efficiency is poorly researched and claims are often weakly
supported on a scientific basis.
-1-

The study investigates some of the technical challenges behind HPGR comminution, in
terms of the efficient use of electrical energy to effect particle breakage and associated
mineral liberation within a laboratory environment. The research aims to determine how
much energy is saved and to better understand HPGR comminution energy efficiency.
The analysis is based on laboratory data that compares the performance of hybrid
HPGR/ball mill circuits against that of conventional grinding mill circuits. The circuits
are evaluated in terms of the total comminution energy, resultant mineral liberation and
the eco-efficient effect of significantly reducing grinding media consumption. Three
different ore types are studied, viz. a lead/zinc ore, a bauxite/aluminum ore and a
platinum/chrome ore. Total circuit energy is compared using both Bond’s third theory and
work index values, and a newly developed methodology of direct energy measurement.
Some of the steps that are necessary in the future to achieve a balanced economic,
sustainable and eco-efficient minerals processing industry are also reviewed to support the
research. The research covers four main areas of interest that were identified to meet the
objectives of the thesis and these areas of interest are reported in Chapters 4, 5, 6 and 7.
Chapter 4
• In a laboratory research environment it is well known that the measurement of the

electrical energy consumed in small scale comminution devices is difficult. The
no-load and gross energy measurements do not normally have a similar
proportionality as is experienced in full scale machines. The assessment of the
energy required for comminution processes often rely on an empirical relationship
more commonly known as Bond’s third theory equation and laboratory work
index. This relationship was originally calibrated against a full scale mill at the
time (1950’s) of development. In 2005 a new digital energy metering device was
used and evaluated as an alternative means of obtaining direct, comparable energy
measurements of the comminution process within laboratory devices. Analyses of
measured and theoretical data on energy measurements made in a Bond mill make
up the content of Chapter 4. Chapter 4 thus verifies that direct energy
measurement of the Bond ball milling process is similar to what Bond stated
during the development of his empirical relationship. These measurement methods
are again used in circuit energy analysis that makes up the content of Chapter 5.
Chapter 5
• Another problem that has plagued research efforts that have considered the issue
of comminution energy efficiency is the fact that comparative analyses are often
flawed because different size distributions are produced in each of the devices
being compared. This research considers evaluating hybrid HPGR comminution
circuits that produce identical products that may then be reliably compared with
conventional comminution processes (Chapter 5). The same digital energy meter
measuring device evaluated in chapter 4 for ball mills is used in the HPGR tests.
The results of the direct circuit energy measurements are compared to Bond’s
methodology of determining the total circuit energy. The erroneous effect of using
the Bond equation in these circumstances is analysed and presented to support the
direct energy meter measurement methodology.
-2-

Chapter 6
• Although the main focus of the study is on the benefits of HPGR technology in
comminution circuits both as a power saving device, and a reduced consumer of
steel grinding media. Other aspects such as the possibility of enhanced mineral
liberation and micro-cracking are also investigated (Chapter 6). A key question is
whether the enhanced mineral liberation and particle micro-cracking are
exaggerated by the use of the HPGR.
Chapter 7
• Chapter 7 investigates and determines which ores/commodities are responsible for

the majority of the comminution energy consumption within a global perspective.
It is shown that of the non-ferrous metal commodities/ores used in this study,
namely aluminum/ bauxite, lead/zinc and platinum were ranked third, fourth and
fifth after gold and copper. Copper and gold are recognized as the most energy
intensive non-ferrous metal commodities, and are the main reason why more
copper and gold mines are considering HPGR. Recent examples are the
Boddington Gold project and Copper’s Cerro Verde mine.
Part of the success of the laboratory based experimental studies revolved around the
ability to validate comminution energy measurements and the development of a new
analysis methodology. As such, a method of direct comminution energy measurement was
investigated and critically compared with the traditional Bond methodology. In order to
confirm the wider application of the methodology and results, three P9N sponsor ores
were used in the investigation. The three ores included an ore from a lead zinc mine (Mt
Isa), a bauxite mine (Worsley) and a platinum mine (Karee-UG2). These ores were
initially selected because of the variation in concentration of the minerals and relative
hardness. In addition to this the grades and mineral dissemination of each of these ores
was different giving more diversity to the study of mineral liberation.
Energy efficiency and new comminution circuit designs

Using three appropriately selected sponsor ores, the energy requirement to reach a
nominated target grind using established crushing and grinding technology namely rod
and ball mills were compared to circuits employing HPGR technology in hybrid
HPGR/ball mill circuits. From this, the potential energy savings as a result of improved
energy efficiency of the comminution device for each ore was determined. The test
program was designed to investigate alternative HPGR flowsheets that are based on
reducing energy and increasing recovery or potentially the economics of an operation.
Enhanced mineral liberation and micro-cracking
Mineralogical characterization and mineral liberation studies are essential tools for
assessing the viability of mineral beneficiation even before ore deposits are developed.
Liberation of valuable minerals from the gangue through comminution remains as an
essential step that determines the effectiveness of downstream hydrometallurgical or
physical separation processes. How the values in the selected ore types liberate during
HPGR treatment is of particular interest and should be related to particle size and applied
energy. Quantitative mineral liberation analysis was used to assess the liberated mineral
and existence of micro-cracking when the ores are subjected to HPGR.
-3-

Direct comminution energy measurement versus Bond’s methodology
The literature review, experimental method chapter and main chapters of the thesis
dedicate substantial effort to confirming the direct measurement of electrical energy and
to compare this to results that are obtained using the empirical Bond equations and other
methods. The outcome being that caution should be exercised when either method is used
to state claims on circuit energy efficiency and the type of energy saving that is expected
should a particular circuit be favored over another.
Thesis Objectives
The study unfolds by reviewing available literature within the context of energy, mineral
liberation and the theory of energy-particle-size relationships in comminution. A key
question to be answered in this thesis is to determine how the introduction of HPGR
technology into comminution circuits will impact comminution practices in the future To
do this, the thesis has focused by setting a few well defined objectives. The thesis
objectives are as follows,
• To quantify the energy efficiency of comminution in circuits employing HPGR by

comparing these circuits to conventional circuits which produce the same product.
• To critically evaluate the Bond methodology of determining energy requirements

for milling, using the locked cycle tests and third theory equation in comparison to
direct energy measurements.
• To test the hypothesis whether HPGR promotes or enhances mineral liberation and
micro-cracking when compared to conventional comminution processes producing
the same product.
• To develop flowsheets that demonstrate step change reductions in energy

requirements, whilst at the same time achieving improved mineral recovery that
could be realised utilising solely HPGR technology for crushing and grinding.
• To better understand comminution energy and sustainability, by reviewing energy

efficiency drivers or “eco-efficiency” within the context of economics “dollar
cost” and energy balances “energy cost”.
• To quantify global comminution energy consumption in respect to the production

of various metaliferous and non-metaliferous commodities so as to identify which
areas within the minerals industry are most likely to be affected by matters
concerning eco-efficiency and increasing “dollar cost” of energy.
-4-

CHAPTER 2
LITERATURE REVIEW
Introduction
The literature review is introduced by defining the term comminution and assessing its
association with energy and mineral liberation. Comminution, or particle size reduction as
it is commonly known as still remains the only physical method by which valuable
mineral components in massive rocks can be liberated for further treatment and recovery.
Comminution processes are very energy intensive and consume vast quantities of
electrical energy.
The energy costs are mainly associated with electrical energy and the direct and indirect
“dollar costs” and “energy costs” coupled with increased use of grinding media. In the
future it is quite possible that a more serious situation may arise when energy costs and
consumption increase rapidly resulting in the energy industry not being able to supply
sufficient quantities of energy, and not being able to maintain the rate at which renewable
energy is generated.
A major trend further compounding energy use in the mineral extraction industry today is
the necessity of having to process more massive, hard, low grade ore bodies (Brown,
2004). As a result a much greater amount of gangue material (waste) is being reduced in
size, and to much finer sizes, due to finer mineralisation, to produce the same amount of
recoverable metal. This research therefore aims to seek alternative mechanisms/processes
in which the minerals can be liberated from the ore resulting in more energy efficient and
economical benefits. For this the high pressure grinding rolls (HPGR) are being evaluated
as an alternative comminution device to that of existing tumbling mills.
From a sustainable viewpoint, the balance between global energy consumption and
economic value (currently 2-4 % of GDP in industrialised nations) is also expected to
change in the future. This means that as the economic cost of energy increases, the
amount of energy required doing the same amount of work increases at a greater rate, and
energy intensive industries such as the mining and mineral processing industry are likely
to be most affected. In particular major processes such as comminution (crushing and
grinding) that consume relatively large quantities of electrical energy generated from coal
or oil are targeted.
It is for these reasons that comminution, which is electrical (coal) energy intensive sub-
process within minerals processing is being researched. The goal being to realise a step
change improvements in energy efficiency and lessened environmental impacts through
reduced equivalent CO2 greenhouse gas emissions.
-5-

Focusing topics.
In order to provide a brief indication of the structure of the thesis, four key topics (A-D)
are considered. Each topic is challenged through a few questions. The questions are
provided to help the reader realise the depth of the thesis topics. These are as follows, but
the reader should note that not every question is answered directly in chapters 4, 5, 6 and
7 that follow.
Topic A: Direct ball mill energy measurement in comparison to Bond’s empirical

method.
1. Does the Bond work index and third theory equation estimate or predict ball
milling energy requirements accurately enough to make comparisons with other
comminution devices and process methodologies?
2. Is the 50 year old Bond work index still valid, (if assumed as a universal theory of
comminution) for the assessment of energy required for grinding in devices other
than tumbling mills?
3. Is Bond’s third theory equation valid when HPGR technology is employed in
comminution circuits? How would the work index be determined?
4. Is the Bond work index an ore property or a process energy descriptor?
Topic B: Bond’s third theory and direct energy measurements to assess circuit
efficiency.
1. Why are tumbling mills said to be grossly in-efficient users of energy?
2. Does the size of the tumbling mill affect the efficiency of the device or breakage
process?
3. What is the true potential of the HPGR in terms of energy efficiency and does the
application apply to all mined and processed mineral ores?
4. Is energy efficiency and the use of energy in comminution processes receiving
adequate attention in relation to the developing needs of the future energy efficient
comminution circuits?
Topic C: HPGR micro-crack and liberation characteristics.

1. Do micro-cracks even exist? If so, is there a difference between HPGR micro-
cracked products and conventionally crushed and milled products?
2. Does increased particle micro-crack density improve mineral recovery?
3. Do micro-cracks play a role in inter-granular or trans-granular fracture?
4. Do micro-cracked particles belong to certain mineralized particles?
Topic D: Comminution energy within a global sustainable development context.
1. Is comminution, a sub-process within the minerals processing, responsible for
most of the carbon emission through the use of electrical energy? What scope is
there to reduce energy consumption and or improve energy efficiency?
2. What are the primary sources of energy (e.g. diesel, electricity, gas, renewable)
used in the mining and minerals processing industry? How much electrical energy
is being consumed globally in relation to comminution processes? Will energy that
drives industry be available as easily accessible and most importantly affordable in
the future?
-6-

3. Are emerging global problems such as climate change and global warming clearly
understood and how are these connected to the minerals industry. Is comminution
important in this respect?
4. What is perceived as the driver of comminution research today, and what type of
mineral resource or commodities are likely to benefit the most from HPGR
technology?
In order to answer these questions in whole or in part, the research covers these as four
sections which make up the body of the thesis (Chapters 4 to 7). Chapters 4 to 6 include
all the technical and scientific aspects of the thesis, whereas Chapter 7 discussed the
implications that the outcomes of the research provide in a broader comminution energy
perspective. The structure is again very briefly described as follows:
• Chapter 4 considers issues relating to the direct measurement of energy in the

Bond ball mill. These are critically evaluated against Bond’s method and third
theory equation of quantifying the energy required to comminute ores from a given
F80 to a product described by the P80.
• Chapter 5 quantifies the direct energy savings resulting from the use of HPGR in
comminution circuits when compared to conventional ball milling circuits.
• Chapter 6 evaluates the liberated products and micro-crack density of the HPGR
included comminution circuits. These main technical topics were investigated
independently.
• Chapter 7 provides an assessment of the impact that the finding of chapters 4, 5,

and 6 has on comminution, in particular with respect to sustainable development
and global comminution energy use.
Modern day comminution circuits are becoming much more sensitive to increasing energy
and grinding media costs, and should these costs rise dramatically the profitability of the
mine could be greatly affected. In such a case comminution circuits would be in need of
major reform in order to reduce the sensitivity to energy. As such, current commercial
HPGR applications have the potential to play a role in the development of future
comminution circuit designs. These mainly economically driven changes exclude any
“eco-efficient” benefits that could be realised should eco-efficiency and sustainability be
incorporated into the decision making process. Eco-efficiency, sustainability and the
impact of this work is reviewed and discussed in more detail in Chapter 7.
Current day understanding revolves more around the theories of comminution rather than
energy efficiency. These theories and the topic of energy efficiency which are both
embroiled in much controversy are reviewed and discussed in the next two sections.
The theories of comminution describe how energy is used with respect to the managed
process of particle size reduction or grind size. The theories revolve around several
empirically developed mathematical equations that have been used for more than 100
years to provide a “measure” of the comminution energy required to perform a specific
size reduction task. Mineral processing engineers use these equations to determine the
energy without having to actually measure it. The Bond methodology and Bond tests
-7-

developed in the 1950’s are still widely used today. In this study a new method of directly
measuring the energy and hence energy efficiency in a laboratory comminution device is
developed. The results are used to compare the direct energy measuring technique to the
results obtained through the use Bond’s empirical methodology. The methodology
comparisons are used in the analysis of the HPGR and conventional circuit comparisons
of Chapter 5.
The energy efficiency debate

Over the past thirty years or so, massive semi-autogenous-mills and ball mills have
evolved as the primary comminution device used in mineral processing. In spite of being
known to be energy inefficient (Johnson et al., 1994), these large mills have remained as
the most common grinding technology used, due to lower capital costs and economies of
scale. Plants are generally designed to achieve a targeted grind size which is ideally
selected on the basis of optimal economics over the life of the mine: energy costs are
rarely incorporated as a major factor, mainly because electrical energy costs have
remained low over the past three decades. This is no longer the case, and aspects of
energy efficiency need to be more clearly understood.
The energy efficiency of a comminution device is difficult to define and quantify as there
is no standard baseline from which to benchmark or reference it against. Much of the
limited energy-efficiency research in minerals processing was aimed towards better
understanding how energy is utilised, which in turn has promoted new comminution
devices such as the rhodax crusher, horomill, Isa mill and HPGR. The use of the most
efficient devices is still limited, and in general the comminution circuit designs have
rarely been built or even operated to minimise energy consumption. Capital cost is the
more common driver; however, this may well change in the near future as new
sustainability issues become more dominant.
Becker, Kwade et al (2001) stated that 0.6% of the energy supplied to a tumbling mill is
used in breakage, whereas Fuerstenau (1992) stated that the energy efficiency of
mechanical size reduction processes is of the order of 2 to 12 %. It is generally agreed that
the energy consumed in the actual breakage operation is low compared with the total
energy consumed. There are many instances in the literature (Fuerstenau and Kapur,
1994; Tromans and Meech, 2004) that repeatedly state that the energy required to actually
break the particles is in the order of 1% of the energy that is typically used in
comminution processes. In summary, comminution is best considered as an indirect result
of the mechanical operation of the mill. The mechanical operation consumes energy and
the size reduction is an indirect result of energy consumption. Stairmand (1975) stated
that in practice, energy losses in the present designs of mills are such that the efficiencies
range from less than 1% for fluid energy mills, to over 80% for roll crushers. Ball and
roller mills, in which the majority of industrial comminution is carried out, have
efficiencies in the range 7-13%.
Fuerstenau et al(1999) in their journal paper ‘The effect of ball size on the energy
efficiency of hybrid high-pressure roll mill / ball mill grinding’ attempted to find ways to
reduce energy consumption in comminution. They used an experimental design that tested
dolomite in a hybrid grinding system comprising a laboratory-scale high-pressure roll mill
and ball mill in series. The work demonstrated that the energy efficiency of the hybrid
system could be increased by reducing ball size, provided the HPGR product contained
-8-

sufficient internal damage to allow for the use of smaller balls. There was an optimum
range of energy partitioning between the high-pressure roll mill and the ball mill in the
hybrid system. The increased efficiency of hybrid grinding was related to the particle
damage imparted by the HPGR, as evidenced by the increased breakage rates of HPGR
product relative to particles that are not subjected to the HPGR. Fuerstenau however
evaluated comminution energy without any further evaluation of downstream process
benefits. His main objective was to highlight energy efficiency, which was
comprehensive.
Schönert (1986, 1988) compared the energy of a ball mill to single particle breakage
mechanisms producing identical products, and concluded that the ball mill process
efficiency was between 10% and 20%. Schönert suggested an alternative definition of
process efficiency, based on the energy consumed in inter-particle crushing (confined bed
breakage) and added that the ball mill energy efficiency relative to an inter-particle
comminution process was 40% to 60% less. In spite of this remarkable reduction in
comminution energy efficiency, the practical application and potential cost benefits were
not evaluated.
Fuerstenau and Abouzeid (2002) went further and put forward an argument that the
comminution efficiency is a technical term that relates some measure of the output from a
comminution machine to the energy input into it. They stated that the term usually
encompassed controversial arguments and that the conflict in reporting comminution
efficiencies had arisen from ill definition of the reference for the output energy. Surface
area and surface energy are widely used in the assessment of comminution efficiency, and
a review of the energy to produce new surface of quartz showed that the comminution
efficiency ranged from 0.1% to 1.0% for various comminution methods. This figure
perhaps is the same figure quoted by the committee on comminution and energy
efficiency (National Research Council, 1981) that quoted that energy efficiency as low as
1% is typical for tumbling mills. Fuerstenau and Abouzeid (2002) refer to these numbers
as being meaningless, mainly due to the ill definition, and mainly because this suggests
that 99% of the energy used during the breakage process is wasted or used elsewhere.
Fuerstenau, in his paper attempts to clarify the abovementioned 1% debate. He showed
that if the energy to produce new surface by single particle breakage is used as the basis
for evaluating efficiency, then the efficiency of ball milling has a more realistic value of
about 15% for the comminution of quartz. In this scenario he suggests that 85% of the
energy is wasted or used elsewhere. A second approach taken by Fuerstenau showed that
if the comminution efficiency is based on comparing the energy to produce some size
distribution parameter of the product from the ball milling with that by single particle
breakage, then the efficiency was likely to be in the range of 25%.
Researchers(Fuerstenau and Kapur, 1994; Fuerstenau and Kapur, 1995; Fuerstenau et al.,
1999; Fuerstenau and VazquezFavela, 1997; Herbst and Lo, 1989; Mahyera, 1988;
Rowland and Erickson, 1984; Stairmand, 1975) have studied the energy efficiency of
tumbling mills and have in some instances proposed HPGR as an alternative technology.
However, the abundant supply and low cost of electrical energy over the past few decades,
along with decreasing commodity prices, have not prompted the use of more efficient
technologies in mineral comminution processes, and as a result research in this area is
lacking.
-9-

Knowledge of the power required to drive mills is necessary for mill design purposes. As
mills have become larger, the cost of rectifying an error in the design and sizing of mill
for particular applications has become more sensitive. The ability to understand and
improve the precision of mill power models has become increasingly important. Research
into the development of mill power models by several authors is relatively mature(Austin
et al., 1992; Bond, 1954; Bond, 1961a; Hogg and Fuerstenau, 1972; Hukki, 1962; Kiyama
et al., 1974; Morrell, 1994) and in recent years there has been a shift in focus towards
online measurement and control of mill charge motion, load and feed rate. Generally mill
control, through consistent operation is used as a way to maximize the power draw of the
mill. When the power draw of a mill is optimised, many operations imply that optimum
mill performance within the tumbling mill device is achieved (Bernhardt, 1996; Pontt,
2004; Radhakrishnan, 1999; Valery and Morrell, 1995). This does not necessarily mean
that the energy efficiency of the comminution process is optimised, since empty mills can
still run at maximum power draw. Rarely, though, has energy efficiency or the prospect of
extremely high energy costs been assessed in respect to the profitability of a mining
operation.
Theories of comminution
Comminution processes are generally expressed in terms of an energy-size relationship.
Several theories have emerged over time, which at the time of their development were
considered as the most advanced understanding of how energy is consumed in relation to
product size. The most commonly used methodology is the well known Bond
methodology. This methodology is used extensively in this study as a comparative method
to assess comminution energy consumption with respect to the comminution device and
direct energy measurement. As such it seemed appropriate to review the origins of the
existing energy size relationships that exist for comminution in general and to reference as
far back as the time when electrical energy in large motors was being introduced to
mechanically drive new comminution devices. The relationship between energy and
product size is not a simple one, and is one which has embroiled several controversies.
The controversies still exist today and some of the background to them is reviewed and
discussed.
Bond, Von Rittinger and Kick

The introduction of electrical energy, during the turn of the 20th century, saw many
researchers dedicating much effort into understanding where and how energy was utilised
in crushing and grinding. This was during a era in which “new technology”, tumbling
mills were being introduced as novel devices. At first sight, the consideration of energy
input as a function of a grinding system was very attractive in terms of trying to
understand the processes within and as a whole. However in the course of this review on
some of the history of comminution with respect to energy consumption, it was found that
it is much more complicated than originally realised. Not all the energy supplied to
tumbling mills for example is dissipated in the breakage of the rock, and the power
consumption of the mill may be independent of the incidence of fracture.
Traditionally, finer grinding is not carried out beyond the point where the mines’ net
return in dollars for increased recovery becomes less than the added operating cost
(Steane, 1976). Wills (1997) also stated that it can be shown using Bond's third “law”
equation that 19% extra energy must be consumed in grinding one screen size finer on a
- 10 -

root-two screen series. This issue is tackled head on in grinding mills, particularly rod
and ball mills which may receive ores crushed down to -12.5 mm in size and which still
require the bulk of the energy to reduce it further to below 75 Om.
Bush in his PhD thesis (Bush, 1967) concluded with “It is considered that in general the
energy-size-reduction relationships do not suitably define the process of size reduction”.
Bush stated that large energy losses occur in the transmission driving the mill, in the form
of friction, heat and sound. While these losses may be measured and determined, Bush
continued by saying that there are other processes within the mill itself that are impossible
to measure or assess. Bush refers to the “other losses” as a characteristic of the
comminution device, which may be defined as:
• Particle friction (heat) without breakage

• Kinetic and potential energy losses
• Elastic and plastic deformation of particles
• Production of sound
Much of the energy losses are dissipated in the form of heat, and do not play any part in
particle breakage. Balancing or minimizing the energy consuming mechanisms within
comminution devices relative to mineral recovery objectives and liberation, may lead to a
better understanding of the comminution “unknowns”.
The literature has paid little attention to the issue of energy efficiency over the past thirty
or forty years (more on this aspect is described later). Electrical energy costs in general
have remained low, and tumbling mill devices have dominated comminution circuits.
Larsen et al. (2001) completed energy related surveys around various comminution circuit
designs employing tumbling mills and concluded that: “Within the accuracy of test
results, there is no significant difference in the total energy requirement for any of the
alternative comminution systems”. Such a finding could easily create a perception that
very little can be done to improve energy efficiency of comminution processes, and that
the energy required in comminution is strongly dependant or governed by the “laws” of
comminution.
Searching the literature, publications by several authors (Austin, 1973; Cleemann, 1994;
Hongqi et al., 1985; Kapur and Fuerstenau, 1987; Stamboliadis, 2004; Walker and Shaw,
1954) refer to what are described at the “laws” of comminution, and these often reference
the work of researchers Rittinger (1867), Kick (1885) and Bond (1952). A thorough
investigation into these so-called “laws” reveals that the original authors of these papers
proposed various theories of comminution and did not describe them as “laws” as such.
Laws are universal in the physical and scientific world (such as the law of gravity),
whereas a theory is defined as a comprehensive explanation of a given set of data that is
repeatedly confirmed by observation and experimentation and gained general acceptance
within the scientific community, but has not been decisively proven. This is particularly
relevant to Bond’s third theory, since the theory is often broadly used to describe all
comminution mechanisms, devices and particle size reduction in terms of the required
energy needed to reduce particles in size.
The basis of the Von Rittinger, Kick and Bond theories was discussed by Walker et al.
(1937) and (Walker and Shaw, 1954) and others (Kapur, 1972; Kapur and Fuerstenau,
- 11 -

1987; Morrell, 2004). The three traditional expressions based on these theories are
commonly stated as variations of the general expression given by equation 1, the rate of
change of applied energy (E) with respect to particle size (x). Other parameters are k, and
n which are constants, specific to the ore and comminution device/mechanism.
dE
= k xn (1)
dx
Rittinger (1867) never actually published his work as a law or hypothesis. His analogy
was to describe energy in terms of new surface area resulting in his general equation
where the energy required for size reduction is stated as being directly proportional to the
change in surface area. In this instance the exponent ”n” is equal to -2. When this
expression is integrated, the relationship between energy and particles size takes the form
of equation 2.
1 1
Er = k r (2)
x2 x1
Where
Er = Specific energy (kWh/t)
kr = Rittinger constant
x2= P80=80% passing size for the product
x1 =F80=80% passing size for the feed.
From a practical point of view, there is no ore characterization test that defines the values
of the Rittinger constant (kr). Nor is it made clear if the size terms x2 and x1 represent the
80% passing size of the product size distribution.
The Kick (1885) equation is also of the same form, where the energy required to reduce a
material in size is stated as being directly proportional to the size reduction ratio. In this
case the exponent ”n” is equal to -1. When this expression is integrated, the relationship
between energy and particles size takes the form of equation 3.
x1
Ek = K k log (3)
x2
Where,
Ek = Specific energy (kWh/t)
Kk = Kick constant
Again in this case, from a practical point of view, there is no current ore characterization
test that defines the values of the Kick constant Kk. Nor is it made clear if the size terms x2
and x1 represent the 80% passing size of the product size distribution.
There were many disputes during the late 1800’s regarding the validity of these two
theories. However after numerous investigations it was generally accepted that the Kick
equation was good for stamp mill devices used at the time to break coarse particles
(Figure 1), and the Von Rittinger equation was used to describe processes where fine
- 12 -

particles were being reduced in size, for example in the production of cement. In the years
1890 to 1915, new technology in the form of tube mills (Figure 2) were introduced, and
many of the ores milled at the time were being ground to a grind size of 850 Om.
The controversy over adequate descriptions of how particle size and energy consumption
are related seemed to re-emerge in the early 1900’s when tumbling mill devices were
introduced into comminution circuits. Unsuccessful attempts were made to describe the
process using the Kick and Von Rittinger expressions, and a new expression was needed.
Literature published in this respect included work by Gates entitled “Kick versus. Von
Rittinger - An experimental investigation of rock crushing performed at Purdue
University”, (Gates 1915), Haultain et al entitled ”A contribution to the Kick versus Von
Rittinger dispute” (Haultain, 1923), and Gaudin, Gross et al. entitled “The so-called Kick
law applied to fine grinding”(Gaudin et al., 1929).
The comminution theory controversy seemed to be solved when Bond (1952) postulated
that the exponent ”n” in Walker’s (1937) equation (equation 1 above) should be equal to -
3/2, and published the “Third theory of comminution” which has the form of equation 4 or
more commonly recognized in equation 5 and 6.
1 1
Eb = kb (4)
x2 x1
Where,
Eb = Specific energy (kWh/t)
kb = Bond work index
From a practical point of view, Bond defines kb as the Bond work index which is
determined by an experimental test procedure or ore characterization test that defines the
value of the equation constant, commonly known as the Bond work index. Bond also
made it clear that the size terms x2 and x1 represent the 80% passing size of the product
size distribution. Equation 4 then reverts to the more commonly used or known forms
given by equations 5 or 6.
(
W = 10 Wi x2 1/ 2 x1 1/ 2 ) (5)
(
W = 10 Wi P801/ 2 F801/ 2 ) (6)
Where,
W = Specific energy (kWh/t)
Wi = Bond work index (kWh/t)
P80=80% passing size for the product
F80=80% passing size for the feed.
- 13 -

Figure 1 shows stamp mill batteries used circa 1897.
Figure 2 shows tube mills at the Randfontein Estates Gold Mines, South Africa in 1911. The caption in the
book said that the tube mills were driven by electricity (Hocking, 1986).
The Bond third theory may have been used because it provides a ball-park figure through
a relatively simple test procedure and mathematical equation or, perhaps, it was Hukki’s
(1962) paper (originally presented in 1959) entitled “The Solomonic settlement between
the theories of Rittinger, Kick and Bond”, and Bond’s (1959) publication “Confirmation
of the Third Theory” (based on new crack length measurements), which suppressed
supporters of other work by Charles (1957) and Holmes (1957) in regards to
understanding the energy-to-particle size relationship in comminution devices. A paper by
Walker and Shaw (1954) entitled “A physical explanation of the empirical laws of
comminution” perhaps clouded the work of Charles and Holmes.
Controversies’ over size energy descriptions (Bond, Hukki, Rittinger, Holmes, Kick
and Charles)
Hukki (1962) proposed an equation where the exponent “n” in Walker’s general equation
takes the form of a function f(x), as shown in equation 7.
dE
= k x f ( x) (7)
dx
- 14 -

Hukki however did not determine what f(x) was, but suggested that it follows the shape of
a curve as suggested in Figure 3. Interestingly, the graph is in log-log format, and defining
the form of such an equation is not mathematically possible. Hukki proposed that each of
the three theories be applied separately depending on the particle size range. Hukki’s
paper does not indicate whether the relationship is affected by the type of comminution
device used, and within the size range.
Figure 3 graphically shows the size energy relationship in comminution from Kelly and Spottiswood (1997)
and Hukki (1962).
In hindsight, Bond’s “Third theory of comminution” really did mark a turning point in
understanding comminution processes. Around the time of his publication many other
publications by himself and other researchers (Agar and Brown, 1964; Bond, 1940; Bond,
1941; Bond, 1947; Bond, 1949; Bond, 1952b; Bond, 1954; Bond, 1957; Bond, 1958;
Bond, 1960; Bond, 1961b; Bond, 1962; Bond, 1963a; Bond, 1963b; Bond and Maxson,
1943; Bond and Wang, 1950; Charles, 1957; Charles and de Bruyn, 1956; Gaudin and
Hukki, 1944; Holmes, 1957; Hukki, 1962; Myers et al., 1947; Prentice, 1946; Rumpf,
1962; Schuhmann, 1960; Taggart, 1945; Tartaron, 1964) all contributed to the
understanding of crushing and grinding in respect to the comminution devices available at
that time. This implies that the Bond theory may not be valid in many other cases where
different sized equipment is used, or certainly not be valid for the description of
comminution energy required in HPGR.
Two publications that stand out, those of Charles (1957) and Holmes (1957) are of
particular interest to this study. These papers provide a more fundamental approach to
understanding energy and particle grind size. Charles stated that Bond chose the
arithmetic mean of n = -2 (Rittinger, 1867) and n = -1 (Kick, 1885) to define the
exponent “n” as a constant at -3/2. The Walker equation, when integrated with the -3/2
exponent, results in the well known Bond third theory equation given by equation 4 and 5
as previously mentioned.
- 15 -

Charles (1957) stated that the exponent “n” in Walker’s equation must be a variable that
depends on the material and the manner in which it is broken. Holmes (1957) too,
contributed by proposing a modified form of Kick’s “law”, where “n” is a variable
associated with the material and the comminution device. The Holmes equation is given
by equation 8, and is derived from a more fundamental approach which considers new
surface area generation. The fundamental approach in fact resulted in a modified version
of the Kick equation in which consideration to the ore character and the comminution
device is made. Holmes claimed that the exponent “n” cannot be fixed and is related to
the comminution device as well as the ore characteristics. The derivation of the Holmes
equation and its application is explained in more detail later in Chapter 5.
n n
P 100
W = Wi 1 (8)
F P
The Holmes (1957) paper is published in Volume 35 of the Transactions of the Institution
of Chemical Engineers (Holmes, 1957). However, the paper was originally presented at a
symposium of the Institution held in London on 7th Nov 1956; the dialogue after the
presentation was discussed, recorded and forms part of the paper which was published in
1957. Two important noteworthy comments in this dialogue, by R.L Brown and W.F
Carey, respectively, sum up the limitations of Bond’s work, and are quoted below as
original excerpts.
• Brown – “ One of the most important accomplishments of this symposium has

been the exposure and clarification of the ‘so-called’ laws of comminution and
what Brown has referred to as engineering results”(Holmes, 1957).
• Carey – “Bond’s technique suggested crushing ores in a ball mill, consequently the
resulting grindability figure, i.e. the amount of energy required to produce a given
degree of reduction, included the losses of energy in the mill”(Holmes, 1957).
Other comments in the discussion praise the work presented by Holmes even though there
is very little evidence in the literature today which supports or even acknowledges this
work. The comments by Brown and Carey merely amplify the argument that Bond’s third
theory is essentially the Bond ball mill energy response to different ore work indices
relative to a 2.4 m diameter wet grinding ball mill. Bond must have essentially calibrated
the “dry grinding locked cycle” tests to an operating 2.4 m pilot/full-scale mill using the
same ore type. These claims are investigated in more detail in chapter 4 by physically
measuring the energy consumption in the Bond ball mill locked cycle test and comparing
this to Bond’s method as well as several other methods that have been applied by other
researchers in attempting to quantify the energy in the test mill. This is important because
if the research is based on energy measurements made with a new energy metering device,
then the validity with reference to what has been accomplished in the past must be
comparable. In the last 30 years or so modified versions of the Bond equations have
emerged in an attempt to more accurately predict comminution, mill energy, and mill
efficiency. The efficiency of the mill is assumed to remain constant as the mill size
increases, however the efficiency may be improved depending on the operating conditions
such as mill charge mass, ball size, lifter design, pulp density and grate and trommel
discharge design.
- 16 -

SAG mill performance in terms of optimum use of energy is very difficult to measure and
control. Should different mill shell liner designs be considered for a plant, it would be
very difficult to trace the improvement based on snap shot mill performance tests because
of the variation in mill performance over time. Plant control strategies are often
inconsistent and operator driven and do not allow for clear evaluation of changes in shell
liners on mill performance. The same argument applies, should HPGR be considered to
compliment or replace conventional tumbling mills in existing or new comminution
circuits. There is no mechanism to compare the advantages of one circuit over another.
Therefore to properly evaluate different comminution devices, all external influencing
process variables such as ore type, grade, and feed size distribution need to be removed
from the experiment. Measuring devices and equipment where possible should be the
same so as to eliminate experimental error based on the measurements made in the
controlled laboratory experiments. This is discussed in more detail in Chapter 3.
Previously Rowland and Kjos (1978) proposed various efficiency factors to adjust the
predicted mill energy as determined by the Bond test and equation when operating
conditions of the mill and design size changed (Rowland and Kjos, 1978). A few years
later other authors (Rowland, 1986; Rowland and Erickson, 1983; Rowland and Erickson,
1984) introduced further scale factors relating to the application of larger ball mills and
the relative grinding efficiency. These factors were necessary as economies of scale were
driving the design size of mills from Bond’s 2.4 m diameter to over 5.5 m. In 1980 it was
thought due to the failures of new mills in Papua New Guinea and Norway that the
maximum size of a ball mill was limited to 5.5 m. However Arbiter and Harris (1982)
helped solve this limiting condition and today, ball mills of up to 7.5 m and 8 m in
diameter with 15.5 MW are being built. SAG mills are major comminution devices that
collectively with ball and rod mills may be described as tumbling mills.
Manufacturers claim that there might be no limit to the size of mills. Table 1 shows
typical increases in ball mill motor sizes and dimensions over time (Lynch and Rowland,
2005). The mills of today are clearly much larger in size than the 2.4 meter mills that were
used as a standard during the time of the development of the Bond test in the 1930’s to
1950’s.
Table 1 shows how ball mill length and diameters have increased in size over the past 100
years(Lynch and Rowland, 2005).
Diameter Length
Year (m) (m) kW
1909 1.2 2.1 11
1912 1.9 2.3 41
1927 2.4 2.4 168
1940 3.1 2.8 447
1963 3.9 5.5 1,491
1970 5.6 6.4 3,169
1990 6.1 9.3 5,593
1997 7.3 10.5 11,440
The comments of Brown and Carey, and the studies of Holmes (1957) and Charles
(1957), highlight a need to understand comminution better in closed systems in terms of
energy efficiency. Aspects of mineral liberation were not considered in Bond’s time,
- 17 -

perhaps due to the fact that acceptable recoveries of the higher grade ores were liberated at
850 microns using the technology at the time (Bond 1949). Since mineral liberation today
is playing a much more important role in the economic feasibility of exploiting massive
mineral ore deposits, it is important to determine if mineral liberation is a function of the
energy input and resultant particle size distribution characteristics or the comminution
device or comminution mechanism used to affect the breakage process.
The past few decades has seen the gradual introduction of fully autogenous (FAG) and
semi autogenous (SAG) mills which have largely replaced secondary and tertiary crushing
and rod milling. Since 1948 economies of scale have driven up the mill size, as shown in
Table 2 (Lynch and Rowland, 2005). Economies of scale have generally dictated this
increase in mill size as mineral grades and commodity prices have decreased.
Table 2 shows increase in SAG mill size since 1948 (Lynch and Rowland, 2005)
Diameter Length
Year (m) (m) kW
1948 3.1 1.0 75
1960 6.7 2.1 1,119
1969 9.8 4.3 5,220
1979 10.4 5.2 9,321
1987 11.0 5.2 11,186
1994 11.0 5.8 13,423
1996 12.0 6.8 19,388
1997 12.2 6.1 20,000
The definition of energy efficiency for comminution devices is defined in numerous ways.
The comminution device has merely provided a suitable environment in which fracture
mechanisms may take place and depending on the particular device used, there are
incalculable energy losses associated within the device that are inherently built into the
machine and become part of the process. Some devices, when compared to others, may
utilise energy more efficiently, and produce different sized products. In these cases it is
very difficult to make an accurate and realistic energy efficiency comparison, particularly
when the comparative devices themselves are operating sub-optimally. The mathematical
description of comminution with respect to energy remains a challenge and is responsible
for the Bond third theory still being used to best describe the energy required to produce a
specified grind size. Morrell (2004) has provided some real advances in this
understanding, which are covered in the next section.
The Morrell and Holmes equations

Morrell (2004) revisited the Hukki (1962) equation (equation 7) and published a paper
entitled "An alternative energy-size relationship to that proposed by Bond for the design
and optimisation of grinding circuits.". In this paper Morrell stated that “The methods by
which circuit energy efficiency is normally assessed are shown to be fundamentally
flawed”. Referring to the Bond equation, Morrell proposed a new equation in which the
exponent “n” in Walker’s equation is replaced by a function f(x) as proposed by Hukki.
Unlike Hukki, Morrell attempts to define the function f(x), and states that “n” is particle
size dependant. This may be true, but only if tumbling mills, in particular, SAG mills and
not other types of comminution devices are being used. Morrell however intended the
function to be applicable to SAG,AG, rod and ball mills.
- 18 -

Morrell points out that the work of Larsen, et al. (2001) is of particular interest, as the
results are based on well run comminution circuits implying that the relative efficiency of
tumbling mill devices is the same, providing the circuit is operated at near optimum
conditions. Comminution circuits that are not optimised and that are not run well will
result in very poor energy efficiencies. Johnson, et al. (1994) have shown that energy
efficiency improvements of up to 25% are achievable when inefficient comminution
circuits employing SAG mills are optimized. These type of scenarios create the
impression that there is no limit to possible energy efficiency improvements in tumbling
mills. Morrell uses these facts to propose a new size energy relationship which is similar
to the equation proposed by Hukki and is applicable to well run tumbling mill devices.
When other “more efficient” devices such as the HPGR are used, then the equations do
not hold, unless a work index equivalent for HPGR’s is developed. Morrell (2006) and
Klymowsky and Liu (1997 b) has already proposed the idea of further developing an
HPGR work index.
Morrell’s new equation considers how the “n” exponent varies with particle size,
particularly in the application of SAG mills, where large particle feed sizes are fed into
the mills. To achieve this, Morrell uses an integrated form of Walker’s (1937) equation.
Morrell (2004) proposed that a more complete description of a general form of
comminution equation is given by equation 9.
d
dE = Cg ( x) f ( x ) (9)
x
where f(x) is a function which describes size reduction, g(x) is a function describing the
variation in breakage properties with particle size and C is a constant related to the
breakage properties of the material. Morrell said further, that from a practical viewpoint,
one of the problems with equation 9 is that the variation in breakage properties with
particle size is not the same for all rocks (Morrell et al., 2001). Hence, there was
unlikely to be a function g(x) that will satisfy all rock types, although there was evidence
that some rocks behave in a broadly similar manner. A general solution to the equation
was therefore unlikely to be found, and so without integrating equation 9, Morrell
suggested that the solution take the form of equation 10, where f(x) and g(x) are
incorporated or lumped together in f(x).
In addition, Morrell stated that the experimental determination of the size-by-size

properties of a rock over the typical range of feed F80 sizes handled in a comminution
circuit (U100 mm +0.1 mm) would be very difficult. This problem is further complicated
by the lack of a proven procedure for carrying out such experiments. As such different ore
characterisation tests for the many different devices have been developed. Typical tests
like the DWT, SMC, SPI, and Bond suit of tests that produce a work index (crusher, ball
and rod). The shape of the function f(x) is given in Figure 4 (Morrell, 2004).
Given this situation, Morrell has proposed an alternative form of the Third theory
equation of Bond, that relates to the specific energy required to reduce a size class to a
specified grind size , given below. The rock breakage properties, as represented by a
comminution index Mi, are assumed to be constant with respect to particle size, leaving
any variation to be taken up in the form of the function f(x).
- 19 -

0
-0.2
-0.4
f(x)
-0.6
-0.8
-1
-1.2
100 1000 10000 100000 1000000
size (microns)
Figure 4 shows the Morrell function f(x) used to describe the variation of the integrated exponent “n” in the
Walker equation (Morrell, 2004).
W = M i K ( x 2f ( x 2 ) x1f ( x1) ) (10)
Where,
W = Specific energy (kWh/t)

Mi = Morrell work index (kWh/t)
P80=x2 =80% passing size for the product
F80=x1= 80% passing size for the feed.
K = constant = 4,
f(x) = -(a+xb), f(x) defines the exponent relative to the particle size.
a ,b = constants
The Morrell (2004) equation (full size range) is compared to the classical Bond equation
in Figure 5 and Figure 6. In Figures 5 and 6 the Bond third theory equation uses the ball
mill work index and applies to a size range from 10mm to 10 Om. This provides a means
for a graphical comparison of the energy requirements given the feed and product size.
This excludes the energy requirements to get the material to that size. The Morrell
function in Figure 4 remains constant at about -0.3 for particles sizes less than 10 mm to
100 microns. There is no indication that this holds true over this size range or even
outside of this range at say below 100 microns. The function (Figure 4) and new equation
(equation 10) are meant to apply for SAG mills where particle size effects have an
influence on the function f(x). This is true when SAG mill energy is considered with
respect to the particle size.
- 20 -

70
Bond Wi = 7
Bond Wi = 11
60 Bond Wi = 14
Bond Wi = 16
Bond Wi = 19
Bond Wi = 22
Net specific comminution energy kWh/t
50 Bond Wi = 29
Morrell Mi = 7
Morrell Mi = 11
Morrell Mi = 14
40
Morrell Mi = 16
Morrell Mi = 19
Morrell Mi = 22
30 Morrell Mi = 29
20
10
0
10 100 1000 10000 100000
Grind size (micron)
Figure 5 shows the variation in size energy relationship as determined by (Bond, 1959) and (Morrell, 2004).
The Morrell (2004) equation is defined on the basis of eighteen industrial scale data sets.
The new equation still relies on the standard Bond work index result conducted with the
ore in a Bond laboratory mill for ball mill size reduction. In this instance the work index is
denoted as the Morrell index Mi . The Morrell equation is a much better estimate in
determining the energy requirement across SAG and ball mills within a comminution
circuit. For SAG mills, Morrell uses another breakage parameter Dwi (known as the drop
weight index) that is modified through the application of another equation to determine a
value of Mi. which is then used instead of the work index Wi.
Where the Bond Work Index and similar tumbling mill tests are used to assess ore
hardness and/or grindability, these are ideal for ball mill applications and do not
necessarily apply for other devices. With the introduction of large diameter SAG mills,
and crushers, the JKRMC studied various methods of characterising the ore for these
devices and developed the JK Drop Weight Test (DWT). The JKMRC took the approach
of decoupling the material and the machine, and as a result the DWT measures the impact
breakage of single particle breakage events. The ore specific parameters “A” and “b” and
the breakage appearance function resulting from the tests are used in JKSimMet to
analyse or mathematically model the performance of the SAG mill and various other
crushing and grinding devices.
From Figure 6, it would appear that the Bond equation has under-estimated the energy for
particles in the size range 75 Om – 100 mm, and over-estimated the energy required for
finer particle reduction of 10 Om-75 Om. However it could be argued that the fine and
coarse size fractions are outside of the Bond and Morrell equation descriptions.
- 21 -

100
Net specific comminution energy kWh/t 90 Bond Wi = 14
80
Morrell Mi = 14
70
60
50
40
30
20
10
0
1 10 100 1000 10000 100000
Grind size (micron)
Figure 6 shows the comparison of the Morrell (2004) equation and the well known Bond third theory equation
for an ore with a work index of 14.
The comparison of the -4 mm to 100 micron “Bond size fraction range” suggests that the
Bond equation under predicts the energy required to comminute an ore with a similar
work index value. This theoretical data representation relates to tumbling mill devices
across the full size spectrum which includes SAG mills. Morrell’s equation is based on
full scale plant data, which is very useful in “Greenfield” plant designs that incorporate
tumbling mill devices. It is not be valid when other grinding equipment such as HPGR’s
or fine grinding devices such as the ISA mill are used.
The Morrell equation was validated using existing plant operating work indices which
produced the Mi value. Morrell suggests that the Mi value is equivalent to the Bond
operating work index. When the operating index is not known, the standard Bond work
indices as generated by the standard Bond locked cycle test may be used.
One very interesting observation of the Morrell equation is that it appears to be a special
form of the Holmes equation. Holmes derived the equation that has the “n” parameter as a
variable is based on the fundamentals of new particle surface area generation, and not two
dimensional particle sizes (Holmes, 1957). Figure 7 shows a direct comparison between
the Bond and Holmes equations. The parameter “n” incorporates the variation of energy
efficiency with respect to the device used. Holmes may have wished to use this to describe
process in-efficiencies in tumbling mill devices, as the HPGR and other new devices had
not emerged when Holmes derived the alternative equation.
- 22 -

Bond Holmes
1 1 P
n
100
n
W = 10Wi W = Wi 1
P F F P
When F is large (Feed in micron)
0.5 n
100 100
W = Wi W = Wi
P P
W = 100 0.5Wi P 0.5

W = 100 n Wi P n
W = 10 Wi P 0.5
Figure 7 compares the equations of Bond(1952) and Holmes(1957).
In the fine size fractions from -4 mm to 100 Om, if the Holmes equation is substituted
with n=0.3 as indicated by Morrell, then the equation is represented as equations 11, 12
and 13. Equation 13 is the equivalent of the new Morrell equation when F80 is large.
W = 100 n Wi P n
(11)
W =100 0.3Wi P 0.3

(12)
W = 3.98 Wi P 0.3
(13)
Interestingly when the feed is large, and a P80 of 100 Om is used in the Holmes and Bond
equations (Figure 7), the equations simplify where W is equal to Wi. Wi is defined as the
work index parameter determined by completing the locked cycle test in the Bond ball
mill. This characteristic is explained in more detail in Chapter 3, in a section titled
“Characteristics of the Bond locked cycle test and third theory equation”.
The next section describes the procedure for completing a standard Bond “locked cycle”
test and a paradox that is associated with it. This is necessary because later in Chapter 4
reference to the measured energy in the mill during a locked cycle test is used to confirm
the origin of Bond’s equation and mill energy. This information is again used in Chapter 5
to demonstrate that the Bond equation data should only be used in circumstances where
full scale wet grinding are being evaluated.
The equilibrium contradiction of the Bond “locked cycle” test

The standard Bond grindability test is a locked-cycle dry grinding and screening
procedure, which is carried out until a steady state condition or equilibrium is obtained.
The test is described in several papers in the literature (Bond and Maxson, 1938;
Magdalinovic, 1989). The contradiction of the test is whether steady state conditions are
- 23 -

achieved when the grindability value changes direction or when the 250% recycle is
achieved. The procedure iterates the cycles until 250% recycle is achieved, but the test is
assumed to have reached equilibrium when the grindability is constant and changes
direction within 3% of the measured value .
Test description
The Bond ball mill test is used to determine the so-called standard work index, which is
defined as the specific energy consumption (in kWh/t) required to reduce a material from
a notional infinite size to a P80 size of 100 Om. The test involves a series of consecutive
batch grinds in a laboratory mill, 30.5 cm by 30.5 cm charged with 285 steel balls which
have a mass of 20,125 g. After each grind the contents which consist of a standardized
700 cc volume of ore, are screened to remove undersize which is replenished with an
equal mass of new feed to make up the mass that of the original 700 cc volume.
Equilibrium reached at 250% recycle.

The length of time or the number of required mill revolutions for each batch grind is set or
adjusted until the mass of the oversize fraction is consistently 2.5 times greater than the
undersize, or the mass of the undersize equals the mass of the new feed. Under these
conditions the test mimics the performance of a 2.4 m mill in closed circuit with a recycle
load of 250 %(Figure 8 LHS). In a practical sense the oversize fraction after each cycle, is
conditioned at equilibrium, and should remain constant in composition and size at the
point when steady state conditions of the locked cycle test are achieved.
Equilibrium reached when grindability changes direction

Equilibrium is also said to be reached when the grindability (equation 14) or net grams of
product produced per revolution stabilizes and changes direction. Here the grinding
period, or number of mill revolutions are adjusted until the net grams of sieve undersize
produced per mill revolution reaches equilibrium At equilibrium the “locked cycle”
process is repeated two more times However, according to Bond, when this is complete,
the mill load (undersize product and circulating load) are screened, and the average of the
last three net grams of final product size generated from the adjusted feed mass (feed
material less the fines in the feed) per revolution (Gbp) is determined. This equilibrium
condition is physically represented by the flowsheet on the right hand side of Figure 8.
The flowsheet on the left hand side (Figure 8) represents the locked cycle tests, and the
flowsheet on the right hand side (Figure 8) represents a typical flowsheet in industry. The
industry flowsheet configuration is preferred because a proportion of the fine material in
the feed never sees the mill, which is not the case of the locked cycle tests. This problem
in industry has also led to the concept of the “phantom” cyclone which makes provision
for the fraction of material is not processed in the ball mill when this flowsheet
configuration is used.
This explains the paradox of the Bond grindability test and is used later in the analysis of
the results in Chapter 5.
- 24 -

Ball mill work index
The ball mill work index (kWh/t) is calculated from the revised empirical equation
(Equation 14) according to Bond (1961a and 1961b). When the experimentally measured
Bond test data is used in equation 14 to determine a work index for the ore, it may then be
used in the third theory equation (Equation 5 or 6) to calculate the energy required to
perform the test mill’s grinding task. Under these circumstances the feed and product size
distributions are used in the equation. The feed and products size distributions are
represented by single point value of the F80 and P80 which is the reason why parallel feed
and product size distributions are necessary for the Bond equation to be valid (Bond
1949,Bond 1952a;Bond 1952b; Bond 1959,Bond 1961a, Bond 1961 b).
Feed
(t/hr)
Feed
(t/hr) Power input Re-circulating
Ball Mill load (Ball mill)
60 Ws per
revolution
Closing
70 rpm screen 70 rpm
Revolutions (1m) Revolutions
(time) (time)
Re-circulating load Closing screen (1m)
Product
(t/hr) Product
(t/hr)
Figure 8 shows two different process configurations that detail the Bond ball mill locked cycle test.
49.1
Wi = (14)
0.82 10 10
P10.23 x Gbp x
P80 F80
Where
Wi = Bond work index (kWh/t)

Gpb = Grindability (g/rev)
P80=80% passing size for the product, (Om)
F80=80% passing size for the feed, (Om)
P1= Closing screen size, (Om)
The equation parameters are measured during the completion of the Bond locked cycle
test. The results of a typical test are given in Table 3. Table 4 shows a summary of the
results and represent the undefined terms in equation 14. Once the work index is
determined, it may then be used with Bond’s third theory equation (Equation 5 or 6) to
determine the theoretical energy required to perform a specified grind, assuming a wet
grinding ball mill of 2.4m is used.
- 25 -

Table 3 lists the data contents of a typical Bond Work index test after each cycle
GRINDING STAGE RESULTS
Grinding Screen Actual Calculated New Final Gross Net Grindability Circ 250%
Stage Oversize Mill Mill feed product product product Load Recycle
Revs Revs # mass in feed
(g) (next cycle) (g) (g) (g) (g) (g/rev) % (g)
1 933.5 150 201 1450 282.4 516.0 233.6 1.557 181 313.6
2 991.6 202 184 516 100.5 457.9 357.4 1.769 217 324.9
3 974.9 184 154 458 89.2 474.6 385.4 2.094 205 321.7
4 1055.2 154 172 475 92.5 394.3 301.8 1.960 268 337.3
5 1058.5 172 185 394 76.8 391.0 314.2 1.827 271 338.0
6 1028.4 185 178 391 76.2 421.1 344.9 1.864 244 332.1
7 1033.1 178 177 421 82.1 416.4 334.3 1.878 248 333.0
Table 4 lists the output data from the Bond locked cycle test that is required to determine
the Bond work index.
Mass of Original Feed (g) 1450
Closing Screen Size (Bm) 150 Average for last 3 grinding stages
Percent -150 um in Feed 19.49 Mass -150 Bm produced per rev (g) 1.856
Feed F80 (Bm) 2160 Circulating Load (%) 254
Product P80 (Bm) 110 Bond work index (BWI) (kWh/tonne) 12.6
To this day the Bond standard (1961) grindability test as described above has provided a
work index that is widely used to estimate the energy required for grinding. There are a
few instances where the efficiency of grinding mills was studied. Over the past few
decades, research has rather focused on the test procedure itself or a fix to the output
result in the form of various efficiency factors. This resulted in various refinements of the
test procedure to limit experimental error, and through the application of efficiency factors
which aim at obtaining a better empirical relationship as opposed to understanding energy
efficiency and ore breakage characteristics. The outcomes of these investigations, test
procedure, equations etc. are still used as a tool in the design of comminution circuits
(Rowland and Kjos, 1978).
As the Bond test specifically relates to the energy required to comminute the ore within a
ball mill, it becomes increasingly difficult to compare the benefits that could be realised
should the ore be treated in so called energy saving comminution devices such as the
HPGR. This further complicates the energy efficiency debate. Fuerstenau and Kapur
(1994) conducted grinding tests on solids of different hardness and showed that, in
general, energy utilisation in a choke feed high-pressure roller mill is greater than in a ball
mill, but lower than in a single particle mode of comminution and slow compression of
loads. This result confirmed the fundamental research of Schönert (1979; 1988) in which
he concluded that, in terms of energy, the most efficient means of comminution is the
breakage of single particles by compression loading. As previously mentioned, HPGR or
comminution circuits employing HPGR are said to be up to 50% more energy efficient. It
is claimed by Brundiek (1994) that the Loesche mill, when employed for the grinding of
clinker and granulated blast furnace slag in several works, has been found to have a
specific energy consumption which lies significantly below that of tube mills and is better
than that of a plant with a tube mill preceded by high-pressure grinding rolls. This type of
experimental evidence suggests that the energy efficiency is not necessarily independent
of the comminution device.
By using the Bond “locked cycle” test data and direct mill energy measurements it is
possible to obtain the specific comminution energy of the Bond ball mill. This is useful in
laboratory experiments as the direct energy is measured and does not rely on having to use
- 26 -

the empirical Bond equations. This is explained later where it will be shown that Bond
calibrated his empirical data to that of the 2.4 m wet grinding mill(Chapter 4) and that
using the Bond equation in these type of experiments results in errors (Chapter 5).
Total circuit energy consumption of new circuits employing HPGR should be compared to
conventional circuits. For this reason the experimental methodologies of the laboratory
scale tests that were required needed to be carefully designed. Details are covered in
Chapter 3. However in the next section an account of the potential benefits of using
HPGR is reviewed followed by a review of comparative methods of analysis that have
been published.
The potential role of HPGR

When considering the various facts presented in the chapter thus far, it is surprising that
the Bond equation is still being applied universally. Bond’s third theory equation has even
been used in the blasting domain as published by authors such as Free, et al. (2004) and in
the domain of SAG mill circuits as published by Li and Xu (1993). This is risky as it was
not intended to be used in this area, and certainly not for other devices such as the HPGR.
However the HPGR does have a role to play in the future of comminution circuit design,
and some aspects of where HPGR should be used in comminution circuits are reviewed
here.
The HPGR is reported to be between 20-50% more energy efficient than conventional
crushing devices or in conventional comminution circuits (Esna-Ashari and Kellerwessel,
1988; Schönert, 1988; Schwechten and Milburn, 1990). This large range in reported
energy savings depends on which circuit arrangement is used and how the energy saving
is defined. For example, 20% when used as a pre-treatment to a conventional ball mill,
and 50% when used in closed circuit with a de-agglomerater and classifier (Morsky, et al
1995; Norgate and Weller, 1991).
HPGR technology has been available for over 20 years. In this time it has become a
standard comminution device for cement clinker grinding. The cement industry in Europe
has implemented the technology mainly because electrical energy costs in Europe are
much higher than for other mining focused countries. Another reason for the fast
adaptation of HPGR technology in the cement clinker grinding industry is the fact that the
variability of the cement clinker in terms of an “ore” characteristic is much more
consistent and the HPGR manufacturers became much more confident in sizing units for
new installation. In contrast to this the minerals industry has ore characteristics that are
much more variable in terms of rock strength, abrasiveness and mineral content. Driven
by incredibly large tonnages, HPGR has since 1996 started to make headway in the iron
ore industry. Very recently HPGR’s have been accepted at Cerro Verde and Boddington
both copper and gold mines. HPGR’s are yet to be accepted in base metals, but has also
recently been installed at a platinum mine in South Africa (Anglo Platinum’s PPL plant).
From a process and engineering perspective, HPGR’s are being seen as competitors to
SAG mills. This topic has recently been published by several authors (van Drunick and
Smit, 2006; Vanderbeek et al., 2006). In the future it is envisaged that the HPGR could
make in-roads into the ball milling domain. Figure 9 shows the first installation in the
world where two installed HPGR’s are substituting ball mills for final grinding in an iron
ore pelletizing plant at CVRD in Brazil. Figure 9 is referenced on the internet at
http://www.goldenqueen.com/gall07.htm which shows these units. The goldenqueen
- 27 -

website lists various aspects of the Soledad mountain feasibility project which
investigated the option of using HPGR’s in the flowsheet that considers gold recovery via
a heap leaching process.
HPGR’s are today competing with mature and established tumbling mill technology (in
particular SAG mills with hard competent rock types). Knowledge of tumbling mill
operations has developed for more than eighty years which makes the introduction of
HPGR very challenging. By comparison SAG and ball mill experts have developed
advanced control systems that are easily adapted to steady circuits treating variable ore
types. In a document released by Mcintosh Engineering entitled Hard Rock Miners
Handbook, several “Rules of Thumb” for designing engineers recommend the following
advice in regards to energy usage in milling and grinding operations.
1. A concentrator (mill) requires up to 3 tons of water for each ton of ore processed.
It is therefore important to operate with the maximum practical pulp density and
minimum practical upward or horizontal movement. The basic philosophy requires
movement over the shortest possible distances between processing units and
makes use of gravity to save on power consumption.
2. A mill at the mine (and related facilities) accounts for approximately 85% of the
total electrical power consumption for an open pit operation, and about 45% for a
typical underground mine.
3. For a typical underground mine, the cost for electrical power for the mill
(concentrator) will be approximately 35% of the total electrical power cost for the
mine.
4. Grinding is a low-efficiency, power-intensive process and typically can account
for up to 40% of the direct operating cost of the mineral processing plant.
5. For purposes of design, it may be assumed that a ball mill will carry a 40% charge
of steel balls; however, the drive should be designed for a charge of 45%.
6. A grate (diaphragm) discharge ball mill will consume 15% more power than an
overflow (open) discharge ball mill even though the grinding efficiencies are the
same.
7. The larger diameter the drum, the more efficient the grinding. However, this
phenomenon stops when the diameter reaches 3.8m. Thereafter, the efficiency
bears no relation to diameter.
Semi-autogenous grinding mills (SAG) have a major impact on the design and economics
of comminution circuits today; however these advantages have been eroded in recent
years and the minerals industry is now faced with the challenge of having to build even
higher throughput plants which are more energy efficient. The HPGR technology has been
seen as a relatively new comminution device that offers higher throughputs, lower specific
energy and potentially improved recoveries (Patzelt et al., 2000). The vital role that
HPGR will play in the future in respect to energy reduction and reduced grinding media
consumption is the main reason for choosing it in this comparative study. It is for the
above mentioned reasons that some technical information on the HPGR technology and
its application within the minerals industry be described in the next few paragraphs
- 28 -

HPGR Overview
Since the HPGR is viewed as a relatively new comminution device, and is a key
component of the research effort it is worthwhile giving a brief description of its design
and operation before further discussion.
The basic machine concept and operation is very simple as described by (Daniel, 2003).
The material is force-fed into the unit by creating a head of material over the machine, as
seen in Figure 9 http://www.goldenqueen.com/gall07.htm. Two counter-rotating rolls
allow the compression breakage to be used in a continuous rather than batch operation.
Nitrogen cylinder Feed

Ore feed chute.
Oil cylinders Moveable roll Fixed roll
Rolls are choke fed
Bearing block, movable roll

Opposite side is fixed roll
Product
Nitrogen Accumulators Oil cylinders Roll Gearbox
Figure 9 shows the main components of a 650 tph Polysius HPGR unit that is being used in an iron ore
peletising plant in Brazil. Web reference at http://www.goldenqueen.com/gall07.htm.
One of the rollers in the HPGR rotates on a fixed axis while the other is allowed to move
linearly with a pressing force applied to the moving roll. The moveable roller is forced up
against the material in the gap between the rollers by a hydraulic oil pressure system. This
oil pressure acts through four or two cylinders (depending on the manufacturer) and
transmits the grinding force over the cross-section of the diameter of the rolls where the
bed has formed. The amount of material in the gap, or compression zone, may be
manipulated to a limited degree to result in optimum operating conditions, but generally,
it is a characteristic of the process ore, roll diameter and surface characteristics.
During processing, the particle bed is compressed to a density of greater than 70% solids
by volume. The material is usually agglomerated into a cake (flake) that may have to be
de-agglomerated before passing on to subsequent processes. This is achieved by either
immersing the product in water in a sump under the discharge end of the rolls or by using
a hammer, impact or ball mill (Schönert, 1988).
The philosophy of the roll design has many factors and is constantly changing due to new
advances being developed. The length / diameter ratio of the roll varies between 0.4 and
0.7 depending on the application and the specific throughput requirements. Special
applications bring the ratio closer to 1.0 when a high throughput is required. The roll
- 29 -

bearings are of a self-aligning roller type that is mounted in large, heavy duty bearing
housings. The bearing blocks are installed in a rigid frame that forms the structure of the
HPGR. Hydraulic cylinders are mounted in the frame to apply the grinding force to the
movable roll (Otte, 1988).
HPGR breakage mechanism

The HPGR breaks particles predominantly in an autogenous way, unlike other
comminution devices such as ball and rod mills. The grinding force is transferred from
one particle to the next, with a small proportion of the particles coming into direct contact
with the rolls.
Schönert's (1988) remarks in his fundamental study of comminution processes was that in
any comminution process the particles are broken by contact forces, which deform the
particle and cause a stress field. As the stress level meets the criteria either of yielding or
fracturing then the particle will be deformed in-elastically or broken, respectively. The
number of contact forces depends on the mode of stressing being either in single particle
mode or multiple particles (Schönert, 1988).
Schönert stated that comminution devices such as crushers, mills and HPGR all stress the
material by compression and shear. Both single particle and bed particle stressing
experiments were conducted as part of his fundamental research and Schönert concluded
that inter-particle bed breakage has a lower efficiency than single particle stressing.
Schönert stated that the efficiency may drop by as much as a factor of two to three
depending on the conditions relating to the number of contact forces (Schönert, 1991).
Schönert (1988; 1991; Schönert and Lui, 1996) mentions that even though the inter-
particle process is less efficient than single particle stressing, they found that when a bed
of particles is compressed and comminuted, the result is that the material is comminuted
more efficiently than in a ball mill. For practical applications of the particle bed
compression principle, Schönert suggested the use of two rolls mounted in a strong frame
and fed in such a way that a particle bed is formed between the rolls under high pressure,
which is greater than 50 MPa as shown in Figure 10 (Daniel, 2003; Farahmand and
Ehrentraut, 1997).
Principle of the HPGR process
“HPGR OBJECTIVE”
Process objective is to
comminute the ore
efficiently (energy)
and liberate as much
of the valuable mineral Compression zone
component. > 50 MPa
F
Fixed roll Movable roll
Figure 10 shows a schematic representation of the basic principles of the HPGR comminution process.
- 30 -

Fandrich (1999) claimed that because bed breakage comminution is different to that of
impact breakage, the process is responsible for improved efficiency. Fandrich focused his
research towards evaluating mineral liberation in bed breakage. The study however does
not consider the energy application or losses as a result of the processes within the
comminution device. Fandrich’s work does not compare liberation effects of conventional
breakage mechanisms against bed breakage, although this was not the objective of his
work. This is explained further in the mineral liberation section later in this chapter.
HPGR used in fine grinding

Schönert (1988) presented results for the fine grinding of cement clinker. The inter-
particle crushing of fine feeds (-2.5 mm) was investigated in HPGR and the results are
shown in Figure 11. Schönert (1988) made the observation that particle comminution
took place up until a certain limiting point, above which increases in the grinding force, or
energy input became disadvantageous. In Figure 11 the mass-related energy (kWh/t) and
the fraction finer than 80 Om are plotted against particle bed pressure (MPa).
Where,
E*M,c = mass related energy,

p = compression,
f = fraction finer than 80 µm,
k = (1/f) = circuit factor,
EM,c = kE*M,c = specific energy of compressing,
EM = total specific energy
Figure 11 the relationship between bed pressure, energy and fraction finer than 80 microns from Schönert’s
first survey of grinding with high-compression roller mills.
This is an unusual way to present data which demonstrates the effective performance of
the HPGR. What is glaringly obvious is that Schönert originally presented the benefits of
HPGR for fine grinding. HPGR thus was an immediate success in the large tonnage
cement clinker grinding industry with more than 400 units installed worldwide.
To date many of the installed mineral industry applications have treated much coarser
feeds producing much coarser products. The data points presented by Schönert (Schönert,
1988) are not tabled and the trends are assumed to be indicative of how the process
behaves. It gives a better understanding of the behavior of particles when subjected to
- 31 -

high compression inter-particle breakage, using particularly fine feeds, and producing a
high proportion of products less than 80 Om.
The minerals industry has to date applied HPGR in the same way as the diamond mining
industry. Rather coarse soft material is usually fed into these units, where the particle
sizes are generally larger than the operating gap. These installations are sometimes
referred to as high pressure rolls crushers (HPRC). However in cement and slag grinding
applications, the feed sizes are much smaller than the operating gap, which results in the
process as being described as HPGR. The concept of the two types of HPGR operating
conditions for roller presses is shown in Figure 12. This is important, as the application of
HPGR technology should be considered as both a fine grinding device as well as a
crushing and grinding device for fine and coarse feeds respectively. Currently the
perception is that the HPGR (crushing and grinding) could replace a SAG mill, but there
is no reason why it could not replace ball mills (grinding only) as well.
Figure 12, a schematic that shows the difference in feed conditions to the HPGR. The
coarse feed process is referred to as HPRC in this study to distinguish the different
process conditions. The efficiency of the HPGR bed breakage mechanism is assumed not
to be affected be it HPRC or HPGR as the transfer of mechanical energy of the rolls is the
same when it is transferred to the rock. Daniel (2003) showed that most of the energy
consumption in HPGR is applied to the material already in the compression zone between
the two rolls, as such any large particles in the feed are thus crushed in a pre-crushing
zone and has been found to consume much smaller quantities of energy. This compression
zone is a very small region between the rolls and is generally not influenced greatly by
particle size.
Figure 12 shows on the left had side a HPGR treating coarse particles, where the feed particle size is greater
than the working gap, where as the HPGR on the right hand side, is a condition where the feed size
distribution is less than or equal to the working gap.
These facts lend themselves to the possibility of operating HPGR’s in series in new
flowsheets where the ore may be processed dry. This may continue up until a point just
before the compacted flakes can be lightly de-agglomerated within small high throughput
ball mills. This concept is expanded on later in Chapter 3. Technically though, from an
energy efficiency standpoint there is no reason to think that there is any difference
between the efficiency of the HPGR treating either coarse feed particles or fine feed
particles.
- 32 -

The next section describes four excellent examples of publications that attempted to
compare comminution circuits and or comminution processes, which involved HPGR or
compressed bed breakage. The comparative methods range from being adequate to poor
and highlight the importance and need for sound comparative scientific methodologies.
The methodologies used in these publications resulted in different product size
distributions that could ultimately affect or be responsible for improved mineral recovery.
Unless a coarser product size distribution yielded a better recovery, it would be difficult to
conclude that the device resulted in improved recovery. Different sample head grades
could also be a factor and should be eliminated in the experimental design phase through
the use of representative “homogenised” head samples. Although this is mentioned in the
Evelin et al (1997) paper described in the next section, it nevertheless highlights the need
for a more accurate and consistent comparative methodology.
Comparative methods for comminution energy efficiency and recovery

Contribution by Klingmann
Klingmann (2005) recently presented a comparative study between the effects on gold
recovery against two types of comminution devices namely the Barmac vertical shaft
impact (VSI) crusher and the HPGR. In this study, the Soledad Mountain project
considered the development of an HPGR flowsheet as opposed to conventional VSI
technology. Tests were done on bulk samples of rhyolite, pyroclastics, quartz laterite and
vein ore obtained from surface and old underground workings between 1997 and 1999.
The ore was crushed in a VSI crusher and screened to produce samples sized to 100 % -
2.37 mm. Results from previous test work using the VSI and HPGR in 1990 were
presented and compared. These results are shown in Figure 13 and Figure 14. Only the
1990 samples should be compared as the ore type and grade could have changed over time
with the 1997 samples and as such should not be directly comparable. When comparing
the 1990 samples, the HPGR appears to have produced finer products below 100 Om.
.
Figure 13 shows the product size distribution of the samples treated in a Barmac VSI crusher and an HPGR
(Klingmann, 2005).
- 33 -

Since the results were not statistically verified under controlled feed conditions, the
improved gold recovery in Figure 14 could possibly have been the result of some
unmeasured process condition. The project however deemed the results valid based on
what was termed a definitive test program (Klingmann, 2005).
Figure 14 shows the bottle roll test gold recovery for the 1990 and 1997 samples treated in a Barmac VSI
crusher and an HPGR (Klingmann, 2005).
Contribution by Appling and Bwalya

Apling and Bwalya (1997) conducted similar laboratory studies into possible liberation
benefits offered by HPGR processing. These tests simulated HPGR processing using a
controlled hydraulic press to load samples in a 25 mm piston and die apparatus.
Agglomerated products of the compression tests were de-agglomerated in a vibratory rod
mill (130 mm diameter × 160 mm length) for four minutes. This procedure involves
energy input and the resultant products may differ significantly. Apling and Bwalya
(1997) used a lead-zinc ore for their comparative investigation. Sample sizes for both
compression breakage and rod milling were approximately 500 g. Compression breakage
tests were conducted in sub-samples of 20 g to 24 g and recombined before
agglomeration. Compression breakage tests were undertaken at loading forces of 50 kN,
80 kN and 100 kN in a piston and die apparatus that simulated HPGR performance.
Conventional milling was undertaken in a vibratory rod mill. Both the compression and
rod mill tests were conducted dry. The conventional milling route was conducted using a
rod mill. The rod mill samples were ground dry for periods of 2, 4, 6, 8 and 10 minutes.
No attempt was made to determine steady state conditions in a closed circuit milling
operation. The Bond equation appears to be the method for which the comminution
energy is determined.
The extent of mineral liberation was characterised by flotation release analysis

immediately after milling, which Apling and Bwalya determined to be the best
methodology for the study. Figure 15 shows the flotation response in terms of lead and
zinc recoveries for the various test conditions. Analysis of the results by Apling and
Bwalya (1997) in terms of particle size and applied energy are given in Figure 16 and
Figure 17 respectively.
- 34 -

100
Maximum Recovery (%) Zn
Maximum Recovery (%) Pb
95
90
Pb & Zn recovery (%)

85
80
75
70
65
60
14.3 13.9 13.06 12.55 11.48 20.7 17.3* 12.6* 7.5*
HPGR 100 kN HPGR 80 kN HPGR 80 kN HPGR 80 kN HPGR 50 kN Conventional Conventional Conventional Conventional
milling 8 min milling 8 min milling 6 min milling 4 min
Comminution device /energy (J/g)
Figure 15 shows energy consumption, recoveries for different conditions (Apling and Bwalya, 1997)
*estimated from size and Bond Ball Mill Work Index
The results in Figure 16 show that, at comparable product size (d80) or grinding energy,
the compression breakage product had a higher maximum recovery than the product of
conventional grinding for both lead and zinc. No replicate data was reported by the
authors. These tests have shown promising results for lead-zinc liberation during HPGR
processing. More work would be required to confirm these results and to determine the
optimum operation conditions for a particular ore.
The results show that a linear relationship exists between the grinding energy and the
maximum recovery of both zinc and lead to significance levels of greater than 99% for the
products of conventional grinding. A linear relationship exits between the grinding energy
and the maximum zinc recovery to a significance level of 96.5% for the high pressure
product. It should be noted that the linear relationships shown in this analysis are
representative of the energy range studied and have been used to show the differences
between the two grinding methods within this range. No data is available on the behavior
of the energy-liberation relationship for high pressure grinding at higher energies or
mineral liberations.
Visual interpretation of the data from Figure 17 showed that the products of high pressure
grinding were likely to exhibit a higher maximum recovery of both zinc and lead than the
products of conventional grinding.
- 35 -

100.0
95.0
Maximum Recovery (%) 90.0
85.0
80.0
75.0
Zn (High Pressure)
70.0
Pb (High Pressure)
65.0 Zn (Conventional Grinding)

Pb (Conventional Grinding)
60.0
0 5 10 15 20 25
Energy (J/g)
Figure 16 shows energy consumption against mineral recovery for different grinding mechanisms and
conditions (Apling and Bwalya, 1997)
95.0
90.0
85.0
Maximum Recovery (%)
80.0
75.0
70.0 Zn (High Pressure)

Pb (High Pressure)
Zn (Conventional Grinding)
65.0
Pb (Conventional Grinding)
60.0
0 100 200 300 400 500 600 700
d 80 (microns)
Figure 17 shows product particle size variation against recovery for different minerals and HPGR conditions
(Apling and Bwalya, 1997).
The high pressure comminution tests conducted by Apling and Bwalya (1997) spanned a
narrow energy range compared to the conventional grinding tests. A broader range of
grinding energies would have given a better insight into the effects of high pressure
comminution on flotation response. The validity of the energy measurements are not
convincing, and could be subject to large experimental error. The data published by
Apling and Bwalya (1997) did provide a valuable benchmark from which to design and
structure comparative investigations which include the use of HPGR when compared with
conventional grinding.
- 36 -

Contribution by Evelin et al
Evelin et al (1997) conducted their study to evaluate heap leaching of a primary sulfide
gold ore. The comparative crushers used were a Barmac vertical impact crusher, a Canica
crusher and a HPGR. The crushing conditions were as follows:
1. Barmac: A 1500 kg (-12.7 mm) sample was fed to a Duopactor 2400. The tests
were carried in open and closed circuit with a 2 mm screen.
2. Canica: A 1500 kg (-12.7 mm) sample was fed to a VSI 125 Canica crusher.
The test was performed in open circuit.
3. HPGR: A 1000 kg samples was processed with varying grinding forces, and a
100 kg control sample was processed at a specific grinding force of 6.0 N/mm2.
The sample was fed into a KHD pilot scale unit of 800 mm in diameter and roll
width of 120 mm. The roll speed was set at 0.4 m/s.
The resultant product size distributions of the different crusher devices are shown in
Figure 18. From these results it would appear that the Barmac crusher selectively chooses
the larger sized particles for breakage and not finer particles as shown by the similar
percentages of fine particles in both the feed and product.
Figure 18 shows how particle size distributions differ depending on the type of comminution devices used. No
indication of the energy expended in these devices is given. (Evelin et al, 1997).
The products of the different crusher devices were leached in column leach tests. The
samples leached in column leach tests had a top size of 2 mm for the Barmac and Canica
products. The HPGR product was not screened before leaching. The results of these tests
are shown in Figure 19.
Evelin et al (1997) stated that the results of Figure 19 showed that the HPGR product
performed better than the products of both the other crushers by 10%. The results would
appear to suggest that the Canica crusher product performed better than the Barmac
product. The differences in leaching behavior could be explained by the size distributions
of the different crusher products shown in Figure 18. The HPGR product had the finest
size distribution, followed by the Canica product, with the Barmac product being the
coarsest
- 37 -

80
Gold reovery
75
Gold recovery (%)

70
65
60
55
50
Lime Cement Dolmag Lime Cement Dolmag Cement Dolmag
Barmac Canica HPGR
. Crusher device/ore type
Figure 19 shows the increased gold recovery with the varied crushing devices.
Evelin et al (1997) carried out tests on HPGR products at a range of specific grinding
forces and a conventional rolls crusher product. The feed sample for these tests had a gold
content of approximately 5 g/t. The results of the tests are shown in Error! Not a valid
bookmark self-reference..
Table 5 column leach test % of gold extraction vs. energy input (Evelin et al, 1997)
Pressure Power Gold Extraction

Test Number
2
(N/mm ) (kWh/t) (%)
HPGR RP1 5.1 3.7 77.0
HPGR RP2 5.2 3.7 78.2
HPGR RP3 4.6 3.7 78.5
HPGR RP4 5.0 3.7 77.1
HPGR RP5 6.7 4.7 81.2
HPGR RP6 6.7 4.7 81.9
HPGR RP7 6.7 4.7 80.9
Evelin et al (1997) suggested that the difference in gold extraction between the HPGR
product at the highest specific pressure of 6.7 N/mm2 could be attributed to the relatively
high grade sample used for the tests. Products of the HPGR at specific grinding force less
than 6.7 N/mm2 showed lower gold extractions and could be attributed to the lower
energy input of 3.7 kWh/t as opposed to the higher energy input of 4.7 kWh/t. These
results are inconclusive as they do not provide an equal platform from which to compare
the results. The size distributions are certainly different. No comment is made in this
regards since individual product size distributions for the tests were not given.
Evelin’s work highlights the need to adopt a more careful and structured approach to
identifying the true effects of improved mineral recovery when alternative process routes
are studied. This thesis study decouples the effects of energy, product size distribution and
comminution device.
HPGR in hybrid grinding systems

Several researchers have conducted test work using the HPGR within hybrid grinding
circuits. Hybrid systems mean that the HPGR is used in conjunction with other
comminution devices to achieve a desired product. An example of a hybrid grinding
system that includes the use of the HPGR has been reported by Fuerstenau, et al. (1999) in
which a circuit employing a laboratory sized HPGR (Figure 20) and ball mill in series
- 38 -

were compared and found to use 70% of the energy to that of a ball mill alone. It was
concluded that the energy efficiency of the hybrid system could be further increased by
reducing the ball size in the ball mill, providing that the HPGR product contained
sufficient internal damage to allow for the use of smaller balls. This is a good example of
how operating conditions with such hybrid comminution systems need to be optimised to
obtain conditions where the best possible use of energy has been achieved. Unfortunately
no attempt was made to compare or consider possible changes to the characteristics of the
products in the two circuits, nor was there any detail as to how energy measurements were
made.
Baseline
Circuit
Feed New Circuit
(t/hr)
Product
Feed (t/hr)
(t/hr)
Ball Mill
Classifying
cyclone Product
(t/hr)
HPGR
Ball Mill
Classifying
cyclone
Re-circulating load
Re-circulating load
Figure 20 shows a comparison of the baseline ball mill circuit and the single pass hybrid HPGR- ball mill
circuit.
Ideally a conventional ball mill in closed circuit would be best compared with an HPGR
unit in closed circuit as described in Figure 21. Some previous exploratory research
conducted by Johnson et al. (1995) showed that obtaining steady state conditions in a
closed circuit HPGR was not easily achieved. They, however, did manage to advance the
HPGR power efficiency debate substantially when they completed a test program that was
designed to compare the energy consumption, and hence efficiency, between Bond test
data (conventional milling) and the energy of a locked cycle HPGR test. Details of the
test procedure were not provided.
Baseline New Circuit

Circuit
Feed
(t/hr)
Product
Feed (t/hr)
(t/hr)
Ball Mill
Classifying
cyclone
Product
HPGR
(t/hr)
Classifying
Re-circulating load cyclone OR
Screen
Re-circulating load
Figure 21 shows the direct comparison between a ball mill in closed circuit, against a single pass HPGR in
closed circuit.
- 39 -

Esna-Ashari and Kellerwessel (1988) conducted a locked cycle test to determine the
leaching behavior of the ore when ground in a HPGR in closed circuit. A re-circulating
load of 84% was achieved with a fresh feed rate of 5.0 t/h and an energy consumption of
4.3 kWh/t. The -1.0 mm product from this test was then used for a standard leaching test
over the full size distribution. This test gave a gold extraction of 94.7% after 18 hours.
Similar tests were not carried out on the conventionally ground ore. Even though gold
recovery is high there is no comparison to conventional crushing and milling. No error
data or replicate tests were shown for the locked cycle test. The size distribution of the
sample for leaching was not given, nor was there any mention about the mineralogy of the
ore. This however is not to say that the ore was not suited to HPGR processing.
HPGR micro-cracking and mineral liberation

Perceived fracture mechanisms taking place in HPGR inter-particle comminution.
Liberation of valuable minerals from the gangue through comminution is an essential step
that determines the effectiveness of downstream hydrometallurgical or physical separation
processes. Traditionally this process has taken place in tumbling mill devices. Critical
tasks within this study include the measurement of comminution energy and the
corresponding mineral liberation in hybrid HPGR/ball mill circuits with reference to
conventional ball mill processing on their own. The study as previously indicated
investigates the difference between the energy expended and the nature of the
comminution with respect to liberation and the type of device or circuit configuration
used. The comparative study assumes that a collection of different breakage processes are
taking place at the same time. The effects of this are that different proportions of minerals
are liberated at different stages. The net result is compared using mineral liberation
analysis (MLA). Some of the perceived breakage processes that may well be taking place
in bed breakage are shown in Figure 22.
Direction of applied force
Pressure/particle
orientation
Preferential mineral breakage breakage
Inter-granular
selective particle Trans-granular
breakage
Inter-particle bed breakage mechanisms in HPGR
Figure 22 illustrates typical fracture processes that may take place within the HPGR.
These breakage processes have previously been defined by King (2004), but the actual
processes themselves are very difficult to quantify and measure. The four main
descriptions given by King are as follows.
- 40 -

• Inter-granular – selective particle breakage along phase grain boundaries
• Pressure localized – orientation dependant fracture
• Preferential breakage of an entire mineral phase within a particle
• Trans-granular breakage – random particle fracture
The main hypothesis here is to establish whether selective breakage and /or preferential
mineral liberation (based on MLA measurements and composition distribution) is
enhanced when using HPGR technology relative to conventional ball milling. These
aspects have been sparsely reported in the literature. Although a few authors have
evaluated liberation (Apling and Bwalya, 1997; Austin et al., 1993; Battersby et al., 1993;
Fandrich, 1997). A limited amount of research in the area of quantifying mineral
liberation is reported, but different particle size distributions are being compared.
Three very similar papers by Fandrich (1997; 1999) and Fandrich et al. (1997)
investigated the liberation characteristics of a binary iron oxide ore using scanning
electron microscopy with some degree of success. They concluded that preferential
liberation did occur in particle bed breakage, and used the data to modify a liberation
model to incorporate the effects of preferential liberation. Prior to this, liberation models
considered breakage of particles during size reduction to be random, and there was no
evidence to suggest that a particular comminution mechanism was responsible to a greater
or lesser extent of preferentially liberated mineral.
Esna-Ashari and Kellerwessel (1988) reported that during HPGR processing, cracks
followed grain boundaries preferentially. If the breakage properties of the mineral grains
and the gangue matrix are appreciably different, this could lead to better liberation at
coarser sizes. In the diamond industry, it is well known that the HPGR process
preferentially liberates both small and large diamond within the working gap of an HPGR,
without damage to extremely high value large diamond. However due to the scarcity of
diamond or its large enrichment ratio, the diamond industry has never statistically proven
this nor published any results relating to this in a scientific journal. Diamonds, due to their
hardness properties are preferentially liberated from the host rock. Possible preferential
liberation of generally softer valuable mineral compounds from metaliferous ores still
requires experimental proof. This preferential breakage may lead to over grinding in some
ores, where the valuable mineral is preferentially liberated and then further broken with
gangue in continued grinding processes, but this has not yet been studied.
Stephenson (1997) and Fandrich et al (1997) have provided a most comprehensive report
in terms of selective liberation and selective breakage in particle bed breakage
mechanisms within HPGR and piston die methods. However in these studies there is no
comparative analysis with conventional milling techniques.
Fandrich et al (1997) reported on the liberation characteristics of a binary iron oxide ore
in confined bed breakage tests. The breakage mechanism in the HPGR is very similar to
that found in a piston press using a confined bed of particles. Fandrich et al (1997)
conducted tests on samples in the size range of 425-600 µm with narrow grade fractions
in some samples and wider distributions of grades in other samples. Their tests indicated
that preferential breakage of the silicate gangue occurred in the binary iron oxide tested.
This was shown by a concentration of silicate gangue in the fine product sizes after bed
- 41 -

breakage and it was the authors’ contention that preferential breakage was promoted by
bed breakage. Similar results were obtained by Klymowsky (1998) at the Cleveland Cliffs
Empire mine.
Sutherland and Fandrich (1996) found that grain boundary or interfacial breakage
occurred during confined bed breakage tests. No indication was given as to whether this
mode of breakage was enhanced by particle bed breakage.
Claims that particle micro-cracking due to bed breakage in a HPGR are extensively
reported in the literature. Literature often claims that the micro-cracking is largely
responsible for improved recoveries. Whilst P80’s of HPGR products and conventionally
crushed products are reported as being the same. The question remains whether it is
perhaps the difference in the entire particle size distribution that is responsible for
improved recoveries.
The test methodology designed to eliminate these effects in this study are explained in
some detail in Chapter 3 where the various circuits that employ HPGR were designed to
produce identical products to the conventional circuits to directly compare them.
MLA data is traditionally very difficult to compare as the MLA sized sub-samples are
rarely mass balanced. Mineral particles are not tracked. (the MLA mineral tracking and
mass balancing capability is currently a separate research topic). The effects of HPGR
processing on the liberation of the valuable mineral components of the bauxite ore, lead
zinc ore and the platinum ore are studied here and are compared directly. The
comparisons are valid because the experiments were designed to produce the same
product size distribution.
A comprehensive range of MLA data was collected and analysed on the liberation
characteristics of the three ore types that were processed in the HPGR circuit and
conventional milling circuit. The data are available on the CD accompanying this thesis at
the back of the volume. The data were used to identify if non-random breakage occurred
during HPGR processing, which may lead to liberation benefits, and to determine the
mechanisms by which these benefits were realised e.g. inter-granular, trans-granular,
selective breakage and preferential breakage as previously described in the literature
review. The aim of this section in the literature review is to give an indication of whether
any advantages for mineral liberation exist when processing an ore with the HPGR and to
confirm the so called claim that HPGR promotes particle micro-cracking. Figure 23
shows the two most famous pictures that claim to verify that micro-cracking in HPGR
does exist. (Knecht, 1994, Kirsch, 2006, Knecht and Baum 1994, Klymowsky et al 2002)
Figure 23 shows the two most famous and widely published pictures that claim that HPGR produces
microcracks in the products (Kirsch 2006, ,Knecht 1994, Knecht and Baum 1994, Klymowsky et al 2002).
- 42 -

Stephenson (1997) conducted bottle roll tests on HPGR and conventionally crushed
material, as did McNab (2006) which is discussed later. In order to ensure that similar
products were tested, Stephenson took products of the conventionally crushed material
and created a reconstituted feed that matched the product size distribution of the HPGR
products. When the same size distributions were compared there was a 2-10% difference
in the total recovery between the HPGR and conventional tests, favouring the HPGR.
Stephenson concluded that this was due to the enhancement of micro-cracking.
Micro-crack pictures of an HPGR processed gold ore that were presented by Stephenson
(1997) are shown in Figure 24, Figure 25 and Figure 26. Stephenson showed HPGR
treated particles and did not present any conventionally crushed particles for micro-crack
observation. Hence no comparative particles were provided, and so, it could be argued
that the cracks could have represented a natural particle cracking phenomena. The
cracking could have resulted when the particles were mounted into the epoxy and polished
for the SEM and MLA analysis.
Figure 24 claims to show inter-granular cracks of two pyretic gold ore particles - 850 Lm 1180 Lm
(Stephenson 1997).
Figure 25 on the LHS shows a siliceous gold ore – 850 Lm 1180 Lm and on the RHS a pyretic gold ore -
75Lm 106Lm (Stephenson 1997).
- 43 -

Figure 26 on the LHS shows a pyretic gold ore – 425 Lm 300 Lm and on the RHS a pyretic gold ore - 850
Lm 1180 Lm (Stephenson 1997).
Stephenson (1997) followed a methodology of measuring the crack lengths using a

computer mouse to track the paths of fifty particles. He subjected both HPGR products
and conventionally crushed products to the same process and completed a statistical
analysis on the comparative lengths of the cracks. Stephenson (1997) concluded that
substantial evidence shows that the density of the micro-cracks in the ore particles in the -
425 + 300 um size range were enhanced by HPGR processing for the two pyretic gold
ores. However it would appear that Stephenson did not show by means of comparative
images any extent of the micro-cracks between the conventionally crushed products and
the HPGR products, nor was there any comparison made between the products of the
HPGR against the conventional crushed material. Stephenson measured, and attempted to
quantify the density of the microcracks with some degree of success.
McNab (2006) conducted bottle roll tests on a gold ore. He tested a range of HPGR
specific pressures, and compared the results with a range of products produced through a
range of different close side settings on a conventional cone crusher. The results are
presented in Figure 27 and Figure 28, with Figure 28 superimposing the results of
Figure 27. Whilst the products look similar with respect to P80 as McNab has claimed, the
bottle roll tests produced quite different recoveries. This is partially as a result of the
difference in product size distributions and the presence of micro-cracks as McNab
suggested. However the microcracks were not measured and were not analysed and so his
suggestion is merely speculative.
- 44 -

Figure 27 shows the product size distributions of the conventionally crushed material and corresponding
calculated gold leach extractions (McNab 2006).
Figure 28 shows the product size distributions of the conventionally crushed material and corresponding
calculated gold leach extractions (McNab 2006).
Figure 29 shows the results of the bottle roll tests (after McNab, 2006) in which the gold
leach extraction rate and recovery have significantly improved through the treatment of
the ore through an HPGR. Figure 29 displays the similar P80’s of the HPGR (7802 Om)
and jaw crusher (8173 Om), whereas Figure 30 (Kirsch, 2006) shows how products
produced in an HPGR and cone crusher may well have the same or similar P80 but clearly
a completely different size distribution.
Hence it is very important to be able to compare like with like, and that the P80 does not
adequately represent a description of the full product size distribution.
- 45 -

Figure 29 shows how gold extraction is claimed to be improved through the use of HPGR in a Gold heap
leaching application. In this example by McNab (2006), The products are described as a jaw crushed product
with a P80 of 8.2 mm and an HPGR product with a P80 of 7.8 mm. The microcracks in these images appear
to show trans-granular cracking, but the severity of the cracks appears to be more prominent in the bright
shaded minerals.
Figure 30 shows how the product size distribution of a conventional cone crusher is compared with the
products of an HPGR. Both product size distributions display a P80 of about 8 mm which highlight the effect of
possible errors should the p80 be used to describe the entire size distribution as required in the Bond
equations (after Kirsch 2006).
Types and mechanisms of mineral liberation

There is a very limited amount of literature relating to the study of mineral liberation in
HPGR processing. The literature suggests several possible beneficial effects of HPGR
processing. These effects are discussed in terms of their effects on product size and shape
characteristics, leaching characteristics, liberation characteristics and the amount of
energy expended in the process (Gerrard et al., 2004; Goeldner, 2003; Tavares, 2005).
Dedicated research into examining how minerals liberate in various types of comminution
devices has not yet been attempted.
- 46 -

There have been other technical issues surrounding the wider use of HPGR’s that
dominated the topics studied. Schönert (1991) investigated particle strengths, and stated
that it was not possible to break all the particles even with a very high pressure. Schönert
said that the stronger particles are surrounded by the fragments of the weaker ones and are
protected. Schönert did not investigate if the stronger particles were associated with any
mineralization. The question remains as to whether the valuable minerals are associated
with stronger or weaker phases within the particles.
A different perspective to this is to consider a 2D DEM simulation of the forces involved

in bed breakage. This is illustrated in Figure 31, Figure 32 and Figure 33 that depict how
the force field (denoted by the black lines) is distributed throughout the bed of particles.
This analysis indicates that some of the larger particles, due to their position and
orientation in the bed are sometimes not exposed to the same magnitude of force that
other particles within the bed are be subjected to.
In Figure 33 two circled areas show how some large particles are not exposed to the
differential stress field, and hence could survive the breakage process. This phenomenon
evident in reality when larger particles are known to survive the breakage process, but, up
until now it claimed that the surviving particles consist of a much harder mineral
component than the surrounding particles (Hilden, 2005). These two dimensional discreet
element method (DEM) images suggest that it is rather the particle orientation and
exposure to the force field that results in particles being broken and that it is the high
pressure/stresses at the individual particle contacts that result in breakage. Clearly if two
dissimilar particles in terms of hardness are forced upon one another, it will be the softer
particle that will break, which does affirm the claim that harder particles survive the
process of being broken. Particle orientation and exposure to force fields is apparently
important if the DEM images of Hilden are taken into consideration.
Figure 31 shows particle orientation in a piston and die, before bed compression(Hilden,2005).
- 47 -

Figure 32 shows force/energy dispersion routes as the force is applied and energy imparted by the contact
forces into the bed of particles(Hilden,2005).
Figure 33 shows localized areas within the bed which due to particle orientation appear not to be exposed to
the energy imparted by the contact forces during bed compression(Hilden,2005).
Stephenson’s contribution
Stephenson (1997) provided a comparative analysis of the properties of the materials that
have been subjected to both HPGR and conventional crushing means. In Stephenson’s
(1997) PhD thesis, an accurate account of selective liberation and the benefits to
downstream processes were studied. Stephenson concluded by recommending that further
work be considered to investigate and quantify the extent to which selective/preferential
liberation takes place and whether it is enhanced in the different types of comminution
devices/mechanisms.
After extensive investigations of micro-cracks induced by HPGR, Stephenson (1997)

came to conclude that there was an increase in micro-crack density when ores are
subjected to HPGR treatment.
Density of micro-cracks in ore particles in the 300Om and 425 Om size fraction was
enhanced by HPGR processing for two gold ores. The enhancement of micro-cracking
during HPGR processing is an important conclusion because many hypotheses for the
- 48 -

mechanisms by which the HPGR may present benefits for downstream operations relied
on this enhancement of micro-cracks by the HPGR.
It was shown that the Work Index (WI) and fracture strength of HPGR products are
reduced with respect to products of a conventional rolls crusher. This was shown to be the
case with four of seven ores tested. These ores were quartz, marble, pyritic gold ore and
siliceous gold ore. No strength change was shown for hematite, lamproite or bauxite.
Stephenson paid a lot of attention to getting statistically representative results. The
comparative tests were analysed using statistical techniques and his results and arguments
appear quite plausible. Stephenson however treats the Bond work index as an “ore
property” rather than the energy required to reduce the ore from a known F80 to P80. In
his analysis Stephenson devised a “special” version of the Bond test, whereby the feed
size distribution of the Bond test taken from a conventional rolls crusher was reconstituted
to match the product size distribution of the HPGR tests that are used as the feed to the
Bond work index test (BWI). These two BWI tests are then compared using identical feed
size distributions, but the origin of the feed has been “pre-treated” in both the
conventional rolls crusher and then the HPGR. Stephenson’s special Bond test then
indicates “an ore property” characteristic change. Stephenson re-constituted the
conventional rolls crusher feed with HPGR products. Using this approach he measured a
decrease in the BWI ranging from 3-5 %. HPGR manufacturers claim to do a similar type
of test, but the conventional products are used to reconstruct the HPGR feed size
distribution. This generally results in a claimed 10-15% decrease in BWI. The
methodology is plausible when consideration to the measurements of the characteristics of
the HPGR products is concerned. It is however important to realise that the reduction in
BWI does not imply a similar percent reduction in energy requirement. These expressions
of HPGR benefits are often quoted as percentages, and are often used to justify the
equivalent energy savings that could be realised should HPGR be considered.
The above mentioned Stephenson’s methodology is described here in detail, so as not to

confuse the application of the Bond locked cycle test methodology that has been used in
this study. Put simply, the BWI is not an “ore characteristic” when feed size and closing
screen size is deviated from the standard procedure. In these circumstances it is rather a
measure of the energy required to do specified work in the ball mill by using the modified
“work index” and third theory equation.
A recent study by Shi et al (2006) used the above mentioned methodology in conjunction
with traditional Bond third theory equations to compare the benefits of using HPGR in a
single pass hybrid HPGR/ball mill circuit when compared to ball milling alone. The study
proved to yield a 20% energy saving in the treatment of a platinum ore. Later it will be
shown that this methodology differs slightly in this study in that the analysis of the Bond
energy is taken from direct measurements.
Similarly work published by van Drunick and Smit (2006) followed a similar approach,
but used the traditional Bond work index and third theory equation to determine the ball
milling energy. The net result is that with all these un-standardized methodologies,
various different calculated energy savings are presented. This study aims to clarify this,
and show why the adopted mythology is accurate and representative of the expected
energy saving that can be realised though the use of HPGR.
Stephenson hypothesised that the micro-crack formed in HPGR products facilitated a
strength decrease of the rock. The micro-cracks then facilitate an easier path for further
- 49 -

breakage which results in reduced energy requirements. When the ball mill is used
downstream of HPGR products, then the visible micro-cracks mostly disappear in the
breakage processes that follow. Stephenson’s aim however was to demonstrate that the
material had weakened and was responsible for a lowering in the Bond work index.
Stephenson completed tests on quartz, marble, pyritic gold ore, siliceous gold ore,
lamproite and bauxite. The lamproite and bauxite ore that did not display a strength
reduction after HPGR processing appeared to have a larger degree of porosity than the
other four ores tested. This supported a hypothesis that micro-crack propagation may be
retarded by voids in the structure of the material. This could mean that the tests conducted
on the bauxite results in circumstances where agglomeration of the ore during HPGR
processing played a part in the lack of strength reduction, and possibly as a result of the
now compacted flake particles being more competent than the original material.
Stephenson’s work confirmed that HPGR processing enhanced the liberation of pyrite
from a silicate gangue. The mechanism by which this liberation enhancement was
achieved was hypothesised to be enhanced breakage along grain boundaries. Phase
specific interfacial area calculations for HPGR and conventional crusher products
supported this hypothesis. The evidence from liberation studies was further supported by
micro structural observations (Figure 24-26) which showed that micro-cracks would, in
some cases, follow grain boundaries in preference to the path they had been following.
This was observed in both HPGR and conventional rolls crusher products.
The observation that cracks may preferentially follow grain boundaries lends support to
the evidence that HPGR processing enhances grain boundary breakage over what is
achieved in a conventional rolls crusher. Stephenson provided an excellent critical review
of the then available literature in 1997 in his PhD thesis where as a summary of the
findings of his work of during the completion of his thesis was published in the public
domain at the Mining Technology Conference in 1996 (Stephenson et al, 1996).
The remainder of the literature review section is thus taken directly from Stephenson’s
(1997) thesis. His review included literature on the downstream benefits of using HPGR,
and he summed up the status quo of potential benefits of using HPGR at the time of
writing his thesis. Stephenson’s content is relevant to this study and is included here.
Stephenson is accordingly acknowledged for this contribution to the literature.
Clarke and Wills (1989) reported on experiments into the liberation of cassiterite in
HPGR products compared with products of a rod mill. These tests showed:
1. A lower production of untreatable fines in the HPGR than the rod mill.
2. A higher recovery of tin into the +45 µm concentrate (+3.3 SG material) for the
HPGR product.
3. Higher concentrate grade in the HPGR product.
4. Higher separation efficiency of the HPGR product.
The tests reported on by Clarke and Wills (1989) were preliminary tests and the authors
acknowledged that further validation was required.
Mörsky et al (1995) conducted tests on the liberation characteristics of a chromite ore

after HPGR grinding, rod milling and ball milling. The chromite was present in a talc-
- 50 -

carbonate gangue which was soft relative to the chromite. These studies showed a
difference in liberation of the chromite between the HPGR and rod mill products with the
HPGR product being more highly liberated than the rod mill product. Very little
difference was shown between the HPGR and ball mill products. The liberation tests were
carried out as density separations and the coarsest size tested was 500 µm. This was
consistent with the beneficiation method being practiced on that ore on an industrial scale.
Battersby et al (1992) discussed the possibility of preferential liberation of chalcopyrite

through grain boundary breakage. This was illustrated by a Scanning Electron Microscope
(SEM) photomicrograph of a copper ore showing micro-cracks in a HPGR product
following chalcopyrite grain boundaries. No comparative photomicrographs were shown
for products of other crusher types on the same ore. Without this comparison, no
statements can be made with any degree of confidence on the effect of the HPGR on the
liberation characteristics of an ore.
Knecht (1994) stated, with respect to oxide and refractory gold leaching, that fine particle
liberation was greatly improved due to the large number of micro cracks produced without
the risk of producing excessive fines. Hard silicified and cherty gangue material was
reported by Knecht (1994) to be shattered easily through HPGR processing with deep
micro cracks penetrating as far as 150 µm into the particles, allowing access to any
contained gold. Gold and gold-bearing veins that are “healed” by silica were claimed to
reopen easily with HPGR processing, allowing easy access to the gold. The investigations
were conducted using a scanning electron microscope.
Knecht (1994) stated that gold extractions were increased by 18% over conventionally
ground ores, while the leach time for optimum gold extraction was decreased by 50% over
conventionally ground ores. Silver extractions increased by 33% over conventionally
ground ores. Knecht (1994) claimed that HPGR processing was particularly effective in
making gold in the coarse +850 µm size fractions available for leaching. This point was
illustrated in Table 6.
Table 6 lists the +0.85 mm Leaching Results (Knecht, 1994)
Polycom HPGR Conventional Crushing

+850 micron fractions
31.8 70.6
(weight %)
Unleached Gold in +850
micron fraction 47.1 83.7
(weight %)
Knecht’s (1994) results indicate that 47.1% and 83.7% of un-leached gold occurred in the
+850 Om fraction for HPGR grinding and conventional crushing respectively. Knecht
claimed that these results showed an increased liberation of gold in the coarse size
fractions. It appears, from the data in the paper, that the decrease in the proportion of the
un-leached gold in the +850 Om size fractions of the HPGR product was less than the
decrease in the proportion of +850 Om material in the HPGR product. This data may, in
fact, have indicated that the gold in the +850 Om size fractions of the HPGR product was
actually less easily leached than that in the product of conventional crushing.
- 51 -

The analysis showed that the concentration of gold in the +850 Om size fraction of the
HPGR and conventional crusher products was 1.5 and 1.2 times the average gold
concentration in each of the samples respectively. Assuming the 18% increase in gold
extraction claimed by Knecht (1994), the crushing methods produced tailings gold
concentrations in the +850 Om size fraction that were equal. This analysis showed that the
HPGR did not improve gold leaching from the +850 Om size fraction. The proportion of
+850 Om material was, however, decreased in the HPGR product which could lead to
improved leaching due to the increase in fines.
Knecht (1994) claimed the following advantages for processing of oxide and refractory
gold ores.
1. Better liberation of native gold from gangue

2. High micro-cracking of coarse particles
3. High liberation of fracture located gold
4. Optimised early gravity separation of coarse gold
5. Better cyanidation of low grade silicified ores
6. Faster leach kinetics
7. Better leach performance of iron oxide rich ores
Mineralogical examinations of the HPGR products were undertaken by KCGM to assess

micro-cracking (Watson and Brooks, 1994). These examinations showed visible micro-
cracking in the hard, brittle components of the ore such as quartz, feldspar and pyrite, but
not in the softer minerals. Where a brittle material was surrounded by softer material, no
fracturing of this harder component was found. This may be due to the softer components
of the ore protecting the harder component that they surround.
Patzelt et al (1997) reported much of the results published previously by Patzelt et al

(1995), along with some further results and discussion. Figure 34 shows the leaching
results for a semi-refractory gold ore from Nevada on a size-by-size basis. No error data
was given for these results.
Figure 34 Leach Results of Semi-Refractory Gold Ore (Nevada) in Size Fraction (Patzelt et al, 1995; Patzelt et
al, 1997).
From the results shown in Figure 34, Patzelt et al (1997) made the following conclusions:
-75 µm size fraction of HPGR product showed over 90% gold extraction substantial leach
improvements were made in the +300 µm size fractions of the HPGR product
- 52 -

Patzelt et al (1997) claimed an increase of 13% gold extraction after HPGR processing
from results of a 70 day column leach test. Patzelt et al (1997) reported that gold
extraction in the -43 µm size fraction was improved by 4-5% after HPGR processing of a
Carlin type refractory gold ore. No further description was given for the ore.
When treating low grade gold ores, Patzelt et al (1997) claimed the following: gold
extraction increases of up to 8% up to 50% better extraction of ultra fine native gold
encapsulated in silicified rock material an increase of 9-12% in physical pre-concentration
of gold by gravity separation from an oxide ore. No data was presented to support these
specific claims.
Baum et al (1994) claimed many benefits for gold leaching using the HPGR. They
reported increases in leach extractions as well as decreases in required leach times. Over
grinding and slimes generation, as experienced with conventional SAG and ball mills, was
claimed to be minimised. The production of coarse middlings was claimed to be reduced,
and the recovery of gold from what middlings were created was claimed to be increased
through micro-cracking of the HPGR product. No data was presented to support these
claims.”
Stephenson concluded that future work should focus in the area liberation characteristics
of HPGR products. He suggested that additional studies into the liberation of HPGR
products would provide a greater understanding of:
• The relationship between HPGR specific grinding force and the liberation of
minerals.
• Any benefits that exist for leaching a fine HPGR product without another milling
stage between the HPGR and leaching.
Scarce literature is available on the liberation of values in HPGR products. What literature
that is available is inconclusive and sometimes contradictory. A detailed mineral
liberation study is required to facilitate a better understanding of this area. Claims by
several authors generally consist of hypotheses that have not yet been confirmed with
appropriate experimental methodologies. The hypotheses generally appear quite feasible,
but the appropriate tests and accurate measurements need to be performed before they are
considered to be well-supported.
Literature review summary

The literature review has provided an overview of the drivers behind the research and has
focused on the controversy of the definition and evaluation of comminution energy
efficiency. The review has identified the technical shortfalls in existing methods and
knowledge regarding the analysis of comminution circuit energy efficiency.
In-efficient comminution circuits often result in excessive generation of ultra-fine

particles (at the expense of large amounts of energy), which are sometimes difficult to
recover downstream. Size reduction in comminution processes is accomplished in a series
of stages in a grinding circuit that is tailored to the properties of the particular mine ore.
- 53 -

Commonly this involves two or more classes of equipment starting with crushers,
followed by SAG mills and ball mills.
Broadly, energy consumption increases with the fineness of the material produced by each
piece of equipment. The efficiency of each device employed to do this may vary
depending on operating conditions. Generally the energy efficiency of tumbling mills is
poor, and may be improved by optimised mill control or through the use of more energy
efficient devices such as HPGR.
Much of the literature involving HPGR has simply claimed that the HPGR is more energy
efficient without actually quantifying this (Lane, 2006). Quantifying general comminution
energy efficiency is sparsely represented in the literature and difficult to define due to the
lack of a standard from which to reference the efficiency (e.g. tumbling mills, breakage
process, motor energy consumption). To complicate matters, the use of the Bond third
theory equation does not apply to HPGR comminution and so using the empirical
equation is prone to error.
The energy efficiency of the HPGR is different mainly because the so-called inter-particle
comminution is totally different to impact breakage, with impact breakage being known to
be more energy efficient in particle breakage than that of inter-particle bed breakage
(Schönert, 1988). In addition to the mechanism or mode of breakage the comminution
device may well utilise the available mechanical energy that has been converted from the
electrical energy in a more efficient way. This aspect of the efficiency of an HPGR has not
been researched.
It is for these reasons that a Bond equivalent “work index” for the HPGR is required. As
of yet, the concept of a work index has been proposed (Morrell, 2006, Klymowsky and
Liu, 1997 b), but to the authors’ knowledge this has not been developed as an industry
standard. Klymowsky ingeniously first proposed the need for such an index, and may well
have developed this since 1997 for the HPGR manufacturers. There is little in the public
domain that details and equivalent work index for HPGR other than that recently
published by Morrell (2006)
The original HPGR research contributions of Schönert completed during the 1980's
provided the “hybrid car” version of a more energy efficient comminution device. Today
there are three manufacturers of HPGR who over the years have developed their own and
proclaimed "standard" processing rules which are used by the vendors to retain
competitive advantages over the technology.
Energy efficiency and mineral liberation associated with HPGR’s remains poorly
understood though lack of rigid test designs and comparative data analysis. The limited
amount of literature is often very conservative and inadequately supported without sound
scientific experimentation. This statement does not intend to criticize the manufacturer’s
ability to conduct test work but rather to highlight the point that research institutions and
universities always follow established and proven experimental procedures, detailed data
analysis and well referenced reporting etiquettes. Literature is always critically analysed.
As such this literature survey has identified that sound comparative experimental tests and
methods are required to produce accurate and meaningful data in the measurement and
analysis of energy and energy efficiency. Analysis of data that that makes use of
- 54 -

physically measuring the energy consumption of the mills must be compared to the
empirical (formula calculated) Bond method as this is remains to be widely accepted by
design engineers and metallurgists. The importance of evaluating and quantifying energy
and thus comminution energy efficiency when HPGR is considered needs to be carefully
considered and evaluated. It is for this reason that a concerted effort was made to
reviewing energy-size distribution relationships. The role that micro-cracking and
possible improved mineral liberation brought about by the use of HPGR is covered to
support the findings presented in Chapter 6.
With respect to the selection and sizing of HPGR for industrial comminution circuits,
there exists a very comprehensive chapter in the SME 2002 Vancouver conference
proceedings’, Mineral processing plant design, practice and control. In particular a sub-
section in Chapter 5 of the proceedings by Klymowsky et al (2002) gives a very
comprehensive account of HPGR in respect describing technically the challenges behind
HPGR applications within the minerals industry. Aspects of energy efficiency, mineral
liberation and benefits with respect to sustainable processing are not covered, which is
what this thesis aims to investigate.
- 55 -

- 56 -

CHAPTER 3
EXPERIMENTAL PROGRAM
Introduction
In general the measurement of the total energy consumption and overall energy efficiency
of a particular comminution device is difficult, and more so in laboratory scale
comminution devices because of the grossly oversized motors that are usually attached to
the devices.
For these reasons the Bond suite of tests (crusher, rod and ball mills) and their associated
work indices have become the most widely known and used measure of a rock’s breakage
property in the crusher and tumbling mill. The empirical Bond methodology as such is
reasonably good at determining the energy requirements but does little to explain how a
tumbling mill device and associated breakage processes utilise the available electrical
energy. Bond’s method provides the required energy of a full scale tumbling mill device
under typical wet grinding conditions to reach a specified grind size.
It is unknown whether Bond’s empirical equations are valid under conditions in which the
study tests were conducted. For this reason the study quantifies the energy use by using
both Bond’s 3rd law equation and direct energy measurements made through the use of a
digital energy meter for the laboratory rod mill, ball mill and HPGR.
Background to experimental design

As mentioned previously there are numerous references in the literature that have claimed
that the ball mill is less than 1% efficient and have concluded that much opportunity still
exists to greatly decrease the amount of energy required to comminute ores. As energy
efficiency is key to the investigation, it was decided to use the Bond ball and rod mills as
the baseline circuit, because it is widely used using the Bond methodology to determine
the comminution energy requirements. Any suggested alternative circuit should therefore
be benchmarked against the baseline circuit.
The findings of the literature review suggested that rigid comparative experimental tests
and methods are required for this type of comparative study involving comminution
energy. The hypothesis of determining whether enhanced or preferential mineral liberation
takes place in comminution circuits employing HPGR was linked to the assessment of
comminution energy in the circuits. A comprehensive test program and experimental
design that would be able to independently compare the liberation characteristics of the
products of an HPGR circuit to that of a conventional comminution circuit was necessary.
As such, it was important to ensure that any liberation benefits were not caused as a result
of higher relative energy input or improved energy efficiency of the device. Ideally, the
use of HPGR should be used to promote two positive benefits namely, to reduce energy
- 57 -

utilisation and to improve mineral liberation. As such the following should be borne in
mind when considering the experimental design plan of the research.
1. Energy efficiency is a function of the comminution mechanism and the

comminution device.
2. Mineral liberation is a function of the ore characteristics, the specific

comminution energy input and the resultant product size.
3. Energy efficient mineral liberation is a function of the comminution device, the

comminution energy input, and liberated mineral content. It could be
independent of the particle size distribution, for example a coarser product
producing more liberated mineral.
The statements above result in a triangle of various experimental measurements being

made in different comminution circuits. The experimental measurements focus on
measured net comminution energy, quantified mineral liberation and process energy
efficiency in relation to the type of comminution device employed for the task (Figure 35).
The net comminution energy measurement is determined using both the traditional Bond
third theory methodology as well as a new methodology of direct comminution energy
measurement using a digital energy meter. Mineral liberation is measured using advanced
mineral liberation analyser data and techniques.
Total net circuit Comminution

specific comminution device,
energy energy efficiency
Liberation /
Enhanced preferential
mineral liberation
Figure 35 shows the interaction between the survey or experimental methodology.
In order to independently investigate the effect of liberation and energy use it was
necessary to complete a series of tests that were designed to:
1. Compare total specific comminution energy to produce an identical product in

hybrid HPGR/tumbling mill comminution circuits versus conventional rod or ball
mill comminution circuits.
2. Compare the mineral liberation and micro-crack density between hybrid
HPGR/tumbling mill comminution circuits versus conventional rod or ball mill
circuits when identical products are produced.
3. Quantify the measured amount of mineral liberated in HPGR only circuits versus
conventional rod or ball mill circuits at possible lower total energy consumptions
and with non identical products.
- 58 -

Methodology
Direct energy measurement in the Bond ball mill.
As energy efficiency is one of the key areas of investigation it is vital to confirm an
appropriate experimental energy measurement technique or methodology. In this case the
Bond methodology and a method of direct energy measurement are being considered.
Direct energy measurement techniques in the Bond rod and ball mills are first compared
with Bond’s empirical formula method and form the basis of Chapter 4.
Circuit energy efficiency

The study’s process circuit energy efficiency is defined as the percentage reduction in
energy after the total circuit comminution energy is measured for both circuits each of
which produce the same product and forms the basis of Chapter 5. As previously
mentioned the definition of energy efficiency has often been embroiled in controversy and
as such these experiments were designed so that the HPGR/ball mill circuits could be
directly compared with ball and rod mills as a baseline. In the analysis both the direct
method and Bond’s empirical method are used to evaluate and quantify the energy saving.
On the basis that direct energy measurements adequately represent the energy
consumption of the devices used in this study, the same measurements could be used to
determine the total energy of the comminution circuit. Since the same device is used to
measure the energy, it is assumed that the energy savings measured from one circuit to
another is as a result of improved energy efficiency. Using the Bond method is sometimes
prone to error and should not be used as will be explained later in Chapter 5.
Enhanced preferential liberation and micro-cracking

The possibility of enhanced selective and/or preferential mineral liberation is another area
of research covered in this study. Precisely how enhanced preferential liberation in HPGR
comminution circuits can be measured is to consider the following two methods. Both
methods are used, although the second method proved to be inconclusive. This is
explained later in Chapter 5.
1. The first method is to measure the comminution energy and mineral liberation in
both conventional ball and rod milling processes and compare these to the
measurements made in hybrid comminution circuits that employ HPGR. The
comparisons are based on the assumption that the same product grind size or size
distribution is achieved. The tests aim to compare the energy utilisation of the
circuit under investigation, and determine if there are any noticeable differences in
the manner in which the valuable mineral component has been liberated through
the use of mineral liberation analysis (MLA). The MLA is able to identify each
type of mineral present in the samples that were tested. The number or degree of
completely liberated or partially liberated minerals is determined using the MLA
data and software package. Comparative mineralogical analysis is difficult because
each sub-size fraction within a samples full size distribution is required to be mass
balanced in respect to valuable minerals reporting to different sub-size fractions.
For this study, since the experiments were designed to produce the same sized
- 59 -

particle distributions so that direct comparisons could be made without the need to
mass balance liberation.
2. The second method is to examine the products of conventional comminution

circuits and to compare them with the products produced with circuits that employ
HPGR only. Here the total energy input to the ore may be measured, and the
amount of liberated mineral determined relative to the amount of energy used in
the HPGR circuit. This could be achieved by
• Completing a series of tests designed to compare the mineral liberation and

total specific comminution energy between HPGR circuits and
conventional rod and ball mill circuits when identical products are
produced?
• Completing a series of tests designed to quantify the amount of mineral

liberated between HPGR only circuits and conventional rod and ball mill
circuits at significantly reduced total energy consumptions with possible
non identical products?
Summary
Precision energy measurements, HPGR circuit comparisons, micro-cracking and mineral
liberation and the impact for potential energy reductions are thus separately studied.
Experimental samples
Three completely different mineralized ores were identified for this study. For each ore
tested, two to three drums of sample containing about 1000 kg of -12.5 mm was received.
The details of the samples are shown in Figure 36.
P9N Comminution Project

Proposed sponsor ores
Xstrata – Mt Isa Lonmin – Eastern Platinum BHP Billiton – Worsley Mine

Lead/ Zinc Ore Platinum/Chrome Ore Aluminum Ore
Mine Site A Mine Site B Mine Site C

Ore sample requirements Ore sample requirements Ore sample requirements
± 1000 kg homogenous ore ± 1000 kg homogenous ore ± 1000 kg homogenous ore
(-12.5 mm) (-12.5 mm) (-12.5 mm)
Figure 36 depicts the origin and quantity of sample required of each of the three ores tested in this study.
- 60 -

Identified samples
Sample No. 1
Origin of ore sample Xstrata, North Queensland
Ore Type Lead/zinc ore
Process stream Feed to the Primary Rod mills
Sample date: November 2004
Sample Mass : Approx 1000 kg (three 220 liter drums)
Sample No. 2
Origin of ore sample BHP Billiton, Worsley Alumina
Ore Type Bauxite ore
Process stream Feed to rod mills.
Sample date: November 2004
Sample Mass : Approx 1000 kg (three 220 liter drums)
Sample No. 3
Origin of ore sample Lonmin, South Africa
Ore Type Platinum/chrome
Process stream Rod mill feed.
Sample date: December 2004
Sample Mass : Approx 1000 kg (two 220 liter drums)
Material from each drum was combined and the entire sample of each of the ores received
was first screened using the JKMRC Gilson screens to determine the as received size
distribution. The + 12.5 mm fraction was crushed in a small jaw crusher to ensure 100%
passing -12.5 mm. This was necessary as the specified top size required for the laboratory
scale HPGR tests is limited to 12.5 mm. The whole sample was then homogenised by
splitting each of the Gilson screened size fractions using a 16-segment rotary splitter
(Figure 37).
6.7 mm
4.75 mm
3.35 mm
2.36 mm
1.7 mm
Figure 37 shows the large rotary splitter on the left, and Gilson screens on the right used to generate
representative homogeneous test feed samples and to de-agglomerate HPGR products.
- 61 -

The now carefully pre-prepared -12.5 mm feed ores were subjected to the various HPGR
and milling processes for the determination of specific comminution energy and liberation
of valuable minerals. It was considered that the feed represented a homogeneous mixture.
The different comminution mechanisms effected in the HPGR and rod and ball mill were
from this point evaluated with respect to mineral liberation and product size distribution
as a measured quantity of energy was being applied to the system/circuit.
Feed size distribution.

The ore sample received from the sponsors (approximately 1 tonne each) was assumed to
have been “shoveled” into the three 220 liter drums. The content of each individual drum
was not considered to be representative of the sample received as size segregation could
have occurred during the loading of the drums. As such, the drum contents were
combined, dried, and prepared using a rotary divider and Gilson laboratory screens as
shown in Figure 37. The representative samples were then crushed according to the test
schedule for the HPGR, Bond rod mill and Bond ball mill test work that followed. These
procedures are detailed in the next few paragraphs. The result was that all of the -12.5
mm and -3.35 mm feed samples were representative of the original consolidated
homogenised feed sample.
Flowsheets- Circuit scenarios

The scope of the study was limited to evaluating six new circuit scenarios. The circuit
scenarios are described later in the section which is being compared to conventional Bond
rod and Bond ball milling. The Bond rod milling comparison is used so that material in
the -12.5 mm to +3.35 mm size range may be compared when the same material is treated
in a circuit employing an HPGR. The same logic applies to the Bond ball milling
comparison where -3.35 mm material is being comminuted down to a minimum size of
100 % passing - 75 Om.
These particle-size-to-equipment-size matches are clarified in Figure 38 which related the

HPGR coarse feed (particles larger in size to that of the working gap) to the Bond rod test,
and HPGR fine feed (particles smaller in size to that of the working gap) to the Bond ball
mill test. The reason for doing this was to be able to make a comparison between the
direct energy measurement for comminution and the empirically based Bond method that
makes use of the respective Bond work indices and third theory equation. In these two
cases the empirically based comminution energy uses the Bond Rod mill for the -12.5 mm
feed and the Bond ball mill for the -3.35 mm feed.
As no current method of determining the work index of an HPGR exists, the comparative
analysis had to rely on the direct energy measurements of a digital energy meter. The
digital energy meter is used on the assumption that the net comminution energy could be
measured. This is shown to be possible in Chapter 4. In order to produce similar products
for energy and liberation comparisons it was necessary to introduce hybrid HPGR/rod mill
circuits to be measured against the rod mill, and hybrid HPGR/ball mill circuits to be
measured against the ball mill as suggested in Figure 38 and Figure 39.
- 62 -

Bond Rod
Bond Ball
HPRC
HPGR
Figure 38 shows how comminution devices may be selected relating to size related classes (courtesy of Metso
handbook, Sandgren et al, 2002).
By using the same electrical energy measuring device, it was possible to create various
flowsheet circuit scenarios that used HPGR as a process to get to a desired grind size
(Figure 39). Feed size distribution and grade variability’s were eliminated from the
circuits by homogenizing and splitting the sample in the feed preparation stage as
previously mentioned. The recovery phase was analysed on the basis that the same
product size distributions were obtained and that sub size fractions of the products would
be analysed using the MLA. In addition to this the circuit scenarios provided an
opportunity to see the effect of the mineral liberation using HPGR technology without the
use of conventional milling technologies.
Rod mill – HPRC comparison Ball mill – HPGR comparison
Rod mill baseline HPRC / Rod mill Ball mill baseline HPGR / Ball mill
Circuit (ROD) Circuits 1-3 Circuit Ball Circuits 4-6

Crushing Crushing Crushing Crushing
Rod Milling HPRC Ball Milling HPGR
Rod Milling Ball Milling
Recovery Recovery Recovery Recovery
Figure 39 shows the studies conceptual HPGR flowsheet comparison structure.
Because the Bond equation is used to determine the comminution energy consumption
energy of the rod mill and the ball mill in conjunction with the direct energy
measurement, it became apparent that the measurement techniques be critically evaluated.
- 63 -

When the HPGR was used, the directly measured specific comminution energy was used
in the analysis as empirical techniques’ such as Bond’s do not exist. The energy
measurements are added to the rod or ball mill energy component of the circuit, to
determine the circuit energy. The direct energy measurement results were used to quantify
the energy of the Bond rod and Bond ball mill devices as compared with the Bond
methodology that uses the work index and third theory equation. The circuit comparative
tests use both methods and are presented in more detail in Chapter 5. Energy efficiency
comparison with respect to SAG mills was not considered in this study due to difficulty in
measuring laboratory equivalent SAG mill performance indicators. As such this study is
restricted to laboratory scale devices and their associated energy efficiency at the size
range in which the tests were carried out.
The particle size ranges were chosen to narrow the scope of the research. As such the
practice of grinding, where it was known that most of the energy is consumed is being
studied (Figure 40 and 41). Since issues relating to scale up of test results using a
laboratory scale HPGR unit have already been confirmed, the HPGR test results could
possibly be used to evaluate the performance of the ores in “Scaled-up” scenarios, where
the feed to an HPGR resembles the product from a primary gyratory crusher producing
feed top sizes of 80- 100mm. Difficult-to-treat ores that require ultra fine grinding of less
than 50-60 microns were not the topic of this study as the ores would require ultra-fine
grinding devices such as tower mills, verti-mills and the horizontal Isa mill.
Study limiting particle size range

The scope of this research was strictly limited to particles less than -12.5 mm (within the
Bond Rod size) and greater than 100 % passing -75 Om (within the Bond ball mill product
size). Grinding to below these sizes in the Bond ball mill is considered to be outside of the
Bond test closing screen size operating range. Another reason for limiting the particle top
size to -12.5 mm is that it is the size that the HPGR manufacturers recommend for use in
the Labwal lab scale HPGR unit.
From an energy saving point of view, according to Hukki (1962,1965) the energy input
required to produce -12.5 mm particles is relatively small (less than 1 kWh/t) by
comparison to the energy requirements needed to produce products less than 75 Om, up to
50 kWh/t. This is illustrated in Figure 40, which is assumed to be the size energy
relationship for tumbling mill devices on softer ores that were processed during Hukki’s
time. Size energy relationships that exist for the HPGR have not yet been measured nor
reported in the literature so far, but could be represented as the dotted line as the dotted
line in Figure 41.
- 64 -

75 12.5 150
1m mm mm
Tower &
Verti mills
Rod Mill
& SAG Blasting
Ball Mill Mill &
Vs. Vs. Primary
HPGR HPRC Crushing
25 kWh/t
1 kWh/t
Figure 40 shows a graphical representation of the well known size energy relationship for tumbling mill
devices after (Hukki 1962).
Tumbling
mill devices
25 kWh/t
10 kWh/t
Hybrid HPGR
1 kWh/t comminution
circuits
Figure 41 shows the path of energy size relationship that could be realised should HPGR be incorporated into
a hybrid comminution circuit.
- 65 -

The six flowsheet scenarios that were investigated treated feed size distributions of -12.5
mm and -3.35 mm (Figure 42). This was chosen because the standard Bond Rod and Ball
mill test methodologies use these as standard feed sizes for the work index locked cycle
tests.
Sponsor ore
Ore sample requirements
± 1000 kg homogenous ore
(-12.5 mm)
Rod Mill Size Fraction Ball Mill Size Fraction

-12.5 mm down to 3.35 mm -3.35 mm down to 75 Bm
Circuit Scenario 1 Circuit Scenario 4
HPRC & Rod mill HPGR & Ball mill

fed with -12.5 mm fed with -3.35 mm
material material
Figure 42 shows the high level structure of the thesis study experimental plan.
Figure 43 and Figure 46 show circuit scenarios 1 (rod mill) and 4 (ball mill), as a direct
comparison between the HPGR in closed circuit. These circuits are difficult to compare as
the products are not always comparable. This was unforeseen as it had been reported by
the manufacturers that HPGR circuits with recycle streams could be evaluated on the basis
of conducting a “Bond equivalent” locked cycle laboratory test. As it turned out running a
laboratory scale HPGR in closed circuit with a classifier in a locked cycle test procedure
is difficult and an equivalent “locked cycle test” for the HPGR does not formally exist.
Tests were conducted to see if it was possible to compare the circuits with direct energy
measurement and to see if similar products could be produced. This would have
eliminated the need to evaluate the grinding technologies using hybrid circuits. Direct
comparisons were not possible since products were completely different. These circuit
scenario’s results are briefly presented in Chapter 5.
Figure 44 and Figure 47 show how conventional grinding circuits and HPGR hybrid
grinding circuit (circuit scenario 2 and 5 ) were compared This provided the bulk of the
research analysis and data. Figure 47,(circuit scenario 5) represents the flowsheet that
produced most of the data and analysis. Most of the circuit comparisons in Chapter 5 and
6 are analysed using data from this circuit. Out of the six circuit scenarios, two circuits,
circuit 3 (Figure 45) and circuit 6 (Figure 48) stem from circuit scenario 2 (Figure 44) and
circuit scenario 5 (Figure 47) respectively. Circuit scenario 3 and circuit scenario 6 were
used to provide data to examine the possibility of being able to use HPGR to
preferentially liberate as much of the valuable mineral component without ball milling.
This was done by taking a sample from the HPGR after the third pass process in circuit
scenarios 2 and 5 and examining the products. The results showed that the particles were
still too coarse to enable sufficiently liberated minerals.
- 66 -

Circuit A
Feed
(t/hr)
Product
Feed (t/hr)
(t/hr)
Classifying
Rod Mill cyclone
Product
HPRC
(t/hr)
Classifying
cyclone OR
Re-circulating load
Screen
Re-circulating load
Figure 43 shows how circuit scenarios 1 was designed to compare a rod mill and a circuit comprising of a
HPGR in closed circuit.
Feed
(t/hr)
Circuit A
HPRC
Product
Feed (t/hr)
(t/hr)
Classifying
Rod Mill cyclone HPGR
Product
Re-circulating load (t/hr)
HPGR
Classifying
Rod Mill
cyclone
Re-circulating load
Figure 44 shows how circuit scenarios 2 is designed to compare a rod mill and a hybrid HPGR/rod mill
circuit comprising of three HPGR’s in series in open circuit followed by a rod mill in closed circuit. The
circuit is designed to produce identical products at different grind sizes.
- 67 -

Baseline Feed New Circuit
(t/hr)
Circuit A
HPRC
Product
Feed (t/hr)
(t/hr)
Classifying
Rod Mill cyclone
HPGR
New Paradigm
Re-circulating load
HPGR
Energy Efficient
Liberation
Product
(t/hr)
Figure 45 shows how circuit scenarios 3 is designed to compare a rod mill and a circuit comprising of a
three HPGR’s in series in open circuit. The circuits is not expected to produce identical products, but the
rather a measure of the extent of liberation at this stage of the process.

Circuit B
Feed
(t/hr)
Product
Feed (t/hr)
(t/hr)
Ball Mill
Classifying
cyclone
Product
HPGR
(t/hr)
Classifying
Re-circulating load cyclone OR
Screen
Re-circulating load
Figure 46 shows how circuit scenarios 4 was designed to compare a ball mill and a circuit comprising of a
HPGR’s in closed circuit.
- 68 -

Feed
(t/hr)
Circuit B
HPGR
Product
Feed (t/hr)
(t/hr)
Ball Mill
Classifying
cyclone HPGR
Product
HPGR
Ball Mill
Classifying
cyclone
Re-circulating load
Figure 47 shows how circuit scenarios 5 is designed to compare a ball mill and a hybrid HPGR/ball mill
circuit comprising of a three HPGR’s in series in open circuit followed by a ball mill in closed circuit. The
circuit is designed to produce identical products at different grind sizes .

Circuit B Feed
(t/hr)
Product
Feed (t/hr) HPGR
(t/hr)
Ball Mill
Classifying
cyclone
HPGR
Re-circulating load
New Paradigm
HPGR
Energy Efficient
Liberation
Product
(t/hr)
Figure 48 shows circuit scenarios 6 that is designed to compare a ball mil circuit to an HPGR circuit
comprising of three HPGR’s in series.
- 69 -

Sample requirements and test design for circuit scenario evaluation
As previously mentioned the scope of the study was limited to evaluating three different
ore types. The reasons for the choice of these ores was briefly described in the Literature
review based on knowledge of knowing where most of the comminution energy is being
expended in the minerals industry. This summary provided good reason why HPGR have
had successes in the cement and iron ore industry, with the technology developing rapidly
in the copper and gold industry. The chosen ores thus represent new areas where the
potential for HPGR exists but has not been fully explored yet.
The proposed sample ores were bulk samples taken from a feed stream to either industrial
rod or ball mills as previously mentioned. Once the initial homogenization preparation
stage was completed, the samples were then prepared for the test work as conceptually
outlined in Figure 42 to Figure 48. This entailed further splitting, and crushing in the case
of the -3.35 mm feed requirement of each sample. The -12.5 mm and -3.35 mm sub-
fractions were rotary divided to produce 16 equal homogenised batch samples
(approximately 30-33 kg each) required for the test work that followed (Figure 49).
The batches were labeled 1/16 through to 16/16. Sub-samples 1-3 were used for HPGR
energy calibration and machine setting tests, 4-6 for single pass HPGR in closed circuit,
7-9 for double pass HPGR test, 10-12 for triple pass HPGR tests and 13-15 for
conventional Bond rod and ball milling tests. Sub-sample 16/16 was kept aside for
separate piston and die ore characterization tests which are not part of the scope of this
study. The 16 identical sub-samples of each ore type and primary feed size are indicated
in Figure 50 and Figure 51. The sample breakdown and tests provided the raw data
required that would re-produce or simulate the proposed circuit scenarios. This included
HPGR duplicate tests and initial HPGR calibration tests.
Mine Site
(-12.5 mm)
±500 kg for ±500 kg for

Rod mill series Ball mill series
Rotary split
1 2 3
4 5 6
Rotary split
7 8 9
Calibration
1 2 3
10 11 12/
1st pass
4 5 6
2nd pass 13 14/ 15 16 7 8 9
Calibration
3rd pass
10 11 12/
P&D 1st pass
Circuit ROD mill
2nd pass
(Rod-feed) Baseline 13 14/ 15 16
3rd pass
Circuit P&D
(Ball-feed) BALL mill
Baseline
Figure 49 shows the breakdown of the three proposed ores and their origin.
- 70 -

Conventional Rod Milling vs HPRC (-12.5 mm)
Rotary split
1/16 2/16 3/16
4/16 5/16 6/16
HPRC energy 7/16 8/16 9/16

calibration HPRC HPRC
1st pass Closed Circuit
Open Locked Cycle 10/16 11/16 12/16
Circuit
HPGR 2ndt pass
Open Circuit
HPRC tests
HPGR 3rdt pass
13/16 14/16 15/16 16/16
Open Circuit
4/16 5/16 6/16 7/16 8/16 9/16 10/16 11/16 12/16 Bond Rod Mill WI P&D
Closed Circuit Masters
Baseline
@ 3.35, 1.18 mm X 2
Bond Rod Mill WI

Closing screen
@1.18 mm X2
Conventional Rod
Milling tests
Figure 50 shows the breakdown of the conventional rod milling versus HPGR test series.

Conventional Ball Milling vs HPGR (-3.35 mm)
Rotary split
1/16 2/16 3/16
4/16 5/16 6/16
HPGR energy 7/16 8/16 9/16

calibration HPGR HPGR
1st pass Closed Circuit
Open Locked Cycle 10/16 11/16 12/16
Circuit
HPGR 2ndt pass
Open Circuit
HPGR tests HPGR 3rdt pass
13/16 14/16 15/16
Open Circuit
16/16
4/16 5/16 6/16 7/16 8/16 9/16 10/16 11/16 12/16

Bond Ball WI Closing screen P&D
@ 300, 150, 106 1m X 2 Masters
Open MLA
HPGR Circuit Bond Ball Mill WI
Locked cycle MLA Closing screen @
@ 850 1m 300, 150, 106 1m X2 Conventional Ball
Closed Circuit + MLA Milling tests
Figure 51 shows the detailed breakdown of the conventional ball milling versus HPGR test series.
- 71 -

Brief description of Laboratory comminution tests
Laboratory HPGR tests
The laboratory HPGR tests were completed in triplicate as a minimum requirement for
error analysis and test repeatability. The tests series was designed so that at least 6
comminution energy measurements could be made for the first pass and second pass tests
results. The triple pass test results relied on triplicate tests. Bond test error measurement is
mentioned in the next section. The most important part of the HPGR tests were to focus
on getting reliable and accurate specific comminution energy measurements that would be
later totaled to the Bond grinding milling test energy to complete the energy comparisons
for the two circuits. The products of the triple pass HPGR tests were kept aside for
mineralogical analysis (Chapter 6), and were used to provide the feed material to the ball
mill. The locked cycle tests continued in the same way as the baseline conventional ball
milling tests. The comparative results are presented later in Chapter 5.
Bond rod mill

Standard Bond rod mill tests were carried out on both circuits at two closing screen sizes
of 1.18 mm and 3.35 mm. Although the standard Bond rod test is carried out with a
closing screen of 1.18 mm, a closing screen of 3.35 mm was also used so that the energy
to produce -3.35 mm product could be measured. The -3.35 mm product would resemble
the feed to the Bond ball mill test. The Bond rod mill tests were not that successful as will
be explained in Chapter 5. Bond’s methodology (empirical equations) and error analysis
in respect to the rod mill test and ball mill test are expanded upon later in this chapter.
Bond ball mill

Standard Bond ball mill tests were carried out on both circuits at three closing screen sizes
of 300 Om,150 Om and 106 Om for the lead/zinc ore, 425 Om, 300 Om, and 150 Om for
the bauxite ore, and 150 Om, 106 Om, and 75 Om for the platinum/chrome ore. This was
necessary to determine how milling energy varies with the target grind size or closing
screen setting. The total net comminution energy to produce these sized products was
determined by adding the net HPGR grinding comminution energy to the ball milling
comminution energy, and to compare this with the baseline ball milling circuit. More
details underpinning this methodology are described later.
Characteristics of the Bond “locked cycle” test and third theory

equation.
The research study uses, as previously mentioned, a new direct energy measurement
methodology that enables the most effective way to evaluate the energy efficiency of the
two types of comminution circuits/devices. This is seen as the preferred approach as the
exact same energy measuring device is used to measure energy consumption and to
compare the circuit. This makes up the content of Chapter 5. In addition to this by simply
completing the test work using the Bond ball mill, the research has the added advantage of
being able to critically evaluate the new energy measurement methodology against the
- 72 -

traditional method used by Bond. This makes up much of the content of Chapter 4 and
expands on the current understanding of the Bond methodology and its limitations.
The new direct energy measurements have recently been made possible due to
advancements in measurements of electrical energy and were applied to both the Bond
ball and bond rod mill tests as shown in Figure 52.
Energy meter
measurements
Bond rod Bond ball
Figure 52 shows how a digital energy meter is hooked up with the rod and ball mill tests.
In order to fully support the new digital energy meter measurements, the measurements
have been compared with those that are determined using the traditional Bond method
Bond’s methodology has several known limitations as described by Herbst and
Fuerstenau (1980) and Austin and Brame (1983), but in spite of this, the Bond locked
cycle test remains very useful and has been a vital component in conducting this research
because it is a test which represents the ball mill in closed circuit with a cyclone or screen
classifier.
The well-known third theory equation of Bond as previously mentioned is still frequently
used by process engineers for the design of ball mills. Other common uses of the Bond
work index are to make comparisons of various ore types. These data are then used to
predict the performance of grinding mills in relation to the power requirement and
throughput.
As a result of the long-standing success of the Bond equation, it is generally easily

understood and accepted. The response of a new ore type is often compared to others that
have already been tested and that now belong to large databases of ball mill work indices
which have accumulated over the years. For this reason the test remained as a standard for
years even though many critics (Armstrong, 1986; Austin and Brame, 1983; Charles,
1957; Free et al., 2005; Fuerstenau et al., 1984; Hosten and Avsar, 2004; Kapur and
Fuerstenau, 1987; Kaya et al., 2003; Laplante et al., 1988; Laplante et al., 1993;
Magdalinovic, 1989; Morrell, 2004; Mosher and Tague, 2001; Nitta et al., 2002;
Stamboliadis, 2004; Tartaron, 1964; Tuzun, 2001; Vedat and Huseyin, 2003) have
- 73 -

claimed to have developed easier, quicker or more accurate versions of the Bond locked
cycle test.
Other advantages, in particular those relating to the Bond rod and ball mill tests, is that the
feed particle sizes in these tests often closely resemble the particles sizes in the full-scale
comminution devices, unlike for example large jaw crushers and SAG mills where the
particle sizes are much smaller in relation to the full-scale environment. It is important to
mention that the Bond equation uses the single value parameters of F80 and P80 to describe
the full size distributions. Bond states in the definition of the test, that the size
distributions when plotted on a log-log scale should represent parallel lines. Under such
circumstances representation of the size distribution as a single value is valid. When
conditions are such that non-parallel lines are observed, then the validity of the Bond
methodology may be questionable.
The ball mill employed in a standard Bond ball mill test is a mill of 305 mm in length and
305 mm in diameter, with a smooth internal surface as shown in Figure 53. The charge
consists of the standard Bond charge of 20,125 g in mass and of similar ball size
distribution. The ball top size is 36.4 mm. The balls occupy approximately 30% of the
volume of the mill. Energy input is measured by the newly acquired digital energy meter
which is described in more detail in the next chapter.
“Bico” Bond mill
Bond ball charge (20125 g)
Figure 53 shows the ball charge in the standard “bico” Bond mill at the JKMRC.
Bond confirmed that the third theory was based on new crack tip length, and showed
various ways in which this could be measured. The Bond locked cycle tests and its
associated empirical equation for the determination of the work index are used in
conjunction with the third theory equation to determine the specific comminution energy.
When these empirical equations are compared to the data generated from experimental
tests, the agreement is good and verify the strong relationship between the experimentally
- 74 -

measured grindability (Grp) and the Bond work index (Wi). Figure 55 shows how well the
grindability parameter (grams per revolution) correlates with the Bond work index for
most ores (based on the JKMRC database) tested at a closing screen of 100 Om. This
relationship is mathematically expressed when the empirically based equations of the
Bond locked cycle tests index equation (Equation 15) is combined with the third theory
equation (Equation 16) to produce equation 17. When the closing size screen is chosen at
100 Om or 150 Om equation 17 is reduced to equation 18 and equation 19 respectively.
49
Wi = (15)
P1
0.23
(G )
bp
0.82
10
1 1
P F
Bond’s laboratory ball mill Work Index equation
1 1
W = 10 Wi (16)
P F
Bond’s third theory energy equation is given again as Equation 16. Bonds (1961)
definition of the work index equation for the ball mills is given by equation 15 and is
numerically defined as the kWh/t required to reduce the material from theoretically
infinite feed size to 80 % passing 100 Om.
Hence the original Bond work index test should be conducted at a closing screen of 150
Om to conform with this definition. The reason is when a closing screen size of 150 Om is
used, the product resembles a P80 of approximately 100 Om. Test results extracted from
the JKMRC Bond ball mill database, showed that for 27 tests where the closing screen
was 150 Om, tests produced P80’s that ranged from 91-120 Om. When the Bond work
index defining screen of 150 Om is used, equation 17 is simplified into equation 18 which
represents the work required (kWh/t) to grind the ore in the bond ball mill as a function of
the test parameter “grindability”. The Bond test parameter “grindability” is defined as the
net grams of closing screen size undersize produced per mill revolution. The
“grindability” parameter does not consider the fraction of fine material smaller than the
closing screen in the feed as material that undergoes grinding(see Bond equilibrium
contradiction in chapter 2). Samples with a high percentage fines in the feed can therefore
be subject to an error. This is explained further by way of examples of the circuit
comparison data analysis of Chapter 5.
49
W= (17)
P
1
0.23
(G )
bp
0.82
16.99
W= (18)
(G )bp
0.82
15.47
W= (19)
(G )bp
0.82
- 75 -

When a closing screens of 106 Om and 150 Om are used, the product size (P80 ) is
approximately 70 Om and 100 Om respectively. Equations 18 and 19 represent Bond’s
theoretical energy requirement for a 2.4 m wet overflow mill as a function of the
measured grindability term measured in the dry Bond ball mill test procedure. When the
assumed “Bond test defined” closing screen of 150 Om is used, the experimental results as
determined by using at least 30 Bond ball mill test data, show that the experimentally
determined work indices do not correlate well with those determined theoretically as
shown in Figure 54.
However, when a closing screen of 100 Om is used, and the theoretical values are
compared with the experimental values for many Bond ball mill work index tests. The
experimental Bond tests using a closing screen of 106 Om correlated well as shown in
Figure 55. This suggests that the work index tests and definition of the work index could
have possibly been based on the energy required to reduce the material to a closing screen
of 106 Om or a P100 of 106 Om and not a closing screen of 150 Om.
45
Bond energy (kWh/t) (ie work required to grind test
material according to Bond's empirical equation)
Equation 26 representing a closing screen of 150 um (Theoretical)

40 Experimental Bond ball tests @ c.s. 150 um (Jk database)
Power (Equation 26 representing a closing screen of 150 um (Theoretical))
Power (Experimental Bond ball tests @ c.s. 150 um (Jk database))
35
30
25
20
15
10
y = 20.2x-0.75
5
y = 15.48x-0.82
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Bond test "Grindability" net grams per mill revolution (Gbp)
Figure 54 shows the Bond work index in relation to the Bond grindability at a closing screen size of 150 Lm.
When a different closing screens are used (106 Om), the outcome of the Bond work index
differs as is shown in Figure 55, where again data from the JK Bond ball mill test
database was used.
- 76 -

30
Bond energy (kWh/t) (ie work required to grind test
Equation 26 representing a closing screen of 100 um (Theoretical)

material according to Bond's empirical equation)
Experimental Bond ball tests @ c.s. 106 um (Jk database)

25
Power (Experimental Bond ball tests @ c.s. 106 um (Jk database))
Power (Equation 26 representing a closing screen of 100 um (Theoretical))

20
15
10
y = 18.0x-0.77
y = 16.99x-0.82
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Bond test "Grindability" net grams per mill revolution (Gbp)
Figure 55 shows the Bond work index in relation to the Bond grindability at a closing screen size of 106 Lm.
Figure 56 and Figure 57 shows similar empirical Bond work index results for several
hundred Bond ball mill test results at various closing screen sizes that have accumulated
at the JKMRC over the past 6 years. The grindability parameter here again is Gbp as
defined as the net grams per bond ball mill revolution.
This analysis of the Bond work index test data and the third theory equation suggest that
in general as the grind size is reduced, comminution energy increases exponentially and
that the comminution energy is bound with in the limits of the variation of these data.
These limits may well be true for the representation of comminution energy in ball mills,
as the relationship was formulated by using a ball mill.
The comminution energy however, must include the energy losses of the milling device
(as suggested by W.F. Carey, in the literature review) that may not otherwise be as great
when other possibly more energy efficient comminution devices or breakage mechanisms
are used. Hence direct measurement of the energy in a rod and ball mills and the HPGR
are a way of actually being able to measure and quantify the relative energy efficiency of
using one device over another. The analysis of this data forms the basis of Chapters 4 & 5.
- 77 -

35
closing screen at 75 microns
Bond's empirical work index (BWI) test result

(Kwh/t)(at different test closing screens)
30
25 closing screen at 300 microns
Bond @ 106 closing screen
Power (closing screen at 300 microns)
20 Power (closing screen at 212 microns)
15
-0.74
y = 26.7 x
10
-0.82
5 -0.75
y = 23.5 x
y = 20.0 x
y = 17.9 x-0.77
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Bond test grindability Gbp (net grams per revolution)
Figure 56 shows the same data as Figure 55 plotted using normal axes.
Figure 57 shows the third theory specific comminution energy in kWh/t against the Bond
work index for all tests, highlighting again that perhaps the 106 Om closing screen should
be used as a standard since the 106 Om closing screen option in the Bond test data
analysis correlates well with the y = x line when Bonds third theory equation is applied to
the Bond mill test grinding. When closing screens of 150 Om, 212 Om and 300 Om are
used the data clearly deviates from the y=x reference. Another interesting fact is that as
the closing screen increases in size, the proportion of fine material in the feed less than the
closing screen increases, and the same mill effectively does the same work on less
material because the grindability parameter is defined as the net grams of fine material
produced per mill revolution. At steady state, with a 250% recirculating load, the net
energy of the mill per revolution should be applied to the fresh feed, which would be a
more accurate representation of the energy used in grinding the material. This is explained
by way of examples in Chapter 5 to confirm the best way of measuring specific
comminution energy directly as opposed to relying on a 50 year old empirical equation
that may not be correct in the first place. Bond’s methodology may well have worked
during the time of its development, but new devices and measuring capabilities have
suggested that there are alternative and possibly much more accurate and representative
methods that can be developed.
Conducting Bond index tests is very time consuming, and there have been many instances
where error analysis and repeatability have been published. In this research project many
Bond tests were conducted, but these were not repeated in order to obtain a measurement
of the error, but rather the error is based on exiting knowledge of the type of errors that
are expected when conducting Bond work index tests.
- 78 -

40.0
bond @ 75 microns
bond @ 106 microns
35.0
y=x
bond @ 150 microns
Bond third theory energy (kWh/t)
30.0 bond @ 212 microns

bond @ 300 microns
Linear (bond @ 106 microns) y = 0.9299x
25.0 2
Linear (bond @ 150 microns) R = 0.8941
Linear (bond @ 212 microns)
20.0 Linear (bond @ 300 microns)

Linear (y=x)
15.0
10.0
5.0
0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0
Bond work index (BWI) (kWh/t)

Figure 57 shows the relationship between the Bond work index and the Bond third theory equation energy at
various closing screen settings.
Error analysis of the Bond rod and ball mill tests

The circuits described in scenarios 1-6 rely on the results of many Bond rod and Bond ball
mill tests. Bond tests require a lot of effort to produce a single result. Multiple or at least
duplicate tests would normally be required to establish the error of a measured quantity in
any thesis. However, in this case because the Bond tests is widely used and accepted in
the mining industry and that many published journal papers (Armstrong, 1986; Laplante et
al., 1988; Laplante et al., 1993; Tuzun, 2001; Vedat and Huseyin, 2003) have previously
reported the magnitude of the expected error, for this study single Bond tests were
conducted assuming that the error would be similar to that which has been reported in the
literature.
Bond Rod mill calibration tests are routinely conducted at the JKMRC using a
standardized calibration sample from Mt Marrow Blue Metal Quarries. These samples are
sent to several ore characterisation laboratories around the world to establish the error that
could be expected when using different units, different initial sample feed mass, different
operators, Ro-taps and sieves. The results of one of these round robin tests are tabled in
Table 7 for Bond rod mill tests.
- 79 -

Table 7 shows the error and variation of standard Bond rod mill locked cycle tests.
Weight of % undersize Work Index

Data Set No. feed (g) F80 (Bm) # P80 (Bm) # in feed g/rev (kWh/t)
1 2,197.2 11,045 893 10.2 6.70 17.1
2 2,179.6 10,671 912 10.5 6.87 17.2
3 No data
4 2,272.4 11,065 818 9.9 4.39 20.9
5 2,283.8 10,693 771 10.1 3.93 21.7
6 2,273.0 10,234 882 9.4 6.44 17.6
7 2,070.0 10,361 743 7.5 2.50 28.5
8 2,200.0 9,577 846 9.2 6.94 16.5
For All Data Sets except Set 7:
Average 2,234.3 10,548 854 9.9 5.88 18.5
Std Dev 47 564 53 0.51 1.3 2.2
Max 2,283.8 11,065 912 10.5 6.94 21.7
Min 2,179.6 9,577 771 9.2 3.93 16.5
Excluding Data Sets 4, 5 and 7
Average 2,212.5 10,382 883 9.8 6.74 17.1
Std Dev 41 631 28 0.64 0.22 0.44
Max 2,273.0 11,045 912 10.5 6.94 17.6
Min 2,179.6 9,577 846 9.2 6.44 16.5
• Closing screen size (Data Sets 1 to 6 & 8) was 1.18 mm

• Closing screen size (Data Set 7) was 1.14 mm
• # F80 and P80 were determined by interpolation in Rosin-Rammer space
• Data Sets 4 & 5 used a Bond size rod mill with no liners; other Data Sets used
wave liners
• Data Set 7 used a 1.14 mm closing screen (instead of 1.18 mm).
The results in Table 7 show that outlier test results are real. However if the same device,
operator, and screens are used, then experimental error resulting from the experimenter
and equipment is eliminated. Depending on how many data sets and repeat tests are
conducted, these results indicate that the rod mill error is in the range of 2.57% to 11.89.
Much care was taken in designing the experiments, the feed mass and F80 were measured
in triplicate at the start of conducting the Bond tests, and the average was used throughout
in Bond’s empirical equation. This technique ensured that any possible error induced by
the possible variation in feed mass and feed size distribution measurement was
eliminated. Thus the experimental error of the Bond rod tests conducted in circuit
scenarios 1, 2 and 3 are expected to be closer to 2.57% rather than 11.89%.
The same round robin tests were carried out using the Mt Marrow Blue Metal Quarries
material for Bond ball mill tests. The results are shown in Table 8.
- 80 -

Table 8 shows the error and variation of standard Bond ball mill locked cycle tests.
Weight of % undersize Work Index

# #
Data Set No. feed (g) F80 (Bm) P80 (Bm) in feed g/rev (kWh/t)
1 1,423.7 2,124 73.0 14.7 1.22 15.0
2 1,369.5 2,143 73.0 14.9 1.19 15.3
3 No data
4 1,266.8 2,111 77.4 16.9 1.19 15.8
5 1,434.7 2,135 80.0 16.4 1.24 15.6
6 1,400.4 2,127 78.2 14.1 1.37 14.2
7 1,389.0 2,007 76.6 15.3 1.21 15.6
8 1,409.0 2,139 105.1 18.3 1.43 15.3
For All Data Sets:
Average 1,384.7 2,112 80.5 15.8 1.26 15.3
Std Dev 56 48 11 1.5 0.096 0.54
Max 1,434.7 2,143 105 18.3 1.43 15.8
Min 1,266.8 2,007 73.0 14.1 1.19 14.2
Data Sets 1, 2, 4, 5 and 7
Average 1,376.7 2,104 76.0 15.7 1.21 15.5
Std Dev 67 56 3.0 1.0 0.021 0.34
Max 1,434.7 2,143 80.0 16.9 1.24 15.8
Min 1,266.8 2,007 73.0 14.7 1.19 15.0
Data Sets 1-7
Average 1,380.7 2,108 76.4 15.4 1.23 15.3
Std Dev 61 51 2.8 1.1 0.068 0.59
Max 1,434.7 2,143 80.0 16.9 1.37 15.8
Min 1,266.8 2,007 73.0 14.1 1.19 14.2
• # F80 and P80 were determined by interpolation in Rosin-Rammer space

• Data Sets 1-7 used a closing screen size of 106 Om or 105 Om
• Data Set 8 used a closing screen size of 150 Om
Using the same argument as that for the Bond rod mill error determination, the Bond ball
mill error range was 2.19% to 3.35%. This is normally quite acceptable for these tests and
is within range of that which has been reported in the literature.
Errors in the measurement of the energy using the digital energy meter in the Bond ball
mill are covered in the analysis of the digital energy meter and are presented in Chapter 4.
This review of the expected experimental errors in conducting Bond rod and ball mill
tests are presented here so that single Bond test results conducted in the various circuit
scenarios described earlier can expect to have similar errors in the measurement. This
analysis thus eliminates the need of having to conduct the Bond type tests in triplicate for
the purpose of error analysis in this thesis.
- 81 -

Description of the Laboratory scale HPGR and test procedure
The HPGR tests were conducted using a recently refurbished 300 mm diameter by 70 mm
wide laboratory scale HPGR unit (Figure 58) and fitted with profiled rolls as shown in
Figure 59. The unit was fitted with a digital energy meter as shown in Figure 58 for the
direct measurement of the net comminution energy.
Feed Hopper
Digital energy meter
HPGR Hydraulics
Figure 58 shows the newly refurbished laboratory scale HPGR at the JKMRC installed with a new digital
energy meter.
Fixed
gap
Cheek
plates
Roll
stops
Profiled roll surface and

autogenous ore layer
Figure 59 shows the condition of the rolls and cheek plates of the JKMRC 300 mm by 70 mm HPGR.
For all HPGR tests, a constant 3.5 mm initial gap setting was applied which gave an
actual working gap which ranged from 3.5 to 10 mm. The bearing housing of one of the
rolls is fixed in the frame, while the other roll may move laterally against the adjacent roll
- 82 -

where a fixed initial gap opening may be set. When material is fed through the rolls, the
movable roll moves outwards and presses against the hydraulic ram system. The so-called
operating gap results when the packed bed of material formed between the rolls reaches
equilibrium as shown in Figure 60. A milling force of up to 150 kN is applied which
represents a bed pressure of between 50 to 300 MPa. Each roll is driven by an
independent 7.5 kW motor.
Fixed roll Movable roll
Cheek plates
Fixed Gap Stops
HPGR Flake
Movable roll
working off
the fixed
stops during
processing
Figure 60 shows the movable roll working off the stops during HPGR processing.
The circumferential roll speed was fixed at the slower speed setting of 0.33 m/s or 21 rpm.
Previous work by Daniel (2003) and Austin, et al. (1993) showed that the throughput is
proportional to rolls speed and in general does not alter the characteristics of the product
size distribution but, rather, the rate at which the unit may deliver energy to the particle
bed.
Three HPGR energy calibration tests were performed with a low energy target, a mid
range target and a high range target to determine the ideal machine settings needed for the
tests that followed. All the other HPGR tests were conducted using the same machine
setting. Details of the HPGR operating conditions for each test are given in the data CD at
the back of this thesis.
Many “energy” related research campaigns have traditionally measured energy using
torque sensors strapped onto the roll shaft of the mill/HPGR. In this research, an accurate
digital energy meter is used to measure the gross energy consumed by the unit in treating a
known quantity of material. The energy meter is used to measure the no-load energy. The
difference between the two measurements (the gross energy and the no-load energy)
provides a measurement of the net comminution energy. A detailed description and
analysis of the digital energy meter capability is provided in Chapter 4.
- 83 -

HPGR test measurements and calculated outputs
Conducting HPGR laboratory scale tests requires the measurement of several process and
ore parameters. The following seven sub-sections provide a short description of these
measurements and the importance of them to the study both as a check to the integrity of
the test results, and to provide the data necessary for this study.
HPGR Pre-test settings – Initial gap and the hydraulic pressure system
For the test series, the initial oil pressures for the laboratory-scale were set at the similar
levels that the ore might experience in a full scale HPGR. For the Lonmin and Mt Isa ores
test series, the specific energy required for the tests was expected to be in the range of 2-3
kWh/t. This energy input for the Bauxite ore ranged from 0.5-2 kWh/t similar the specific
energy that has been experienced in pilot scale operations.
To achieve the above, the tests were completed with an initial stop gap setting of 3.5 mm
and an initial nitrogen pressure of 11 kPa (low nitrogen pressures). The initial oil pressure
was 35 kPa so that resultant working pressures of between 35- 110 kPa could be achieved
at working gaps in the region of 3.5 mm to 8 mm.
The “spring” response of the HPGR at these initial gap and pressure settings are given in
the test data CD at the back of this thesis. Important and relevant HPGR test
measurements and calculated outputs are given in Table 9 and Table 10 respectively.
Table 9 List of the HPGR measured experimental test data.

Item Data category Test data (Experimental measurements)
1 Measured input Sample mass (kg)
Roll diameter (D),
Roll width (L),
Roll speed (U)
Bulk "compacted" density ( c),
Feed size distribution,
2 Measured output Working gap (xg),
Flake thickness (xgf ),
Flake density ( g)
Product size distribution (measured),
Batch process time (sec)
Working pressure (pw),
Energy (Wh)
- 84 -

Table 10 List of the critical HPGR process calculated outputs.
Item Data category Calculated output
3 Calculated output Measured throughput (Qm),
Calculated throughput (Qcalc)
Specific Energy (Ecs),
Specific Force (Fsp) – based on spring response
Specific Force (Fsp) – based on measured working pressure
Critical gap (xc),
Product size distribution (calculated)
Calculated HPGR test throughput as a check for measured throughput

The HPGR throughput model utilises a simple plug flow model which is represented by
equation 20. Throughput is calculated from various measurements made during the
process (rolls speed, rolls width, and working gap) and of the material in the working gap
(flake density). Measurement of the flake density in the gap at the time of processing is
impossible. Reliable measurements of the gap are difficult in the case of studded and
profiled roll surfaces.
Qcalc = 3600uL g xg (20)
Where,
u = rolls peripheral speed (m/s)
L = roll length (m)
g = flake density of the material within the working gap (t/m3)
xg = measured rolls working gap (m)
Qcalc = calculated throughput (t/h)
The performance of the calculated throughput should be unaltered when both the flake
density and thickness are used to determine the calculated throughput. For the studded and
profiled rolls, the average working gap was difficult to measure due to the uneven roll
surfaces and the corresponding autogenous layer built up between the studs or profile. A
better and more practical model to describe throughput is then given as equation 21.
Qcalc = 3600u L x gf f (21)
Where,
f = Flake density of the flakes produced (t/m3) – flakes measured in air and water.
xgf = Measured flake product thickness (m)
The calculated throughput is determined during all HPGR experiments and compared
with the measured throughput. This provides a simple method of confirming the validity
of the test and test measurements.
HPGR measured throughput

The mass of each batch sample and the time to process it were measured to the nearest
gram and second respectively. The process time was measured by counting the number of
- 85 -

one second intervals that were produced each time the data logger recorded a power,
pressure or gap measurement. The time to process the batch at steady state was
determined along with accurate mass measurements of the batch to ensure accurate
throughput rates. The measured throughput given by equation 22 for batch experiments
was used to compare the experimental throughput to that of the model’s calculated
throughput.
Test Sample mass (kg )

Qm = 3.6 (22)
Time to process(sec)
Where,
Qm = Measured throughput (t/hr)
Measurement of throughput in continuous operational industrial scale units are typically

calculated from total tonnes treated (daily/shift weightometer readings) over the particular
time period (shift/day) and given by equation 23.
Weightometer tonnes per time period (tons )

Qm = (23)
Time period (hours )
The laboratory scale HPGR unit is fitted with a hopper sluice gate which controls the test
start time and initial choke feed to the rolls as shown in Figure 61. The HPGR is also
fitted with an enclosed chute surrounding the edges of the rolls which eliminates the
possibility of material by-pass during processing. These are called “cheek” plates in
industrial applications.
Hopper sluice
gate to control
test start time
Initial “fixed”
stop gap
Profiled rolls
Roll edges
sealed to
limit material
by-pass
Feed hopper
Figure 61 shows the JKMRC HPGR feed hopper and “profiled” rolls surface.
Usually, the experimental working gap is determined using displacement sensors mounted
between the rolls. The instruments are first calibrated and produce a signal which is
converted to represent the effective gap measurement in millimeters. The effective
working gap in this research was measured using the measurements of the thickness of the
resultant flakes that were produced.
- 86 -

In order to determine the average working gap of the rolls during the period it took to
process each sample a simple procedure was adopted whereby a grab sample of about 6-8
flake pieces was obtained as shown in Figure 62. The minimum and maximum
thicknesses of these flake sections were measured using a vernier caliper. The
measurements were averaged to represent the mean working gap as shown in the data
sheets provided in the data files provided on the CD.
HPGR “Flake”
segments
Figure 62 shows the compacted HPGR “flake sections” used to determine the equivalent gap thickness.
Since actual flakes were measured to represent the working gap, and not the direct
measurement of the gap, the error caused by the expansion of the flake due to relaxation
after exiting the compression zone is not considered. The existence of flake expansion is
known to change the actual measured gap width to the measured flake thickness, and is
known to differ from between 10-25% (Lim and Campbell, 1996). The working gap
measurement using this procedure should then be biased towards being greater than the
actual working gap.
Under maximum compression, the flake density in the gap must be very close to the ore
density and a technique to measure it has yet to be found. The flake material in the
product is then normally used to estimate the density of the flake within the gap at the
point of highest compression.
Measured specific comminution energy

The specific comminution energy or Ecs (kWh/t) and (Wh/kg) for the laboratory scale
HPGR, ball and rod mill tests is calculated directly from the net measured energy (Wh) of
the test divided by the mass of the test sample (kg). The rod and ball mills also use the
empirical Bond equations to do this, which is compared as part of the data analysis
(Chapter 4 and 5).
The gross power draw (Pgross) for any rotating comminution device is normally defined by
equations 24 and 25.
Pgross = Pshaft + Pn 0 load (24)
- 87 -

Where,
Pshaft is the net shaft power or power available for comminution (kW)
Pno-load is the unit no-load power draw (running empty) (kW)
Pgross is the gross power draw for the unit (kW)
Pgross = E cs Qm + Pn 0 load (25)
Where,
Qm is the measured throughput (t/h)

Ecs is the specific energy (kWh/t)
Pgross is the gross power draw for the unit (kW)
The experimentally measured specific comminution energy (Ecs) is then normally

determined from the difference between the gross power (Pgross ) minus the no-load
power (Pno-load) divided by the measured throughput (Qm) as expressed in equation 26.
( Pgross Pno load )

E cs = (26)
Qm
Traditionally, before digital energy meters became available, the shaft power (Pshaft) was
calculated from the experimentally measured torque ( ) on the rotating shaft along with
the circumferential speed of the rotating shaft and was expressed as given in equation 27.
2u
Pshaft = (27)
D
Where,
= shaft torque (Nm)

u = circumferential speed (m/s)
D = rolls diameter (m)
Pshaft = net shaft power (kW)
The no-load power (Pno-load) in laboratory comminution devices is generally very high in
proportion to the gross power (Pgross). In industrial scale comminution devices the no-load
power (Pno-load) power is normally within 5-10% of the gross power (Pgross) leaving most
of the available energy for the device to comminute the rock. For this laboratory research,
the specific energy is measured using accurate measurements of both the gross power and
no load power which are integrated using the digital energy meter device (This is
explained further in Chapter 4). Thus for the experimental laboratory mill the net mill
energy (energy at the shaft) is determined from the measurement of the gross mill energy
minus the no-load energy (equation 28). The specific comminution energy is then
determined from the net energy (Wh) divided by the test sample mass (kg) to give Wh/kg
which is equivalent to kWh/t as shown by equation 29.
E net = E shaft = ( E gross E no load ) (28)
- 88 -

Where,
Egross =s the gross energy consumption of the device for the duration of the test.
Eno-load = the gross energy consumption of the device under no load conditions for the
equivalent of the duration of the test.
Enet = the net energy consumed for the purpose of rock breakage.
( E net )(Wh )
E cs = or (kWh / t ) (29)
test sample mass(kg )
Since the no-load power (Pno-load) normally represents 5-10% of the gross power, the
experimentally measure net energy should approximate gross energy.
In the conventional Bond ball milling tests and the hybrid HPGR/Bond ball milling tests,
a digital energy meter was installed to directly measure energy consumption. The energy
meter measures the total energy in Watt hours (Wh) consumed by the comminution
device. Energy consumption of the devices operating under no-load conditions were
measured. The difference between the two Wh values determines the net energy imparted
into the ore when divided by the sample mass, the specific comminution energy is
determined as was described for the HPGR.
Bulk “compacted” density of feed.

Representative samples of the feed were placed in a 2000 ml measuring cylinder and
tapped 20 times in order to measure the bulk “compacted” density. Particles compacted in
this way represent the situation where the voids between the particles are assumed to be
minimised. The tapping method that was followed was the same method that is used to
ensure consistent packed volumes of material required for Bond ball mill work index
tests. This procedure is followed in laboratory scale HPGR tests, but the convention in
industry is to use “loose” bulk density or free fill bulk density to calculate the nip angle in
industrial scale HPGR units. Bulk density of fine particles in laboratory scale HPGR tests
are more sensitive to “loose” bulk densities and effect the performance of the HPGR
model as described below.
The "compacted" bulk density ( c) is an important measured ore property which is used in
the HPGR model. The bulk “compacted” density of the material represents the character
of the material at the interface between the pre-crusher zone and the compression zone.
The bulk compacted density is used to track variations in the measure of flake densities.
Often high flake densities are measured as a result of higher “bulk” compacted densities
and not HPGR process.
HPGR “Flake” density determination

The measured density of the flakes usually confirms that the density of the flake is in the
order of 80-85% of the rock density. Measurement of the flake is nevertheless important
as it is an indicator on how well the test was performed. The measured density of the flake
material is used in the throughput model. The throughput in turn is important because it is
used to calculate the specific comminution energy.
- 89 -

The density of the flakes is determined by using the measurement of the mass of the flake
in air, compared with the measurement of the mass of the flake when immersed in water.
Since the flakes are normally quite friable and are not water resistant, a simple procedure
of coating the flakes with a thin layer of lacquer paint both strengthens and makes the
flakes water resistant (Figure 63). The change in mass of the flake due to the coating may
be considered as being negligible and was previously measured to be of the order of 1.5%
(Daniel, 2003).
HPGR coated flake product
Figure 63 shows the labelled and coated flakes that are used to determine flake density.
Product size distribution measurement procedure

The entire product from each of the HPGR tests were subjected to full sieve sizing down
to 850 Om. The sample is de-agglomerated in the screening process using a deck of Gilson
screens as previously shown in Figure 37. Representative sub-samples of the -850 Om
material was split and screened down to 38 Om using standard laboratory screens and Ro-
tap screen shakers. Figure 64 shows material passing through the HPGR and the
compacted flakes that require de-agglomeration.
- 90 -

Figure 64 shows the Mt Isa ore passing through the HPGR
Summary
The experimental program has shown that careful planning was required when conducting
these comparative comminution tests, especially when new comminution energy
measurements are being made. The application of the Bond work index and associated
third theory equation requires special attention in as far as the limitations of its use. Bond
results are expected to change should the standard test conditions deviate from the
standard procedure.
Several sensitive process measurements are required when conducting HPGR tests.
Measurements are often used to confirm the validity of the test runs, and to provide
additional process data that could be used for modelling purposes at a later stage. The
HPGR process models have previously been developed by Tondo (1999) and Morrell et al
(1996, 1997) and validated by Daniel (2003). These models are used in the processing
simulator software JKSimMet. Other models developed by Klymowsky and Liu (1997 a)
at the same time as Tondo are used by Klymowsky to size and specify HPGR units for one
of the HPGR manufacturers.
- 91 -

- 92 -

CHAPTER 4
MEASUREMENT OF ELECTRICAL ENERGY

CONSUMED DURING A BOND BALL MILL
LOCKED CYCLE TEST.
Introduction
This chapter presents the results of the direct energy measurement technique used to
determine how much energy is consumed in the Bond ball mill. The results are used to
evaluate a new digital energy meter with reference to Bond’s empirical energy
calculations. Since one of the objectives of the study is to compare the energy required to
grind ores with the HPGR against conventional grinding sound energy measuring
instruments were needed, and these results should ideally conform to Bond’s
methodology. At the same time, in order to be able to conduct the comparison in an
unbiased fashion it was necessary to have the same device for measuring the energy
required to break the ore in both systems.
The Bond ball and Bond rod mill are used as the standard method for determining the
energy required for breakage in these devices. The experimentally determined Bond work
index and the Bond third theory equation are combined to determine the amount of energy
required to do the work in the device. The Bond methodology provides an accurate work
index value provided the standard procedure is followed. When a deviation in the test
procedure occurs, then the work index value needs to be treated with caution. The Bond
work index or operating work index does not apply for HPGR. as the mode of breakage
and the efficiency of the breakage process in the HPGR is totally different to tumbling
mills, particularly the ball and rod mills.
The digital energy meter was evaluated as a means of measuring the energy required to
break ore in both the Bond mills and the laboratory scale HPGR. The suitability of the
digital energy meter for measuring the energy required in comminution devices is
investigated. The measurements are confirmed for accuracy and repeatability and are used
to total the circuit energy requirements in the hybrid HPGR ball mill circuits. The circuit
energy is then compared with the conventional milling energy requirements. Another
objective of using the digital energy meter was to generate data that could support a
measured size-energy relationship for hybrid HPGR/ball mill circuits.
The mill energy obtained using the digital energy meter is possible through advances in
fast computer chip technologies which numerically integrate current and voltage signals
as they are measured This capability enables direct energy measurement that can be used
to determine the energy efficiency of the comminution device. The theory behind the
digital energy measuring device is described in more detail in the paragraphs that follow.
One of the key objectives of the thesis is the assessment of comparative comminution
circuits. To do this it was necessary to establish the baseline circuit energy consumption.
The ball mill was used in this instance since the Bond ball mill and Bond equation are still
widely used to determine the required comminution energy. Existing outdated rod mill
- 93 -

installations are being replaced by newer more efficient technologies such as water flush
cone crushers, SAG mills, and HPGR’s. Any device or circuit which is subsequently
tested as a possible alternative to ball or rod milling would then benchmarked against the
ball-mill / ball-mill-circuit in terms of its use of electrical energy and hence energy
efficiency.
Energy measurements and definitions

The comminution process is defined as the process of reducing the size of ore particles in
order to liberate the valuable mineral component/s. In Chapter 6 the extent of the liberated
values of the ores tested in this study are presented. At this stage it is unclear as to how
mineral liberation is affected by varied amounts of energy used to break the particles.
When different devices are used, each device will utilize the available energy in other
processes not related to breakage, such as heat, noise and grinding media wear. As such
the energy efficiency of each device is different. In the past, measurement of energy in
small devices was not possible which lead to the development of empirical models such
as Bond’s to determine the net energy requirement in mill. Today the quantity of energy
consumed in a laboratory device can be measured directly with the digital energy meter.
As such it could be possible to determine if mineral recovery is a function of the energy
used in a particular device and the characteristics of the liberated products. If the
efficiency of the devices tested are different, then it is important to obtain identical
products with which to make a comparison between the energy efficiency of the devices
or circuits that are being compared.
Traditional methods of power and energy measurement in comminution devices are

described in an example by Fletcher (1990). Fletcher defines comminution energy
terminology as the measurement of specific energy consumption (kWh/t) and specific
comminution energy (kWh/t) for various crushing and milling devices.
Fletcher’s definition of the specific comminution energy (kWh/t) is the energy imparted
into the rock. It is calculated by dividing the measured net energy by the mass of the rock
crushed. The net energy during crushing is calculated as the difference between the gross
energy during the crushing period and the gross energy when the crushing device is
running idle (no-load conditions) for the same period.
Fletcher’s definition of specific energy consumption (kWh/t) is the total energy imparted
by the device into the rock per unit mass. It is calculated by measuring the total units of
energy used during crushing (kWh) which includes the no-load component losses,
divided by the rock mass crushed (t).
For Fletcher to obtain representative power measurements, batches of 250 kg of rock

material were processed. An ammeter was fitted to the crusher motors and the energy
consumption was determined by measuring both the voltage across each phase and the
current. Fletcher took phase or line current readings at 30 second intervals and the mean
power values were calculated over the test period. A power factor of 0.8 was assumed in
the calculation. Energy was determined from the integration of the power over a defined
period of time. The method of using an ammeter is not accurate, but has been the method
most commonly used through lack of available and accurate energy measuring devices.
Normally, the accuracy of power measurement is diminished, especially when the device
is connected to a three phase electrical supply as one phase is measured and is assumed to
- 94 -

be balanced over the remaining two phases. The line current is assumed to equal the phase
current. The power draw is then calculated using the following expression (equation 30),
in units of kW.
3 line current. x voltage x power factor

Power Draw = (30)
1000
The new digital energy metering device is capable of measuring both single phase
applications and three phase applications. In three phase applications, the digital energy
meter a measure each phase independently and totals the energy of each of the three
phases. The energy meter can therefore determine if the motor windings are balanced or
not. The new energy meter is currently the only one known to have this capability and has
enabled new ways of measuring specific comminution energy and the energy efficiency of
a particular comminution device. Very small quantities (1- 20 kg) of rock can be crushed
or milled, and a measurable differentiation between the load and no-load energy
consumption is possible. A description of how the energy meter achieves the accurate
measurements is given in the next section.
The new digital electrical energy metering device

The new meter uses a combination of innovative software and hardware design and
common off-the-shelf components to produce a simple to use, accurate power and energy
measuring instrument. Before an explanation is given on how this is achieved, it is useful
to define some of the units and terminology used to describe electrical energy.
Electrical energy or electromagnetic energy is a form of energy related to the position of

an electric charge in an electric field. The electrical energy of a charge Q situated at the
electric potential V equals to the product QV. If V is a potential difference, the same
expression gives the energy transformed when the charge moves through the potential
difference. Electrical energy is identical to the electromagnetic fields surrounding a
transmission line or propagating in space as electromagnetic radiation. The quantity of
electrical energy surrounding a conductor or traveling through a waveguide may be
calculated by integrating the cross product of the magnetic and electric field vectors
throughout the volume.
The joule (symbol J, also called Newton meter (Nm), or coulomb (C) volt (V)) is the SI
unit of energy and work. Thus
1J=1N×1m =1W×1s=1C×1V
One joule is the work required to exert a force of one Newton for a distance of one meter,
so the same quantity may be referred to as a Newton meter. However, to avoid confusion
the Newton meter is usually used as a measure of torque, not energy. It is interesting to
point out that many researchers have until now used and relied on torque measurements at
the shaft of mills to determine the specific comminution energy.
The watt (W) is the SI derived unit for power. It is equivalent to 1 joule per second (1 J/s),
or in electrical units, 1 volt-ampere (1 V·A).
- 95 -

W = J·s-1 = N·m·s-1 = kg·m2·s-3
Energy is a quantifiable state function of every physical system. Energy allows physical
system to do work or to exchange a quantity of heat. The terawatt hour (TWh) is a unit for
measuring energy and corresponds to 1,000,000,000 kilowatt hours (kWh). The kilowatt-
hour (kWh) is a common unit for measuring energy and corresponds to one kilowatt (kW)
of power being used over a period of one hour.
The kilowatt-hour is most commonly used for electrical energy, since it is easier to
understand in a practical context than the proper SI unit for energy, the joule, which is a
watt-second (Ws). The joule is a comparatively small unit, making numbers quite large.
For example 1 TWh = 1,000,000,000 kWh = 3,600,000,000,000,000 J.
These definitions are useful since global energy consumptions are measured and
expressed in terawatt-hours, whereas energy consumptions in laboratory test mills are
expressed in terms of joules or watt-seconds. Considering that millions of tons of ores are
crushed and milled annually in the mining industry, this amounts to the very large TWh of
electrical consumptions as previously defined.
The primary display of the new energy meter is the total energy used over time in watt
hours (Wh). The internal circuitry is protected against voltage spikes, surges, etc, and the
self calibrating circuitry compensates for temperature and other factors. The energy
metering device has been factory tested and calibrated using a laboratory grade Yokogawa
Power Meter (Model 2533) to ensure accurate power measurements regardless of load or
power supply.
A common misconception is that power is simply the product of voltage and current,
implies that any ammeter can be used to determine power usage. This is true in a few
special cases. Voltage and current interact together to produce power. The digital energy
meter utilises a combination of analog and digital signal processing to calculate power. In
alternating current (AC) systems, voltage and current go from positive to zero to negative
generally 50 times a second (50 cycles or 50 Hz). If voltage and current are "high" or
"low" at the same time, then they are said to be "in phase." A common term to describe
the time relationship between voltage and current is called the "power factor." When
voltage and current are "in phase," the power factor is regarded as unity. When the current
peak follows the voltage peak, this is called "lagging power factor"; when current leads
voltage, this is called "leading power factor." When voltage and current are exactly
opposite (when voltage is at a peak and current is zero, or vice versa), the power factor is
zero.
To find the "real power", it is common to determine the product of the volts, amps and
power factor. The power factor on the other hand is determined by the load, that is, the
device using the power. A common induction motor, such as those found in laboratory
comminution devices stores some of the electrical energy in a magnetic field. This energy
storage causes the voltage and current to be high at different times, and the resulting real
power usage can be a small fraction of the product of the voltage and current.
The new digital energy meter accurately measures the energy used by electrical devices
which are commonly affected by the changing power factor conditions. It does this by
measuring current and voltage approximately 4000 times per second. These values are
- 96 -

processed digitally and the power used is updated 3 times every second. The energy meter
compensates for changes in frequency, voltage and current supply. How this is achieved is
described in the next section.
Active power calculation

Electrical power is defined as the rate of energy flow from source to load. It is calculated
as the product of the voltage (equation 31 a) and current (equation 31 b) waveforms. The
resulting waveform is called the instantaneous power signal and is equal to the rate of
energy flow at every instant of time. The unit of power is the watt or joule/sec. Equations
32 and 33 give expressions for the instantaneous power signal in an AC system.
(t ) = 2 V sin( t ) (31 a)
i (t ) = 2 I sin( t ) (31 b)
Where,
V = root mean square (rms) voltage

I = root mean square (rms) current
p (t ) = (t ).i (t ) (32)
p (t ) = VI VI cos(2 t ) (33)
The average power over an integral number of line cycles (n) is given by the expression in
equation 34.
nT
1
P= p (t )dt = VI (34)
nT 0
Where,
T = the line cycle period.

P = the active or real power.
The instantaneous power signal p(t) is generated by multiplying the current and voltage
signals. This process is illustrated graphically in Figure 65.
- 97 -

Figure 65 depicts the waveforms of electrical energy used in the active power calculation.
Energy calculation
As stated above, power is defined as the rate of energy flow. This relationship is
expressed mathematically in equation 35.
dE
P= (35)
dt
Where,
P = power
E = energy.
Conversely, energy is given as the integral of power (equation 36):
E = Pdt (36)
The new digital energy meter achieves the integration of the active power signal by
continuously accumulating the active power signal in a 40-bit active energy register. This
discrete time accumulation or summation is equivalent to integration in continuous time.
Equation 37 expresses the relationship:
"
Lim
E = p(t )dt = p (nT )T (37)
T !0 n =0
Where,
- 98 -

n = the discrete time sample number
T = the sample period.
The discrete time sample period (T) for the accumulation register in the energy meter
electronic chip is 1.1 Os. As well as calculating the energy, this integration removes any
sinusoidal components that may be in the active power signal. This is normally referred to
as the power factor.
Figure 66 shows a graphical representation of this discrete time integration or

accumulation. The active power signal in the waveform register is continuously added to
the active energy register. This addition is a signed addition; therefore negative energy
will be subtracted from the active energy contents.
Figure 66 depicts the digital energy integration mechanism in the new energy metering device.
The Bond test methodology and ball mill power models

In Bond’s time, researchers did not have the luxury of modern day computers or the
electronic instruments that are readily available today. Nor did Bond have the ability to
accurately measure the electrical energy consumed in his small laboratory scale
comminution device. Bond however did have knowledge from the laws of physics. For
this reason Bond “invented” the standard lab-scale locked cycle test and the empirical
equations from which the well known “Work Index (WI)” or more specifically the “Bond
Work Index (BWI)” is determined. The work index for either rod milling or ball milling
is used in his so called “third theory” (equation 16 on pg 94 or equation 38 on pg 119) to
determine the specific comminution energy to grind an ore to a specified size.
Over the years, several hundred Bond ball mill tests have been completed at the JKMRC.
The BWI of these tests on various mineral ores have been determined and classified into
four broad categories, soft, medium, hard and very hard. These data are displayed and
compared with UCS values in Table 11(Napier-Munn et al., 1996).
- 99 -

Table 11 lists the ranges over which the Bond work index and UCS are characterised.
Property Soft Medium Hard Very Hard
UCS (Mpa) 50-100 100-150 150-250 >250
Bond WI (kWh/t) 7-9 9-14 14-20 >20
The statistical average of more than 1000 other tests completed over several years at AR
MacPherson Consultants is given as 14.6 kWh/t as shown in Figure 67 from (Mosher and
Tague, 2001).
250
Mean = 14.6 kWh/t Mosher
200 and Tague
Frequency
150
100
50
0
5.2 7.4 9.2 11 12.8 14.6 16.4 18.2 20 21.8
BWI (kWh/t)
Figure 67 represents a statistical histogram of the Bond test work indices that were based on more than 1000
test results. (Mosher and Tague, 2001).
Bond’s so-called “Third theory” equation is expressed again as;
Wi 10 Wi 10
W= (38)
P80 F80
Where again,
W = specific energy
Wi = Bond ball mill work index
P80 = 80% passing size for the product in microns
F80 = 80% passing size for the feed in microns
The Bond equation (equation 38) is in fact the classical definition of the size – energy
relationship as displayed in Figure 68 and Figure 69. Figure 68 shows that an exponential
increase in energy arises, particularly when the grind size is less than 75 Om.
Ball mill energy consumptions are by comparison to crushing devices extremely high, and
using the Bond method to predict mill performance can sometimes be erratic or produce
erroneous results. This is caused mainly by deviations from the original test procedure,
- 100 -

and secondly because the BWI is being used with the Bond equation outside of the range
in which it was designed to be used (This was covered at the end of the last chapter).
Figure 69 highlights the size range where the Bond equation is most applicable.
200
Bond Wi = 7
180
Bond Wi = 11
Bond Wi = 14
160
Bond Wi = 16
Bond equation comminution energy kWh/t
140 Bond Wi = 19
Bond Wi = 22
120 Bond Wi = 29
100
80
60
40
20
0
0.1 1 10 100 1000 10000
Grind size (micron)
Figure 68 shows the full range of theoretical particle grind sizes versus specific comminution energy when
Bond equation represents all particles sizes and comminution devices.
30
Bond Wi = 7
Bond Wi = 11
Bond eqn - ball mill comminution energy kWh/t
25 Bond Wi = 14
Bond Wi = 16
Bond Wi = 19
Bond Wi = 22
Bond Wi = 29
20
15
10
0
100 1000 10000
Grind size (micron)
Figure 69 shows the limited range of particle grind size (10 mm – 100 Lm) versus specific comminution energy
in ball mills for which the Bond equation was originally designed to represent.
- 101 -

Mill power and energy consumed per mill revolution
It is well understood that the required mill power is driven largely by the size and load of
the mill (Figure 70). Several excellent power models have been developed in the past by
Morrell (2004) and Hogg and Fuerstenau (1972) to mention a few (Figure 71) and these
models are commonly used to facilitate the mill design process. This is to say that given
the required throughput and grindability of the ore, the required power may be predicted.
From the power requirements, the mill load and size are determined. Morrell claims to
have developed the most accurate mill power model (Figure 70), although the model data
by Arbiter and Harris (1982) (Figure 71) is also very good. Morrell’s “C” model is
relatively easy to understand and use provided that the appropriate mill dimensions and
mill operating conditions are known (Morrell 1993).
4500
4000 Gross KW
3500 Linear (Gross KW)
Gross Power (kW)
3000
2500
2000
1500
1000
500
0
0 50 100 150 200 250 300 350
Ball Mill Mass (Tonnes)
Figure 70 shows how a direct relationship between ball mill mass and gross power exists (Morrell, 1993)
4
2.5 10
4
2 10
Predicted Power Draw (kW )
4
1.5 10
4
1 10
Arbiter & Harris
Rose & Evans

5000 Austin
Bond
0
4 4 4 4
0 5000 1 10 1.5 10 2 10 2.5 10
Measured Po wer Dra w (kW)

Figure 71 shows the predictive performance of various power models as developed by Arbiter & Harris, Rose
& Evans, Austin and Bond (after Morrell, 1993).
- 102 -

The predictive power of these models was applied to the laboratory scale 1 ft. Bond ball
mill so as to compare the energy measurements with the model prediction. The results are
presented later in this section. The specific comminution energy in the Bond ball mill as
described in section 3 is calculated (assuming drive train losses as negligible) using the
Bond equation, the experimentally determined Bond work index, the feed F80 and
product P80. Bond (1961a), states that the kWh/t as determined by the third law equation
is the specific energy consumption of a full scale mill. The Bond laboratory mill is used to
mimic the specific comminution energy of the full scale mill. Direct measurement of the
gross power is not used since small mills such as these have by comparison to full scale
mills much larger motors and drive systems. Full scale drive systems are normally much
more efficient in the use of energy than small laboratory devices. Estimated drive train
losses of full scale SAG and ball mills is usually in the range 5-10% of the installed
power, whereas HPGR units usually have a 2-3% no-load power component on full scale
units. Thus the Bond equation may well have the drive train losses built into the empirical
method. This needs to be taken into consideration when actual measurements of the Bond
laboratory ball mill are made.
Bond (1961 b) stated that the energy per mill revolution was 60 J or 60 Ws, but no
evidence was given to support this, nor if it related to balls in the mill or balls and ore in
the mill. However in Bond’s “Confirmation of the Third Theory” paper of 1959, Bond
again stated that the net work input to the grindability tests mill was “about” 60 Ws/rev,
but this time evidence of how this numerical value was determined was referenced to his
original “Standard Grindability Tests Tabulated” paper (Bond, 1949). Bond (1949) stated
here that, “it has been calculated that the ball mill does 52.3 joules of useful work on the
ore per revolution, with 93 joules input to the mill.” This information is vital as it would
appear that the 60 J/rev is approximated from the 52.3 J/rev. Bond (1949) did not specify
if the energy in J/rev was for ore and balls or for balls only, but he did specify the mass of
the ball charge being 285 balls of mass 20125 g which is the same mass and number used
in the test procedure and is consistent with current day practice. Bond (1949) states the
energy input to the mill as 93 joules or Ws/rev, and when combined with the mill rotating
at 70 rpm is equivalent to an energy input or net power draw of 108.5 W or 108.5 Wh per
hour. At the JKMRC, the measured mill rotational speed is 71.67 revolutions per minute
which equates to a mill power draw of 111.1 W. It appears that the Bond specific
comminution energy for the Bond test conditions may be determined using the steady
state mill revolutions multiplied by the approximated 60 J/rev, which is converted to
Wh's, and is then divided by the new feed mass in tonnes as depicted in Figure 72.
Mill
revs @
60 J/rev 100% Product
100 % Feed
size distribution passing a limiting
IN screen size
OUT
Mill load = 500 ml
Figure 72 depicts new feed at steady state conditions where Bond estimated the mill energy to be 60 J/rev(See
also Figure 8’s left hand side flowsheet, the same conditions where Figure 72 represents when steady state
condition during a locked cycle test.
- 103 -

Another way of describing this is given by equation 39 where the specific energy
consumption is a function of grindability and is expressed in units of kWh/t. This equation
is very similar to Bonds equation as represented in equation 18, except the grindability
term does not have the 0.82 exponent. This alternative approach merely confirms the
simplistic nature of Bond’s empirical equation that is used to provide the energy size
relationship that is commonly known and used for the description of ball mill
comminution processes.
W .s
Energy per revolution rev = W .s = 60 = 16.667 (kWh / t ) (39)
Wball mill = =
Grindability (Gbp ) g Gbp 3.6 Gbp Gbp
rev
In Figure 73, the measure of the specific comminution energy as determined by using a 60
J/rev is compared to the specific comminution energy obtained from the Bond work index
and third theory equation for more than 250 Bond ball mill test results. As may be seen
the agreement is very good. From this remarkable and convincing correlation it would
appear that the energy required to mill the ore from a given feed size to product using the
Bond work index and third theory equation is the same as obtaining the energy from
calculating the energy from the number of steady state revolutions multiplied by the 60
J/rev which is then divided by the mass of the new feed (at steady state) to the mill. Figure
74 shows the variation in the measured variables that constitute the measured specific
comminution energy based on the test results of more than 250 Bond work indices.
Using this valuable information it would have been expected that the measured Ws per
revolution as determined by the new digital energy meter would be of the order of 60 J/rev
or 60 Ws/rev. However the measured value was 91.44 Ws/rev. This suggests that Bond
may have estimated the energy of the locked cycle test mill from the then full scale 2.4 m
diameter wet grinding mill data. Searching the literature, it was discovered that Bond
stated that the locked cycle test mill consumed in the order of 93 J/rev which is
comparable to the measure value of 91.4 J/rev (Bond 1949) When a value of 93 Ws/rev is
used as the actual energy input to the locked cycle ball mill test unit, then equation 40 is
assumed to hold true.
A detailed analysis of the overall effect of this on the thesis data is given later, but points
out that the measured energy in the HPGR and the Bond ball mill should be used as the
common energy measuring device that is ultimately used to determining the energy
efficiency of the circuits that are being compared. If the Bond equation methodology was
to be used, then the two effective energy measuring devices would be different and would
not reflect the actual energy savings that are possible by using HPGR.
W .s
Wbond = rev = W .s = 93 = 25.83 (kWh / t ) (40)
ball mill
g Gbp 3.6 Gbp Gbp
rev
From the evidence of the Bond’s original historical data, and from equations 39 and 40, it
is realised that according to Bond, there is a built in scale-up factor of 0.645 between the
Bond ball mill test energy and the 2.4 m mill which Bond used to calibrate his model.
- 104 -

30.00
Bond test database

y = 1.0041x
y=x 2
R = 0.9866
Linear (y=x)
25.00 Linear (Bond test database)
Bond energy @ 60 J/rev (kWhr/t)
20.00
15.00
10.00
5.00
0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Bond Equation kWhr/t
Figure 73 shows the relationship between the Bond specific comminution energy based on 60 J/rev and Bond’s
specific comminution energy using the empirical equation and work index value.
10000
Feed mass
1000
Closing screen
Bond test output value
% fines in the feed
100 Feed F80
Product P80
Grindability Grp
10
Mill revolutions at S.S.
Circ Load
1 BWI
0.1
0 50 100 150 200 250 300
Bond test data sorted
Figure 74 shows the characteristics of the raw Bond data as assembled from a 250 tests database.
Figure 75 shows the thesis Bond test data results plotted along with the JKMRC Bond test
database, confirming that the experimental data is in good agreement with the database.
- 105 -

Figure 76 and Figure 77 shows the same data as presented in Figure 75, but the results are
plotted on a log-log and log normal scale against the Bond grindability, Gbp parameter.
Co-incidentally the fitted curve to the data represents equation 39.
Clearly the linear shape of the data plot in Figure 76 represents an exponential increase in
energy requirements as the measured grindability decreases. This indicates that some
other factor other than just ore hardness is driving the overall process of producing fine
particles. It is suspected that the generation of new particle surface area is the probable
cause. This is not new idea, as is forms the basis of the Rittinger hypothesis which
suggests that it is new particle surface area that describes the particle grind size versus
energy relationship as previously discussed in the literature review. Holmes(1957)
contribution too considers an alternative size energy relationship which was derived from
the concept of reducing a one cubic meter of rock in consecutive cubic splits
(incorporating theoretical new particle surface area generation) which is expanded on in
more detail at the end of Chapter 5.
30
Bond Energy @ 60 J/rev ground product
25
y = 0.9982x
2
20 R = 0.9782
15
JKMRC Bond test database
Phd test data - Mt ISA lead-zinc ore

10
Phd test data - Bhpbilliton bauxite ore
Phd test data - Lonmin platinum ore

5
Linear (JKMRC Bond test database)
0
0 5 10 15 20 25 30
Bond (Third theory) Equation Energy (kWh/t)
Figure 75 shows how the Bond third theory equation energy is linked to a 60 J/rev net energy consumption in
the Bond locked cycle test.
- 106 -

45
Bond Eqn energy @ C.S. 75um
40
Bond third theory Eqn Energy (kWh/t)
Bond Eqn energy @ C.S. 106 um
35 Bond Eqn. energy @ C.S. 150 um
30 Bond Eqn energy @ C.S. 212 um
Bond Eqn energy @ C.S. @ 300

25 um
20
Power (JKMRC Bond test
database)
15
10
y = 16.6 x-0.98
5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Bond test Grindability (Gbp) net g/rev
Figure 76 shows the relationship between the Bond specific comminution energy, BWI and Gbp (Grindability) –
Equation 19.
100
Bond Third theory Eqn Energy (kWh/t)
10
Bond Eqn energy @ C.S. 75um
y = 16.6 x-0.98
Bond Eqn. energy @ C.S. 150 um
Bond Eqn energy @ C.S. @ 300 um
Power (JKMRC Bond test database)
1
0.1 1 10
Bond test Grindability (Gbp) g/rev
Figure 77 shows the log- log representation of the data in Figure 86.
Levin’s methodology
Levin (1984,1989) used the results of 248 Bond tests and the resultant Bond equation
energy to establish a value for the energy per mill revolution. Levin calculated this value,
and determined that it did not remain constant, but varied depending on the ore type. On
average the value was quoted as 198x10-7 kWh/rev which is 71.2 J/rev or 71.2 Ws/rev or
- 107 -

0.020 Wh/rev. This is equivalent to a net power draw of 85.1 W when the mill is
operating at 71.67 rpm. Levin (1984,1989) stated that the mill power ranged from 77.4 to
94.6 W, which at a glance could seem possible due to possible varied frictional effects of
the ore in the mill, as well as the variation in the mass of the 700 cc of ore due to density.
However, when Levin’s work is examined more closely, the method of determining the
energy per revolution is based on using Bond’s third law equation in conjunction with the
expression for grindability, given that the BWI is known or has been measured using the
Bond test. If for whatever reason the Bond equation (equation 16) or the grindability
equation (equation 15) are flawed, then the resultant energy per revolution calculation will
be flawed. Levin combined equations 15 and 16 (Page 75) and proposed a mathematical
term for the calculation of the net energy per mill revolution (B), given by equation 41 .
0.0049 Gbp0.18
B= (41)
P10.23 (100 U )
Where,
B = net energy per mill revolution (kWh/rev)

Gbp = net grams of undersize produced per revolution in the Bond grindability test mill.
P = product size (closing screen in microns)
U = percentage undersize in the feed
By Levin’s method or equation 41 (Figure 78 and Figure 79), it is implied that the mill
energy varies according to the closing screen size, the ore grindability and the percent of
fines less than the closing size in the new feed. The variation in calculated mill energy per
revolution ranges from 50 to 110 Ws/rev. From a practical point of view this is nonsense
and the Levin method should be dismissed. If the mill energy was to vary, then the
variation would be attributable to the frictional characteristics or stickiness of the ore
when milled. These frictional interactions between the ore and the balls and the mill liner
could change the lift angle, and hence the energy input required to rotate the mill.
In addition to this mill energy variations could be attributed to a slight mass increase or
decrease in the average load when materials of different densities are used. The Bond mill
standard uses 700 cc of ore in the charge, and this could range from 900 g to 2300 g over
and above the ball charge mass of 20,125 g. This effect on power is clearly illustrated in
the data Tables A1-A3 of the Hogg and Fuerstenau (1972) model in Appendix A. The
tables show how the mill energy is sensitive to the lift angle, but that 35 degrees is a good
default value to use. The Bond mill however operates at a much higher critical speed, and
hence the lift angle should be expected to be slightly higher than this. The lift angle may
be difficult to measure in practice, but it can be estimated if the mill energy and operating
conditions are experimentally measured. Such results are given in the next section.
- 108 -

130
120
110
Energy per revolution (W.s/rev)
100
5
90
10
15
20
80
25
70
60
50
40
0 1 2 3 4 5 6 7 8 9 10
Net grams of undersize material produced per mill revolution
Figure 78 shows the variation in mill energy per revolution as determined by Levin where the percent product
in the feed is varied from 5% -25%, and the closing size is fixed at 106 Lm.
120
110
100
Energy per revolution (W.s/rev)
90
75
106
80 150
212
300
70
60
50
40
0 1 2 3 4 5 6 7 8 9 10
Net grams of undersize material produced per mill revolution
Figure 79 the variation in mill energy per revolution as determined by Levin where the closing size is varied
from 75 Lm to 300 Lm, assuming 10 % of the product in the feed.
Djordjevic (2004) built a DEM model of the Bond ball mill (Figure 80) and determined
that the net power draw of the ball mill charge was 76 W, which translates to 63.8 J/rev or
63.8 Ws/rev for a mill operating at 71.67 rpm. Bond specifies a mill speed of 70 rpm, and
the DEM model resulted in a power draw of 73.7 W or 63.2 J/rev when operated at 70
rpm, rather than 71.67 rpm (Figure 81). These values are however sensitive to the value of
the coefficient of friction and the coefficient of restitution that is selected in DEM
- 109 -

modelling. In this particular case 0.4 was used for the coefficient of friction. These
coefficients are difficult to measure in practice as described by (Chandramohan and
Powell, 2004). Figure 80 shows the visual of the 285 balls with a weight of 20,116 g in
the Bond ball mill with a single 5 mm proud lifter which represents the door of the Bond
mill.
e: Bond Mill - Balls Only
Figure 80 DEM representation of the Bond Ball Mill with the specified charge of 285 ball of 20,116 g.
View Title: Bond Mill - Balls Only

power (W) x10^2
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
1.0 2.0 3.0 4.0

time (sec)
Figure 81 shows the DEM output power profile of the rotating Bond ball mill, which averages out at 73.7 W
for a mill rotating at 70 rpm.
Morrell’s mill power model.

Morrell (1993) developed a power model which has been proven using a wide range of
data. The model uses a simplified description of the grinding charge motion and
incorporates the effects of the slurry pool which forms in all overflow mills and in some
grate discharge mills at high flow rates. The net power drawn by a Bond ball mill has
- 110 -

previously been estimated (Morrell and Man, 1997) as 86.65 W, again since the Bond mill
operates at 70 revs per minute, this translates the Morrell power to a value of 74.3 J/rev or
74.3 Ws/rev. This particular Bond ball mill value was determined using a modified
version of the Morrell power model (Morrell 2004) which takes into consideration mills
which have no internal lifters. The internal surface of the Bond ball mill is smooth.
The standard Morrell power model which incorporates lifter effects was tested on the
Bond mill conditions. When the calibration constant is set as 1, the resultant mill power is
81.4 W or 68.8 W.s/rev which is comparable to the balls only energy of Bond. Morrell
compared the theoretical model output to measured full scale wet tumbling mill data, and
determined that a calibration constant of 1.26 should be applied to correct the mill power
model. When a value of 1.26 is used, then the mill power increases to 102.6 W or 91.4
W.s/rev which is comparable to the assumed balls and ore component of Bond. The
results are given in Table A4 of Appendix A .
The above mentioned examples showed that there exist many different ways of
determining the amount of energy consumed in a Bond ball mill. These examples are used
to back up the results obtained from the digital energy meter. The results of the energy
measurements made using the digital energy meter are presented in the next section, with
a summary table of all methods at the end of the chapter.
Measured Bond ball mill energy results

Digital energy meter results
Figure 83, Figure 84 and Figure 85 show three different methods of data display for the
results obtained using the digital energy meter during the completion of 12 Bond ball mill
tests. The special tests for this exercise used the same ore, but the closing screen was
varied from 300 Om down to 45 Om. The ore was found to have a very high Bond work
index; hence the test required a large number of revolutions to achieve the steady state
condition. Other energy measurements taken include conditions where the mill was empty
and where the mill was filled with the ball charge (20.125 kg). The data was then
classified into four different categories, and the measured energy plotted against mill
revolutions. The data categories are classified as :
• Ore (700 cc) + Balls (20.125 kg) - Bond test cycles 1-4
• Ore (700 cc) + Balls (20.125 kg) - Bond test cycles 5-7/8
• Balls (20.125 kg) - Standard Bond ball charge
• Empty mill, no-load condition
These results are displayed in Figure 83 and Figure 85. Figure 83 gives the average energy
rate per revolution of the ore and balls in the mill for cycles 1-4 and 5-7/8 of the Bond test
procedure. This gives an average measured total energy of 0.1331 Wh/rev. The no-load
total energy was measured as 0.1077 Wh/rev. Thus the specific comminution energy is
determined using 0.0254 Wh/rev (difference between 0.1331 and 0.1077) or 91.44
Ws/rev. This equivalent to a power draw of 106.76 W for the mill with ore and balls and
is comparable to Bond’s (1949) stated mill energy of 93 Ws/rev or 93 joules/rev. When
the identical process as described above is applied to the balls in the mill, then the result is
66.2 J/rev or 66.2 Ws/rev which represents a power draw of 79.1 W for the mill which is
- 111 -

comparable to Bond’s (1949) stated available useful energy of 52.3 joules/rev or 52.3
Ws/rev and Bond’s (1959) stated energy of “about” 60 joules/rev or 60 Ws/rev.
The slope of each of the fitted linear curves is given in Figure 83, Figure 84 and Figure
85. These linear coefficients represent the rate at which energy is used per revolution.
Figure 83 and Figure 84 show the results of two sets of energy measurement data that was
generated using two independent but similar digital energy meters (Figure 82). The energy
meter used in Figure 83 had a watt-hour resolution of 0.1 Wh (unit on the left hand side of
Figure 82), whereas of the energy meter used in Figure 84 had a watt-hour resolution of
1 Wh(unit on the right hand side of Figure 82). Essentially either unit could be used,
however, when the locked cycle tests involved 10-50 revolutions or when the HPGR tests
had less than 10 kg of sample, and then the resolution could have an effect on the
accuracy of the results. The regression R coefficients of the linear fits to the data resulted
in very good results, with a small variation in the regression co-efficient of the load, and
no-load conditions. The HPGR unit is fitted with three phase current motors; hence the
three phase “one meter” must be used in this type of application. The Bond ball mill is a
single phase motor. The two digital energy meters as depicted in Figure 82 were used to
determine if there was any measurement bias in respect to the digital energy meter itself.
The results were similar, and thus propagation of measurement errors due to the
instrument can be ignored in this set of data, although the “one” meter which measures
three phases simultaneously was the unit used in all of the HPGR tests.
Energy meter Unit No. 2

Brand electronics single
phase power meter
Energy meter No. 1,

Energy meter Unit No. 1 accurate to 0.1 Wh
Brand electronics 3 phase Energy meter No. 2,
“one meter” accurate to 1.0 Wh
Figure 82 shows the two digital energy meters that were used to test the repeatability and accuracy of the
totalises Wh measurements made by the device.
- 112 -

150
y = 0.1326x y = 0.1337x
Ore + balls (Bond cycles 1-4)
140 2
R = 0.9984
2
R = 0.999
empty mill
130
mill+balls only (20 125 g)
120
Energy (Whr) - Unit Number 1
Ore+balls(Bond cycles 5-8)

110 Linear (Ore+balls(Bond cycles 5-8)) y = 0.1077x
2
Linear (empty mill) R = 0.9978
100
Linear (Ore + balls (Bond cycles 1-4))
90
80
70
60
50
40
30
20
10
0
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Mill revolutions
Figure 83 shows the energy consumption against mill revolutions using digital energy meter number 1.
100
y = 0.1276x
Empty mill (Unit Nr.2)
Energy consumption (Whr) - Unit Number 2
2
R = 0.9999
90 Balls only 20125 (Unit Nr.2)
Balls + Lonmin ore (Unit y = 0.1231x
2
80 Nr.2) R = 0.9999
Linear (Empty mill (Unit
Nr.2))
Linear (Balls + Lonmin ore
70 y = 0.0995x
2
R =1
60
50
40
30
20
10
0
0 100 200 300 400 500 600 700 800
Mill revolutions
Figure 84 shows the energy consumption against mill revolutions using digital energy meter number 2.
Figure 85 and Figure 86 display the instantaneous mill power (W) based on the energy in
Wh and the number of mill revolutions. It is noted that the no-load power is high, 474 W
compared to the balls only mill power of 535 W and the balls and ore mill power of 578
W. The difference between the no-load and the balls and ore mill power is equivalent to
104 W, also known as the net power. In this particular case, the no-load power accounts
for 82.1 % of the total power input. This is very high compared to the 5-10 % experienced
in full scale ball mills.
- 113 -

Bond milling time (min)
0 5 10 15 20 25 30 35 40 45 50 55 60 65
600 Empty mill (Unit Nr.2)
y = 0.1268x y = 0.1276x
Energy consumption (Whr) / Mill power (W)
R2 = 0.9981 R2 = 0.9999
550 y = 0.1231x Balls only 20125 (Unit Nr.2)
R2 = 0.9999
500 Balls + Lonmin ore (Unit Nr.2)
y = 0.1077x
R2 = 0.9978 Empty mill (Unit Nr. 1)
450
y = 0.0995x Balls only 20125 (Unit Nr.1)
400 R2 = 1
Balls+platinum ore (unit Nr.1)
350
Linear (Balls+platinum ore (unit Nr.1))
300 Linear (Empty mill (Unit Nr.2))
250 Linear (Balls + Lonmin ore (Unit

Nr.2))
200 Linear (Balls only 20125 (Unit Nr.2))
Linear (Empty mill (Unit Nr. 1))

150
Linear (Balls only 20125 (Unit Nr.1))
100
50 Measured Bond ball mill speed = 71.67

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Bond mill revolutions
Figure 85 shows the comparison between the two digital energy meter measurements and the error that is
expected should different devices be used to measure the energy.
However in practice as grinding mills increase in size, the no-load component remains in
the region of 5-10%. The balls “only” in a mill dominate the energy requirement and are
governed by the mass of the charge as illustrated previously in Figure 70 (Page 102). The
shape of the balls “only” instantaneous energy data appear to drop as the revolutions
increase. After several experimental runs it was found that as the electric motors warmed
up, the motors became more efficient and consumed less energy. The ball only experiment
data presented here illustrates this phenomenon, and represents “cold start” conditions. As
a result all subsequent tests were run after the mill had been in operation for at least 30
minutes.
625
empty mill
600
y = -0.0101x + 577.77
2
mill+balls only
575 R = 0.0668 (20 125 g)
Instantaneous Mill Power (W)
550
"energy ore +
balls all cycles
y = 0.006x + 535.26
2
R = 0.0092
525
Linear
(mill+balls only
500 (20 125 g))
Linear (empty
475 mill)
y = -0.0097x + 474.17
2
450 R = 0.1081
Linear ("energy
ore + balls all
cycles)
425
0 200 400 600 800 1000 1200
mill revolutions
Figure 86 show the Bond mill test data represented as instantaneous power for different mill load conditions.
- 114 -

These results also confirm that small mills such as the Bond ball test mill are grossly
power inefficient, and have resultant drive train losses of up to 80.74 %. This very high
value of mill in-efficiency is verified using the Hogg and Fuerstenau model as shown in
Table 12 and in the examples given in the Appendix. This condition is responsible for the
meaningless specific energy values that result when measuring only the gross power of the
laboratory comminution device. Measurement of accurate net energies has been
problematic.
Another observation of the Hogg and Fuerstenau data in Table 12, which is not
surprising, is that most of the energy is used in moving the ball charge. For effective
comminution to occur the balls need to transfer or impart all or most of the energy they
receive back into the ore particles to effect particle breakage. Much of the energy transfer
is unproductive and is consumed by noise, heat and the actual wearing away of the
grinding media. Overall this accounts for the very low energy efficiency of ball mills.
Crushers on the other hand are perceived to be known to be more energy efficient since
the mechanical energy of the moving liner is transferred directly to the ore particles.
However most crushers have proportionally a very large no-load power, where it could be
argued that the crushers are in fact not that efficient. However, when the absolute energy
between a crusher and a grinding ball mill are considered, the crusher uses relative to the
ball mill far less energy. The crusher duty is to generally prepare the material for the ball
mills. Crushers are limited in their ability to continuously reduce particle sizes in that as
the particles become smaller the crusher is unable to provide a sufficiently large enough
number of breakage events at the right level of energy intensity. As particles become
smaller, the number of particles that require size reduction increases and becomes more
difficult to accomplish due to the increase new surface area generation (more on this in
Chapter 5). The results of the measured digital energy (kWh/t) plotted against the Bond
equation's net comminution energy are shown in Figure 87. The off-set in the measured
versus predicted Bond value could be the effect of scale when using the Bond ball mill.
The measured scale-up in the Bond case was found to be 0.645 as previously suggested.
80
70
Measured Net Comminution Energy (kWhr/tonne)
60
50
40
Bond @ 60 J/rev
30
Measured Energy kWhr/tonne
y=x
20
JK Bond ball mill database
Linear (y=x )
10
Linear (Measured Energy kWhr/tonne)
0
0 10 20 30 40 50 60 70
Bond Equation -Net Comminution Energy
Figure 87 shows a comparison of the measured digital energy (kWh/t) plotted against the Bond equation's net
comminution energy and the calculated energy based on 60 J/rev.
- 115 -

Table 12 summarises the results of the Bond mill conditions calculated by the use of the
Hogg and Fuerstenau model (1972).
GRINDING TASK : HOGG & Fuerstenau power model (1972)
Ore Work Index, kWh/ton (metric) 14.00 Specific Energy, kWh/ton 13.57
Feed Size, F80, microns 2900 Net Power Available, kW 0.111
Product Size, P80, microns 75.0 Number of Mills for the Task 1
Total Plant Throughput, ton/hr 0.01 Net kW / Mill 0.111
Total Plant Throughput, kg/hr 8.20
Total Plant Throughput, g/s 2.28
Total Plant Throughput, g/rev 1.91
Total Plant Throughput, g/min 136.74
MILL DIMENSIONS AND OPERATING CONDITIONS :

0.3048
Diameter Length Mill Speed Charge Balls Interstitial Lift
ft ft % Critical Filling,% Filling,% Ore Filling,% Angle, (°)
1.00 1.00 93.55 19.30 19.30 35.00 40.00
L/D rpm
1.00 71.66
Charge Mill Charge Weight, tons Apparent
% Solids in the Mill 100.00 Volume, Ball Ore Density
Ore Density, ton/m3 2.80 m3 Charge Interstitial above Balls ton/m3
Slurry Density, ton/m3 2.80 0.004301 0.020128 0.001686 0.000000 5.1
Balls Density, ton/m3 7.80
Charge Mill Charge Weight, kg Apparent
Volume, Ball Ore Density
m3 Charge Interstitial above Balls ton/m3
0.004301 20.128 1.685904 0.000062 5072
Model outputs - specifically ralating to the bond ball mill
Mill Mill Mill Energy Mill Energy Mill Energy
Power, kW Power, W Whr (Whr) per rev J/rev

0.1027 Balls 102.70 102.70 0.02389 85.99
0.0000 Overfilling 0.00 0.00 0.00000 0.00
0.0086 Ore 8.60 8.60 0.00200 7.20
0.1113 Net Total 111.30 111.30 0.02589 93.19
80.7400 % Losses
0.5779 Gross Total 577.88 577.88 0.13440 483.86
Energy for milling (ore only)% 1.489

Energy for milling (ball and ore)% 19.26
Drive train losses% 80.74
Figure 88 highlights the fact that the energy measured in the Bond test mill equates to
Bonds stated laboratory scale energy consumption. However, the Bond empirical equation
assumes an energy input of 60 J/rev which is the “calibrated” energy requirement that
matched the performance of 2.4m wet grinding mills as measured in Bond original
experiments. Hence Bond effectively built in a scale factor (0.645) into the well known
empirical equations that takes into account the performance of the then common full scale
(2.4 m) diameter mills. The empirical equation as such has a built in correction factor by
assuming the net useful work done in the mill is equivalent to 60 joules/rev. This
observation needs to be considered carefully when measurements are made in scientific
experiments to quantify the specific comminution energy of milling in laboratory scale
units such as the Bond ball mill. However Bond’s stated 93 joules per revolution, which is
defined as the energy input to the mill is identical to the measure energy to the mill as
shown in the linear relationship in Figure 88 and Figure 89.
- 116 -

80 80
Bond @ 93 J/rev
Measured Energy kWhr/tonne
70 70
Linear (Measured Energy kWhr/tonne)
Linear (Bond @ 93 J/rev)
Calculated specific comminutiuon enegy (kWh/t)

Measured specific comminution energy (kWh/t)
(using Bond's (1949) mill energy @ 93 J/rev)

60 60
(using the digital enegy meter)
50 50
40 40
30 30
20 20
10 10
0 0
0 10 20 30 40 50 60
Bond Equation Mill Energy (kWh/t)
Figure 88 show the measured digital energy (kWh/t) – left axis and Bond’s (1949) mill energy of 93 J/rev –
right axis, plotted against the Bond equation net comminution energy for the same data set. This confirms that
the digital energy meter achieves the same result as Bond stated in 1949 more than 55 years ago.
80
y = 0.969x
Bond energy @ 93 J/rev
R2 = 0.9103
70 vs Measured
Linear (Bond energy @
Measured specific comminution energy (kWh/t)
60
(using the digital enegy meter)
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80
Calculated specific comminutiuon enegy (kWh/t)

(using Bond's (1949) mill energy @ 93 J/rev)
Figure 89 shows the specific comminution energy using Bond’s (1949) mill energy versus the measured
specific comminution energy using the digital energy meter (kWh/t)
From the results it may now be confirmed that, using a digital energy meter, direct
measurement of the energy used to comminute the ore is possible. From this, it is further
proposed that a new way of determining the specific comminution energy required to
produce the steady state products in the Bond mill be given. The method is simple, and
involves the direct measurement of the energy required to mill the new feed to 100%
passing the selected closing size at steady state. The gross energy measurement may be
subtracted from the no-load energy measurement to determine the net specific
comminution energy. The specific comminution energy can then be calculated on the
basis of the total net Joules or W.s or Wh at steady state, divided by the mass of the new
- 117 -

feed. From this data it is possible to determine the net energy per revolution of the Bond
ball mill, and the products may then be compared to the energy consumed in other devices
to achieve the same product size or liberated mineral.
The results of all of the methods investigated to determine the energy consumption in the
Bond ball mill are summarised in Table 13. The Morrell (1993) power model example is
given in the Appendix.
In Table 13, 70 rpm* represents ideal conditions of a standardized Bond mill operating at
the required 70 rpm. The 71.6 rpm represents the actual measured mill speed of the Bond
ball mill located at the JKMRC, and is the same unit used in the experimental tests. Any
discrepancies in the measured energy were thought to be influenced by the slight variance
in mill rotational speed, and hence were evaluated in the DEM study by Djordjevic.
Table 13 Summary of the Bond mill power as determined by various methodologies

Bond vs Digital Mill energy per rev Mill power (W) @
Balls Only (20,125 g)
energy meter W.s/rev or J/rev 71.7 r.p.m.
Bond (1961) - estimated mill energy
Bond (Literature) 60 71.7
(Assumes Bond's est. ball energy)
Bond (1949) - useful work on ore
Bond (Literature) 52.3 62.47
(Author assumes ball only energy)
Digital energy meter (balls only)
Digital energy meter 51.1 61.1
-instantaneous power method
Digital energy meter (balls only)
-linear coefficient method Digital energy meter 66.2 79.1
Djordjevic (2004) (Balls only) 70 rpm*
DEM (unpublished) 63.2 73.7*
(sensitive to friction coefficients)
Djordjevic (2004) (Balls only) 71.7 rpm
DEM (unpublished) 63.8 76.1
(sensitive to friction coefficients)
Hogg and Fuerstenau (1972) Power model
86 102.7
(Balls only -component) (text example)
Bond vs Digital Mill energy per rev Mill power (W) @
Balls (20,125 g) + Ore (700 ml)
energy meter W.s/rev or J/rev 71.7 r.p.m.
Bond (1949) - energy input to the mill Bond (Literature) 93 111.1

(Author assumes ball and ore energy)
Digital energy meter (balls + ore)
Digital energy meter 86.8 103.6
-instantaneous power method
Digital energy meter (balls + ore)
-linear coefficient method Digital energy meter 91.44 109.2
Mill energy using various literature Mill energy per rev Mill power (W) @
Method W.s/rev or J/rev 71.7 r.p.m.
models
Levin (1989) -LOW Bond back-calculation 66.3 79.2
Levin (1989) -Ave Bond back-calculation 71.3 85.2
Levin (1989) -HIGH Bond back-calculation 81.2 97
Morrell and Man (1997) - Literature Power model
72.5 86.65
- Levin comparison to Morrell (literature)
Morrell model -bond mill K=1.26 Power model
85.9 102.6
(Balls + ore) (appendix)
Hogg and Fuerstenau (1972) balls + ore Power model
93.19 111.3
(sensitive to lift angle - not measured) (text example)
- 118 -

Summary of results
The specific comminution energy for milling is traditionally based on the Bond method.
The method appears to be based on 60 J/rev, which defines the net energy required in the
Bond ball mill to realise the same conditions in a 2.4 m wet grinding mill. Bond stated
that the net energy input to the mill is 93 J/rev which is comparable to 91.44 J/rev as
measured using the new digital energy meter. This value has been measured and estimated
using various other means such as the Levin method, DEM, and the power models of
Morrell and Fuerstenau. The results are in good agreement, with the exception of the
Levin method which achieves similar results, but for different reasons.
In general each of the mill power models and the energy measurements are in agreement
with the Bond and digital energy meter. This with the exception of Bond’s (1949) paper
comments which stated that the “useful work on ore” was 52.3 J/rev. This is being
interpreted as the “Ball only” component, but could mean the total energy absorbed by the
ore which excludes the losses due to the ball mill itself. It is not understood how Bond
might have measured this or have drawn this conclusion as it is not adequately explained
in the literature.
In addition to this, an exception to the modelled prediction of the balls only energy by
Hogg and Fuerstenau at 86 J/rev could be as a result of the in-accuracies of the model due
to variations in the lift angle as exemplified in Table 12, and again in the Appendix where
lift angles ranging from 30 degrees to 35 to 40 were used under similar conditions. The
resultant mill power was calculated to be 80.0 W, 91.8 W and 102.8 W for the three
respective increasing lift angles. The 30 degree angle is equivalent to 67 J/rev, which is
comparable to the 60 J/rev; however the lift angles in the experiments were never
measured.
Overall the Bond energy for the ball prediction appears to have the greatest variation,
where-as measurements on the ball charge and ore are similar. The slight variations in the
results could be attributed to the experimental Bond mill operating at 71.66 rpm as
opposed to the specified 70 rpm.
The new digital energy meter has confirmed that small electrical energy measurements in
motors/mills are possible, and that the specific comminution energy is measurable. This
device has effectively opened up a new window of comminution research, especially
where energy consumption and energy efficiency are important.
- 119 -

- 120 -

CHAPTER 5
ENERGY EFFICIENCY OF HYBRID HPGR

CIRCUITS COMPARED WITH
CONVENTIONAL ROD AND BALL MILLS
Introduction
In this Chapter, the results of direct electrical energy measurements made during Bond
ball milling and hybrid HPGR/ball mill circuits are presented and analysed. The analysis
includes a comparative analysis of the direct energy measurement results with Bonds
equivalent empirical equation result and again provides a critical assessment of the Bond
methodology including the Bond work index. The results support the hypothesis of
Holmes (1957) which stated that the exponent “n” in Walker’s equation that defines the
Bond, Kick and von Rittinger equations must be a variable associated with the
comminution device. Morrell’s new equation appears to have the same form as the
Holmes equation.
The laboratory scale HPGR tests used the same digital energy meter device that was used
for measuring the specific comminution energy in the ball and rod mills (chapter 4). The
HPGR test results are summarised and presented with an error analysis of the test energy
measurements.
The “energy efficiency” of the comparative circuits or the expected absolute energy
saving that is realised is presented should HPGR be considered in the comminution
circuit. The circuit that revealed the most energy saving was the conventional ball mill
grinding circuit that is compared with the hybrid HPGR/ball mill circuit. This circuit
incorporated HPGR technology in a triple pass (circuit scenario 5) Figure 47.
Chapters 2, 3 and 4 introduce and discuss aspects of two techniques used to determine the
specific comminution energy of the ball mill. To recap, methods are described as:
1. To use Bond’s empirical work index and third theory equation to calculate the ball
mill energy.
2. To use a new method of direct energy measurement through the use of the digital
energy meter.
In the analysis that follows, comparative circuit analysis uses both of the above mentioned
energy measuring techniques. The difference in the net comminution energy required in
the circuits investigated is quantified using both methods. This was done because the
Bond method is commonly used and reported in the literature. In this chapter it is shown
that the two methods produce entirely different results. Each circuit that was tested in
closed circuit with the ball mill to produce three grind sizes are used in the comparative
analysis. The data and product are used to confirm whether enhanced preferential
liberation occurs should HPGR be used in a comminution circuit. This is discussed in
Chapter 6.
- 121 -

The analysis of the test results shows that the HPGR utilises the available electrical energy
delivered to the device more efficiently for the purpose of breakage than the energy
delivered to a ball mill. The overall efficiency of the hybrid HPGR/ball mill circuit was
measured and was shown to be significantly more energy efficient than ball milling.
Bond mill energy test evaluation methodology

The literature review highlighted some of the present and past definitions and research
approaches that have been followed in an effort to understand the particle size / energy
relationships and whether the relationship is dependant on the comminution device. The
energy efficiency debate surrounding different devices and different comminution
processes such as single particle breakage, abrasion and bed breakage or what is
sometimes also referred to as inter-particle breakage complicates the independent study of
the device and the breakage processes. This study was no exception and during the
experimental design it became apparent that the devices and processes needed to be
isolated from one another in order to made relevant comparisons. So to recap, in order to
compare circuits and /or comminution devices such as HPGR and ball mills it was
necessary to:
• Produce the same product size distribution

• Measure net comminution energy in the device (not just the breakage mechanism)
• Compare the Bond third theory equation results to measured energy results.
• Measure the degree of mineral liberation via e.g. the JKTech MLA analysis.
The two methodologies described above were chosen to evaluate in particular the energy
savings that may be realised in absolute terms. This differs to energy savings as a
percentage which often results in confusion as to whether the entire circuit energy is
considered to benefit from the quoted amount or just a device within the circuit.
At this point it is however important to state that the analysis involves net energy
measurements made in laboratory scale ball mills and hybrid HPGR/ball mill circuits. The
sizes of the feed particles used in the study encroach into the size fraction that is clearly
within the full scale “ball milling domain”. Nevertheless the same energy measuring
device is being used to reduce a representative sample of feed to a similar product size via
different flowsheets as described by flowsheet scenarios 1-6 in Chapter 3.
In order to expand the description of the Bond test methodology and explain the finer
details of the locked cycle test within the context of this research, Figure 90 to Figure 93
are used show how a batch locked cycle ball mill test at steady state is used to represent a
continuous ball mill in open circuit or a continuous ball mill at steady state within a
closed circuit. The circuit may be closed with a screen or classifying cyclone. These are
the conditions which plant operator or plant designers aim to achieve or understand.
Whether the ball mill is used in closed-circuit with a cyclone or another classifying device
is really irrelevant, because for the purpose of understanding the overall circuit balance it
is necessary to know the difference between the feed characteristics and the final products
and the energy required to do this.
Figure 90 represents a basic ball mill in open circuit, which at steady state will ensure that
the feed tonnage is equivalent to the product tonnage. Essentially the Bond locked cycle
test aims to characterise the ore in terms of its expected performance in the ball mill by
- 122 -

indexing it, but also provides a methodology of actually determining the energy required
to mill a specified known feed to a specified known product size. In the bond calculations
the feed and product sizes are represented by a single parameter, the F80 and P80, which
gives the reason why the test is valid with the feed and product size distributions are linear
and parallel when plotted on a log-log scale. This is important to note as the test for
parallel feed and product sizes become very important in the analysis of the data in
Chapter 5.
Steady
State
Ball Mill
System
Feed Product
(t/hr) (t/hr)
Feed (t/hr) IN = Product (t/hr) OUT

Figure 90 shows the equivalent of the Bond locked cycle test at steady state.
Figure 91 shows the same mill, albeit a full scale mill or a laboratory scale mill in
continuous operation at steady state. Supposing a thin slice of the feed and product were
obtained, along with the measurement of the feed rate and the mill energy consumption,
then the specific energy of the ore could be calculated. However doing this in practice
often requires the mill to be in steady state, and requires accurate measurement of the
energy consumption and mill feed rate. This is further complicated when researchers are
required to make energy measurements of steady state mill conditions in the laboratory.
Batch tests are inappropriate in these instances, which prompted Bond to develop the
locked cycle test.
Feed Product
(t/hr) (t/hr)
Ball Mill
Mill load
Feed In (t/hr) = Product out (t/hr)
Figure 91 shows an equivalent continuous ball mill at steady state, provided the feed rate is controlled to
provide steady state conditions.
A graphical description of the Bond test is given in Figure 92. The Bond test by definition
achieves steady state conditions when the recirculating load reaches 250% of the mill
charge. This roughly equates to approximately 200 ml of the fresh feed representing the
feed rate at steady state conditions the mass of the product produced is equivalent to 200
ml of product passing a closing size screen. This was previous discussed in the paradox of
the Bond locked cycle test in Chapter 2.
- 123 -

Fresh feed Product, passing
200 ml closing screen
200 ml
“Bond”
Ball Mill
Mill ore load = 500 ml
Figure 92 shows the equivalent of Figure 91, but represented as the Bond ball mill locked cycle test.
The energy to grind the new feed and 250 % circulating load was accurately measured
using the digital energy meter as described in Figure 93 . The measured energy is used to
calculate the specific comminution energy of the laboratory ball mill under the actual dry
grinding batch test conditions. This approach is the same that was used for the energy
measurement calculation in Chapter 4, where an assumed energy consumption per mill
revolutions was 60 J/rev. For the analysis of this work, the actual measured energy is used
in the comparative analysis of the hybrid HPGR ball mill circuits when compared with the
conventional milling circuits.
Feed in + Breakage in Product out + Breakage out
“Bond” Energy input is

the number of
Ball Mill mill revolutions
at steady state
Mill load = 500 ml
Figure 93 shows the locked cycle test at steady state 500 ml of conditioned ore plus 200 ml of fresh feed
totalising 700 ml of ore as a standard, with the energy input represented as the number of mill revolutions.
It should be noted that during the Bond locked cycle test when steady state conditions are
reached, the test bears no significance to the coarse material remaining in the mill with the
ball charge as the calculations do not consider this mass or size distribution. This
graphically represented by the approximately 500 ml of ore as depicted in Figure 94 and
Figure 95. Getting the recirculating load into equilibrium or steady state, requires the
Bond test locked cycles to continue up until at least seven cycles(sometimes less than
seven), at which point, the new feed added to the re-circulating is ground at precisely the
correct number of revolutions (Energy) to produce an equivalent mass of product which
represents 100% passing the closing size screen setting.
- 124 -

At steady state the mass of the new feed equals the mass of the fines generated, and
represents 250% of the total ore charge of the mill. At this point the net grams per
revolution is constant (Gbp), and is the measured locked cycle test parameter that defines
the Bond work index. The standard Bond methodology procedure states that the locked
cycle test reaches equilibrium when the grindability parameter is measured within 3%
error and changes direction. It is in fact possible to reach this condition during the test
without actually reaching mass equilibrium as defined by the 250% recirculating load.
“Bond”
Ball Mill
Mill load = 500 ml
ore + 20125 g balls
Figure 94 shows the composition of the mill load(approximately 500 ml of conditioned ore, where at steady
state the new feed approximately 200 ml representing a circulating load of 250% is added.
The material does however act as a fine particle inventory that is used to assist particles
breaking in and out of various size classes. When the ball mall is considered under these
conditions the interaction between the balls and the ore in inventory play a role in
comminution as represented in Figure 94 as well in a DEM schematic of the Bond ball
mill in Figure 95 (Djordjevic,2004).
Figure 95 shows a DEM representation of the mill charge at steady state, which consists of balls (20125 g)
and 500 ml of conditioned ore. The ore constitutes grinding media and plays a part in the generation of new
surface area in the milled products.
The conditioned ore acts as a permanent inventory of “rock media” which aids in breaking
fine particles between the ball media and fresh feed. This is essentially the same as SAG
milling, but using much finer particle sizes. The significance of this is that if the surface
area of the media is taken into consideration during the “reaction” between the media and
the ore, then the “reaction” could be skewed, since the ore particle surface area is an
integral part of the comminution process. Using surface area to describe comminution
energy has already been proposed by von Rittinger (Rittinger, 1867). This is expanded on
later in the Chapter using theoretical Bond test data to explain the concept.
- 125 -

Comminution energy measurement in HPGR
Data from the HPGR, Bond ball and Bond rod locked cycle tests were used to construct
the flowsheet scenarios as presented in Chapter 3. This resulted in many HPGR tests
being conducted with identical feeds size distributions and with single, double or triple
passes through the HPGR unit. Overall, the individual results of each test were assembled
in such a way to determine the error in the energy measurements of the HPGR. The results
are shown in Table 14. The columns labels HPRC represent the coarse feed (-12.5 mm)
tests, and the column labels HPGR represent the fine feed HPGR tests (-3.35 mm). These
different feed sizes were used in flowsheets that compared the Rod milling results and the
Ball milling results respectively.
Table 14 give a summary of the HPGR specific comminution energy measurements (HPRC
represents coarse feed (-12.5 mm) and HPGR represents fine feed (-3.35 mm).
HPRC HPRC HPRC HPGR HPGR HPGR
Xstrata BHPBilliton Lonmin Xstrata BHPBilliton Lonmin
Test ID kWh/t kWh/t kWh/t kWh/t kWh/t kWh/t
First pass 7a 2.32 0.90 2.94 1.10 1.45 1.34
First pass 8a 2.29 1.00 2.90 1.13 1.67 1.29
First pass 9a 2.39 0.90 2.92 1.08 1.39 1.33
First pass 10a 2.21 0.96 3.00 1.12 1.43 1.28
First pass 11a 2.13 0.98 2.95 1.07 1.38 1.32
First pass 12a 2.16 1.02 3.00 1.08 1.40 1.32
Ave 2.25 0.96 2.95 1.10 1.45 1.32
Std Dev. 0.10 0.05 0.04 0.02 0.11 0.02
Seond pass 7b 1.64 0.71 1.61 0.90 1.35 1.13

Seond pass 8b 1.58 0.69 1.58 0.93 1.30 1.15
Seond pass 9b 1.58 0.70 1.57 0.87 1.42 1.16
Seond pass 10b 1.56 0.62 1.52 0.88 1.45 1.16
Seond pass 11b 1.50 0.69 1.54 0.86 1.43 1.09
Seond pass 12b 1.56 0.60 1.49 0.87 1.43 1.12
Ave 1.57 0.67 1.55 0.88 1.40 1.14
Std Dev. 0.05 0.05 0.04 0.02 0.06 0.03
Third pass 10c 1.03 0.59 1.23 0.70 1.36 0.96

Third pass 11c 0.97 0.62 1.15 0.65 1.38 0.84
Third pass 12c 0.99 0.60 0.93 0.72 1.40 0.91
Ave 1.00 0.60 1.10 0.69 1.38 0.91
Std Dev. 0.03 0.01 0.16 0.03 0.02 0.06
Cumulative
energy for three 4.82 2.23 5.61 2.67 4.23 3.36
passes
The results show that the energy measurement is repeatable with a very low error which
ranged from 1.66 % to 14.5 %, but most of the errors were low, with two high results of
7.58 % and 14.5%. It is interesting to see the variation in the totaled energy results for the
three different ores. The HPGR was set up in the same way for most tests, variations are
as a result of the different ore characteristics as opposed to the variation in the HPGR unit
set-up conditions for the coarse feed and fine feed tests. Another interesting trend
observed was that in each case the comminution energy intensity decreased after each
pass, which suggests that as the feed gets finer, the ability of the particle to absorb energy
for comminution decreases, which could be as a result of the greater new particle surface
area that is being generated. This concept is considered in more detail later in this
Chapter.
- 126 -

Bond rod mill circuit comparison test results
Duplicate Rod mill tests were conducted using the Worsley Bauxite samples. The results
are shown in Table 15 and Figure 96. The error in the rod mill work index with a
standard closing screen size of 1.18 mm is small. However, the work index for tests
completed at 3.35 mm was higher at 4.08%. The 3.35 mm closing screen was specifically
chosen because it would allow for the measurement of the energy required to comminute
the ore from -12.5 mm down to -3.35 mm, which is the feed size for the Bond ball mill.
The ball mill work index result would be used to determine the energy required to
comminute the ore from -3.35 mm to the closing screen sizes of 300 Om, 202 Om and 150
Om (presented later).
Table 15 gives the results of duplicate Bond rod mill work index test results for the
Worsley bauxite ore.
Rod mill Rod mill

work work
index index
Closing
screen 3.35 1.18
(mm)
Test 1 17.9 12.0
Test 2 16.9 12.1
Ave 17.4 12.1
Std dev 0.71 0.07
Bond rod tests are normally completed with a 100% circulating load using a 1.18 mm
screen. Rarely is the closing screen size changed. As mentioned above, a closing screen of
3.35 mm was also tested. The results in Table 15 show that the duplicate tests gave very
similar results, whereas the results of the closing screen test resulted in a work index
variation from 12.1 kWh/t for the 1.18mm closing screen to 17.4 kWh/t for the 3.35 mm
closing screen.
By definition the work index should remain constant for an ore, regardless of the closing
screen. This is on the assumption that the ore strength and micro structural characteristics
are the same. For these bauxite ore tests that result is different, and is possibly caused by
the strength of larger naturally sized minerals within the ore. This type of work index
variation was observed when a platinum ore was subjected to bond ball mill work index
tests in which the closing screen size was changed from 300 Om to 45 Om as shown in
Table 16.
- 127 -

Table 16 shows the variation in Bond work index test results when the closing screen is
changed while treating a platinum ore. (Shi, Lambert and Daniel, 2006).
Conventional Bond ball Closing Bond WI Specific

mill work index test screen energy
material pre-treatment (µm) (kWh/t) (kWh/t)
300 24.0 16.2
150 24.1 25.6
HPGR crushing 106 24.4 32.8
75 29.4 55.5
45 25.7 53.7
300 28.2 23.6
150 26.5 33.7
Conventional crushing 106 27.0 40.0
75 30.2 63.0
45 30.2 70.0
100
Product (1) - 3.35 mm
FEED (2) closing at -3.35 mm

Cum % Passing
FEED (1)closing at -3.35 mm
Product (1) -3.35 mm
10
0.1 1 10
Size (mm)
Figure 96 shows similar and parallel product size distributions of the duplicate Bond rod mill tests.
The energy of the Bond rod mill tests and the Bond ball mill test could then be totaled to
determine the total energy to get from -12.5 mm to the final grind size allocated. The
parallel representation of the Bond rod mill test results indicate that the third theory
equation does apply and that the feed and product size distributions is represented as
single parameters F80 and P80. This requirement for testing the validity of the Bond
methodology is considered for all subsequent test analysis.
- 128 -

Comminution circuit scenarios comparison results
This section discusses the results of the six previously defined comparative comminution
circuits Figure 43 to Figure 48 that were presented in Chapter 3. Out of these Circuit
scenario 4 proved to provide most of the comparative data that makes up the bulk of this
Chapter. The comparative tables presented later use the comparative tools of the Bond
third theory equation for energy as well at the directly measured energy for the sake of a
comprehensive comparative analysis.
Evaluation of circuit scenarios 1 and 4 of direct unit process comparisons.

Circuit scenario 1 (Figure 43) and circuit scenario 4 (Figure 46) were problematic, as it
was difficult to achieve a steady state condition and quantified re-circulating load. The
recirculating load needs to be known as well as the amount of fresh feed to the circuit.
These are used with the measured energy to determine the specific comminution energy.
This was not possible when completing tests in an HPGR in closed circuit with a screen
although some authors Johnson et al 1995 have previously claimed to have achieved
steady state in HPGR tests, but this has not been verified.
At the Randol Pacific Metallurgical Forum, Johnson (2005) again re-presented and
elevated the HPGR power efficiency debate of 1995, and stated that a 50% power saving
could be achieved using a HPGR in a closed-circuit configuration. The energy saving is
measured against the energy needed to achieve the same result using conventional
comminution devices such as ball mills. The energy required for the ball mill was
estimated using the Bond’s equation and the Bond work index value for the particular ore
in question. The stated 50% power saving is based on a single locked cycle HPGR test
which was not analysed in a statistical sense; Johnson confirmed at the Randol forum that
his test work did not actually obtain steady state. High experimental errors are possible
due to the oversize mass fractions being measured in weight and not balanced against
applied and controlled energy as is the case in the Bond locked cycle test. The use of the
Bond equation is an indicator of the desired result. It is based on a long standing equation
with no actual measurements being made other than the Bond work index, which is
determined by an empirical equation. The validity of the 50% is subject to much doubt
without a comprehensive experimental program.
It was found that when conducting locked cycle tests in an HPGR the energy input to a
known quantity of ore could not be adjusted as is the case in the locked cycle bond ball
mill tests. In the locked cycle Bond ball mill test, the number of mill revolutions (energy
input) required at the end of each test cycle is adjusted so that a steady state condition is
reached and evaluated at a standard re-circulating load of 250%.
A procedure for a locked cycle test for HPGR processing with precise mass balancing and
a firm circulating load is not yet defined as a standard and is not described in the
literature. The manufacturers claim to routinely use such a test, but the recirculating load
is not controlled in proportion to the applied energy. Hence obtaining steady state
conditions are rarely achieved. In comparison to the Bond test which is a precision tests,
much more research is required to define a standard test procedure for the HPGR using an
appropriately sized particle size distribution.
- 129 -

This limitation which was identified during the study was, if developed as a standard test
was seen to widen the scope of this study rather than focus it and was therefore not
pursued in any great detail. An outcome of the tests showed that the feed size distribution
changes after each cycle, and stabilizes to a limiting size distribution. Since the closed
circuit resembles a single pass unit, the first pass results collectively describe the variation
in the products passed either through a single pass or the results of the attempted locked
cycle test where the feed size distribution changes. The results of these HPGR tests using
the coarse feed Mt Isa lead zinc ore (-12.5 mm), Worsley bauxite (-12.5 mm) and Lonmin
ore (-12.5 mm) are given in Figure 97, Figure 98 and Figure 99 respectively. In each case
the narrow error of the repeatability tests in the 2nd and 3rd pass circuit products are
shown. Bearing in mind that the first pass product size distribution resembles the mill
output and not re-circulating load and final circuit product. This again demonstrates the
difficulty of conducting an HPGR locked cycle test. This initial attempt of comparing
circuit scenarios 1 and 4 nevertheless highlight the usefulness of the methodology as
determined in circuit scenarios 2 and 5. Circuit scenario 2 had shortcomings which are
described in the next section which leaves circuit scenario 3, 5 and 6 as the remaining
circuits to be analysed.
Circuit scenario 3 is similar to circuit scenario 6. The coarse feed in circuit 3 (HPRC –F1,
F2 and F3) and the fine feed in circuit scenario 6 (HPRG – F1,F2 and F3) results in
products (HPRC P1-3 and HPGR P1-3) that is similar after three passes using either the
finer feed or the coarser feed. These results are summarized for the Lonmin and Mt Isa
ores in Figure 100 and Figure 101 respectively.
From these results it would appear that product fineness controls the overall breakage and
energy required for breakage in the small laboratory scale unit. Closer examination reveals
that the energy breakage profiles differ from each ore type and process condition, as there
is no energy trend in Table 14 that follows the characteristics of the products. This too
was observed during the completion of the tests, where the “bulk” properties of the
processed ore appeared to influence the outcome of the test in terms of the smoothness of
the test run. Smoothness referring to the sound that the ore makes as it passes through the
rolls. The sound may be described as either quiet, gritty or crunching which also appears
to affect the quality and competency of the flake product. Other process observations such
as varied working gaps as well as fluctuations (steady or oscillating) in the oil pressure
gauges that measure working pressure. Relatively high total energy consumptions were
observed in the coarse feed tests after three passes. This as compared with conventional
rod milling masked the energy benefits of HPGR and so it is assumed that some of the
tests were carried out well above the energy saturation points. The energy level during
testing at which no further comminution benefit is realized with increased energy.
- 130 -

100
Average Feed
90
Feed circ load 1
Feed circ load 2

80
Ave prod 1 st pass
Ave prod 2 nd pass

Cumulative % Passing
70
Ave prod 3 rd pass
60
50
40
30
20
10
0
0.1 1 10
Particle Size (mm)
Figure 97 shows the change in feed size distribution and products when circuit scenario 1 was evaluated in
comparison to the products after two and three passes (Mt Isa Ore).
100
Average Feed
90
Feed circ load 1
Feed circ load 2

80
Ave prod 1 st pass
Ave prod 2 nd pass

70
Ave prod 3 rd pass
60
50
40
30
20
10
0
0.01 0.1 1 10
Particle Size (mm)

comparison to the products after two and three passes (Worsley Bauxite ore).
An important aspect of circuit scenario 1 and 4 are that the products are similar to the
single pass open circuit products however the total net energy imparted on the ore is
different as the process rates of the unit in relation to the fresh feed are different. The
- 131 -

control of the unit at steady state is subjective and is not currently defined as a precision
test. Interestingly the multiple passes in the case of the Bauxite ore (Figure 98) did not
respond that well to the HPGR in comparison to the lead zinc ore and platinum ore
(Figures 97 and 99) suggestion that not all ore types are suitable for HPGR processing.
100
Average Feed
90
Feed circ load 1

80
Feed circ load 2

70
Ave prod 1 st pass

60
Ave prod 2 nd pass

50
Ave prod 3 rd pass
40
30
20
10
0
0.01 0.1 1 10
Particle Size (mm)

comparison to the unique product size distribution shape after two and three passes with the Lonmin ore.
100 HPGR-F1 Lonmin

Platimun
HPGR-F2 Lonmin
90
Platimun
HPGR-F3 Lonmin
80 Platimun
HPGR P- T10 Lonmin
Platimun
Cumulative % passing
70
HPGR P- T11 Lonmin
Platimun
60
HPGR P- T12 Lonmin
Platimun
50
HPRC-F1 Lonmin
Platimun
40 HPRC-F2 Lonmin
Platimun
30 HPRC-F3 Lonmin
Platimun
HPRC P- T10 Lonmin
20 Platimun
HPRC P- T11 Lonmin
10 Platimun
HPRC P- T12 Lonmin
0 Platimun
0.1 1.0 10.0
Particle Size (mm)
Figure 100 shows the similar products of the three pass tests for both the HPGR(coarse feed) and HPGR(fine
feed) tests that represent circuit scenarios 2 and 5 (Lonmin ore).
- 132 -

100
HPGR-F1 Xstrata-Mt Isa
90
HPGR-F2 Xstrata-Mt Isa
80 HPGR-F3 Xstrata-Mt Isa
HPGR P- T10 Xstrata-Mt

Cumulative % passing
70 Isa
60 Isa
Isa
50
HPRC-F1 Xstrata-Mt Isa
40 HPRC-F2 Xstrata-Mt Isa
30 HPRC-F3 Xstrata-Mt Isa
HPRC P- T10 Xstrata-Mt

20 Isa
10 Isa
Isa
0
0.10 1.00 10.00
Particle Size (mm)

Figure 101 shows the similar products of the three pass tests for both the HPGR(coarse feed) and HPGR(fine
feed) tests that represent circuit scenarios 2 and 5 (Xstrata Mt Isa ore).
For all tests, the products were adequately screened so that they were representative of the
HPGR process. Since all samples were de-agglomerated using a standardised time and
screen set using the Gilson apparatus, the possibility of agglomerated flake particles in the
product was minimised. Flake character differed, and could not be predicted, for example
the Bauxite sample readily produced flake products.
Evaluation of circuit scenarios 2 and 3, HPGR and rod milling.

Initially the intention was to evaluate HPGR within a hybrid rod mill circuit with that of
the rod mill alone. However it became apparent that the products of the triple pass HPGR
alone were finer than the products produced in a rod mill locked cycle test, and as such
the test using the products of the triple pass HPGR for the Bond rod mill test that followed
seemed unwarranted. For this reason circuit scenario 2 in Figure 44 could not be
completed as planned, however circuit scenario 3 in Figure 45 could be directly compared
since the products were somewhat similar (Figure 102). The measured energy input and
Bond Rod mill work index comparison method (Table 17) is similar to the analysis of
Chapter 4 for the Bond ball mill. However for the closing screen of 3.35 mm the energy
determined by the Bond method and direct energy measuring method are similar but differ
somewhat when the standard closing screen of 1.18 mm is used. This result trends the 60
Joules/rev approximation result for the Bond ball mill and 90 Joules/rev for the measured
energy of the in the Bond ball mill.
The measured energy of the rod mill at steady state is on average 3.5 kWh/t at a closing
screen of 1.18 mm, this is in comparison to the 2.3 kWh/t (Table 18) supposedly at steady
state and producing a product at a closing screen of 1.18 mm. However this comparison is
prone to error as the methodology has not as yet been verified as previously pointed out.
The tests results represent a closing screen of 3.35 mm which represent the cross over
particle size from Rod mill testing to ball mill testing.
- 133 -

100
Rod mill prod (1) Close
90 screen 3.35 mm
80 Rod mill feed (2) close

screen 3.35 mm
70
Cum % Passing
60 Rod mill feed (1) Close

screen 3.35 mm
50
Rod mill prod. (2) close
40 screen 3.35 mm
30 Average 3 pass HPRC

product
20
10 HPRC feed
0
0.01 0.1 1 10
Particle size (mm)
Figure 102 shows how the products of the three pass HPGR were finer than the products of the rod mill,
resulting in circuit scenario 2 to not be evaluated as planned.
Table 17 give the energy comparisons of using the Bond equation and Rod mill work
index in comparison to direct energy measurements made on the Rod mill.
Bond Equation Measured

Test Description (Rod mill baseline) Bond Rod mill locked (Rod Mill) specific
Rod mill work index test Rod mill cycle test specific comminution
work closing screen comminution energy (Digital
Test ID index size (mm) energy (kWh/t) meter kWh/t)
Rod mill circuit (Worsley bauxite) 15/16 12.0 1.18 2.7 3.4
Rod mill circuit (Worsley bauxite) 14/16 12.1 1.18 2.7 3.6
Rod mill circuit (Worsley bauxite)* 15/16 17.9 3.35 1.4 1.3
Rod mill circuit (Worsley bauxite)* 14/16 16.9 3.35 1.4 1.3
Table 18 give the energy comparisons of using HPGR in a locked cycle arrangement
against the multiple pass scenarios.
HPRC
locked Measured
Test Description (HPRC baseline) Multiple cycle test specific
pass circulating load test closing comminution
screen energy (Digital
Test ID size (mm) meter kWh/t)
HPRC closed circuit (Worsley bauxite) 6 1.18 2.38
1st pass HPRC - cumulative 4 1.18 0.98
2nd pass HPRC - cumulative 7,8,9 1.18 1.65
3rd pass HPRC - cumulative 10,11,12 1.18 2.25
Evaluation of circuit scenario 5 hybrid HPGR/ball mill vs. conventional ball milling.
The remainder of the analysis in this chapter focuses on the results of the test work of
circuit scenario 5 as shown again in Figure 103 below, where the hybrid HPGR grinding
circuit is compared with conventional ball milling. Ball milling is the most energy
intensive comminution process and consumes grinding media which not only evaluated in
terms of the “dollar” cost but also in terms of “energy” cost. The hybrid HPGR circuit
thus aims to reduce the energy required for comminution and also significantly reduce the
- 134 -

amount of grinding media required to achieve a targeted grind size. Previous results by
Shi, Lambert and Daniel (2006) showed that limited energy benefits could be realised
with the use of a single HPGR in the hybrid circuit. The triple pass HPGR option was
chosen since the comminution benefits appeared to improve as the number of passes
increased. Particle breakage or comminution benefit however appeared to diminish after
three passes and coincided with a corresponding decrease in the amount of energy
absorbed by particles. The decrease in specific comminution energy in the HPGR when
treating finer feed sizes as displayed in the results in Table 14 is probably due to the
much greater new particle surface required to be generated per tonne of fine feed material.
It is envisaged that when a full scale HPGR unit are used to treat ever increasing finer
material the energy made available for breakage is limited resulting in what appears to be
minimal comminution benefit. These observations highlight the possibility of reverse
scale-up effect that could possibly influence the comminution observed in a laboratory
scale unit in relation to a full scale unit treating fine feed material. This reverse effect is
also possibly caused in a non-liner relationship behavior observed in scale-up studies that
measure ratios between particle size, working gap and roll diameter. These non-linear
observations were previously reported in a HPGR model verification and scale-up study
completed by the Author in 2003 (Daniel 2003).
Feed
(t/hr)
Baseline Circuit Hybrid HPGR Circuit
HPGR
Product
Feed (t/hr)
(t/hr)
Ball Mill
Classifying
cyclone HPGR
Product
HPGR
Ball Mill
Classifying
cyclone
Re-circulating load
Figure 103 shows the flowsheets of the two circuits that were compared as part of this study.
The comparison of the energy efficiency or the measure of the absolute energy consumed
and saved in each of the two circuits in Figure 103 are based on the comparative analyses
of both the Bond equation methodology as well as the new direct energy measurement
methodology. The direct energy measurements in the HPGR were achieved through the
use of the newly acquired digital energy meter as described in Chapter 4. In this case the
standard Bond locked cycle tests were completed using the product of the three pass
HPGR process. Similarly, the Bond locked cycle test was used with the original feed for
the determination of the baseline energy requirement (i.e. baseline circuit energy
requirement) as shown by the circuit on the left hand side of Figure 103.
For the purpose of completing the HGPR hybrid circuit analysis (right hand circuit of
Figure 103), triplicate HPGR tests were conducted after each pass. Energy measurements
for the first pass, second pass and third pass cycles in the HPGR were then averaged and
are summarised in Table 14. The reason for doing this was to have confidence in the
HPGR energy measurement, which would then be followed by direct comparable Bond
test data and energy measurements. These summarised energy measurements are tabled in
- 135 -

the results as single entry specific comminution energy values. Figure 104 shows the
results of the product size distributions for the lead/zinc ore which was treated in the
HPGR. The error bars in Figure 104 represent the variability of the product size
distributions for each of the tests, where as Table 14 show the error in the energy
measurement.
It is interesting to note that the results presented in Figure 104 show that after each pass,
the resultant product size distributions were significantly finer. This is in contrast with the
usual HPGR product size distributions that are produced in a single pass with increasing
pressure and energy input. The HPGR has a well known characteristic of producing
product size distributions which remain largely unchanged when operating pressures and
energy are increased. The energy level at which this occurs is commonly known as the
energy saturation point (Daniel, 2003).
100
90
80
70
60
Average Feed
50
Ave prod 1 st pass
40
Ave prod 2 nd pass
30 Ave prod 3 rd pass
20
10
0
0.01 0.1 1 10
Particle Size (mm)
Figure 104 shows the product size distributions of single, double and triple pass HPGR processing for the
lead/zinc ore, and the confidence at which these products were obtained through triplicate tests.
The remaining HPGR results are summarized in Figure 105, Figure 108 and Figure 111
for all the ore types under investigation, namely the lead/zinc ore, the bauxite ore and the
platinum/chrome ore. In these Figures the product size distribution of the third pass
through the HPGR is presented in the graph which then constitutes the feed to the ball
mill in the hybrid HPGR circuit.
Description of how the results are presented.

In the analysis of the results the three different ore samples tested, a set pattern of tables
are used to explain the results. In total there are six tables per ore type presented. The first
two tables show how the comminution energy may be determined using Bond work
indices and the third theory equation (first table) and the method of using 60 J/rev (second
table). The results of the first table and second table were found to be very similar
(confirming the findings of Chapter 4).
The fourth and fifth tables present the energy measurements as determined by the digital
energy meter. The fourth table considers only the coarse fraction within the feed that is
- 136 -

milled, whereas the fifth table presents the whole mass fraction of the new feed at steady
state conditions. These results are also similar, but the in the fourth table, the energy
consumption represents the actual milling process within the locked cycle grinding tests.
The comparison of the energy of the hybrid HPGR circuit against the conventional circuit
energy (energy efficiency -fifth table) using the Bond results of the first and second tables
were found to produce negative energy efficiency. By following the Bond energy
calculation methods, erroneous results and meaningless numbers confirm why the Bond
method should be used with extreme caution.
The results of the fourth table that represent the grinding process are then used in the sixth
table to show the direct energy saving that was measured. The table provides a summary
of the energy efficiency of the circuits as compared at different grind sizes. This system of
table presentation provides a template or a sequence that describes how the results are
presented for the three ore types under investigation. As such the results of each ore are
presented in the following three sections.
Results of the Mt Isa ore.

Figure 105 summarises the results for the lead/zinc ore, where the feed to the baseline and
hybrid HPGR circuits is presented along with the product size distributions of both
circuits. The triple pass HPGR product that makes up the feed to the ball mill in the
hybrid HPGR circuit is also shown. For both the baseline circuit and the HPGR hybrid
circuit the products of the ball mill tests were obtained at closing sizes of 300 Om 150 Om
and 106 Om. The objective of comparing these two circuits in this fashion is to ensure that
the products produced by the two circuits were identical.
The total energy to achieve these individual product size distributions is determined by
two different methods, namely the well known Bond third theory method and direct
energy measurement (digital energy meter). The digital energy meter measurements were
previously investigated (Daniel, 2004), and found to be comparable to what Bond had
originally intended the locked cycle test to be used for. In order to utilise the Bond
methodology it is imperative to ensure that the feed and product size distributions of the
Bond test are parallel when plotted on a log-log scale.
Figure 106 presents the lead/zinc feed and product size distributions and it may be
concluded based on the visual appearance of the graphs that the Bond methodology may
be applicable in this case.
For the lead/zinc ore, the measured energy results are presented in Figure 107. The two
coefficients used are 0.1335 and 0.1077. The difference between these two coefficients
when converted to Watt second per revolution or joules per revolution amounts to an
average of 93 J/rev.
- 137 -

100 Product of ball
mill 150 um
90
Product of ball
mill 300 um
80
Feed to conv.
70 ball mill
Cum % Passing
60 Product of ball
mill 106 um
50
Feed to ball
mill, HPGR
40 circuit
Product HPGR
30 circuit 150 um
20 Product of
HPGR circuit
300 um
10 Product of
HPGR circuit
0 106 um
0.01 0.1 1 10
Size (mm)
Figure 105 show the summarised results of the lead/zinc ore, and includes the feed size distributions to both
circuits, with the intermediate product of the triple pass HPGR circuit making up the feed to the ball mill in the
hybrid HPGR-ball mill circuit.
100
Feed-150um
Cum % Passing
Product 150 um
10
Feed 300 um
Product 300 um
Feed 106 um
Product 106 um
1
0.01 0.1 1 10
Size (mm)
Figure 106 shows a log-log plot of the feed and product size distributions of the lead zinc ore, as observed
during the Bond ball mill tests.
- 138 -

80
y = 0.1268x
empty mill 2
R = 0.9981
70 mill+balls only (20 125 g) y = 0.1335x
2
R = 0.9989
Mt Isa ore +ball
60
Linear (empty mill)
Linear (mill+balls only (20 125 g))
50
Energy (Wh)
Linear (Mt Isa ore +ball)
40 y = 0.1077x
2
R = 0.9978
30
20
10
0
0 100 200 300 400 500 600
mill revolutions
Figure 107 shows the measured mill energy under conditions of no-load, balls, and balls and ore. The linear
coefficients are used to determine the net energy per mill revolution for the lead zinc ore.
Table 19 summarises the total net comminution energy required for the conventional ball
mill circuit compared with the high-pressure grinding rolls hybrid circuit, when using the
Bond third theory equation. As shown in Table 20, the Bond energy may be calculated
using the steady state mill revolutions and an estimated 60 joules per revolution.
The net energy required to produce the new product using this method produces very
similar results to that which is produced using the Bond third theory equation. This
method was presented previously by Daniel (2004).
Table 19 gives the standard BWI and third theory equation energy for the lead /zinc ore.
Bond Equation - Calculated Energy

Net
Grind size Standard Comminution
100% Bond test energy
passing Bond test P80 Bond test F80 WI Bond Eqn
Ore Pre-treatment (Bm) (Bm) (Bm) (kWh/t) (kWh/t)
Conventional ball milling 300 214 2284 17.2 8.16
HPGR+ ball milling 300 214 1106 15.1 5.78
- 139 -

Table 20 gives the equivalent third equation energy of the lead/ zinc ore when calculated
using a mill energy input of 60 J/rev.
Bond Energy - Calculated @ 60 J/rev

Energy at Effective
Grind size Steady State Measured Bond coarse feed Net Energy
100% Bond mill Bond test subjected to req. to produce
passing revolutions at literature (grindability) milling (g) new product
Ore Pre-treatment (Bm) steady state 60 J/rev g/rev (grindability) kWh/t
Conventional ball milling 300 231 13860 1.8 416 9.3
HPGR+ ball milling 300 94 5640 2.7 258 6.1
The results of Tables 22 and 21 are not comparable as the results represent the Bond mill
equilibrium contradiction as described in Chapter 2 (Page 25). The mill, during the Bond
locked cycle test physically does the work of the flowsheet described on the left hand side
of Figure 8, where as the Bond empirical equations follow the flowsheet on the right hand
side of Figure 8. As a consequence, direct energy measurement results should be based on
the results of Table 22 (fine material in feed included) and not Table 21 (fine material in
feed excluded – phantom cyclone). This pattern is followed throughout in the analysis of
the other samples evaluated in this manner.
Table 21 presents the same results based on direct energy measurements using the digital
energy meter. This is achieved by multiplying the Bond ball mill steady state revolutions
by the average measured Joules per revolution. The average measured energy
consumption per mill revolution is determined from the difference between the coefficient
of the ball mill energy which includes the ore and the balls to that of the mill running
empty (From Figure 107). Assuming that the locked cycle test has reached steady state,
this implies that the new feed (excluding fines) which is measured has a mass of 413 g,
465 g, and 4476g as shown in Table 21, representing the right hand flowsheet in Figure 8.
Whereas in Table 22, and on the basis that the locked cycle test has reached steady state,
implies that the new feed (including fines fraction) which is measured has a mass of 492
g, 494 g, and 493 g as shown in Table 22, representing the left hand flowsheet in Figure 8.
The discrepancy between the Bond test methodology and direct measurement of the
locked cycle test has been known to exist as reported in the literature. Since the object of
this test work was to directly measure the energy to produce identical products, it would
seem reasonable to accept the energy comparisons of Table 22 and not 21 nor the results
of the method which uses the Bond equations. These results showcase the Bond mill
equilibrium contradiction as described in Chapter 2. This pattern is followed throughout
in the analysis of the other samples.
- 140 -

Table 21 gives the calculated energy for the lead/zinc ore based on the measured energy
per mill revolution and the effective coarse fraction of the new feed that is subjected to
milling.
Digital energy meter - Measured Energy

Net
Grind size comminution Effective Net Energy
100% Bond mill energy per Energy at coarse feed req. to produce
passing revolutions at rev (Ws/rev)or Steady State subjected to new product
Ore Pre-treatment (Bm) steady state J/rev Ws milling (g) kWh/t
Conventional ball milling 300 231 93 21455 413 14.4
HPGR+ ball milling 300 94 93 8731 266 9.1
Table 22 gives the calculated energy required for the lead/zinc ore based on the measured
energy per mill revolution and the effective new feed as determined at steady state when
the fraction of new feed equals the 250% recycle fraction
Digital energy meter - Energy based on new feed

Measured
Grind size Energy Energy Net energy kWh/t
100% No load Full load comminution Effective Mill based on new
passing Empty mill (Balls + ore) energy per rev NEW feed rate feed at steady
Ore Pre-treatment (Bm) (Wh) (Wh) (Wh/rev) (g) state
Conventional ball milling 300 24.9 30.8 0.02580 492 12.1
HPGR+ ball milling 300 10.1 12.5 0.02580 475 5.1
An aspect of the direct measurement technique is that it is used to consider the “real”
measured energy that is used to grind the coarse material. This is the portion of the new
feed that is greater than the closing size. Alternatively the energy consumed should be
considered as the energy required to grind the new feed at steady state. In the Bond
locked cycle test this is equivalent to the 250% recirculation load. If the Bond
methodology is used then as the closing size is changed, then the mill effectively has more
material to grind (as the closing size becomes finer) which is then included into the
calculations. An example of this as previously described was shown in Table 21, where
the effective coarse feed subjected to milling changes from 413 g to 465 g to 476 g when
the closing size decreases from 300 Om to 150 Om to 106 Om.
Table 23 summarises the comparison between the two circuits using the conventional
Bond work index and Bond third theory equation. In this particular instance it appears that
with closing screen sizes of 300 Om and 150 Om the conventional ball milling circuit is
more efficient than the hybrid HPGR/ball mill circuit. There are a number of factors
which may have contributed to these results.
- 141 -

• Firstly the feed and product size distributions of the Bond tests (Figure 106) can be
viewed as not being parallel and hence it may be questioned whether the Bond
methodology was valid in this instance even though it was initially accepted as
being valid.
• Secondly as described above in the previous paragraph the Bond methodology
looks at the coarse portion of the feed which is used in the calculation to determine
the grind ability of the ore and hence the Bond work index.
• Thirdly it is possible that the HPGR tests were conducted under conditions where
excessive pressing force was used at the expense of much larger quantities of
energy. Under these circumstances it is quite possible that hybrid HPGR/ball mill
grinding circuits are less efficient than conventional grinding circuits. The
tabulated discrepancies in the application of the Bond methodology have resulted
in the preferred use of the direct energy measurement technique.
Table 23 gives the calculated energy comparison for the lead zinc ore between the two
circuits, determined through the use of the Bond third theory equation and the Bond work
index as measured during the Bond locked cycle test.
Bond Eqn. energy circuit comparison

HPGR Ball mill
Grind size Component Component Total circuit Circuit
100% (Net energy) (Net energy) energy energy
passing (3 passes) (Measured) measured saved % Circuit
Ore Pre-treatment (Bm) kWh/t kWh/t (kWh/t) (kWh/t) energy saved
Conventional ball milling 300 0.00 8.16 8.16 - -
HPGR+ ball milling 300 2.67 5.78 8.45 -0.3 -3.5
HPGR+ ball milling 106 2.67 11.58 14.25 0.2 1.6
Table 24 compares the measured energy consumptions between the baseline circuit
(conventional ball milling) and the hybrid HPGR/ball milling circuit. Table 24 includes
the measured energy consumption of the ore when subjected to three passes through the
HPGR. The ball mill energy components, as described above and in Table 22 and Table
23, are included. Table 24 includes a column which provides the absolute circuit energy
saved in kilowatt hours per tonne.
Table 24 gives the measured energy comparison for the conventional and hybrid
HPGR/ball mill circuits for the lead/zinc ore, based on the measured energy per mill
revolution and the effective new feed as determined at steady state when the fraction of
new feed equals the 250% recycle fraction.
Measured energy circuit comparison

HPGR Ball mill
The final column of Table 24 calculates the percentage of the energy saved between the
two circuits. The measured energy savings for the hybrid HPGR/ball mill circuit is
- 142 -

significant with 35.8%, 26.1% and 27.8% being achieved for a 100% passing a closed
screen size of 300 Om, 150 Om and 106 Om respectively. It should be noted that these
results represent dry grinding of the lead/zinc ore in laboratory scale conditions.
Results of the Worsley Bauxite ore

The results of the second ore tested, the bauxite ore, are presented in Figure 108, Figure
109 and Figure 110. This ore was tested under identical conditions to that of the lead/zinc
ore. Once again almost identical product size distributions were produced at the closing
screen size of 425 Om, 300 Om and 212 Om.
100
Product 425 um
90
Product 300 um
80
70 Feed 212 um
Cum % Passing
60 Product 212 um
50
Product of HPGR
Circuit
40
Product HPGR
30 circuit 425 um
20 Product of HPGR
circuit 300 um
10 Product of HPGR
circuit 212 um
0
0.01 0.1 1 10
Size (mm)
Figure 108 shows the summarised results of the bauxite ore, and includes the feed size distributions to both
circuits, with the intermediate product of the triple pass HPGR circuit making up the feed to the ball mill in the
hybrid HPGR-ball mill circuit.
One interesting observation that is presented in Figure 108 is that the product size
distribution of the bauxite having passed through three cycles of the HPGR was similar in
the fine size fraction region to that of the original feed. This suggests that during the
HPGR processing, coarser particles were selected preferentially for breakage.
The term “selective breakage” is specifically used when particles of a certain size are
either chosen or not chosen for breakage. Rod mills are traditionally known to create
conditions conducive to the process of selective breakage. The observed shape of the
bauxite size distribution suggests that selective breakage is taking place, and is discussed
further in Chapter 6. Figure 109 shows the feed and product size distributions of the
bauxite ore on a log-log scale. Visually the feed and product sizes are virtually parallel
and therefore the Bond methodology should be applicable for this ore type.
Table 25 and Table 26 compare the energy results of the ball mill locked cycle test and
the hybrid HPGR/ball mill circuit in which the Bond third theory is applied. The results
using the Bond equation (Table 25), and 60 joules per revolution (Table 26) are very
comparable. As mentioned before the direct energy measurement with the effective new
mill feed rate should be used for the overall circuit comparison.
- 143 -

100
Cum % Passing Feed-425um
Product 425 um
10
Feed 300 um
Product 300 um
Feed 212 um
Product 212 um
1
0.01 0.1 1 10
Size (mm)
Figure 109 shows a log-log plot of the feed and product size distributions of the bauxite ore, as observed
during the Bond ball mill tests.
Table 25 gives the standard BWI and third theory equation energy for the bauxite ore.

Net
Table 26 gives the equivalent third equation energy of the bauxite ore calculated using a
mill energy input of 60 J/rev.

Energy at Effective
Table 27 shows the net comminution energy per revolution which was determined using
the difference between the linear coefficients as shown in Figure 110 (difference between
0.1374 and 0.1077). The energy per revolution is calculated to be 107 J/rev after the
appropriate unit conversions have been made.
- 144 -

40
y = 0.1268x
empty mill 2
y = 0.1374x R = 0.9981
2
35 mill+balls only (20 125 g) R = 0.9992
Bauxite +Balls
30 Linear (empty mill)
Linear (mill+balls only (20 125 g)) y = 0.1077x

2
Linear (Bauxite +Balls) R = 0.9978
25
Energy (Wh)
20
15
10
0
0 50 100 150 200 250 300
mill revolutions
coefficients are used to determine the net energy per mill revolution for the bauxite ore.
energy per mill revolution and the effective coarse fraction of the new feed that is
subjected to milling.

Net
energy per mill revolution and the effective new feed as determined at steady state when
the fraction of new feed equals the 250% recycle fraction.

Measured
- 145 -

Table 29 summarises the energy comparisons for the bauxite ore based on the Bond
equation. As mentioned previously the feed and product size distributions displayed
parallel lines when plotted on a log-log scale. The results show that marginal energy
savings of between 0.7 and 1.1 kWh/t are possible when grinding the bauxite ore to a
closed screen size of 425 Om, 300 Om and 212 Om. The preferred method of using the
measured energy results (as determined by the digital energy meter) are summarised in
Table 30.
Energy savings in absolute terms of between 4.0 and 4.3 kWh per tonne are achievable
when comparing conventional ball milling with hybrid high-pressure grinding rolls. These
energy savings in absolute terms are lower than in the lead/zinc ore case study; however in
this particular case (bauxite ore) the circuit energy savings are higher in percentage terms
at 41.8%, 34.7% and 31.0% for circuit closing screen sizes of 425 Om, 300 Om and 212
Om, respectively.
Table 29 gives the calculated energy comparisons for the bauxite ore between the two
circuits, determined through the use of the Bond third theory equation and the Bond work
index that was determined using the data from the Bond locked cycle test

HPGR Ball mill
Table 30 gives the measured energy comparison for two bauxite comminution circuits,
based on the measured energy per mill revolution and the effective new feed as determined
at steady state when the fraction of new feed equals the 250% recycle fraction.

HPGR Ball mill
- 146 -

Results of the platinum/chrome ore
The third and final ore, the platinum/chrome ore, was tested and produced an unexpected
result. The feed and product size distributions are given in Figure 111, and are clearly very
different to the feed size distributions of the bauxite and lead/zinc ore
100
Product 106 um
90
Product 150 um
80
Feed bond ball
70 conventional
Cum % Passing
60 Product 75 um
50 Product of
HPGR Circuit
40
Product HPGR
circuit 106 um
30
Product of
20 HPGR circuit
150 um
10 Product of
HPGR circuit
75 um
0
0.01 0.1 1 10
Size (mm)
Figure 111 show the summarised results of the platinum/chrome ore, and includes the feed size distributions
to both circuits, with the intermediate product of the triple pass HPGR circuit making up the feed to the ball
mill in the hybrid HPGR-ball mill circuit.
It is not surprising that the products of the three passes through the HPGR produced a
product size distribution which was self similar to the feed size distribution: The self-
similarity effect is a well known characteristic of the HPGR process and has previously
been reported in the literature on numerous occasions (Fuerstenau and Kapur, 1995;
Gutsche et al., 1993; Kapur, 1972).
Under these circumstances it appears that the fine size fraction of the feed to the locked
cycle ball mill test when closing at 150 Om produced a final product which is finer than
the product which was produced using the original feed. Clearly the two products
produced by the two circuits are not identical when closing at 150 Om. A similar result
was obtained when the closing size setting was limited to 106 Om. However it appears
that when the closing size setting was reduced to 75 Om the products from the locked
cycle test are similar regardless of the feed size distribution (whether it is produced from
the original feed or from the product of three passes within the HPGR.
The unusual feed size distribution of the platinum/chrome ore, when plotted on a log-log
scale as shown in Figure 112, clearly shows two non-parallel lines. This result raises
doubts about whether the Bond test methodology will be applicable in this particular case.
Another point to note as observed in Figure 111, is that the product size distribution after
three passes in the HPGR is quite close (in position) to the fine size fraction that
represents the final products of the conventional ball mill test with a closing size setting
of 150 Om.
- 147 -

100
Feed-150um
Cum % Passing
Product 150 um
10
Feed 300 um
Product 300 um
Feed 106 um
Product 106 um
1
0.01 0.1 1 10
Size (mm)
Figure 112 shows a log-log plot of the feed and product size distributions of the platinum/chrome ore and the
non-parallel lines as observed during the Bond ball mill tests.
The calculated energy in Table 31 and Table 32 once again appear to indicate a plausible
outcome when using either the Bond third theory equation or the Bond energy calculated
at 60 joules per revolution.
Table 31 gives the BWI and third theory equation energy for the platinum/chrome ore.

Net
Table 32 gives the equivalent third equation energy of the platinum/chrome ore when
calculated using a mill energy input of 60 J/rev.

Energy at Effective
- 148 -

Table 33 and Table 34 illustrate the two different methods of determining the measured
circuit comminution energy. In the first method the measured energy is calculated on the
basis of the effective coarse fraction of feed that is subjected to milling and the
corresponding measured net comminution energy per revolution. For the platinum/chrome
ore, the net comminution energy per revolution was measured at 95 J/rev (as determined
by the difference between the coefficient of 0.134 and 0.107) and shown in Figure 113.
measured energy per mill revolution and the effective coarse fraction of the new feed that
is subjected to milling.

Net
In the second method the circuit comminution energy is based on the measured joules per
revolution and the effective new feed rate (i.e. the 250% recycle fraction as a whole).
Using the second method the energy calculation is based on an effective new feed of 516
g and 511 g for the conventional ball milling and hybrid HPGR ball milling circuits,
respectively. When the Bond methodology is used as shown in Table 33, there is a large
discrepancy in the quantity of coarse material subjected to milling. For example the
effective coarse feed of the plus 150 Om size fraction reporting to the conventional ball
milling circuit was found to be 366 g, compared to the coarse feed subjected to milling of
232 g for the feed to the ball mill in the hybrid HPGR circuit. As a result of these large
differences, the calculation results in very similar overall net energies as required by the
ball mill. In this particular test, the error associated with using this particular method is
amplified.
measured energy per mill revolution and the effective new feed as determined at steady
state when the fraction of new feed equals the 250% recycle fraction.

Measured
These results confirm that the digital energy meter measurements and corresponding
effective new feed rate should be used as the basis of calculation for the circuit specific
comminution energy. The overall difference between the two methodologies is similar in
- 149 -

magnitude with the platinum/chrome ore (Table 34) to the results that were obtained for
the lead/zinc and bauxite ores in Table 22 and Table 28.
Table 35 and Table 36 summarises the results of the platinum/chrome ore . Table 35
presents the overall energy comparisons based on the third theory equation. As mentioned
previously, the feed and product size distributions did not display parallel lines when
plotted on a log-log scale. The results show that the energy input to the hybrid HPGR
circuit was more intensive than into the conventional ball milling circuit. At a closing
screen size of 75 Om was a very small energy saving of 2.5% found, which is likely to be
in error as negative energy savings of -25.9% and -53.0% for closing screen sizes of 106
Om and 150 Om respectively were observed.
120
empty mill y = 0.1341x

2
R = 0.999
100 mill+balls only (20 125 g)
Lonmin Ore + balls
Linear (empty mill)
80
Linear (mill+balls only (20 125 g))
Energy (Wh)
Linear (Lonmin Ore + balls)

y = 0.1288x
60 2
R = 0.9993
40 y = 0.1077x
2
R = 0.9997
20
0
0 100 200 300 400 500 600 700 800
mill revolutions
coefficients are used to determine the net energy per mill revolution for the platinum/chrome ore.
Table 35 gives the calculated energy comparison for the platinum/chrome ore between the
two circuits as determined through the use of the Bond third theory equation and the Bond
work index measured during the Bond locked cycle test.

HPGR Ball mill
The preferred and chosen method of using the measured energy results (as determined by
the digital energy meter) are summarised in Table 36. The measured energy savings in
this particular case are questionable as a result of there being a negative energy saving in
- 150 -

the 150 Om closing size fraction. The remaining two circuit energy comparisons at 106
Om and 75 Om produced energy savings of 4.5% and 23.6% respectively.
Table 36 gives the measured energy comparison for two platinum/chrome comminution
circuits, based on the measured energy per mill revolution and the effective new feed
determined at steady state when the fraction of new feed equals the 250% recycle fraction.

HPGR Ball mill
There are three possible reasons for the platinum/chrome data anomalies. The first, as
shown in, Figure 111, is that the final products of the ball mill with a closing screen size
at 106 Om and 150 Om were not identical.
Figure 111 shows that the products of the hybrid HPGR/ball mill circuit are much finer
than in the conventional milling circuit. This could mean that more energy has been
utilised to produce these finer products and as a result, when the energies are compared,
the absolute energy savings are -1.5% and 0.8 % for the closing screen size of 150 Om and
106 Om respectively.
The second possible reason for these discrepancies is that during the HPGR three pass
processing, the initial oil pressure of the high-pressure rolls was set at 50 bar as opposed
to 30 bar for the previous two ore samples. At this higher initial oil pressure, the
cumulative measured energy for the HPGR was found to be 3.36 kWh per tonne. Under
these conditions it is possible that a higher proportion of the available energy was
converted to heat, and that the overall energy efficiency of the HPGR operating in this
pressure is lower than when the HPGR was operated at the lower oil pressure. This effect
is well-known: it has previously been reported in the literature by Daniel (2003) and
Tondo (1997) that an unlimited amount of energy may be applied to the HPGR without
any additional benefits in comminution. The point at which this occurs is known as the
energy saturation point, and once this energy level is passed, the energy is converted
mainly into wasteful heat energy.
The third reason could be as a result of a suspected high level of chromites in the platinum
ore which could distort the breakage characteristics in the size region below or above 500
Om to 1 mm, as indicated by the peculiar non linear behavior or “knee” in the feed and
product size distributions as shown in log normal and log-log Figure 111 and Figure 112
respectively.
- 151 -

Specific energy test results consolidated, compared and summarised
A summary of the two methods used for the comparison of energy consumption and
energy efficiency is given below. These methods as described and presented above are the
direct energy measurement method using a digital energy meter and the Bond
methodology using the Bond third theory equation and Bond work index.
The summarised Bond equation method rests on Bond’s statement that the mill consumes
60 J/rev, and grinds the coarse fraction of the feed above the closing screen setting. This
method and its implications have been reported previously (Daniel 2004). The test results
however confirm that Bond effectively calibrated the laboratory mill test results to that of
a 2.4 m wet grinding mill. Figure 114 shows how the comparative techniques for the
three ores tested are consistent with the observations made when a similar exercise of data
analysis is conducted on a large database of Bond ball mill test results.
The summarised direct digital energy meter measurements are shown in Figure 115 and
Figure 116. The best results obtained in terms of energy savings may be summarised by
choosing the maximum energy difference that was measured at a particular closing size.
This means that an overall energy saving of 35.6% at a closing size of 300 Om is
achievable with the lead/zinc ore, a 41.8% energy-saving at a closing size of 425 Om is
achievable on the bauxite ore and a 23.7% energy-saving is achievable on the lead zinc
ore at a closing size of 75 Om. The energy savings quoted represent single points on the
graphs shown in Figure 116. Figure 116 represents a graphical summary of the tests
completed in the final size fraction grinding tests.
30
Bond Energy @ 60 J/rev ground product
25
y = 0.9982x
2
20 R = 0.9782
15
Phd test data - Mt ISA lead-zinc ore
10
Phd test data - Bhpbilliton bauxite ore
Phd test data - Lonmin platinum ore

5
Linear (JKMRC Bond test database)
0
0 5 10 15 20 25 30
Bond (Third theory) Equation Energy (kWh/t)
Figure 114 shows the correlation of the Bond energy determination methodology which is based on a constant
mill energy consumption of 60 J/rev.
The results represent a direct measured energy size relationship that may be compared
with the relationships derived by the researchers Rittinger, Kick, Bond, Hukki and
Morrell. However choosing the grind size on the basis of maximum energy savings may
- 152 -

not provide maximum mineral recovery, nor will it result in the optimal economic
extraction of the valuable mineral component. Economics, energy usage and mineral
recovery all need to be analysed in order to define the optimal process route for optimal
exploitation of the resource or sustainable resource processing
30
Conventional milling-bauxite
HPGR + milling-bauxite
25
Conventional milling -mt isa

Energy Input (kWh/t)
20 HPGR+milling Mt ISA
Conventional milling Lonmin
15 HPGR+milling-Lonmin
Power (HPGR + milling-bauxite)
10
Power (Conventional milling-bauxite)
Power (Conventional milling -mt isa)
5
Power (HPGR+milling Mt ISA)
Power (Conventional milling Lonmin)

0
10 60 110 160 210 260 310 360 410 460
Grind size (um) or Closing screen size (um)
Figure 115 shows the measured energy size relationship for three different ores that were subjected to a
conventional milling circuit and a hybrid HPGR grinding circuit. Suspect platinum/chrome ore data points
have been omitted .
Figure 115 shows the same data that is presented in Figure 116, however in this instance a
power curve is fitted to the data. Interestingly, the ball mill test data fits a curve which has
an exponent of -0.58 which is very similar to Bond’s Third Theory exponent of -0.5. The
hybrid HPGR/ball mill circuit displays an exponent of -0.71. The expression constants
differ at 295 for Bond milling, and 424 for the hybrid HPGR/ball mill circuit. The results
support the hypothesis of Holmes(1957) which stated that the exponent “n” must be a
variable associated with the comminution device. The results presented in Figure 116
summarises the energy efficiency and new size energy relationship that exists for hybrid
HPGR ball mill circuits when compared with conventional ball milling. This is one of the
more important conclusions of this thesis. Another important outcome of the results of
Chapter 4 and Figure 116 is that the fitted power curve for the ball mill confirms that
Bond’s third theory equation and ball mill work index is applicable to tumbling mill
devices limited within certain particle size ranges, and that the comminution energy is
dominated by what appears to be the generation of new particle surface area.
- 153 -

30
Conventional milling data-all ores

25
Energy Input (kWh/t)
20
Hybrid HPGR/ball milling -all ores
15
y = 295 x-0.56
R2 = 0.9534
Power (Hybrid HPGR/ball milling -all ores)
10
5
Power (Conventional milling data-all ores)
y = 424 x-0.71
R2 = 0.9851
0
10 60 110 160 210 260 310 360 410 460
Grind size (um) or Closing screen size (um)
Figure 116 shows the measured energy size relationship for three different ores that were subjected to a
conventional milling circuit and a hybrid HPGR grinding circuit, and that show the fitted functions to the
data.
In the next section, the modified Kick equation of Holmes (1957) confirms that perhaps
the Morrell equation is the most advanced grind size energy relationship that exists today,
and that comminution is dominated by new particle surface area generation, the efficiency
of which is dependant on the type of comminution device used.
Holmes theoretical approach to comminution and the relation to Bond energy

Holmes (1957) proposed a modified version of the Kick hypothesis, as previously
described in the literature review. The basis for the derivation of equation 11 in the
literature review is that the equation relies on the theoretical basis of new particle surface
area generation, and allowing the exponent “n“ to remain as a variable. In order to
simplify the concept, so that mass is retained, particle size and surface area defined the
following assumption was made in relation to the comminuted particles. A one meter
cubic block of ore of a known density is subjected to a size reduction process whereby
each block is cleaved along the planes which would produce eight equally sized cubes.
The process is repeated for each and every cube that makes up the original cubic meter as
shown in Figure 117.
- 154 -

1m
0.5 m
250
6 m2 12 m2 24 m2 125 mm
62.5 mm
1 m in length 0.5 m in length

Surface area 6 m2 Surface area =
8 x 6 x 0.5 x 0.5 =12 m2
Figure 117 shows a schematic of how a one meter cubic block of ore of a known density could be subjected to
a size reduction process whereby each block is cleaved along the planes which would produce eight equally
sized cubes.
Table 37 displays the effect of the process described in Figure 117. These results in a
pyramid effect with values for the new surface generated as the particle decrease in size
by halving each time. The numbers of equivalent sized particles that represent 357 g of
new feed in the Bond ball mill test are given along with the equivalent calculated surface
area of that number of size particles. Table 37 shows that 17 754 particles of size 1.95
mm would have a mass of 357 g and a total estimated surface area of 0.41 square meters.
This could be representative of a typical Bond test feed. If the closing screen was 122 Om,
then at steady state the equivalent mass of 357 g would be reduced to 72,719,023 particles
that had a surface area of 6.5 square meters. The theoretical data of Table 37 are
presented in Figure 118 which shows that at below the 300 Om size, the surface area
begins to increase exponentially. This is at about the same size that the energy in Figure
116 was observed to increase rapidly, suggesting that new surface area generation is the
more dominant factor controlling size reduction and energy consuming processes within
comminution. At the same time it is important to point out that the energy consumption
relates to the energy that is consumed in tumbling mill devices using balls as the contact
medium between the device and the rock particles.
- 155 -

Table 37 shows how the new surface area is generated from the systematic theoretical
comminution of a 1 cubic meter block into small cubes as described in Figure 119
Charge mass (g) 1250
New feed mass (g) 357
Block size Total surface area Bond particle number Bond new feed Block size
3 2 2
(mm) Block numbers per m (m ) per new feed mass surface area (m ) (Bm)
1000 1 6 1,000,000
500 8 12 500,000
250 64 24 250,000
125 512 48 125,000
63 4,096 96 62,500
31 32,768 192 31,250
16 262,144 384 15,625
8 2,097,152 768 7,813
4 16,777,216 1,536 2,219 0.20 3,906
2.0 134,217,728 3,072 17,754 0.41 1,953
1.0 1,073,741,824 6,144 142,029 0.81 977
0.5 8,589,934,592 12,288 1,136,235 1.63 488
0.2 68,719,476,736 24,576 9,089,878 3.25 244
0.1 549,755,813,888 49,152 72,719,023 6.50 122
0.06 4,398,046,511,104 98,304 581,752,184 13.00 61
0.03 35,184,372,088,832 196,608 4,654,017,472 26.01 31
0.015 281,474,976,710,656 393,216 52.01 15
0.008 2,251,799,813,685,250 786,432 104.03 8
0.0038 18,014,398,509,482,000 1,572,864 208.05 4
0.0019 144,115,188,075,856,000 3,145,728 416.10 2
0.0010 1,152,921,504,606,850,000 6,291,456 832.20 1
10,000,000,000 30
Bond new feed mass particle
numbers against particle size
1,000,000,000
Bond ball mill new feed particle
surface area against particle size 25
100,000,000
10,000,000
Particle surface area m2
20
Number of particles
1,000,000
100,000 15
10,000
10
1,000 y = 1E+14x-3
100
5
10
y = 793 x-1
1 0
10 100 1,000 10,000
Theoretical particle cube size (micron)
Figure 118 shows the number of particles and particle surface area increase as the particle size decreases.
When the same Bond estimated surface area data in Figure 118 is plotted with extensive
Bond third theory equation comminution energy data, the data appear to coincide (Figure
119). The vertical axes of surface area (m2) and comminution energy (kWh/t) are the
- 156 -

same, suggesting that the Bond third theory and locked cycle tests are strongly linked to
the generation of new particle surface area.
100 100
Particle surface area (m2) vs particle size (microns)
Bond eqn energy (kWh/t) vs particle size P80- JK Bond test database
Bond third theory equation energy (kWh/t)

Power (Bond eqn energy (kWh/t) vs particle size P80- JK Bond test database)
2
Particle surface area m
10 10
1 1
0 0
10 100 1,000 10,000
particle size (micron)
Figure 119 shows the co-incidence of the bond test surface area and comminution energy on a log -log scale.
Summary
The summarised direct digital energy meter measurements as shown in Figure 116
represent a direct measured energy size relationship that may be compared with the
relationships derived by the researchers Rittinger, Kick, Bond, Hukki and Morrell.
Overall, the best results obtained in terms of energy savings may be summarised by
choosing the maximum energy difference that was measured at a particular closing size.
This means that an overall energy saving of 35.6% at a closing size of 300 Om is
achievable with the lead/zinc ore, a 41.8% energy-saving at a closing size of 425 Om is
achievable on the bauxite ore and a 23.7% energy-saving is achievable on the lead zinc
ore at a closing size of 75 Om. This does not provide any information on the performance
of the recovery process that follows. The circuit product streams at the varied closing
sizes were sent for mineral liberation analysis, and are analysed in the next chapter.
The lead zinc ore and the bauxite ore display a similar energy size relationship to that
which was proposed by a Bond, Hukki and Morrell. The results confirm that energy size
relationships are not independent of the comminution device used. The properties of the
ore in respect to the mineralisation and grain boundaries must have an effect on how
energy is utilised in the size reduction or fracture process.
The study confirms that particle size/energy relationships are not simple and the theories
developed by Charles and Holmes may provide a deeper insight to the problem. Both
- 157 -

Charles and Holmes stated that the size/energy relationships are dependent on the
characteristics of the ore as well as the characteristics of the breakage process that the
different types of comminution devices impart on particles.
Understanding energy efficiency and the measurement of energy savings in absolute terms
in comminution circuits/processes are complicated. There are well established
methodologies that produce quite different results to the direct measurement results.
However within the context of the literature review, the methodologies presented in this
Chapter are supported with the experimental results, and highlight the limitations of each
of the methodologies followed.
This research however provides a baseline (in terms of energy consumption) from which
to work from, in order to make further evaluations and assessments with regards to the
liberation characteristics of the ores when they have been subjected to HPGR processing.
The fact that controlled transport and stressing is featured in HPGR processing is the main
reason why hybrid HPGR circuits are more energy efficient than conventional ball milling
circuits. This results in a high proportion of available energy being used solely for the
purpose of stressing the material.
In conventional mills, the material transport and stressing inside the active volumes of the
mill between the balls occurs randomly. This often allows particles to move out of
position resulting in unproductive collisions between the grinding media or the media and
the liner wall within the mill. This mode of energy input is inherently wasteful because of
the hit-and-miss nature of the process, whereas the HPGR targets all of the material.
Finally, the results exemplify the importance of energy efficiency in comminution, which
clearly needs to be understood within the context of sustainable mineral extraction and
especially when new comminution circuits are being designed on the basis of energy
efficiency and energy costs. Energy efficiency may well become one of the more
dominating factors for the design of future mineral processing comminution circuits and
may well be unconventional if compared to the concepts that have governed the industry
for over 100 years.
- 158 -

CHAPTER 6
RESULTS- ENHANCED PREFERENTIAL

MINERAL LIBERATION
Introduction
The results of the mineral liberation analysis (MLA) are presented in this chapter. The
objective was to determine if enhanced mineral liberation was possible through the use of
HPGR technology as opposed to conventional technologies. Mineral liberation was
measured for each of the various circuits by individually measuring sub-size fractions
within the products. The analysis concluded that any possible enhancement in the
characteristics of the liberated mineral through the use of HPGR technology was not
founded when the same product size distribution is compared, but widespread evidence of
particle micro-cracking was observed in HPGR products. These micro-cracks result in
particle weakening and are responsible for part of the energy savings that are realised in
downstream processes such as ball milling and improved leaching of HPGR crushed
products. Ball mill processing used to achieve similar sized products largely destroyed any
remnants of the visible microcracks. These observations are presented in the chapter.
As mentioned in the literature review, the aim of comminution in mineral processing is to

liberate valuable minerals. This must be achieved economically at the coarsest possible
size suitable for a chosen separation technique, such as leaching, gravity separation or
flotation. If it is shown that the liberation characteristics of an ore is enhanced using the
HPGR processing, then this could lead to appreciable financial benefits for minerals
processing operations in the sense that less material will be required to be milled to a
specified grind size. These benefits as previously mentioned could be realised by savings
in comminution energy and costs through alternative technologies and reduced grinding
media consumption. The prospect of improved mineral liberation at a coarser grinding
size and increased mineral recoveries may be considered as a double bonus if realised on a
commercial basis.
The MLA results represent the analysis of samples produced in the experiments of circuit
scenario 5 and 6 (Figure 47 and 48). The energy efficiency of these circuits were
evaluated and presented in Chapter 5. The same three study ores were again analysed. The
first presented is the BHP-Billiton bauxite ore, followed by the Mt Isa lead zinc ore, and
finally the Lonmin platinum/chromite ore.
The JKMRC MLA system

The JKMRC MLA system and software package (Figure 120) was used to produce the
particle liberation data. The samples were prepared into several sub-size fractions that
were individually measured. The back-scatter images were used to see particle micro-
cracking. The standard MLA software images do not show this characteristic. As was
- 159 -

explained in the literature review, particle micro cracking seen in micrograph images may
be an artefact of the particles embedded into the epoxy resin mount that is normally
polished in preparation for the scanning electron microscope. Currently there is only one
example known to the author where products of ball mills and HPGR’s have been
compared (Klymowsky and Cordes 1998), the results however were not analysed using
mineral liberation analysis (MLA).
More than 100 samples made up the three different grinds (closing size screen) for each of
the three sponsor ores. The data for these were all independently produced and analysed
using the MLA software suit. Selected images and graphs are used to support the
arguments in this chapter, but the full set of data are available in the CD’s provided at the
back of this thesis. The next two sections report the results of micro-cracking and mineral
liberation.
Figure 120 shows a picture of the JKMRC MLA unit along with the mineral identification image on the right
and the black and white back scatter image on the left. The insert picture shows the sample mount shuttle
capable to analysing fourteen samples.
Existence of particle micro-cracking

Since the products of the HPGR process, conventional ball milling and hybrid HPGR/ball
mill circuits were produced independently using an identical feed, they may be directly
compared. The back scatter images produced some remarkable images. These are shown
alongside the MLA mineral interpreted images that follow in the results. At the present
time, software does not exist that interprets the extent of the micro-cracks nor is it able to
identify the minerals in which micro-cracks are more prominent.
As this has not yet been developed, a subjective analysis has been performed on the
images that were examined. These are broadly categorized in to four categories, namely
none, few, some, and many that without software confirm the extent of microcracks.
The observations however can conclusively confirm that HPGR readily promotes particle
micro-cracking. Depending on the ore, some size fractions appear to be less prone to
micro-cracking. There is unquestionable evidence based on the observations that ball mill
products to not contain particle micro-cracks. This was observed for all three ore types.
The extents to which microcracks are visible in the Hybrid HPGR/ball mill circuit varied,
- 160 -

and suggest that the microcracks produced by the HPGR are to a degree destroyed in the
ball milling process. However there exist a smaller number of particles that show micro-
cracking. The degree to which the HPGR selects minerals of interest is very difficult to
assess, due to the lack of measurement technique. However it is shown later in the
preferential mineral liberation section that no real difference in liberated characteristics
was observed.
The next three sub-section provide the MLA results of the three ores investigated. These
are of the form of visual pictures of the back scatter image micrographs showing the
microcracks alongside the equivalent MLA interpreted mineral image. A subjective
analysis of the extent of the microcracks for the ball mill, HPGR and Hybrid HPGR/ball
mill circuits is provide in the corresponding section tables.
Worsley Bauxite ore MLA micro-crack results

The results of the Worley bauxite sample clearly showed a presence of microcracks in the
HPGR treated samples, no evidence of microcracks were observed in the ball mill
processed material. The extent of the microcracks are subjectively summarised in Tables
38 and 39)
The Bauxite ore is a soft ore and may not be readily suited for HPGR processing. This is
motivated by the low energy absorption in the coarse feed HPGR processing, and higher
energy absorption in the fine feed HPGR process. What appears to be a higher internal
resistance of the bauxite material causes higher than expected energy consumption in the
HPGR. It is for these reasons that the Bauxite material may not be suitable for HPGR
processing, however this fact should be verified through pilot scale test work. In addition
it appears that the energy does not transfer to the fine size fractions below 200 Om as
observed in Figure 108. Bauxite shows a tendency for larger particle being selectively
chosen for comminution.
Table 38 shows the extent of microcracks for the bauxite ore of a hybrid HPGR/ball mill
circuit against a ball mill circuit producing three different closing screen size distributions
and multiple MLA sub-size fractions.
HPGR- HPGR-
Ball Ball Only Ball Ball Only HPGR-Ball Ball Only
Closing Closing Closing Closing Closing Closing

screen @ screen @ screen @ screen @ screen @ screen @
Size Class 425 1m 425 1m 300 1m 300 1m 212 1m 212 1m
-425+300 Few None X X X X
-300+212 Few None Few None X X
-212+106 Few None Few None Few None
Not Not Not Not Not
-38 Analysed Analysed Analysed Analysed Not Analysed Analysed
- 161 -

Figure 121 shows the comparison between the MLA mineral image analysis against the
equivalent MLA back scatter image. This is of the ball mill circuit and shows no evidence
of particle micro-cracking in the -425 + 300 Om size range. The gibbsite and other mineral
constituents appear to be liberated, but the mineralogy remains relatively complex.
Gibbsite, due to the nature of the mineralized ore is very abundant within the samples.
Figure 121 shows MLA identified mineral particles within the bauxite sample on the left hand side, and the
corresponding back scatter image on the right hand side. These images represent the ball mill circuit showing
the -425+300 Lm size fraction. No microcracks are seen.
Figure 122 shows the comparison between the MLA mineral image analysis against the
equivalent MLA back scatter image with a fully liberated Gibbsite particle circled. This is
of the hybrid HPGR/ball mill circuit and shows clear evidence of Gibbsite particle micro-
cracking in the -425 + 300 Om size range.
Figure 122 shows MLA identified mineral particles within the bauxite sample on the left hand side, and the
corresponding back scatter image on the right hand side. These images represent the hybrid HPGR/ball mill
circuit showing fully liberated gibbsite particles within the -425+300 Lm size fraction. The liberated gibbsite
particle shows evidence of microcracks as seen in the back scatter image.
- 162 -

The extent of micro-cracking in the HPGR circuit and feed material is summarised in
Table 39. The individual micrographs are not shown here, but are available for viewing in
the data CD at the back of this thesis. An interesting feature of the micro-crack
observations of these sub-samples was the apparent disappearance of the presence of
microcracks in the finer size fractions. This could support the idea that energy is not
transferred to the fine sized particles, as discussed previously.
Table 39 shows the extent of microcracks for the bauxite ore for the HPGR circuit against
the feed sample to both circuits at multiple MLA sub-size fractions.
HPGR Only HPGR Only Feed only Feed only
HPGR open
circuit +850 1m HPGR open circuit - Feed +850 Feed -
size fraction 850 1m size 1m size 850 1m size
Size Class only fraction only fraction only fraction only
+850 Few Few None None
-850+600 Some None
-600+425 Some None
-425+300 Some None
-300+212 Few None
-212+150 Few None
-150+106 None None
-106+75 None None
-75+53 None None
-53+38 None None
Mt Isa results- lead zinc ore MLA micro-crack results

The Mt Isa lead zinc results were quite conclusive in that the main zinc and lead minerals,
sphalerite and galena were abundant and quite easily identifiable in-and-amongst the
gangue particles. Minerals were clearly bi-modal in many cases and clearly liberated at the
chosen experimental grind sizes, which made the assessment of micro-cracking more
relevant to the hypothesis of being able to determine if HPGR promotes or enhances
mineral liberation. Figure 123 shows a fully liberated sphalerite particle in the ball mill
circuit that shows no evidence on any particle micro-cracking.
Figure 123 shows MLA identified mineral particles within the Mt Isa Lead Zinc ore sample on the left hand
side, and the corresponding back scatter image on the right hand side. These images represent the ball mill
circuit showing liberated sphalerite and galena particles within the -150+106 Lm size fraction. The liberated
particles do not show any evidence of microcracks as seen in the back scatter image.
- 163 -

The same applies for a liberated galena particle in the MLA -150 Om +106 Om sub-size
fraction that was processed in a ball mill. There is no evidence of micro-cracking as
shown in Figure 124.
Liberated Galena
No Microcracks
Figure124 shows an MLA identified galena mineral particle within the Mt Isa Lead Zinc ore sample on the
left hand side, and the corresponding back scatter image on the right hand side. These images represent the
ball mill circuit showing liberated particles within the -150+106 Lm size fraction. The liberated particles do
not show any evidence of microcracks as seen in the back scatter image.
Figure 125 provides an example of a typical MLA data image showing many thousands of
particles. Up to 3000 particles may be analysed and which may be interpreted in several
ways. The mineralogical interpretations are explained in more detain in the next section.
However, Figure 125 identifies a single particle that is examined for microcracks and
mineral identification in Figure 126.
Figure 126 shows complex mineralogy of the Mt Isa Lead Zinc ore sample. On the left
hand side three different minerals are identified (Circled), with the corresponding back
scatter image on the right hand side. These images represent the HPGR circuit (ie after the
triple pass HPGR) showing particles within the -425+300 Om size range. The un-liberated
particles show evidence of visible trans-granular and inter-granular microcracks as seen in
the back scatter image
- 164 -

One of thousands of particles
Figure 125 shows MLA identified mineral particles that are individually examined for liberated mineral
content as well as the extent of the microcracks. The figure illustrated that the MLA block mounts contain
thousand of particles, each of which either display features of microcracks if they have been subjected to
HPGR and no-microcracks when subjected to conventional comminution technologies.
Visible Microcracks
Complex mineralogy
Figure 126 shows MLA identified mineral particles within the Mt Isa Lead Zinc ore sample on the left hand
side, and the corresponding back scatter image on the right hand side. These images represent the HPGR
products (ie after the triple pass HPGR) and show how complex the mineralogy is within the -425+300 Lm
size fraction. The un-liberated particles show evidence of visible trans-granular and inter-granular
microcracks as seen in the back scatter image.
Figures 127 shows the back scatter images of the Mt Isa lead zinc ore as a feed sample to
two comparative circuits under investigation. Whereas the previous Figures 124, 125 and
126 showed the presence or absence of microcracks, Figure 127 shows that no micro-
cracks are visible in the feed samples that were prepared using conventional laboratory
crushing equipment. This further confirms that the micro-cracks that are visible are as a
result of the HPGR process. The large +850 Om particle on the right hand side of Figure
127 clearly shows two un-liberated mineral phases within the particle, and shows the
bright finely disseminated grains that represent the valuable mineral of interest.
- 165 -

No Microcracks
No Microcracks
Figure 127 shows large particle back scatter images of the Mt Isa Lead zinc ore. These images represent feed
particles to both comminution circuits. Complex unliberated particles within the -425+300 Lm size fraction
are show in on the LHS, whilst particles within the -3350+850 Lm size fraction are show in on the RHS.
These un-liberated particles do not show any evidence of microcracks whatsoever as seen in the back scatter
image
Table 40 summarises the micro-crack observations of all the lead zinc ore MLA sub-size
fractions for the ball mill circuit compared against the hybrid HPGR/ball mill circuit. A
few images were presented; however it does confirm that no microcracks are seen in the
ball mill product sample, whereas a few micro-cracks were observed in the hybrid
HPGR/ball mill circuit. The observation that a few micro-cracks lead to the possible
conclusion that some of the micro-cracks formed in the HPGR are destroyed in the ball
milling circuit that follows the HPGR. Never-the-less there is still some evidence that
micro-cracking is present, which was clearly not the case when the ball mill product were
examined.
Table 40 shows the extent of microcracks for the lead-zinc ore with the hybrid HPGR/ball
mill circuit compared against the ball mill circuit producing three different closing screen
size distributions and multiple MLA sub-size fractions.
HPGR- HPGR-
screen @ screen @ screen @ screen @ screen @ screen @
Size Class 300 1m 300 1m 150 1m 150 1m 106 1m 106 1m
-150+106 Few None Some Few X X
-106+53 Few Few Some Few Few None
-53+38 Few Few Some Few Few Few
Not Not Not Not Not
Figure 128 shows a large +850 Om particle from the triple pass HPGR product. HPGR
products are known to contain large particles, these may exist either as edge material
particles or bed particles. In this particular case extensive micro-cracks are seen, and so
the particle is likely to be represented by one that is closer to the edge material than to the
- 166 -

compressed bed. The particle does not appear to display inter-granular or grain boundary
cracking.
Microcracks
Figure 128 shows a single large particle back scatter image of the Mt Isa Lead zinc ore. These image
represents products of the triple pass HPGR. Complex unliberated minerals within the+850 Lm particles are
partially liberated by the presence of microcracks. The microcracks appear to be random and are described as
trans-granular.
Figure 129 is an image from the same particle sub-size fraction as in Figure 128 (triple
pass HPGR product). Finely disseminated “bright” galena mineralisation is visible.
Another mineral phase is present and there does appear to be some degree of inter-
granular micro-cracking through the particle’s visible phases. The quantification of the
extent of the micro-cracking cannot at this stage be determined.
Microcracks
Figure 129 again shows a single large particle back scatter image of the Mt Isa Lead zinc ore. These images
are similar to those in Figure 128 and represent products of the triple pass HPGR .
In Figure 130 shows complex un-liberated lead zinc ore particles within the -600 + 425
Om sub-size fraction. The “bright” phased particles represent the mineral of interest and
appear to be extensively micro-cracked, more so than the surrounding gangue minerals. It
could be possible that the HPGR has preferentially selected these particles during
comminution, due to their assumed softer character. However as previously described,
some of the particles may be representative of edge effect material, in which case far less
- 167 -

micro-cracks would be visible. A quantitative assessment in this case is again difficult to
determine without the appropriate software.
Microcracks
Figure 130 shows multiple particles of the Mt Isa Lead zinc ore particles within the -600+425 Lm size
fraction. These images are similar to those in Figure 128 and 129, finer in size and represent products of the
triple pass HPGR.
In Figure 131 shows liberated minerals that appear to be preferentially micro-cracked,

implying possible preferential liberation within the size fraction -425 + 300 Om The
brightly shaded minerals represent galena and sphalerite and appear to have extensive
visible microcracks. Some of the microcracks in these images appear to be to show trans-
granular cracking as opposed to the claimed inter-granular breakage, however the micro-
cracks are extensive, and appear to randomly transgress throughout the particle.
Visible Microcracks Microcracks
Figure 131 shows multiple particles of the Mt Isa Lead zinc ore particles within the -425+300 Lm size
fraction and represent products of the triple pass HPGR . Valuable unliberated binary mineral particles
appear to be selectively chosen for micro-cracking.
Table 41 summarises the micro-crack observations of all the MLA sub-size fractions. A
few images have been presented here, but the main point is that no microcracks could be
seen in the feed sample in all size fractions above 106 Om. Very few particles in the -106
Om feed showed micro-cracks, however it could be possible that marginal inter-particle
comminution could be taking place in the feed preparation stage, however clearly not to
the same extent as in the HPGR products in all size fractions.
- 168 -

Table 41 shows the extent of microcracks for the lead-zinc ore for the HPGR circuit
against the feed sample to both circuits at multiple MLA sub-size fractions.
HPGR Only HPGR Only Feed only Feed only
HPGR open
circuit +850 1m HPGR open circuit - Feed +850 Feed -
size fraction 850 1m size 1m size 850 1m size
Size Class only fraction only fraction only fraction only
+850 None Some None None
-850+600 Many None
-600+425 Many None
-425+300 Many None
-300+212 Many None
-212+150 Many None
-150+106 Many None
-106+75 Many Few
-75+53 Many Few
-53+38 Many Many
Lonmin platinum ore MLA micro-crack results

The Lonmin platinum results were conclusive, but for a very different reason. The
platinum minerals that naturally float are rare and are difficult to find in their liberated/
un-liberated state when a few MLA block mounts are being used. However, as was
observed in Figure 111, the products of two of the coarser size closing screen ball mill
tests were not comparable. Several reasons for this were given in chapter 5, but the
mineralogical results presented here are able to shed more light on the difference in
product size distribution between the two circuits. This is explained towards the end of
this section.
Within the platinum concentrate, there is an abundance of Chromite, and a very small
percentage of sulphide floating minerals such as sphalerite, chalcopyrite and pyrite. These
particles are rare, but not as rare and finely disseminated as the platinum mineral. As such,
a few of these minerals were identified in the MLA images. Chromite was abundant and
quite easily identifiably in and amongst the other gangue minerals. Chromite is claimed
not to be associated with the platinum mineral, and so it would be undesirable to have to
comminute the very hard chromite at the expense of energy and the possibility of
negatively affecting flotation response of the platinum.
Figure 132 shows an example of the clearly identified rounded liberated Chromite
particles along with a single fully liberated sphaletite particle in the -425 +300 Om size
fraction. The sample represents particles produced through three passes of the HPGR.
Some gangue minerals appear to be micro-cracked; however the liberated sphalerite
particle is not. It is believed that this particle has not been subjected to the centre
compresses bed zone, but rather a particle that has passed through the HPGR towards the
edge zone.
- 169 -

Coarse Fully
Sphalerite particle
liberated coarse
Sphalerite particle
ROUND Liberated Chromite

Minerals liberated some microcracked
Figure 132 shows multiple particles of the Lonmin platinum ore, particles are within the -300+212 Lm size
fraction and represent products of the triple pass HPGR .
Figure 133 shows the products of the ball mill circuit. Particles represent -150 +106 Om
size fraction and shows a single identified chalcopyrite particle (circled) that is not micro-
cracked. Some rounded chromite particles are present as well as several shard/needle
shaped chromite particles all of which do not appear to have any micro-cracks.
Chalcopyrite Mineral NOT Micro-cracked Chalcopyrite Mineral NOT Micro-cracked
Figure 133 shows multiple particles of the Lonmin platinum ore which have undergone ball mill processing.
The particles represent the -150+106 Lm sub-size fraction. Semi-valuable liberated chalcopyrite mineral
particles appear to have no micro-cracking.
Figure 134 shows the products of the triple pass HPGR circuit. Particles represent the -
300 +212 Om size fraction and shows a single identified pyrite particle (circled) that is
severely micro-cracked. Many rounded chromite and other gangue mineral particles are
present all of which appear to show evidence of micro-cracks. Although the shard-like
shaped particles are not present, the micro-cracks indicate that particle weakening has
taken place within the hard chromite particles. This suggests that further breakage of the
harder chromite particles may require less energy. This was discussed earlier in chapter 5.
- 170 -

Large
Pyrite
particle
Liberated
Chromite
ROUNDED
Pyrite mineral severely micro-cracked
Figure 134 shows multiple particles of the Lonmin platinum ore which have undergone hybrid HPGR /ball
mill processing. The particles represent the -300+212 Lm sub-size fraction. Semi-valuable liberated
chalcopyrite mineral particles have severe micro-cracking.
The difference in the shape of the product size distribution as observed in Figure 111 can
partly be explained by the preferential breakage/ micro-cracking of the hard chromite. As
these particles are further comminuted at closing screen sizes of 106 Om and 150 Om, the
resultant size distributions are different, and could be attributed to the preferential
breakage of the chromite particles within the ore, which at this size range are considered
as a bi-modal ore. Since the HPGR circuits produce the micro-cracks and particle
- 171 -

weakening, it is not surprising that the HPGR circuit materials consistently produced a
finer product size. However, when a closing screen size of 75 Om is sued, all the
chromite particles again are comminuted predominantly by the ball milling process of the
two circuits, and the product size converges and become quite comparable again.
Table 42 summarises the micro-crack observations of the Lonmin platinum ore MLA sub-
size fractions for the ball mill circuit compared against the hybrid HPGR/ball mill circuit.
A few images are presented, however point again confirming that no microcracks could be
seen in the ball mill product sample in all size fractions (some images were not available),
whereas a few micro-cracks were observed in the hybrid HPGR/ball mill circuit. The
observation of a few micro-cracks leads to the possible conclusion that some of the micro-
cracks formed in the HPGR are destroyed in the ball milling circuit that follows the
HPGR. Never the less there is still some evidence that micro-cracking is present, which
was clearly not the case when the ball mill product were examined.
Table 42 shows the extent of microcracks for the Lonmin platinum ore with the hybrid
HPGR/ball mill circuit compared against the ball mill circuit producing three different
closing screen size distributions and multiple MLA sub-size fractions.
HPGR- HPGR-
screen @ screen @ screen @ screen @ screen @ 75 screen @
Size Class 150 1m 150 1m 106 1m 106 1m 1m 75 1m
-150+106 * Few X X X X
-106+90 Few None Few None X X
-75+53 Few * Few None Few None
-53+38 Few * Few * * None
Not Not Not Not Not
* No XBSE image provided
The liberation and grinding characteristics of the chromite in its natural form, has often
caused problems for Lonmin when it come to designing plant mill sizes (Goodall, 2005).
As such the Bond work index is not normally used to determine mill grinding
performance. The chromite particles in question are shown in Figure 135 as un-liberated
particles in the -850 + 600 Om size fraction and fully liberated (Figure 136) in the -425
+300 Om size range. This bi-modal effect of the properties of the ore at this size is the
cause of the kink in the size distributions of the platinum ore at the 425 Om size. On a log-
log scale the size distribution is not linear. As previously discussed (Figure 112, Page
146) the P80 and Bond test methodology should not be used.
- 172 -

UN-Liberated Chromites - ROUNDED
Figure 135 shows multiple particles of the Lonmin platinum ore which have undergone HPGR processing.
The particles represent mainly edge effect HPGR material in the +850 Lm size fraction. A very large semi-
valuable liberated sphalerite mineral particle is identified within the sample mount in and amongst the
rounded un-liberated chromite mineral phases.
Liberated Chromites - ROUNDED
Figure 136 shows thousands of Lonmin platinum ore particles which have undergone HPGR processing. The
particles represent material in the -425 +300 Lm size fraction. Most of the chromites are perfectly liberated at
this size range providing a reason for the kink in product size distribution at this size. Grinding the chromites
to sizes below this size results in shard/needle shaped particles.
- 173 -

Figure 137 shows large +850 Om particles in the HPGR products. The enclosed spherical
chromite particles are clearly visible in the back scatter image, and do not show evidence
of particle micro-cracking. These particles are most likely to have survived the HPGR
compressed bed zone, through being processed at the edge zone within the HPGR.
This is not the case in Figures 139 that show particles from the same sample as in Figure
137, but with micro-cracking clearly visible. Both images of 138 show trans-granular and
inter-granular fracture with respect to the liberation of the chromite particles.
Figure 139 shows four similar back-scatter images of yet again the same sample, but the
severity of the particle breakage is astonishing. Both the chromite and the particle matrix
are severely comminuted. These particles are possibly agglomerated particles that clearly
show the effect of the inter-particle HPGR breakage mechanism. The shard/needle shaped
chromite particles are seen, and again confirm the reason for the difference in the product
size distribution for the platinum ore with the closing screen at 150 Om and 106 Om as
previously discussed.
Edge material +850Bm

No Microcracks
Figure 137 shows a single particle of the Lonmin platinum ore which have undergone HPGR processing. The
particle is thought represent mainly edge effect HPGR material in the +850 Lm size fraction and has no
micro-cracking visible.
- 174 -

Centre bed material
+850Bm
Microcracks
Inter-granular
Trans-granular
Centre bed
material
+850Bm
Microcracks
Inter-granular
Trans-granular
Figure 138 shows two images of single particles of the Lonmin platinum ore which have undergone HPGR
processing. The particles are thought to represent a partial bed compression zone within HPGR processing.
The particles are of the +850 Lm size fraction and show evidence of trans-granular and inter-granular
particle micro-cracking.
- 175 -

Edge material Centre bed
+850Bm material
No Microcracks +850Bm
Microcracks
Centre bed
material
+850Bm
Microcracks
Centre
material
+850Bm
Microcracks
Centre bed
material
+850Bm
Microcracks
Figure 139 shows four single particles of the Lonmin platinum ore which have undergone HPGR processing.
The particles are thought to represent the centre bed compression zone region of HPGR processing. The
particles are from the +850 Lm size fraction and show evidence of severe trans-granular and inter-granular
particle micro-cracking.
- 176 -

Enhanced preferential mineral liberation
The aspect of enhanced preferential mineral liberation within this study was difficult
quantify. The results on micro-cracking in the previous section showed that preferential
breakage was taking place, but this does not quantify the extent to which the enhanced
preferential mineral liberation is taking place. The MLA data analysis was used to try and
quantify the liberation characteristics of the products of the two process circuits that were
compared.
Part of the reason for the difficulty in quantifying the liberation characteristics is that
MLA mineralogical data is difficult to mass balance if mineral particle tracking is
attempted. Mass balancing is necessary because the sub-size fractions used in the mounts
do not contain the proportional mix of the various masses that made up the original
sample. Fortunately in this studies experimental design, the product size distributions
were similar, and so it is reasonable to make direct comparisons of the mineral content
within MLA sub-size fractions. This is the logic followed in the comparative analysis of
the samples.
Each of the individual sub-size fractions as summarised in Tables 39 to 42 were subjected

to MLA image analysis. An outcome of the data that is generated from the MLA software
is that within a known particle size range, a new size distribution is measured due to the
stereological effects of viewing two dimensional images of three dimensional particles
within sample mount. The MLA size distribution is not a true reflection of the actual
particle size distribution. The MLA generated particles size distributions are nevertheless
useful in this case as it represents a way of checking the integrity of the data and the
sample. Figure 140 shows the sets of comparable size distributions of the Mt Isa lead zinc
ore for the various MLA particle sub-size fractions. As presented, the size distributions
are similar, giving confidence in differences that are observed in respect to mineral
content. Most of the other size distributions produced similar results with the odd
exception of two size distributions being slightly mismatched. These anomalies were
assumed to be as a result of sample preparation error or because of slight variations in the
actual product size distributions.
Figure 141 shows the MLA grade recovery results for the Worsley bauxite samples. These
grade recoveries represent different size fraction, that all show very similar results. This
indicates that minerals that exist in abundant quantities such as gibbsite in bauxite ore are
not affected by HPGR in this instance. However as was previously suggested in the
discussion in Chapter 5, bauxite is seen as a mineral that is not conducive to HPGR
processing, and the effects of bed breakage appear to not influence the liberation
characteristics. A similar result was obtained for the Mt Isa lead zinc ore in Figure 142
and Figure 143 shows the grade recovery composition curves for lead and zinc
respectively in the single size sub-fraction -53+38 Om. As observed there is a variation in
the grade recovery results. If this is evaluated and balanced over the entire particle size
distribution then the results do not show an appreciable difference in liberated
characteristics.
- 177 -

Figure 140 shows the 2D particle size distribution generated by the MLA data for the Mt Isa lead zinc ore. The comparisons represent the ball mill products against the hybrid HPGR/ball
mill products.
- 178 -

Figure 141 shows the grade versus recovery composition for Aluminium as generated by the MLA mineralogical analysis. The analysis represents three different particle sub size fractions
- 179 -

Figure 142 shows the grade versus recovery composition for lead (Pb) as generated by the MLA mineralogical analysis. The analysis represents a single +38 Lm particle sub size fraction.
- 180 -

Figure 143 shows the grade versus recovery composition for lead (Pb) as generated by the MLA mineralogical analysis. The analysis represents a single +38 Lm particle sub size fraction s
- 181 -

Summary
The chapter has presented results that show conclusively that HPGR produces particle
micro-cracking and weakening. This has the effect of significantly reducing the energy
required to grind the material to a specified size as previously reported in Chapter 5.
However in as far as trying to significantly alter the characteristics of the liberation
distribution of the valuable minerals within the size distribution, this was not founded.
With exception, the Lonmin platinum ore, the results did show a difference in liberation
characteristic with respect to the chromite within the ore. This may not necessarily be the
case for the platinum mineral content, nor does it mean that there will be an advantage or
disadvantage. These aspects need to be investigated further.
- 182 -

CHAPTER 7
THE IMPORTANCE OF COMMINUTION

ENERGY AND ENERGY EFFICIENCY FOR
THE DEVELOPMENT OF SUSTAINABLE
MINERAL PROCESSING
Introduction
The chapter considers the impact of the study with respect to the current understanding of
efficient comminution and mineral liberation within the context of emerging global
phenomena such as climate change, energy resources, “peak oil” and environmental
degradation. This chapter provides an insight to the areas in which HPGR technology
could be used and implemented into the minerals industry, and identifies the commodities
that should be targeted for HPGR technology.
An energy balance on existing comminution practices has revealed that the “embodied
energy” of producing the grinding media in the first place is very high. Reducing or
eliminating their use in processes that consume large quantities of grinding media are not
eco-efficient and these processes should consider HPGR as an alternative sustainably
attractive technology.
An account of global comminution energy, global grinding media consumption and global
mineral grades and their relation to crushing and grinding processes within the mining and
minerals industry is reviewed here. The stand-alone results as presented are impressive.
However, the impact that HPGR has on the minerals beneficiation industry in general is
not yet realised.
Future sustainable development – “Energy efficient mineral liberation”

Comminution processes in minerals processing account for between 50-70% of the
electrical energy used in mining (DOE, 2001; National Research Council, 1981) and
sometimes a substantial portion of the direct operational costs. The potential of significant
increases in energy costs, capital costs and possible reduced revenue (as a result of lower
grades and not necessarily commodity prices) are factors that could possibly unbalance the
profitability of many mining operations in the future. This unbalancing could be
counteracted by researching how “energy costs” and “dollar costs” are utilised in
comminution processes. Then evaluating how the valuable mineral components are
liberated from the host ore using alternative technologies and circuit designs.
Sustainable development has the following consequences. Renewable energy, climate

change, environmental degradation, so-called “peak oil”, population growth and poverty
are currently highly controversial and debatable global topics. Does meaningful data exist
to substantiate scientific analysis to draw conclusions to these issues? Do current day
- 183 -

research projects provide adequate data that could have a major impact in partially solving
such problems?
When scientists speak of the future, what do they mean, next year, one or two decades
hence, the end of the twenty-first century, the end of the third millennium or forever?
When the future is examined within the context of “sustainable development”, generally
the idea is aimed at trying to understand how lifestyles will have to change to meet the
challenges posed by emerging global events such as climate change, peak-oil, limited non-
renewable energy resources and continued environmental and ecological degradation.
Fragments of “sustainable development” concepts have emerged in the form of labels
such as “sustainable society”, “sustainable energy” and “economic sustainability” to
mention a few. General consensus is that economics principles and business always has
and always will remain central to the transformation process of change and development.
The author’s view is that the boundaries of the roles and responsibilities of business in
society have shifted. Is business fulfilling its part of its contract with society and the
environment, and what exactly should the role of business be?
The current understanding of mineral processing is based on the following:
• Mining and plant performance is measured on tonnage throughput and quantity

of metal produced which by default is economically viable.
• Throughput in the plant is managed through various size reduction processes
which indirectly result in mineral liberation.
• The material size reduction process is managed through the operation of various
stages of crushers, mills, screens and classifying cyclones.
• The overall plant performance defines recovery factors which may affect the
mine plan and definition of the ore reserve.
• Unless optimised (usually in economic terms), the process will result in sub-
optimal exploitation of the mineral resource.
• The traditional perception of indiscriminate size reduction, as applied by most
conventional comminution processes over the past 100 years often result in
excessive generation of ultra-fine particles (both mineral and gangue material at
the expense of large amounts of energy), which are sometimes either difficult to
recover or store in downstream processes.
Focusing on effective resource consumption, reduction of waste and eco-efficiency

harnesses the business concept of creating value and links it to the environment.
Real issues facing the mining industry

Delegates at a recently held “Resourcing the future” conference of the global mining
initiative were asked in a survey to identify the three most significant issues facing the
mining and metals industry in the geographic area where the respondent was based. More
than 500 key industry people responded and identified the following as the top three:
1. Environmental concerns
2. Social & community concerns
3. Financial & economic performance of the industry
- 184 -

Other ranked issues identified in the global mining initiative survey are shown in Figure
144. (Pricewaterhousecoopers, 2002).
Figure 144 shows how key people representing the mining industry have prioritised the main issues facing the
industry.
The environmental concerns are almost certainly directly linked to energy resources and
the availability of (net) energy, as well as the environmental degradation caused by having
to mine more massive low grade ore deposits. These real and complex problems facing
the mining industry have spurred sustainable mining interests in energy efficiency and
reduced green house gas emissions. In the mining process as a whole, the energy intensive
process of comminution remains a focus for driving technological change, and research in
this area. Two very important aspects of comminution are energy consumption and
grinding media consumption. Both have a significant impact on the overall “dollar cost”
of the processes, and both are strongly linked with the sustainable drive to reduce the
overall “energy cost” or energy consumption per tonne treated or per tonne metal
produced. If the “energy cost” of producing the grinding media is factored into the
sustainable objectives, then reducing media consumption, or eliminating it altogether
could have a large “eco-efficiency” benefit to the overall process. Eco-efficiency being a
term defined as being able to do the same with less.
“Dollar cost” and “Energy cost” of mineral extraction.

Table 43 (Norgate and Rankin,2002) shows economical ore grades required for six major
metal commodities in 2000. Commodities at the time of writing are booming with much
higher prices being obtained for most non-ferrous metals. This cycle is reported to be
mainly driven by the demand for raw materials by China and India.
In the past year the price of many commodities has risen by more that 100%, and in the
case of copper up to 300%. These prices have the effect of making low grade ore deposits
economical, and a surge in existing plant throughput expansion projects. Overall “dollar
costs” to produce each commodity are well known and controlled because the “dollar
costs” are closely linked to the financials (Figure 145), but the associated “energy costs”
are not very well documented and understood from a sustainable perspective.
- 185 -

Table 43 Economic grades, reserves, production rates and years of supply (Norgate and
Rankin, 2002).
Economic Production Years of

ore grade Reserves in 2000 Years of reserve
Metal (%w/w) (Mt of metal) (Mt/y)1 supply 2,3 estimation
Iron/steel 30-60 65000 842 77 1995
Aluminium 27-29 3910 24.2 162 1997
Copper 0.5-2 320 14.8 22 1997
Lead 5-10 65 6.6 10 1997
Zinc 10-30 142 8.9 16 1995
Nickel 1.5-3 47 1.1 43 1995
Notes: 1. Includes primary and secondary metal.
2. Assumes consumption rate closely balanced to total production rate
3. Assumes no recycling
Figure 145 shows the relationship between the ore grade of several commodities and the “dollar cost” of
producing them.
Fortunately mineral resources are abundant in Australia and provide for a large percentage
of the export wealth, which favors the economist’s standpoint. In 2002/3, minerals
generated $42 billion in exports, representing 35 % of Australia’s total merchandise
exports (http://www.austrade.gov.au). Earnings from mineral and energy exports are
expected to account for 75 % of export revenue, at a record $84.8 billion, in
2005/6(Irvine, 2005). The “dollar cost” data of producing every commodity is well
documented because of the economic system under which it is managed. Each producing
- 186 -

mine may be ranked against the specified commodity’s annual global production as shown
in Figure 146.
Mine A (2004) 263.7 kt @ 38 Mine B (2004) 237.6

c/lb, 1.67 % of global copper kt @ 59.4 c/lb
Mine C(2004)
176.4 kt @ 31 c/lb
Figure 146 shows how global copper cash operating costs for the year 2004 are ranked for each operating
mine. Costs range from between 8 - 98 c/lb, producing a total of 15,7 million tonnes of copper.
This provides accurate information that is used mainly for investment decisions on the
various commodities that are traded in a volatile price market. Unfortunately “energy
costs” are not factored into the decision making process.
“Energy costs” are likely to increase in terms of the quantity of energy consumed as new
mineral processing plants become more energy intensive due to lower head grades and in
terms of the higher “dollar cost” of the energy. The increased “dollar cost” of the energy is
driven as a result of increased oil prices which in turn are being driven by the increased
“energy cost” required to produce the energy. Only new energy efficient processing
technologies can reverse this trend.
The minerals industry is a very energy negative industry, unlike the mining of energy
minerals such as coal, gas and petroleum where, although the “dollar cost” may escalate,
the “energy output” cannot exceed the “energy input” or “energy cost”. In the long term
the requirement for more energy efficient and cost effective devices will need to be
considered for the mining and minerals industry as well as energy resource industries.
A trend compounding the increasing “energy cost” of producing minerals today is the
necessity of having to find improved technologies to process massive, hard, low grade ore
bodies. In particular open pit copper mines are currently processing more than 200,000
tonnes per day which is four times the rate as in the 1960’s (Figure 147) (Brown, 2004).
As a result much larger amounts of gangue material or waste is required to be reduced,
and to much finer sizes than before due to finer mineralisation.
- 187 -

250000
Escondida
Grasberg
200000
NNES PER D AY
Chuquicamata
Bingham Canyon
150000
Batu Hijau
TONNES
100000
Bougainville
TO
Palabora
Collahuasi
Mount Tom Price
50000 Morenci
Andina
0
1900 1920 1940 1960 1980 2000 2020 2040
YEAR
Figure 147 shows the evolution of increased tonnes per day being processed in several large copper mines.
This increased production owes much to the adage “economies of scale” that have been
prompted by lower grades and relatively “cheap” and “readily accessible energy”. The
figure on the left hand side of Figure 148 shows how copper ore deposits grades have
decreased steadily over the past few decades where the minerals today are much more
finely disseminated within the host rock.
Other factors such as increased mechanisation and increased capacities of equipment such
as blast hole drills, shovels, loaders, trucks, crushers and mills have kept costs low. These
trends bring with them a range of management and engineering challenges, and one area
in particular which is being greatly affected is the mineral preparation stage of
comminution.
The figure on the right hand side of Figure 148 shows how mineral concentrations vary in
abundance within the earths crust and in the size of the ore body in which those minerals
are found. Over time the smaller rich deposits were mined out. Large massive ore-bodies
of relatively low grades (less than 0.5% in the case of copper) now remain. These mineral
resources are currently mined in both open pit and underground mines in a select number
of countries whose economies are often described as resource based economies. Australia,
Chile and South Africa are good examples of such countries and are expanded on briefly
in the next section.
For copper ores, the size of the ore body increases exponentially as grade decreases as
shown in Figure 149, which backs up Brown’s comments and Figure 147 on the trend of
increasing tonnages in mass mining. The typical lower grade trend for copper is
synonymous for most other mineral ore resources today. This coupled with increased
tonnages, is increased “energy cost” of extraction as Figure 150 shows. Figure 150
expresses the total energy in MJ/kg of metal produced that increases exponentially as the
ore grade is decreased. Economies of scale as previously mentioned have counteracted
this trend so far, all within a long cycle of depressed commodity prices (Chapman and
Roberts, 1983).
- 188 -

Figure 148 shows in the figure on left hand side how Copper resource grades have declined over the past 100
years after Ruth, (Ruth, 1995) and where the abundance of the various commodities as they exist in the earths
crust is found relative to estimated reserves in the figure on the right hand side (Kesler and Macmillan,
1994)
Figure 149 shows the how the size of a copper ore body increases with decreasing copper head grades (Kesler
and Macmillan, 1994)
Figure 150 shows the energy costs of mineral extraction with respect to ore grade composition in a study by
(Chapman and Roberts, 1983) for copper and nickel ores.
The total energy input per kg of product generally applies in the production of all
commodities, including the crushing and production of bulk industrial materials such as
aggregates, gypsum, limestone, potash and phosphate rock, and in the grinding of other
- 189 -

industrial bulk commodities such as cement and coal. An analysis of the “energy
intensity” and associated global “energy cost” in comminution of these bulk commodities
as well as ferrous and non-ferrous metals is given later in the chapter.
Comminution energy within a mining operation as a whole.

Mining is a very energy intensive industry and comes a close second to the petroleum
industry as shown in Figure 151. Energy is an important input in the mining industry,
where two major energy carriers, that of electricity and diesel fuel are predominantly used.
The diesel fuel energy is used for the transport of personnel, material and ore via mining
production machines and the electrical energy for the ore processing equipment. In
underground mining operations electrical energy is utilised in pumping water, ventilation
and refrigeration, with much of the electrical energy used in comminution processes.
A textbook view by Wills (1997) stated that grinding is the most energy intensive
operation in mineral processing. Wills quotes a survey by Joe (1979) of the total plant
energy consumed in a number of copper concentrators showed that the energy
consumption was on average 2.2 kWh/t for crushing, 11.6 kWh/t for grinding and 2.6
kWh/t for flotation. These energy consumptions are based on actual measured plant
energy of installed tumbling mill devices. The high grinding energy intensity of 11.6
kWh/t is similar to the energy intensity for most ore types that are being processed in large
copper porphyry mines today, with the exception that lower copper grades and often finer
mineralisation are becoming more common. Wills (1997) further stated that all ores have
an economic optimum grind size which depends on many factors, including the mineral
dispersion in the gangue and the subsequent separation processes. Since grinding is the
greatest single operating cost, the ore should not be ground any finer than is justified
economically.
gy Intensity vs. Energy Consumption
1000
Energy-Intensive Industries
Petroleum
Energy Intensity (Thousand Btu/$ GDP)
100
Primary Metals Paper
Chemicals
Mining Minerals
Textiles/Apparel
Wood
10
Food Processing
Tobacco/Beverages
Plastics/ Fabricated Metals
Furniture Rubber
Transportation
Leather Printing
Machinery and Computers
Miscellaneous Electrical
1
10 100 1000 10000

Energy Consumption (Trillion Btu)
Sources: EIA 2001, 1998 Manufacturing Energy Consumption Survey; U.S. DOE 2002, Energy and Environmental Profile
of the U.S. Mining Industry
Figure 151 shows the energy intensity of various industries, of which mining is one of the most intensive.
From an energy perspective, it is generally accepted that at least 50% of the US mining
energy bill (102 trillion British thermal units -btu) is used for mineral preparation. This
amount includes petroleum, diesel, electricity/coal and natural gas. (Figure 152). The
“dollar cost” of energy for comminution as a percentage of the total mining process costs
- 190 -

is much lower. In the future energy balances should be taken into consideration, as there is
a complex link between economic and energy cycles.
Figure 152 depicts the annual energy use in the US mining industry (US Department of Energy, 2001).
In South Africa a study by Kilani (2002) estimated that the mining industry accounted for
32 TWh of electricity which represents 18% of total national consumption. Electrical
energy costs accounted for 11.4% of the working costs in the Gold mining industry, and
8% in the coal mining industry. In terms of diesel fuel consumption, 515,231 kilolitres
was consumed which represents 8.7% of total national consumption. Bearing in mind that
diesel fuel costs have increased dramatically over the past two years and hence are likely
to increase the cost of mining coal, which in turn will increase the cost of electrical
energy.
In Australia, the mining industry accounts for 22% and 10% of the national diesel and
national electrical energy consumption (210 TWh) respectively (Trewin, 2005). The diesel
and electricity is consumed in predominantly truck-and-haul and mineral processing
activities. These high quantities of energy are typical of resource based economies, but
later it will be shown that the crushing and grinding component of global energy for the
processing of non-ferrous metals is relatively low.
These quoted numbers are useful because they provide some perspective of where energy
is used in mining and what percentage mining energy contributes to national energy use
for two major resource-based countries. Even though the data does provide a perspective
of energy consumption it is not detailed enough to provide a comparison on what type of
impact a new comminution devices will have on an energy.
Eco-efficiency and sustainable resource processing are important emerging criteria that
are driving new comminution technologies, but it should not be confused with the still
relatively low percentage that comminution energy in “dollar costs” have in respect to the
overall mining and recovery process costs. The new forms of energy accounting or
“energy footprints” and equivalent green house gas emissions are being factored into mine
management and new plant designs as exemplified in Figure 153.
- 191 -

This form of “sustainable mineral processing” is important because even though there
aren’t immediate large “dollar cost” benefits (capital and operational) in choosing a more
energy efficient comminution circuit that uses HPGR. An HPGR circuit when analysed on
the basis that little or no grinding media is used looks much more attractive when the so
called “unaccounted energy cost” is included into the equation. The “unaccounted energy
costs” are the total energy input that accounts for the production of the grinding media
itself, both in mining the ore and producing the iron. These “unaccounted energy costs”
and not the “dollar costs” should be factored into the decision making criteria on new
plant designs and comminution processes.
Figure 153 shows an example of the energy footprint being used nowadays by mining groups to manage
energy consumption and equivalent green house gas emissions (Angloplatinum, 2005).
Many major mining multinationals are today, as part of their sustainability objectives, are
aiming to reduce the total energy input per tonne or per ounce of metal produced as
exemplified in the aluminum industry (Figure 154). Interestingly the aluminum industry
has several energy intensive processes other than comminution which has led to
Aluminum being almost environmentally unfriendly. HPGR processing should rather
target comminution intensive processes where fine grinding and high grinding media
consumption is concerned. This represents specifically the hard, low grade ores, and if
“unaccounted energy cost” and not “dollar cost” be considered, then the case for HPGR
could be much more convincing. More on this justification is given later in the review.
- 192 -

Megawatt hours per tonne of aluminium produced
16,200
Commodity - Aluminuim
16,000
15,800
15,600
15,400
15,200
15,000
1988 1990 1992 1994 1996 1998 2000 2002 2004 2006
year
Figure 154 shows how the total energy consumption per tonne of Aluminium produced is steadily decreasing
as a result of improved efficiency and technologies (Marks, 2006).
In the next two sections, global comminution energy consumption is analysed and
presented. Major mineral and industrial commodities that are responsible for mainly
crushing and grinding processes are identified, and in particular the type of ore resource
that should be studied with respect to HPGR possibilities. This analysis is based on data
detailing global electrical energy generation, global comminution energy and global
commodity production. This data is then compared with a similar energy based analysis
that hinges on the estimated global grinding media consumption and the associated ore
tonnages that are treated in the various types of grinding media consuming comminution
devices such as SAG, rod and ball mills.
Comminution energy within a global context.

Comminution is known to be extremely energy intensive, but placing it in the context of
global energy use has been very difficult due to the lack of coordinated or regulated data
capture. Several researchers (Fuerstenau, 1983; Fuerstenau, 1992; Fuerstenau and
Abouzeid, 2002; Fuerstenau and Kapur, 1995) have previously stated that the electrical
energy consumed globally for comminution processes was of the order of 3–4%. These
cited figures have been commonly based on a report published by the committee on
comminution and energy consumption in 1981(National Research Council, 1981) At
about this time it was estimated that in the US alone, the energy consumption for crushing
and grinding of solids was in excess of 29 TWh, a figure that was re-quoted for
comminution energy consumption in 1997 (DOE, 2001).
In a recent edition of “State of the World – Progress towards a Sustainable Society, 2003”
(Sampat, 2003), it is claimed that electrical energy consumption in the mining industry
accounts for 7-10 % of the global electrical energy of 4,900–6,600 TWh). Although
Figure 155 shows that the global net electrical energy consumption in 2004 was much
higher at 15,500 TWh or 15,500 billion kWh. The global comminution energy should
however be estimated using the 4,900-6,600 TWh consumption figure which may exclude
electrical energy transmission losses. If it is assumed that 50%-70% of the electrical
energy consumption within the mining industry is attributable to mineral preparation or
- 193 -

more specifically comminution processes (DOE, 2001) then globally, about 540 to 775
TWh of electrical energy is consumed in comminution processes. This then represents
3.5-5 % of the total global electrical consumption based on a consumption of 15,500 TWh
and represents all comminution processes. The increase in global energy consumption in
comminution processes from 3-4% (1981) to 3.5-5% (DOE, 2001) may be directly
attributed to the mining of larger, more massive and often lower grade ore bodies and the
fact that the 3-4 % quoted in previous years was based on data from before 1981. The 3.5-
5% however, does not reflect any associated increase in global electrical energy which has
risen from 7500 TWh in 1980 to 15,500 TWh in 2004. As such the mining industry is
now using far more energy in 2004, than it did before 1981.
16,000
15,000
14,000
TWh or Billion Kilowatthours
13,000
12,000
11,000
10,000
9,000
8,000
7,000
World total net electricity consumption-IEA 2006
6,000
1980 1984 1988 1992 1996 2000 2004
Year
Figure 155 show how world “net” electrical energy consumption has doubled since 1980(EIA, 2004)
No-where in the literature is the split between crushing and grinding energy consumption
ever mentioned. This leads to the assumption that grinding processes are responsible for a
large proportion of the global electrical energy consumption in comminution because
grinding is known to be far more energy intensive than crushing. In the next section an
attempt is made to provide an analysis of energy split between crushing and grinding
energy consumption based on steel ball grinding media consumption.
Global comminution energy based on grinding media consumption.

Since this study focuses on aspects relating to quantification of the energy that is
consumed in comminution circuits that employ HPGR and the effects that HPGR
processing has on mineral liberation. The potential to reduce energy consumption and
improve mineral recovery seem very attractive from an economical and sustainable energy
perspective. However part of the cost analysis involves the quantification of the
consumption of grinding media in conventional circuits to that of HPGR circuits which do
not require grinding media, other than the roll liners.
So if the quantity of global grinding media production is known, and if a grinding media
consumption rate for SAG and ball mills is known, then it could be possible to estimate
the quantity of material that is subjected to grinding/milling processes. If the tonnage of
- 194 -

each processed commodity ore is known, and an estimate of the respective energy
intensity is made then it is possible to estimate the amount of energy that will be
consumed globally for each commodity in crushing and grinding.
Searching the web, estimates of global grinding media production were sourced and are
given in Figure 156 and based on data presented Horsburgh (2005b) of Smorgon steel, a
manufacturer of steel ball grinding media. An analysis of the data presented in Figure 156
shows that global steel grinding medial production according to Smorgon steel is
equivalent to 2.5 million tonnes of steel per annum(Horsburgh, 2005a; Horsburgh,
2005b).
The estimated global steel production in 2005 was 1004 million tonnes which is
illustrated in Figure 157.(http://www.steelonthenet.com/pdf/Consline_0405.pdf). Based
on the data shown in Figure 156, the 2.5 million tonnes of steel ball grinding media
production and consumption thus represents 0.249 % of global steel production. This is
negligible in relation to the total amount of steel used in industry. Later in this chapter it
is shown that grinding energy consumption by comparison to global comminution energy
consumption is much lower than the quantity of energy used in crushing which includes
the crushing of non-ferrous and non-metallic commodities such as cement, coal, and
building material aggregates.
Grinding media costs are largely dictated by global steel prices, which in the case of the
mining industry are governed largely by the fixed steel price contracts that the mining
companies have with steel producing blast furnaces in various parts of the world.
Producing the media of different sizes and strength specifications requires an added cost
factor during the manufacturing process of the balls which is estimated to be of the order
of 1.4 – 1.8 times the price of the steel on the global market.
0.8
0.7 Global grinding media
Million tonnes
0.6 production 2.5 million

0.5
0.4
0.3
0.2
0.1
0
a
C ro
O g
s
C V ar w
ta el
er x
a
in Me l
rS .
l
Sm (Va us)
nt )
l
ee
Fo tee
A An
de t l
)
th ag e e
Sa dau
er
t in
ps
au
go us
ca
D ste
st ce
in
th
st
t
io
ta tte
as
or r io
rm in p/S
a
S
ca rja
B ng
lt a
n
o
G
ri
o
el
r(
C
ee
(
or
a
or M
y-
IS
ol
or
ng
F o Ch
M
er
N
E
gy
N
er
En
Figure 156 shows the cumulative production of world steel ball grinding media, and the companies that
represent the global production market (Smorgon-Steel-Group, 2005).
Grinding media price depends largely on the country in which the media is manufactured
and the specifications (size and hardness) and quality (chrome and carbon content) of the
media. Rapidly increasing steel prices could be expected in the future which is being
driven mainly by increased demand for steel in China as shown in Figure 157. This
- 195 -

overall knock on effect will eventually lure large low grade iron ore, copper and gold
operations to consider HPGR as an alternative to large media consuming tumbling mills
in the future.
Figure 157 quantifies global steel production and shows the steady increase in production.
Other mineral commodities may consider HPGR from a media cost perspective, but will
focus mainly on potential recovery benefits that are realised, especially with high value
commodities such as the platinum group metals and base metals. In the next section, the
potential reduction in steel media consumption is evaluated from a sustainable and eco-
efficient perspective as an option to consider in the future.
Eco-efficiency and the “energy cost”(Emergy) of steel grinding media

The wear of mill liners and grinding media add to the “dollar cost” and “energy cost”,
sometimes referred to as emergy. Emergy defined as the “embodied energy”, that is, the
quantity of energy necessary for the fabrication of a specific material. When measuring
embodied energy, all energy inputs, from raw material extraction, to transport,
manufacturing, assembly, installation and others are considered. Embodied energy as a
concept seeks to measure the true energy cost of an item. The need for continual cost
reduction and increasing regulatory and social pressure relating to greenhouse emissions
create a strong need for improvements in comminution energy efficiency and associated
sustainable development initiatives.
New eco-efficient and sustainable development initiatives are not necessarily linked to
“dollar cost” savings, but rather the indirect “energy cost” savings through a targeted
significant reduction in the consumption of grinding media. This is attractive because of
the high energy intensity required to produce the steel in the first place. Typical
consumption figures for coal and coke is 700 kg per tonne of finished steel in the U.S.
which is based on an ore grade of 62.5% iron. This coal consumption rate is equivalent to
an approximate energy usage of 5,320 kWh per tonne of finished steel. The consumption
- 196 -

of oil, natural gas and electrical power in addition to coal usage is 710 kWh per tonne
finished product. This results in a total “energy cost” of 6030 kWh per tonne finished
steel (Figure 158).
In terms of eco-efficiency this unaccounted “energy cost” implies the use of 6 kWh per kg
of steel grinding media. In a typical SABC grinding circuit the average grinding media
consumption rate is 0.7 kg of steel per tonne of ore treated which is equivalent to an
additional 4.2 kWh/t of indirect and un-accounted energy. The 4.2 kWh/t represents about
33% of equivalent additional energy effort in a comminution circuit that has an
operational work index of say 12.6 kWh/t (SvensktStal, 2006).
Figure 158 depicts an energy balance over the production of each tonne of the steel grinding medial
Share of comminution energy in crushing and grinding

Outside of increasing “dollar costs” and “energy cost” of grinding media, it is interesting
to back-calculate the amount of ore being ground in milling processes globally, and thus
estimate the corresponding quantity of electricity that is consumed in milling processes.
As a rule, media consumption at copper and gold processing circuits using carbon alloy
steel balls is of the order of 0.3 - 0.35kg of steel /tonne of ore milled in SAG mills, and
0.4 - 0.6kg of steel/tonne of ore milled in ball mills. If it is assumed that on average 0.7 kg
of steel is consumed per tonne of ore milled in a conventional SABC grinding circuit, then
approximately 3286 million tonnes of ore would be processed each year based on the 2.3
million tonnes of grinding media produced.
If on average the total comminution energy intensity is 25 kWh/t, then the total specific
energy to mill the ore would be 82 TWh. Assuming a 5% no-load power inefficiency
(large mills) then the total specific power required would be 86 TWh that is used in non-
ferrous metal comminution processes. Total global electrical energy consumption is
15,500 TWh (Figure 155), hence the global electrical energy consumed in grinding media
mills is of the order of 0.55 % which is significantly lower that the energy consumption
percentage of 3.5 – 5 % quoted in the previous section.
In an attempt to clarify the low milling share of comminution energy, a summary of major
global mineral commodities has been assembled as shown in Table 44 and Table 45. In
these tables the energy intensity of the comminution operation is estimated and the total
comminution energy is estimated based on estimated volumes and grade of each
commodity. These ore volumes are based on the average grade for each commodity as
previously referenced in Table 43.
Not surprisingly it is shown in Table 44 that even though all non-ferrous metal
production process are assumed to use tumbling mills and grinding media, the total
electrical energy is estimated to total 87 TWh for all commodities listed. The 87 TWh
corresponds closely to the 86 TWh based on grinding media consumption. From Table 45
- 197 -

it is clear that other bulk commodities such as steel, cement, aggregates and coal are
produced in massive quantities with much lower energy intensity due to crushing
requirements only. The large tonnages amount to a very large comminution energy
consumption of 225 TWh. The total comminution energy of both bulk materials and
metals production is 312 TWh which is equivalent to 2.01 % of global net electrical
energy consumption at 15,500 TWh. Table 44 and Table 45 confirm that milling
processes, although much more energy intensive, consume much less energy than do
crushing processes overall. HPGR should target milling processes rather than crushing
processes because crushing process are reasonably efficient. Milling processes on the
other hand are less efficient and consume costly grinding media.
This analysis is important because is re-confirms why HPGR technology has already
successfully been applied into the cement and iron ore industries. The results highlight
why HPGR technology is currently seriously being considered in the copper industry with
the installation at the Cyprus Sierrita copper mine in Arizona (Thompsen et al., 1996),
and more recently at the Cerro Verde copper project in Peru (Vanderbeek, 2006). The
gold industry is exploring with HPGR with an installation in Russia (Kirsch, 2004), and
more recently with an installation at the Bendigo gold mine in Australia (Pyke, 2006;
Johansen et al, 2005) and Newmont’s Boddington gold project (LeVier et al 2004)..
The data in Table 44 confirms that approximately 0.56 % (87 TWh) of the global net
electrical energy consumption of 15,500 TWh is used to crush and grind non-ferrous ores
to produce most major metal commodities. This comes from crushing and grinding an
estimated 3360 million tonnes of ore per annum based on the total annual production of
the metals and is comparable to the 3285 million tonnes of ore processed per annum based
on the global steel grinding media consumption.
In the bulk industrial mineral industries of coal, iron ore, cement and aggregates,
comminution energy intensities, with the exception of cement, are very much lower than
those for metal commodities. The tonnages are however much larger and account for
more than 2.5 times as much electrical energy (225 TWh) as is consumed in comminution
processes within the metal commodity industries. This represents 1.45 % of the global
electrical energy consumption. The analysis of estimated comminution energy
consumption of all major non-ferrous ores is useful so that potential HPGR applications
might be placed into perspective where the impact to global energy consumption is
concerned.
- 198 -

Table 44 show where energy is being consumed in the global materials and minerals
industry, and the status of where HPGR technology is being applied.
Average
Areas of Crushing Grinding estimated
HPGR circuit circuit Bulk ore comminution Estimated
Grinding applications (jaw and (SABC) processed energy Comminution Mineral ore Annual global
Study ore sample type media in the cone) energy energy (millions of intensity energy @ ave grade / mined comodity
and organisational Major Metal used minerals intensity intensity tonnes per (kWh/t) for energy bulk production
origin commodity (yes/no) industry (kWh/t) (kWh/t) year) process intensity (TWh) material tonnes
Tin yes Not used 0.5-2 10-25 9 20 0.2 08-1% 90,000
Nickel yes Not used 0.5-7 10-25 65 20 1.3 1-3% 1,500,000
Manganese yes Not used 0.5-3 10-25 20 20 0.4 45-60% 9,790,000
Xstrata, Mt Isa (Pb,Zn) Lead yes Not used 0.5-5 10-25 85 20 1.7 7.50% 3,280,000
Xstrata, Mt Isa (Pb,Zn) Zinc yes Not used 0.5-6 10-25 70 20 1.4 15% 10,100,000
BHPBilliton, Bauxite Aluminium yes Not used 0.5-3 10-25 165 20 3.3 25-30% 31,200,000
Exploring, two
Platinum
Lonmin, (UG2 ore) yes installations 0.5-8 20-35 90 30 2.7 3-4 g/t 218
/Chrome
planned
Exploring, two
Chrome /
Lonmin, (UG2 ore) yes installations 0.5-8 10-25 36 20 0.7 40-60% 18,000,000
Platinum
planned
External to this study, Exploring, two

yes, major
Boddington and Gold
consumer
installations, 4 0.5-8 20-35 960 30 28.8 3-4 g/t 2,450
Bendigo expansion planned
External to this study,
Exploring, one
Cyprus sierrita and yes, major
Cerro Verde expansion
Copper
consumer
installation 4 0.5-4 20-30 1860 25 46.5 0.5-1% 14,900,000
planned
projects
TOTAL -major
commodies
3360 87.0
Table 45 show where energy is being consumed in the global bulk materials industry, and
the status of where HPGR technology is being applied.
Average
Areas of Crushing Grinding Bulk mined estimated
HPGR circuit circuit ore comminution Estimated
Major mined Grinding applications (jaw and (SABC) processed energy Comminution Mineral ore Annual global
Industrial media in the cone) energy energy (millions of intensity energy @ ave grade / comodity
mineral used minerals intensity intensity tonnes per (kWh/t) for energy bulk production
commodity (yes/no) industry (kWh/t) (kWh/t) year) process intensity (TWh) material tonnes
Established
Diamonds no 40 units 0.5-7 5-15 240 8 1.9 50 cts/100t 24
installed
Potash no Not used 0.5-4 N/A 31 3 0.1 100% 31,000,000
Phosphate no Not used 0.5-4 N/A 148 3 0.4 100% 148,000,000
Gypsum no Not used 0.5-2 5-10 110 4 0.4 100% 110,000,000
Limestone no Not used 0.5-2 5-10 128 4 0.5 100% 128,000,000
Established
Cement clinker
yes more than 0.5-4 30-50 2200 57 125.4 100% 2,200,000,000
grinding
400 units
Established
Iron ore/steel yes more than 20 0.5-2 5-16 1300 9 11.7 40-62.5% 1,000,000,000
units
Agregate
no Not used 1-6 N/A 6000 4 24.0 100% 6,000,000,000
crushing (est)
Coal grinding
power station
yes Not used N/A 10-18 5500 11 60.5 100% 5,500,000,000
and blast
furnace (est)
TOTAL -all
materials
15657 225.0
From the analysis, it would be deemed reasonable that HPGR technologies should be
targeted towards gold and copper mining in the future because of the very large low grade
tonnages involved that require both crushing and grinding. The analysis has indicated
other important mineral ores that should consider HPGR as an alternative energy saving
comminution device would be commodities such as platinum, aluminum and lead/zinc,
which co-incidentally were the chosen ores for this study.
- 199 -

The data assembled in Table 44 and Table 45 are sourced mainly from the US geological
survey (USGS) web site( http://minerals.usgs.gov/minerals/pubs/commodity/ ) (USGS,
2006). Data representing the platinum/chrome ores overlap which depends on whether
chrome is seen as the product or platinum. The same logic applies to copper and gold ores
where some copper mines produce gold as a by-product. These multi-mineral ore are
assumed not to affect the content of Table 44 with respect to estimating ore tonnages
processed and corresponding comminution energy consumptions. The same applies to
comminution energy intensity where a lack of co-ordinated data capture has led to best
estimates of the energy used in the respective comminution processes.
Summary
Mines in the future may be required to produce products within a new framework of
sustainable and eco-efficient development that include aspects of global energy and
environmental restriction policies. These formidable changes and the ever increasing cost
of energy (particularly if generated from non-renewable energy resources), global
warming and climate change, will eventually force the mining industry to seek alternative
sources of electrical energy coupled with greatly improved energy and eco-efficient
comminution strategies
Greater eco-efficiency can be realised in the reduced wear of mill liners and grinding
media. Though the “dollar cost” of comminution is accounted for as a cost consumable in
the process, it is rarely considered for its “energy cost” or the “embodied energy” of
having to manufacture the steel in the first place. Eco-efficient and sustainable
development initiatives are not always required to be linked to “dollar cost” savings.
Rather the direct and indirect “energy cost” savings and the impact on the environment
should be targeted. Reducing grinding media consumption is attractive from a sustainable
viewpoint because the high energy requirements to produce the steel are eliminated. These
aspects are reviewed and it is shown that 0.56% (87 TWh) of the global net electrical
energy consumption of 15,500 TWh per annum is used to crush and grind non-ferrous
ores, and of this 33% and 53% of the energy is required to process gold and copper ores
respectively. This suggests that the HPGR should be targeted at gold and copper mining
operations in the future to be effective in reducing carbon emissions. As such new eco-
efficient flowsheets that use multiple HPGR’s in series could be considered. HPGR would
be employed as primary comminution devices followed by small ball mills.
The step change improvement in the overall eco-comminution process efficiency offers a
real potential of playing an important role in reducing the rate of current energy resource
consumption and depletion, whilst new renewable energy resources are exploited and
developed. This could be achieved through a renewed approach to comminution, through
improved energy efficiency, improved mineral liberation, and through overall energy
conservation. The study has the potential to reduce environmental impact by reduced
water consumption and also by changing the properties (particle size) of the waste
products produced in the mining and minerals processing industry, but these aspects are
beyond the scope of this work.
- 200 -

CHAPTER 8
CONCLUSIONS & RECOMMENDATIONS

Introduction
The drivers of change should be explored in most applied research projects so that the
impact of the research is ascertained and placed into perspective. This enables the work to
be placed within the context of change and progress. Energy and mineral resources are
fundamental to human well-being. They are integral parts of virtually every sector of the
economy and provide the essential raw materials needed to build modern infrastructures.
The management of energy and mineral resources is often a controversial topic because of
the potential for negative environmental and social impacts stemming from nonrenewable
resource production and depletion. Economists focus on the threat to the economy if
extraction does not occur, while environmentalists focus on the threat to the natural
environment if it does.
This thesis has investigated the potential for improved, eco-efficient processes within the
mining industry. In particular the use of a new technology, HPGR was studied with a view
to using the device as an alternative to existing tumbling mill devices such as SAG and
ball mills. The use of the HPGR resulted in focused experimentation which was carefully
planned and executed. The results were supported through the use of existing
measurement techniques and literature which has resulted in convincing arguments that
show that significant improvements in terms of energy efficiency are possible.
Energy use and efficiency will remain as the key components of the sustainable
development debate as suggested by several researchers (Jefferson, 2006; Marechal et al.,
2005; Voorspools, 2004). However it is the authors opinion that a new eco-efficient
comminution standard could challenge conventional wisdom, and could result in new
flowsheet designs that have different objectives. Although it is envisaged that the
proposed flowsheet will look familiar, in terms of energy consumption both direct and
indirect will be very different. Should indirect energy be incorporated into the decision
making process the new flowsheets could possibly eliminate existing tumbling mill
devices and costly grinding media within the minerals processing industry.
Direct Bond ball mill energy measurements

The specific comminution energy for milling is traditionally based on the Bond method.
The method appears to be based on 60 J/rev, which defines the net energy required in the
Bond ball mill to realise the same conditions in a 2.4 m wet grinding mill. Bond stated
that the net energy input to the mill is 93 J/rev which is comparable to 91.44 J/rev as
measured using the new digital energy meter. This value has been measured and estimated
using various other means such as the Levin method, DEM, and the power models of
Morrell and Fuerstenau. The results are in good agreement, with the exception of the
Levin method which achieves similar results, but for different reasons.
- 201 -

In general each of the mill power models and the energy measurements are in agreement
with the Bond and digital energy meter. The Bond energy for the balls prediction appears
to have the greatest variation, where-as measurements on the ball charge and ore are
similar. The slight variations in the results could be attributed to the experimental Bond
mill operating at 71.66 rpm as opposed to the specified 70 rpm.
The new digital energy meter has confirmed that small electrical energy measurements in
motors/mills are possible, and that the specific comminution energy is measurable. This
device has effectively opened up a new window of comminution research, especially
where energy consumption and energy efficiency are important.
The summarised direct digital energy meter measurements as shown in Figure 118
represent a direct measured energy size relationship, which may be compared with the
relationships derived by the researchers Rittinger, Kick, Bond, Hukki and Morrell.
Energy efficiency – Hybrid triple pass HPGR/ball mill circuit

Overall, the best results obtained in terms of energy savings using the direct energy
measurements in the Bond mill (93 J/rev) and triple pass hybrid HPGR flowsheet may be
summarised by choosing the maximum energy difference that was measured at a
particular closing size. This means that an overall energy saving of 35.6% at a closing size
of 300 Om is achievable with the lead/zinc ore, a 41.8% energy-saving at a closing size of
425 Om is achievable on the bauxite ore and a 23.7% energy-saving is achievable on the
lead zinc ore at a closing size of 75 Om. This of course does not provide any information
with regards to the performance of the recovery process that follows, only the liberated
mineral characteristics were measured using the MLA.
The lead zinc ore and the bauxite ore display a similar energy size relationship to that
which was proposed by Bond, Hukki and Morrell. The results verify that energy size
relationships are not independent of the comminution device used. The properties of the
ore in respect to the mineralisation and grain boundaries must have an effect on how
energy is utilised in the size reduction or fracture process.
Particle grind size to energy relationship is not simple. Results provide evidence that the
theories developed by Charles and Holmes are relevant and provide a better understanding
of the behavior of grind size to energy requirements. Both Charles and Holmes stated that
the size/energy relationships are dependent on the characteristics of the ore as well as the
characteristics of the breakage process that the various different types of comminution
devices impart on the ore particles during size reduction.
The results confirm that energy efficiency and the measurement of energy savings in
absolute terms in comminution circuits/processes are complex. There are well established
methodologies that produce quite different results to the direct measurement results.
However within the context of the literature review, the methodologies presented in this
report are supported with the experimental results, and highlight the limitations of each of
the methodologies followed.
This research however provides a baseline (in terms of energy consumption) from which
to work from, in order to make further evaluations and assessments with regards to the
liberation characteristics of the ores when they have been subjected to HPGR processing.
- 202 -

HPGR micro-cracking and mineral liberation
HPGR does produce particle micro-cracking, this was a conclusive result that has up until
now been claimed as an artifact of HPGR. In this analysis several direct comparisons
between ore particles that have been exposed to HPGR and those not, were clearly
presented. The thesis hypothesis of trying to determine weather HPGR promotes or
enhances preferential mineral liberation was not founded as part of this investigation.
Although in this study preferential mineral liberation at fine size fractions was not
evident, there certainly exists scope for further investigations as to preferential grain
boundary liberation as coarser particles sizes with the use of HPGR. In light of this
statement further research in to liberation characteristics of HPGR processed ores should
be conducted in the future.
Main thesis conclusions
• HPGR is more energy efficient and may reduce overall grinding energy between
20-40% depending on the target grind size.
• HPGR readily promotes particle micro-cracks, which is suitable for heap leaching
ores, and the reduction in downstream mill energy requirements.
• Generally micro cracks are destroyed in downstream ball mills and could affect
flotation response and reagent consumption.
• HPGR products are more easily milled downstream due to particle weakening and
observed microcracks.
• Preferential liberation not easily quantified.
The direct energy measurements of today were not available at the time when Bond
developed his empirical equations. As such, over the past 100 years, comminution
performance has often been based on particle size energy relationships. Modern methods
of texture analysis and mineral liberation analysis using a scanning electron microscope
have enabled innovative techniques for the measuring of energy -liberation approach. In
the future the targeted comminution process should be to maximise the mineral liberation
as opposed to the traditional method of grinding all the ore to a pre-defined particle size.
Future HPGR installations into the minerals industry may well be driven by imminent
changes to energy and environmental policies and the perceived gradual or sharp increase
in the cost of energy. In addition to this HPGR may well become much more attractive
owing to the reality of having to mine and process larger, more complex, lower grade ore-
bodies. These factors will eventually force the mining industry to adopt energy efficient
technologies such as the HPGR.
- 203 -

Recommendations
It is recommended that the direct energy measurement method developed in this thesis be
used in future comparative comminution energy studies. The more efficient triple pass
HPGR circuit that is closed with a small ball mill should be investigated further at pilot
scale. Comminution energy that was shown to be dominated by new surface area
generation should be further investigated with respect to applying and confirming the-
Holmes and Morrell particle size-energy relationships.
The energy costs are mainly associated with electrical energy and the direct and indirect
“dollar costs” and “energy costs” coupled with increased use of grinding media. In the
future it is quite possible that a more serious situation may arise when energy costs and
consumption increase rapidly resulting in the energy industry not being able to supply
sufficient quantities of energy, and not being able to maintain the rate at which renewable
energy is generated. A futuristic approach to mineral beneficiation could be based on the
following:
• The mine managed under “optimal exploitation of the resource”, and

corresponding profit/revenue ratios within the context of “sustainable mining”.
• Mineral liberation processes are managed through optimal energy (cost /
sustainability) and recovery processes (revenue).
• Optimisation of the measured liberation and energy parameters will ensure
optimal exploitation of the ore resource from a comminution perspective.
• The energy efficient liberation approach emphasizes the importance of value
tracking the mineral in the liberation process. Any improvement in the
efficiency of the comminution and liberation process will impact directly on the
control of environmental damage, conservation of energy and the economics of
the operation.
• This approach to comminution will ensure that the comminution energy is
utilised efficiently for the sole purpose of mineral liberation, and will provide
mines with data to substantiate sustainable process routes that may be
economically tailored to the mineralisation of the ore-body.
• Therefore the challenge is for the mining industry to continue to meet the
demand for minerals in a climate of declining head grade and price, whilst
ensuring the industry’s commitment to sustainability is not compromised.
Mineral liberation analysis in terms of particle tracking and MLA data mass balancing
should be considered in the future. This is currently a topic of research within the JKMRC
,at the University of Queensland. This type of analysis is a means to confirm if there is
indeed a process that promotes or enhances preferential mineral liberation.
Finally the concept of sustainable mineral processing and reduced energy consumption
should be investigated further by considering a pilot scale triple pass HPGR project. The
results of such a project should include an analysis of the processing benefits with respect
to the economics of the process concept and with respect to current emerging global
phenomena such as global warming, carbon dioxide emissions and energy resource
depletion.
- 204 -

REFERENCES
Agar, G.E. and Brown, J.H., 1964. Energy requirements in size reduction.
Angloplatinum, 2005. Sustainable development report, 2005.
http://www.angloplatinum.investoreports.com/angloplatinum_ar_2005/downloads/ang
loplatinum_ar_2005_br_full.pdf: pp 50.
Apling, A. and Bwalya, M., 1997. Evaluating high pressure milling for liberation enhancement
and energy saving. Minerals Engineering, 10(9): 1013-1022.
Arbiter, N. and Harris, C.C.,1982, Scale-up and dynamics of large grinding mills – a case study.
In :A.L. Mular and G.V. Jergensen II (Editors), Design and Installation of
Comminution Circuits. AIME, New York, Ch. 26, pp. 491-508.
Armstrong, D.G., 1986. An alternative grindability test. An improvement of the Bond procedure.
International Journal of Mineral Processing, 16(3-4): 195-208.
Austin, L.G. and Brame, K., 1983. A comparison of the Bond method for sizing wet tumbling
ball mills with a size--mass balance simulation model. Powder Technology, 34(2):
261-274.
Austin, L.G., 1973. A commentary on the Kick, Bond and Rittinger laws of grinding. Powder
Technology, 7(6): 315-317.
Austin, L.G., Hilton, W. and Hall, B., 1992. Mill power for conical (Hardinge) type ball mills.
Minerals Engineering, 5(2): 183-192.
Austin, L.G., Weller, K.R. and Lim, I.L., 1993. Phenomenological modelling of the high pressure
grinding rolls, 18th International Mineral Processing Congress. Published by
Australasian Inst of Mining & Metallurgy, Parkville, Australia, pp. 87.
Battersby, M.J.G., Kellerwessel, H. and Oberheuser, G., 1992. Important advances in the
development of high pressure rolls comminution for the minerals industry. Extractive
Metallurgy of Gold and Base Metals, Kalgoorlie, Australia, pp 159-165
Battersby, M.J.G., Kellerwessel, H. and Oberheuser, G., 1993. High pressure particle bed
comminution of ores and minerals. World Cement, 24(11): 19-23.
Baum, W., Patzelt, N. and Knecht, J., 1996. The use of high pressure roll grinding for
optimisation of copper leaching. Proceedings SME Annual Meeting, Phoenix,
Arizona, Preprint 96-68.
Becker, M., Kwade, A. and Schwedes, J., 2001. Stress intensity in stirred media mills and its
effect on specific energy requirement. International Journal of Mineral Processing,
61(3): 189-208.
Bernhardt, P., 1996. Investigations into the process dynamic and control of a laboratory agitator
bead mill. International Journal of Mineral Processing, 44-45: 629-640.
Bond, F.C. and Maxson, W.L., 1938. Grindability and grinding characteristics of ores, American
Institute of Mining and Metallurgical Engineers - Meeting, Feb 1938. American
Institute of Mining and Metallurgical Engineers - Technical Publications -- Mining
Technology. American Institute of Mining and Metallurgical Engineers, New York,
NY, United States, pp. 22.
Bond, F.C. and Maxson, W.L., 1943. Standard grindability tests and calculations, American
Institute of Mining and Metallurgical Engineers - Meeting, Feb 1943. American
Institute of Mining and Metallurgical Engineers - Technical Publications. American
Institute of Mining and Metallurgical Engineers, New York, NY, United States, pp. 11.
Bond, F.C. and Wang, J., 1950. New theory of comminution. American Institute of Mining and
Metallurgical Engineers - Mining Engineering, 187(8): 871-878.
- 205 -

Bond, F.C., 1940. Wear and size distribution of grinding balls, American Institute of Mining and
Metallurgical Engineers - Meeting, Feb 1940. American Institute of Mining and
Metallurgical Engineers - Technical Publications. American Institute of Mining and
Metallurgical Engineers, New York, NY, United States, pp. 12.
Bond, F.C., 1941. Measuring surface area in grinding, American Institute of Mining and
Metallurgical Engineers - Meeting, Feb 1941. American Institute of Mining and
Metallurgical Engineers - Technical Publications. American Institute of Mining and
Bond, F.C., 1947. Standard grindability tests tabulated, American Institute of Mining and
Metallurgical Engineers - Meeting, Mar 1947. American Institute of Mining and
Metallurgical Engineers - Technical Publication. American Institute of Mining and
Bond, F.C., 1949. Standard Grindability Tests Tabulated. AIME Trans, Vol. 183,: pp 313.
Bond, F.C., 1952a. Crushing and grinding calculations. Pit and Quarry, 45(5): 118-119.
Bond, F.C., 1952b. The third theory of comminution. Transaction AIME (Mining), 193: 484-494.
Bond, F.C., 1954. Crushing and grinding calculations. Canadian Mining and Metallurgical
Bulletin, 47(507): 466-472.
Bond, F.C., 1957. Comminution exposure constant by third theory. Mining Engineering, 9(12):
1372-1376.
Bond, F.C., 1958. Grinding ball size selection. Mining Engineering, 10(5): 592-595.
Bond, F.C., 1959. Confirmation of the third theory. American Institute of mining, metallurgical
and petroleum engineers, Allis Chalmers 07R8731A.
Bond, F.C., 1960. Three principles of comminution. Mining Congress Journal, 46(8): 53-56.
Bond, F.C., 1961a. Crushing and grinding calculations. British Chemical Engineering, 6(6, 8).
Bond, F.C., 1961b. Crushing and grinding calculations. Allis-Chalmers Publication,
No.07R9235B.
Bond, F.C., 1962. New ideas clarify grinding principles. Chemical Engineering, 69(3): 103-108.
Bond, F.C., 1963a. Particle size reduction -- Theory and practice. Canadian Mining Journal,
84(6): 43-48.
Bond, F.C., 1963b. Some recent advances in grinding theory and practice. British Chemical
Engineering, 8(9): 631-634.
Brown, E.T., 2004. "Keynote address" JKMRC International Student Conference,. Brisbane 6-7
September, 2004.
Brundiek, V.H., 1994. Loesche mill for comminution of cement clinker and inter-ground
additives in practical operation. ZKG International, Edition B, 47(4): 179-186.
Bush, P.D., 1967. The Breakage of Mineral Particles in Ball Mills. PhD Thesis, JKMRC,
University of Queensland, |Australia.
Chandramohan, R. and Powell, M.S., 2004. Measurement of Particle Interaction Properties for
Incorporation in Discrete Element Method Simulation. University of Queensland,
JKMRC International Student Conference,, Brisbane 6-7 September 2004,: pp 379-
398.
Chapman, P.F. and Roberts, F., 1983. Metal resources and energy. Butterworth Scientific,
Borough Green, Kent.
Charles, R.J. and de Bruyn, P.L., 1956. Energy transfer by impact. Mining Engineering, 8(1): 47-
53.
Charles, R.J., 1957. Energy-size reduction relationships in comminution. Transactions of AIME,
Mining Engineering, 9(1): 80-88.
Clarke, A.J. and Wills, B.A., 1989. Enhancement of cassiterite liberation by high pressure roller
comminution. Minerals Engineering, 2 (2): pp 259-262.
- 206 -

Cleemann, J.O., 1994. Comprehensive comminution theory. ZKG International, Edition B, 47(4):
187-191.
Daniel, M.J., 2003. HPGR Model verification and scale-up. Masters thesis, University of
Queensland, JKMRC, Australia.
Daniel, M.J., 2004 . Measurement of electrical energy consumed during a Bond ball mill locked
cycle test. AMIRA International P9N, "The optimisation of Minerals Processes by
Modelling and Simulation(2nd Technical Report): pp2-28.
Daniel, M. J. and S. Morrell. 2004. HPGR model verification and scale-up. Minerals Engineering
17(11-12): pp1149-1161
Daniel, M.J., 2005. Particle bed compression comminution using a piston-die to predict the
performance of a HPGR. Randol Perth Forum, Perth August 21-24, 2005.
Daniel, M.J., 2006a. Measurement of electrical energy consumed during a Bond ball mill locked
cycle test., XXVIII IMPC, 3-8 September 2006, Istanbul, Turkey.
Daniel, M.J., 2006b. The researcher who rummaged through Bond’s toolbox., Second
international JKMRC student conference, Brisbane, 7-8 March 2006.
Djordjevic, N., 2004. Discrete element modelling of the Bond ball mill,. JKMRC, University of
Queensland, unpublished report.
DOE, 2001. Mining- Industry of the future. Office of industrial technologies (OIT), Office of
Energy Efficiency and Renewable Energy, USA, doe/GO-102001-1157,(February
2001): www.oit.doe.gov/mining.
EIA, 2004. International energy outlook.
Esna-Ashari, M. and Kellerwessel, H., 1988. Inter-particle Crushing of Gold Ore Improves
Leaching. Randol Gold Forum 88, Scottsdale, USA : 141-146
Evelin, S.S., Souza, D.R., Fonseca, C.A.P. and Nardi, R.P., 1997. Evaluation of non-
conventional grinding systems for the treatment of a Brazilian primary gold ore. XX
International Minerals Processing Congress, Aachen, Germany,, 2: pp 165-171.
Fandrich, R.G., 1997. Mineral liberation model for confined particle bed breakage. Particulate
Science and Technology, 15(2): 133.
Fandrich, R.G., 1999. Mineral liberation through confined particle bed breakage, PhD Thesis,
University of Queensland, Brisbane, 321 pp.
Fandrich, R.G., Bearman, R.A., Boland, J. and Lim, W., 1997. Mineral liberation by particle bed
breakage. Minerals Engineering, 10 (2), pp 175-187
Farahmand, A. and Ehrentraut, G., 1997. Erzmetall,. No.3/March 1997: pp 201-210.
Fletcher, A., 1990. "Crushing mechanisms and mineral release". Department of Mining and
Minerals Engineering, University of Leeds, UK(PhD Thesis,).
Free, K.S., McCarter, M.K. and King, R.P., 2004. Evaluation of a new method for work index
estimation using single particle impact tests, 2004 SME Annual Meeting Preprints,
Feb 23-25 2004. 2004 SME Annual Meeting Preprints. Society for Mining, Metallurgy
and Exploration, Denver, CO, United States, pp. 597-601.
Free, K.S., McCarter, M.K. and King, R.P., 2005. Evaluation of a new method for work index
estimation using single particle impact tests. Minerals and Metallurgical Processing,
22(2): 96-100.
Fuerstenau, D.W. and Abouzeid, A.-Z.M., 2002. The energy efficiency of ball milling in
comminution. International Journal of Mineral Processing, 67(161-185).
Fuerstenau, D.W. and Kapur, P.C., 1994. New approach to assessing the grindability of solids
and the energy efficiency of grinding mills. Minerals & Metallurgical Processing,
11(4): 210-216.
Fuerstenau, D.W. and Kapur, P.C., 1995. Newer energy-efficient approach to particle production
by comminution. Powder Technology, 82(1): 51-57.
- 207 -

Fuerstenau, D.W. and VazquezFavela, J., 1997. On assessing and enhancing the energy
efficiency of comminution processes. Minerals and Metallurgical Processing, 14(1):
41-48.
Fuerstenau, D.W., 1983. Comminution and Simulation Using Rod and Ball Mills., United States.
Fuerstenau, D.W., 1992. Comminution: past developments, current innovation and future
challenges, Proceedings of the International Conference on Extractive Metallurgy of
Gold and Base Metals, Oct 26-28 1992. AusIMM Extractive Metallurgy Conference.
Published by Australasian Inst of Mining & Metallurgy, Parkville, Aust, Kalgoorlie,
Australia, pp. 15-21.
Fuerstenau, D.W., Kapur, P.C., and Gutsche, O., 1993. Comminution of single particles in a
rigidly-mounted roll mill Part 1: Mill torque model and energy investment. Powder
Technology, 76, pp 253-262
Fuerstenau, D.W., Lutch, J.J. and De, A., 1999. Effect of ball size on the energy efficiency of
hybrid high-pressure roll mill/ball mill grinding. Powder Technology
Fuerstenau, D.W., Venkatarman, K.S. and Williams, M.C., 1984. Simulation of closed circuit
mill dynamics by locked cycle grinding of mixtures., CONTROL '84:
Mineral/Metallurgical Processing (1st International Symposium on Automatic Control
in Mineral Processing and Process Metallurgy). Held in conjunction with the 1984
AIME Annual Meeting. Soc of Mining Engineers of AIME, New York, NY, USA, Los
Angeles, CA, USA, pp. 49-53.
Gates, A.O.,1915., "Kick versus. Von Rittinger: An experimental investigation of rock crushing
performed at Purdue University." Trans. AIME: 111.
Gaudin, A.M. and Hukki, R.T., 1944. Principles of comminution -- Size and surface distribution,
American Institute of Mining and Metallurgical Engineers -- Meeting, Feb 1944, New
York, NY, United States, pp. 17.
Gaudin, A.M., Gross, J. and Zimmerly, S.R., 1929. The so called Kick Law applied to fine
grinding. Mining and Metallurgy(10): p 447.
Gerrard, M., Costello, B. and Morley, C., 2004. Operational experiences and performance
assessment of HPGR technology at argyle diamond mine, The AusIMM New Leaders'
Conference, Apr 21-22 2004. Australasian Institute of Mining and Metallurgy
Publication Series. Australasian Institute of Mining and Metallurgy, Carlton, Australia,
Ballarat, Victoria, Australia, pp. 63-73.
Goeldner, D., 2003. Energy Reduction Objective Creates New Role for Dry Grind Rollers.
AusIMM Bulletin(4): 30-34.
Goodall,C.,2005, Lonmin metallurgical services, personal comminucation.
Gutsche, O., Kapur, P.C. and Fuerstenau, D.W., 1993. Comminution of single particles in a
rigidly-mounted roll mill part 2: product size distribution and energy utilization.
Powder Technology, 76(3): 263-270.
Haultain, H.E.T., 1923. A contribution to the Kick versus Von Rittinger dispute. Trans. AIME,
69: 183.
Herbst, J.A. and Fuerstenau, D.W., 1980. Scale-up procedure for continuous grinding mill design
using population balance models. International Journal of Mineral Processing, 7: 1-31.
Herbst, J.A. and Lo, Y.C., 1989. Grinding efficiency with balls or cones as media. International
Journal of Mineral Processing, 26(1-2): 141-151.
Hilden, M., 2005. DEM bed breakage simulations. JKMRC, University of Queensland,
unpublished report.
Hocking, A., 1986. Randfontein Estates-The first hundred years, Bethulie, pp 104 pp.
Hogg, R. and Fuerstenau, D.W., 1972. Power relationships for tumbling mills., 252(4): 418-423.
Holmes, J.A., 1957. Contribution to study of comminution -- Modified form of Kick's law.
Institution of Chemical Engineers -- Transactions, 35(2): 125-141.
- 208 -

Hongqi, Y., Bingquan, X. and Zhongshang, Z., 1985. New equation of comminution work - An
energy yield equation. Zhongnan Kuangye Xueyuan Xuebao/Journal of Central-South
Institute of Mining and Metallurgy(2): 121-126.
Horsburgh, R., 2005a. The Battle for Growth, UBS raw materials conference, Sydney, pp. pp 10-
11.
Horsburgh, R., 2005b. Prospering in an era of globalisation, consolidation and capacity addition.
In:http://www.smorgonsteel.com.au/content/speeches/docs/SSX_Nov05_GS_Steel.pdf
(Editor), Goldman Sachs Global Steel CEO Forum, New York.
Hosten, C. and Avsar, C., 2004. Variation of back-calculated breakage rate parameters in Bond-
mill grinding. Scandinavian Journal of Metallurgy, 33(5): 286-293.
Hukki, R.T., 1962. Proposal for a solomonic settlement between the theories of Von Rittinger,
Kick and Bond. Trans. AIME (Mining), vol 223: 403-408.
Hukki, R.T., 1965. Autogenous mills operated at supercritical speeds. Mining Congress Journal,
51(9): 120-125.
Irvine, J., 2005. Mineral export to Jump 23%, Sydney morning herald.
Jefferson, M., 2006. Sustainable energy development: performance and prospects. Renewable
Energy, 31(5): 571-582.
Joe, E.G., 1979. Energy consumption in Canadian mills. CIM Bulletin(June 72.).
Johansen, G., Lane, G. and Gardula, A., 2005. Bendigo utilises high pressure grinding rolls to
maximise gravity recovery. AusIMM Bulletin, March/April: pp 54-58.
Johnson, G., Tondo, L.,Morrell,S., 1995. The power efficient use of High Pressure Grinding
Rolls, Randol Pacific Gold Forum, Perth, 1995, pp 39-43.
Johnson, G., 2005. Presentation presented at the Randol Pacific Metallurgy Forum, Perth, August
22-24.
Johnson, G., Hunter, I. and Holle, H., 1994. Quantifying and improving the power efficiency of
SAG milling circuits. Minerals Engineering, 7(2-3): 141-152.
Johnson, G., Tondo, L.A. and Morrell, S., 1995. The power efficient use of high pressure
grinding rolls. Randol pacific gold forum: 39-43.
Kapur, P.C. and Fuerstenau, D.W., 1987. Energy-size reduction "laws" revisited. International
Kapur, P.C., 1972. Self-preserving size spectra of comminuted particles. Chemical Engineering
Science, 27(2): 425-431.
Kaya, E., Fletcher, P.C. and Thompson, P., 2003. Reproducibility of bond grindability work
index. Minerals and Metallurgical Processing, 20(3): 140-142.
Kelly, E.G., and Spottiswood, D.J.,1982.,Introduction to mineral processing, John Wiley and
Sons, p116.
Kesler, S.E. and Macmillan,1994, Mineral resources, economics and the environment: 391 pp.
Kick, F., 1885. Des Gesetz der Proportionalem Widerstand und Seine Anwendung Felix,
Leipzig.
Kilani, J., 2002. The business sense of energy management in mining. Global mining initiative
conference(http://www.icmm.com/gmi_conference/439JohnKilani.pdf).
King, R.P., 2004. Modelling and Simulation of Mineral Processing Systems. Butterworth,
Heinemann, pp69-71 pp.
Kirsch, S. and Beloborodov, V., 2004. Worldwide first industrial scale HPGR applied for a gold
ore processing. In: B.A. Wills (Editor), Comminution 04, Perth, Western Australia.
Kirsch, S., 2004. Practical Experience Gathered with a Roller Press Set Up in the Gold Deposit
“Zapadnoye”, 5th IIR Crushing and Grinding Conference, Townsville.
Kirsch,S.,2006,Workshop Codelco/VDMA, Santiago, 25 May 2006.
http://www.camchal.com/esp/polysius.pdf
- 209 -

Kiyama, H., Majima, H. and Fujinaka, Y., 1974. Driving power of tumbling mills., 67(749): 62-
70.
Klingmann, H.L., 2005. Soledad mountain project - flowsheet development and benefits of the
HPGR. Randol Perth Forum 2005
(http://www.goldenqueen.com/pdf/HPGR%20Technical%20Paper.pdf): pp 105-106.
Klymowsky, R. and Cordes, H., 1999. The modern roller press - practical applications in the ore
and minerals industry. Aufbereitungs-Technik, 40(8): 387-396.
Klymowsky, I.B and Liu, J. 1997 a. Modelling of the comminution in a roller press. In the
proceedings of the XX International Mineral Processing Congress(IMPC), Aachen,
Germany, September 1997.
Klymowsky, I.B, Liu, J. 1997 b. Towards the development of a work index for the roller press.
Chapter 14 in Comminution Practices (Edited.) S. Komar Kawatra (SME)
ISBN-10: 0873351495, ISBN-13: 9780873351492 Publisher: Society for Mining,
Metallurgy & Exploration, Incorporated - January 1997 Textbook Paperback, 362pp.
Klymowsky, I.B.,Patzelt, N.,Knecht,J.,Burchardt, E., 2002, Chapter 5, Selection and sizing of
high pressure grinding rolls, Conference Proceedings, SME, Mineral processing plant
design, practice and control. Edited by Andrew L Mular, Doug N. Halbe, and Derek J
Barratt, Volume 1,Vancouver , British Columbia, Canada, October 2002, pp 636-668
Knecht, J., 1994. High-pressure grinding rolls - A Tool to optimise treatment of refractory and
oxide gold ores. Fifth Mill Operators' Conference, Roxby Downs SA, pp 51-59,
AusIMM
Knecht, J., and Baum, W., 1994. Optimising refractory and oxide gold ore operations with high-
pressure grinding rolls. Proceedings SME Annual Meeting, Albuquerque, New
Mexico, Preprint 94-12
Lane, G.,2006. How do you choose between a SAG mill and HPGR based comminution
circuits?, 7th IIR Crushing and Grinding conference, March 2006, Townsville.
Laplante, A.R., McIvor, R.E. and Finch, J.A., 1988. Error analysis for bond work index
determinations part 1: Accuracy and reproducibility. Minerals Engineering, 1(2): 113-
125.
Laplante, A.R., Prasad, U., McIvor, R.E. and Finch, J.A., 1993. Error analysis for bond work
index determinations part 2: A phenomenological basis for the analysis of systematic
errors for the rod mill test. Minerals Engineering, 6(5): 509-521.
Larsen, C., Cooper, M. and Trusiak, A., 2001. Design and operation of Brunswick's AG/SAG
Circuit. Mining and mineral process engineering, University of British Columbia, Vol
IV: pp 350-367.
LeVier, K.M., Logan, T.C., Patzelt, N.,Klymowsky, I.B., 2004, Latest developments in HPGR
Technology - results of a field trial. Metplant 2004 conference, Perth, WA, 6-7
September 2004, pp 55-63.
Levin, J., 1984. A Proposed Test for the Determination of the Grindability of Fine Materials,
Mintek Report M177, Council for Mineral Technology, South Africa.
Levin,J.,1989 Observations on the bond standard grindability test and a proposal for a standard
grindability test for fine materials, Journal of the South African Institute Mining and
Metallurgy 89 (1989) (1), pp
Li, G. and Xu, X., 1993. Experimental investigation of the energy-size reduction relationship in
comminution using fractal theory. Minerals Engineering, 6(2): 163-172.
Lim, W.I.L. and Campbell, J.J, 1996, Report on high pressure rolls experimental work with
Argyle ore, Application of high pressure grinding rolls in mineral processing, AMIRA
Project P428, Report P428/07
- 210 -

Lynch, A.J. and Rowland, C.A., 2005. The history of grinding. SME, Littleton, Colorado, 209
pp.
Magdalinovic, N., 1989. A procedure for rapid determination of the Bond work index.
International Journal of Mineral Processing, 27(1-2): 125-132.
Mahyera, A., 1988. Energy-Efficient Concept for Fine Size Reduction of Rocks and Ores.,
United States.
Marechal, F., Favrat, D. and Jochem, E., 2005. Energy in the perspective of the sustainable
development: The 2000 W society challenge. Resources, Conservation and Recycling,
44(3): 245-262.
Marks, J., 2006. Global Aluminum Industry Response to Energy and Climate Change. CSD 14
Business and Energy Tuesday 9 May 2006, International Aluminum Institute, Page 6
(http://www.wbcsd.org/web/projects/climate/csd14/biz_energy_csd14_jerry.pdf).
McNab, B., 2006. Exploring HPGR Technology for Heap Leaching of Fresh Rock Gold Ores,
7th IIR Crushing and Grinding conference, Townsville.
Morrell, S.; Lim, W.I.L.; Tondo, L.A.; David, D;1996, Modelling the high pressure
grinding rolls, Mining technology conference, pp 169-176
Morrell,S., Shi,F., Tondo,L.A., 1997, Modelling and Scale-up of High Pressure Grinding
Rolls, In the proceedings of the XX International Mineral Processing Congress (IMPC),
Aachen, Germany, September 1997.pp 129-140.
Morrell, S. and Man, Y.T., 1997. Using modelling and simulation for the design of full scale ball
mill circuits. Minerals Engineering, 10(12): 1311-1327.
Morrell,S.,1993, The prediction of power draw in wet tumbling mills, Ph.D Thesis, University of
Queensland, Australia.
Morrell, S., 1994. Power draw of grinding mills - its measurement and prediction, Proceedings of
the 1994 5th Mill Operators' Conference, Oct 16-20 1994. Australasian Inst of Mining
& Metallurgy, Roxby Downs, Aust, pp. 109.
Morrell, S., 2004. An alternative energy-size relationship to that proposed by Bond for the design
and optimisation of grinding circuits. International Journal of Mineral Processing,
74(1-4): 133-141.
Morrell, S., 2006. Rock characterisation for High Pressure Grinding Rolls Circuit Design.
Conference proceedings, SAG 2006: IV-267.
Morrell, S., Valery, W., Banini, G. and Latchireddi, S., 2001. Developments in AG/SAG Mill
Modelling Proc. International Autogenous and Semi Autogenous Grinding
Technology., vol. IV, University of British Columbia: pp. 71–84.
Mörsky, P., Klemetti, M. and Knuutinen, T., 1995. A comparison of high pressure roller mill and
conventional grinding. XIX International Minerals Processing Congress, Littleton,
Colorado, USA, 2, pp 55-58
Mosher, J.B. and Tague, C.B., 2001. Conduct and precision of bond grindability testing.
Minerals Engineering, 14(10): 1187-1197.
Myers, J.F., Michaelson, S.D. and Bond, F.C., 1947. Rod milling -- Plant and laboratory data,
American Institute of Mining and Metallurgical Engineers -- Meeting, Mar 1947.
American Institute of Mining and Metallurgical Engineers -- Technical Publication.
American Institute of Mining and Metallurgical Engineers, New York, NY, United
States, pp. 11.
National Research Council, 1981. Committee on Comminution and Energy Consumption.,
United States. Bureau of Mines.,
National Science Foundation (U.S.), Report on comminution and energy consumption. National
Academy Press, Washington, D.C., 283 pp.
- 211 -

Nitta, S., Bissombolo, A., Furuyama, T. and Mori, S., 2002. Relationship between Bond's work
index (Wi) and uniformity constant (n) of grinding kinetics on Tower mill milling
limestone. International Journal of Mineral Processing, 66(1-4): 79-87.
Norgate, T.E. and Rankin, W.J.;2002; Proceedings, Green Processing 2002, International
Conference on the Sustainable Processing of Minerals, May 2002, pp 49-55. Report
accessed on web at http://www.minerals.csiro.au/sd/CSIRO_Paper_LCA_Sust.pdf.
Norgate, T.E. and Weller, K.R., 1994. Selection and operation of high pressure grinding rolls
circuits for minimum energy consumption. Minerals Engineering, 7(10): 1253-1267.
Otte, O., 1988. Polycom high pressure grinding principles and industrial application. Third Mill
Operators' Conference, Cobar NSW, AusIMM: pp 131-136.
Patzelt, N., Knecht, H. and Baum, W., 1995, Case made for high-pressure roll-grinding in gold
plants. Mining Engineering, 47 (6), pp 524-529, June 1995.
Patzelt, N., Knecht, H. and Baum, W., 1997. The metallurgical potential of high-pressure roll
grinding. XX International Minerals Processing Congress, Aachen, Germany,
September 1997, vol 2, pp 155-164
Patzelt, N., Knecht, J., Burchardt, E. and Klymowsky, R., 2000. Challenges for high pressure
grinding in the new millennium, 7th Mill Operators Conference, Oct 12-14 2000.
Australasian Institute of Mining and Metallurgy Publication Series. Australasian
Institute of Mining and Metallurgy, Kalgoorlie, Australia, pp. 47.
Pontt, J., 2004. MONSAG: A new monitoring system for measuring the load filling of a SAG
mill. Minerals Engineering, 17(11-12): 1143-1148.
Prentice, T.K., 1946. Crushing and grinding efficiencies. Institution of Mining and Metallurgy of
London -- Bulletin(477): 35.
Pricewaterhousecoopers, 2002. Executive Summary PricewaterhouseCoopers’ , Survey of GMI
Conference Delegates May 2002. International council on mining and metals,
http://www.icmm.com/gmi_conference/456GMIDelegateSurvey.pdf.
Pyke, P., 2006. Application of HPGR technology in processing gold ore in Australia. Conference
proceedings, SAG 2006: IV-80.
Radhakrishnan, V.R., 1999. Model based supervisory control of a ball mill grinding circuit.
Journal of Process Control, 9(3): 195-211.
Rittinger, P.R., 1867. Lehrbuch der Aufbereitungs Kunde. Ernsr and Korn, Berlin.
Rowland, C.A., 1986. Ball mill scale-up diameter factors., Advances in Mineral Processing: A
Half-Century of Progress in Application of Theory to Practice. Proceedings of a
Symposium Honoring Nathaniel Arbiter on his 75th Birthday. Soc of Mining
Engineers, Littleton, CO, USA, New Orleans, LA, USA, pp. 605-617.
Rowland, C.A.J. and Erickson, M.T., 1983. Large ball mill scale-up factors to be studied relative
to grinding efficiency., Mini Symposium - Mill Design and Grinding Performance of
Large Ball Mills I. Papers presented at the Society of Mining Engineers of AIME Fall
Meeting. Mini Symposium - Society of Mining Engineers of AIME. Soc of Mining
Engineers of AIME, Littleton, CO, USA, Salt Lake City, UT, USA, pp. 27-29.
Rowland, C.A.J. and Erickson, M.T., 1984. Large ball mill scale-up factors to be studied relative
to grinding efficiency. Minerals & Metallurgical Processing, 1(2): 165-172.
Rowland, C.A.J. and Kjos, D.M., 1978. Rod and ball mills. Mular, Bhappu, (eds), Mineral
Processing Pland Design. SME: pp.239-278.
Rumpf, H., 1962. Fundamental physical problems in comminution. Grundlegende physikalische
Probleme bei der Zerkleinerung. Chemie-Ingenieur-Technik, 34(11): 731-741.
Ruth, M., 1995. Thermodynamic constraints on optimal depletion of copper and aluminum in the
United States: a dynamic model of substitution and technical change. Ecological
Economics, 15(3): 197-213.
- 212 -

Sampat, P., 2003. Scrapping mining dependence, State of the world - Progress towards a
Sustainable Society - 2003, pp. pp 110-129.
Sandgren, P.E., Berglind, B. and Modigh, S., 2002, Basics in Minerals Processing, Metso
Handbook. Web site accessed 2007,
www.metsominerals.com/inetMinerals/mm_segments.nsf/WebWID/WTB-041213-
2256F-43973 - 9k
Schönert, K. and Lui, J., 1996. Modelling of inter-particle breakage. International Journal of
Mineral Processing, 44-45: 101-115.
Schönert, K., 1979. Energy aspects of size reduction of brittle material. Zement-Kalk-Gips
(Translation of No. 1/79): 40-44.
Schönert, K., 1986. On the limitation of energy saving in milling. In: 1. World Congress Particle
Technology,, part II(comminution): page 1.
Schönert, K., 1988. A first survey of grinding with high-compression roller mills. International
Schönert, K., 1991. Advances in comminution fundamental, and impacts on technology. XVII
International Mineral Processing Congress, Dresden, Volume 1,: pp 1-21.
Schuhmann, J., R., 1960. Energy input and size distribution in comminution. Mining
Engineering, 12(2): 161-164.
Schwechten, D.;Milburn, G.H. ; 1990, Experiences in dry grinding with high compression roller
mills for end product quality below 20 microns., Minerals Engineering, Vol.3, No. 1/2
, pp. 23-34
Shi, F., Lambert, S., Daniel, M.J., 2006, Measurement of the effect of HPGR treating platinum
ores. SAG 2006, Vancouver, 23-27 September 2006.
Smorgon-Steel-Group, 2005. Financial results 2004/5 and strategy update.
Stairmand, C.J., 1975. Energy efficiency of milling processes. A review of some fundamental
investigations and their application to mill design. pts A/1-A/2, for European
Symposium on Comminution, 4th, Zerkleinern, Sep 15-17 1975, 79(1549-1588 pt a/1):
1-17.
Stamboliadis, E.T., 2004. Energy distribution in comminution: A new approach to the laws of
Rittinger, Bond and Kick. Canadian Metallurgical Quarterly, 43(2): 249-258.
Steane, H.A. ,1976, Coarser grind may mean lower metal recovery but higher profits, Canadian
Min. Journal.
Stephenson, I, Guerney, P., Napier-Munn, T.J., Soh, C.,;1996, High Pressure Grinding Rolls –
Downstream Processing, Mining Technology Conference, pp 177-187.
Stephenson, I.;1997, The downstream effects of high pressure grinding rolls processing, PhD
Thesis, Julius Kruttschnitt Mineral Research Centre, Department of Mining and
Metallurgical Engineering, University of Queensland.
Sutherland, D.N. and Fandrich, R.G., 1996. Selective fracture and liberation of minerals.
Proceedings Chemeca 96, Sydney, 3: pp 83-88.
SvensktStal, 2006. Energy and material balance. http://www.ssab.com.
Taggart, A.F., 1945. Handbook of Mineral Dressing. Wiley and Sons, New York.
Tartaron, F.X., 1964. Foundation of general theory of comminution. Society of Mining Engineers
of AIME -- Transactions, 229(2): 120-125.
Tavares, L.M., 2005. Particle weakening in high-pressure roll grinding. Minerals Engineering,
18(7): 651-657.
Thompsen, L., Patzelt, N. and Knecht, J., 1996. High-pressure grinding for copper at Cyprus
Sierrita. Mining Engineering (Littleton, Colorado), 48(9): 23-26.
Tondo, L.A.;1997,Phenomenological modelling of a high pressure grinding roll mill, MSc
Thesis, Julius Kruttschnitt Mineral Research Centre, Department of Mining and
Metallurgical Engineering, University of Queensland.
- 213 -

Trewin, D., 2005. 2005 Year book Australia. Australian Bureau of Statistics, ABS Catalogue No
1301.0: p 530.
Tromans, D. and Meech, J.A., 2004. Fracture toughness and surface energies of covalent
minerals: theoretical estimates. Minerals Engineering, 17(1): 1-15.
Tuzun, M.A., 2001. Wet bond mill test. Minerals Engineering, 14(3): 369-373.
U.S Department of Energy;2001, Mining- Industry of the future, Office of industrial technologies
(OIT), Office of Energy Efficiency and Renewable Energy, February 2001, doe/GO-
102001-1157, www.oit.doe.gov/mining.
USGS, 2006. Commodity statistics and information.
http://minerals.usgs.gov/minerals/pubs/commodity/.
Valery, W.J. and Morrell, S., 1995. Development of a dynamic model for autogenous and semi-
autogenous grinding. Minerals Engineering, 8(11): 1285-1297.
van Drunick, W. and Smit, I., 2006. Energy efficient comminution - HPGR experience at Anglo
Research. Conference proceedings, SAG 2006: IV-124.
Vanderbeek, J.L., Linde, T.B., Brack, W.S. and Marsden, J.O., 2006. HPGR Implementation at
Cerro Verde. Conference proceedings, SAG 2006: IV-45.
Vedat, D. and Huseyin, O., 2003. A new approach to bond grindability and work index: Dynamic
elastic parameters. Minerals Engineering, 16(3): 211-217.
Voorspools, K., 2004. Sustainability of the future; rethinking the fundamentals of energy
research. Renewable and Sustainable Energy Reviews, 8(6): 599-608.
Walker, D.R. and Shaw, M.C., 1954. A physical explanation of the empirical laws of
comminution. Transactions AIME, Mining Engineering: pp 313-320.
Walker, W.H., Lewis, W.K., McAdams, W.H. and Gilliland, E.K., 1937. Principles of Chemical
engineering. New York.
Watson, S. and Brooks, M., 1994. KCGM Evaluation of High Pressure Grinding Roll
Technology. Fifth Mill Operators' Conference, Roxby Downs SA, pp 69-83
Wills, B.A., 1997. Mineral Processing Technology. An Introduction to the Practical Aspects of
Ore Treatment and Mineral Recovery. Butterworth and. Heinemann, 486 pp.
- 214 -

APPENDICES
All data accessed and used in this thesis is located on the CD provided
on the back page
APPENDIX A
Tables A 1-3 Hogg and Fuerstenau (1972) model, illustrating the sensitivity to the lift
angle in the mill from 30 -40 degrees.
GRINDING TASK :
Total Plant Throughput, g/min 58.844879 Mill Mill Mill Energy Mill Energy
MILL DIMENSIONS AND OPERATING CONDITIONS : Power, kW Power, W Whr (Whr) per rev
0.3048 0.07999348 Balls 79.99347980 79.99 0.01861
Diameter Length Mill Speed Charge Balls Interstitial Lift 0.00000000 Overfilling 0.00000000 0.00 0.00000
ft ft % Critical Filling,% Filling,% Slurry Filling,% Angle, (°) 0.00553801 Slurry 5.53801014 5.54 0.00129
1.00 1.00 93.55 19.34 19.34 30.00 30.00 0.08553149 Net Total 85.53148993 85.53 0.01989
L/D rpm 81.05000000 % Losses
1.00 71.66 0.45135351 Gross Total 451.35350889 451.35 0.10498
% Solids in the Mill 100.00 Charge Mill Charge Weight, tons Apparent energy for milling (ore only)
Ore Density, ton/m3 2.70 Volume, Ball Slurry Density energy for milling (ball and ore)
Slurry Density, ton/m3 2.70 m3 Charge Interstitial above Balls ton/m3 drive train losses
Balls Density, ton/m3 7.80 0.004309 0.020165 0.001396 0.000000 5.004

Volume, Ball Slurry Density
0.004309 20.164649 1.396014 0.000000 5004.000
GRINDING TASK :
0.3048 0.09176475 Balls 91.76475014 91.76 0.02134
1.00 1.00 93.55 19.34 19.34 30.00 35.00 0.09811769 Net Total 98.11769439 98.12 0.02282
L/D rpm 81.05000000 % Losses
1.00 71.66 0.51777147 Gross Total 517.77147433 517.77 0.12042

0.004309 20.164649 1.396014 0.000000 5004.000
- 215 -

GRINDING TASK :
0.3048 0.10283764 Balls 102.83763534 102.84 0.02392
1.00 1.00 93.55 19.34 19.34 30.00 40.00 0.10995716 Net Total 109.95716394 109.96 0.02557
L/D rpm 81.05000000 % Losses
1.00 71.66 0.58024889 Gross Total 580.24888621 580.25 0.13496

0.004309 20.164649 1.396014 0.000000 5004.000
Table A- 4
Morrell (1993) Bond ball mill model with the calibration constant equal to either 1. The
equivalent power is given when the default calibration constant of 1.26 is used.
Morrell Ball Mill Power Estimator

Morrell
Mill / Identifier Bond ba ll mill
"Bond"
Power
model Dry grind
Mill Da ta Tota l Power
dia me ter (m) 0.3048 134 W 71.67 rp m Mill Spee d
be lly length inside line rs (m) 0.3048 20.125 kg Ba ll ma ss
Feed Cone Angle 0 Estima te d 1370.0 g Ore ma ss
Discha rge Cone Angle 0 No-Loa d Power 81.42 W Net power
trunion dia me ter (m) 0.00001 53 W
fra ction of crit. spe ed 0.9354
ba ll volume (%) 19.3355 For 1 hours ope ra tion
tota l filling of cylindrica l section(%) 19.3355 Net power 81.42 Whr Energ y in o ne ho ur
Void Filling Fra c tion (1=slurry fills loa d) 0.295 81.4 293119 W.s Energ y in o ne ho ur
ore sg 2.7 4300 revs Re volutiuo ns in 1 ho ur
liquid sg 0.00001 1.89 Whr Energ y p e r 100 re vs
Discha rge slurry % solids 100
Discha rge Me cha nism gra te
ca libra tion consta nt 1 68.2 J/ rev Energy per rev
BOND TEST VITAL STATISTICS

Bond test Closing sieve (Bm) 300
Ave Bond Revs as steady state 353
Ave Energy (Whr)-as measured by "one meter" 46.9
Ave Grindability (g/ev) - last three grinds 0.877
Ave %250 recycle- last three grinds 251
Ball charge mass (g) ± 7 g 20125
Ore charge mass (g)- mass of 700 ml of ore 1334
% passing closing screen 18.64
Feed F80 (Bm) 2983
Product P80 (Bm) 202
Total Mill Charge mass (g) 21459
Effective Mill feed rate (g) 381
Effective Mass subjected to milling (g)(less fines ) 310
Energy No load Empty mill (Whr) @ 0.1077 Whr/rev 38.0
Net comminution energy per rev (Whr/rev) 0.02516
Energy to move balls and ore per rev (Ws/rev)or J/rev 90.6
Morrell power model (J/rev)calib cont=1.26 85.9
Morrell power model (J/rev)calib cont=1.00 68.2
Bond Literature energy per rev @ 60 J/rev 60.0

"one meter" J/rev divided by g/rev = J/g or Ws/g 103.3
Whr/kg 28.7
Ws/g using Bonds 60 J/rev 68.42
kWhr/tonne using Bonds 60 J/rev 19.0
Morrell power model Comminution energy(kWhr/tonne) 21.6
measured "one meter" comminution energy kWhr/tonne 28.7
"BONDS" BASE LINE COMMINUTION ENERGY REQUREMENT

Net Comminution energy Bond Eqn (kWhr/tonne) 14.7
Bond WI (using bonds eqn) 28.3
Diff Bond @ 60 J/rev and Bond eqn -4.3
- 216 -

Energy Efficient Mineral Liberation Using HPGR Technology: Valid Solutions Space

Uploaded by

Copyright:

Available Formats

Energy Efficient Mineral Liberation Using HPGR Technology: Valid Solutions Space

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Energy Efficient Mineral Liberation Using HPGR Technology: Valid Solutions Space

Uploaded by

Copyright:

Available Formats

What is the thesis about?

What is the thesis about?

What are the original contributions of the thesis?

What are the original contributions of the thesis?

Energy efficient mineral liberation

using HPGR technology

Michael John Daniel

A thesis submitted at the University of Queensland in fulfillment of the requirements for

Julius Kruttschnitt Mineral Research Centre,

• To quantify the energy efficiency of comminution in circuits employing HPGR by

• To critically evaluate the Bond methodology of determining energy requirements

• To develop flowsheets that demonstrate step change reductions in energy

• To better understand comminution energy and sustainability, by reviewing energy

• To quantify global comminution energy consumption in respect to the production