How To Set Up and Develop A Geometallurgical Program

Geological Survey of Finland
Economic Minerals Unit

Espoo 11.5.2020 GTK Open File Work
Report 72/2019
How to Set Up and Develop a

Geometallurgical Program
Simon Michaux
Louisa O’Connor
Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Espoo  Kokkola  Kuopio  Loppi  Outokumpu  Rovaniemi
www.gtk.fi  Puh/Tel +358 29 503 0000  Y-tunnus / FO-nummer / Business ID: 0244680-7
ii
GEOLOGICAL SURVEY OF FINLAND DOCUMENTATION PAGE

Date: 11/05/2020
Authors Type of report
Simon Michaux GTK Open File Work Report
Associate Professor Geometallurgy Commissioned by
Geological Survey of Finland/Geologian tutkimuskeskus
BATCircle Project – Business Finland

Louisa O’Connor
Adjunct Senior Research Fellow. Western Australia school
of Mines, Curtin University.
Adjunct Associate Professor, WH Bryan centre, University
of Queensland,
Honorary Fellow, Rio Tinto Centre for Mining Automation,
Australian Centre for Field Robotics
Title of report
How to Set Up and Develop a Geometallurgical Program
Abstract
Geometallurgy is a comparatively new discipline in mining that involves the fundamental integration
of multiple conventional technical areas of the industry, where the outcome is a genuinely new
methodology. It can be a daunting prospect to be tasked with designing and justifying a
comprehensive geometallurgical program. The task could be starting a program from scratch or
collating existing information to re-analyze as a pre-cursor to a program extension.
Commonly it is seen as purely the relationship between metallurgical testing and process mineralogy,
although this is not incorrect, it is only part of the whole program. The discipline covers many areas
across the Mine-Value-Chain. Geometallurgy is an appreciation of the prediction of rock behavior,
being subject to an engineering outcome; this is in context of the engineering processes.
This paper discusses setting of strategic geometallurgy goals, how they could be used and maps out a
step by step procedure experimental design. Different methods for characterizing ore at different
scales is discussed. The different kinds of ore test characterization methods are discussed in context
of conventional tests and the smaller scale proxy comparative tests, for comminution, flotation and
leaching.
Various methods of data analysis to bring all these different technical disciplines together are also
discussed. There are many valid definitions and approaches to this methodology. This paper
represents just one approach to this subject.
Keywords
Geometallurgy, comminution, flotation, hydrometallurgy, mining, SEM automated mineralogy, data
analysis, campaign, core imaging, ore characterization, ore body knowledge, batch flotation, impact
breakage, Bond Ball Work index, leaching, comparative test, experimental design.
ISBN Number
ISBN 978-952-217-409-3

iii
Map sheet
N/A
Other information
N/A
Report Type Report serial
GTK Open File Work Report 72/2019
Total pages Language Price Confidentiality
245 English N/A Public Domain
Unit and section Project code
MIM Ore Geology and Mineral Economics,
Mineral Intelligence
Signature/name Signature/name
Associate Professor/ Apulaisprofessori GTK BATCircle Project Manager

iv
”It’s a long way to the top if you want to rock ‘n’ roll”
Bon Scott ACDC 1975

v
Contents
1 Introduction 1
2 Methodology and Results 6
2.1 Campaign planning of objective goals 6
2.2 Phases in a Geometallurgy campaign 12
2.3 Planning and experimental design 14
2.4 Sampling from site 15
2.5 Orientation study 20
3 Ore Mineralogical characterisation 21
3.1 Characterization of Meso-textures 24
3.2 Core Imaging 26
3.3 Hyperspectral characterization 31
3.4 Bulk mineralogy characterization 32
3.4.1 Quantitative X-Ray Diffraction (QXRD) 33
3.4.2 X-ray Fluorescence measurement (XRF) 33
3.4.3 Hand Held XRF measurement 34
3.5 Characterization of Micro-texture 35
4 Conventional process characterization tests for process behavior 39
4.1 Blasting in Geometallurgy 40
4.2 JKMRC Drop Weight Impact Breakage Test 43
4.3 JKMRC Abrasion Test 48
4.4 High Pressure Grinding Roll (HPGR) 49
4.5 Grinding Bond Ball Work Index 52
4.6 Fine Grinding – Laboratory scale Isa Mill 53
4.7 Selection of grind size for the whole campaign 56
4.7.1 Electrodynamic fragmentation 61
4.8 Flotation 69
4.8.1 Batch Flotation Test Characterization 73
4.9 Hydrometallurgy Leaching 77
4.9.1 Column Leaching Characterization 78
4.10 Gravity Separation 80
4.10.1 Strong Gravity Separation 82
4.10.2 Weak Gravity Separation 83
4.11 Magnetic Separation 87
4.11.1 Low-intensity magnetic separation - ferrite magnet LIMS 90
4.11.2 Medium intensity magnetic separation - permanent magnet MIMS 90

vi
4.11.3 High intensity magnetic separation – electromagnetic HIMS 90

4.12 Ore Sorting 91
4.13 Pyrometallurgy 93
5 Small Scale Proxy Tests 94
5.1 EQUOtip test 95
5.2 Point Load Index Measurement 98
5.2.1 Diametral test 100
5.2.2 Axial test 100
5.2.3 Block and irregular lump tests 100
5.2.4 Anisotropic rock samples 100
5.2.5 Collecting samples 101
5.2.6 Use of the Point Load Index 101
5.3 Impact Breakage SMC test 102
5.4 RBT Impact Breakage test 104
5.5 Piston and Die Compress Bed Breakage (CBB) 104
5.6 Grinding - Batch Operating Work Index 108
5.7 SGS SPi test 110
5.8 Mergan Work Index 110
5.9 Comminution Index (JKCi) 112
5.10 Comparative Comminution Hardness Test Geopöyrä 118
5.11 Comparative Comminution Hardness Index Tester HIT 122
5.12 Comparative flotation test JKMSi 130
5.13 CSIRO proxy diagnostic leaching test 131
5.14 Establishing what comparative tests are appropriate for the target deposit 131
6 Orientation Study Design 133
7 Orientation Study data analysis 136
7.1 Experimental design of the Orientation Study 137
7.1.1 Step 1 – Sample selection. 137
7.1.2 Step 2 – Sample preparation. 139
7.1.3 Step 3 – Characterization of each Orientation Study sample. 139
7.1.4 Step 4 – Selection of grind size for the whole campaign. 141
7.1.5 Step 5 – 1st pass process separation for each Orientation Sample 142
7.1.6 Step 6 – Data Analysis of the 1st pass process separation for each Orientation Sample
144
7.1.7 Step 7 – 2nd pass process separation for each Orientation Sample 144

vii
7.1.8 Step 8 – Data Analysis of the 2nd pass process separation for each Orientation Sample
145
7.1.9 Step 9 – Comparison process paths for each Orientation Sample 146
7.1.10 Step 10 – Selection of the most effective process path for each Orientation Sample 147
7.1.11 Step 11 – Comparison of Process Paths Between Orientation Samples 147
7.1.12 Step 12 – Trial of best Combination Process Separation for the Deposit 149
7.1.13 Step 13 – Selection of the most effective process path for the Deposit 149
7.1.14 Step 14 – Selection of the mineralogical mapping signatures of the most effective
process path 150
7.2 Method to Mathematically rank and compare process recoveries 150
7.3 Example A1 – What mineralogy controls grinding? 152
7.4 Example A2 – What mineralogy controls comminution footprint 155
7.5 Example A3 – Process Circuit Performance Estimation Economic Modelling of each End
Member Sample 156
7.6 Economic modelling and business model development 158
7.7 Conclusions of orientation study 159
8 Building the geometallurgy data set 160
9 Multivariate Analysis of Ore Deposit data 164
9.1 Statistical data domain definition with Principle Component Analysis (PCA) 168
10 Process behavior ore type Class definition 171
10.1 Process modelling Class by Class 181
10.1.1 Example Case Study P 184
10.2 Engineering Simulation 187
11 Hypothesis Sample study 188
11.1 Hypothesis Sample experimental test work 189
12 Mapping Study Process Domaining across drill core sections 190
12.1 Geometallurgy Mapping Study Experimental Design 192
12.2 Mineral signatures that control process response 192
12.3 Proxy test approach to mapping process domains 192
12.4 Use of Proxy Test Outcomes to Map Process Domains 195
12.5 Domaining behaviour and quantifying variability 197
12.6 Chemical Assays and CuSUM statistical method 199
13 Assessment of Geometallurgy Campaign 203
13.1 Assessment for potential sampling gaps in the study 203
13.2 Assessment on whether the study addressed the original experimental aims and
hypotheses 205
14 Possible Uses of the Geometallurgy Outcomes 206

viii
14.1 Use of Geometallurgical Analytical Outcomes to Define Sampling Strategies for Full Scale
Laboratory Testing for Process Design 206
14.2 Mine and Process Plant Design 207
14.3 Geometallurgical Block Model Generation 210
14.4 Economic Modelling and Business Model Development 212
14.5 Corporate problem solving with the Theory of Constraints 213
14.6 Corporate decision making for the executive board 217
14.7 Geometallurgy to support due diligence 217
15 Development of production protocol 218
16 Conclusion and Summary 221
17 References 225

Geological Survey of Finland 1/237
1 INTRODUCTION
Geometallurgy is the characterization of a family of ore types in context of process performance. As such, it
requires the integration of geological, mining, metallurgical, environmental and economic parameters. It
also requires the genuine integration and evolution of the paradigm of each of these technical areas into a
new holistic paradigm. The ultimate goal is to develop a more sophisticated operational design, streamline
and optimize the mining cycle and maximize the economic value of an ore body, while minimizing technical
and operational risk
Geometallurgy is an important addition to any mining operation, large or small as it aims to improve the Net
Present Value (NPV) of an ore body (Dominy & O’Connor 2016). It is known to increase site stakeholder
communications and collaborations (Jones & Morgan 2016 and Suazo et al 2010). Creating the right kind of
environment for knowledge sharing whilst concentrating on improved data acquisition and interrogation,
will result in confident integration of data into the Mine-Value-Chain. All these aspects create better business
optimization, better utilization of staff and better targeted key performance indicators (Chibaya 2013 and
O’Connor 2016).
The benefits of a geometallurgical program is to ‘de-risk’ or allow for better risk management processes,
either technical or operational (Chibaya 2013 and Vann et al 2002). The risk being the lack of knowledge of
your raw materials (ore) and the expected performance and recoveries of those materials. This risk could
translate into an overrun of the Capital Expenditure (CAPEX), or the money required to build and commission
a mining operation to the point where it can start producing a saleable metal concentrate. Table 1 below
shows an example of several recent CAPEX overruns (Noort and Adams 2006). These CAPEX overruns are a
consequence of not understanding what was genuinely required to economically feasible produce revenue
from these deposits.
Table 1. Recent major mining project CAPEX overruns (Source: Noort and Adams 2006)
Feasibility Actual/forecast
Company
Project budget cost cost overrun
Ravensthorpe/Yabilu
BHP Billiton A$1.4 billion 30%
Expansion
Spence (Chile) BHP Billiton US$990 million 10%
Telfer Mine Newcrest A$1.19 billion 17.50%
Stanwell Magnesium AMC A$1.3 billion 30%
Boddington Newmont A$866 million 100%
Goro Project
Inco US$1.45 billion 15%
(Indonesia)
Prominent Hill Oxiana A$350 million 51%

CAPEX blowouts are just one risk that has the potential to destroy a mining operation. Figure 1 shows the
rate of achievement of design capacity of plants at startup over time (McNulty 1998 & 2014). The curves
shown are classified as Types 1, 2, 3, and 4. They show four types of operations. Type 1 is an operation that
achieves the designed for production throughput easily within 12 months of commissioning. This is the ideal
objective of all mining operations. Type 4 is an operation that never reaches the designed production
throughput and takes a long time to reach its much lower maximum output. This is the worst case scenario
for mining operations, as produced revenue is much lower than planned for and takes much longer to
consolidate.
Figure 1 Start-up curve classification showing % of design capacity vs. time

(Source: Plant Ramp-Up Profiles an Update with Emphasis on Process Development, by T. McNulty, from
Proceedings of the 2014 Conference of Metallurgists. Reprinted with permission of the Canadian Institute of Mining,
Metallurgy and Petroleum).
Figure 1 above is a very real problem for capital investment due to the use of the Net Present Value tool
(NPV), where possible future produced revenue is discounted as a financial risk. The longer an operation
takes to pay its CAPEX loan off, the less valuable it is to investors. This is what distresses venture capitalists
the most.

This however, is a common problem. A study done by the Chemical Bank (now part of JP Morgan Bank)
drawn from a project database of 18 mining projects, examined the performance of these projects (Castel
1985). Questions asked and examined were:
 Was the mine completed on time?
 Was the mine completed at target budget?
 Did the mine eventually produce at design capacity?
 Was the mine producing expected cash flows?
The conclusions were as follows:

 12 of 18 had construction delays
 5 more than 2 years late
 4 never satisfied their ‘option’ test
 12 of 18 had CAPEX overruns
 10 mines had greater than 20% CAPEX overruns
 14 of 18 had difficulty operating at capacity
 6 persisted for more than 3 years
 12 of 18 yielded cash flow below original estimate
 7 of 18 mines never achieved budgeted operating levels
 7 of 18 had negative cash flows
 9 of 18 were subject to lower commodity prices
In a survey of 258 Public-Private Projects (Flyvbjerg 2002), 86% had capital overruns, with a CAPEX average
overrun of 28%. Gypton 2002 surveyed 60 mining projects over 21 years:
 Capex average overrun 22%
 No correlation between who complied original estimate
 Problems with in-house and “blue-chip”, highly-reputed consultants
 1st tier companies were no better than 3rd tier companies
 No influence from project size and location
 Bank consultants (due diligent audits) failed to identify “red flags” or CAPEX overruns
 Feasibility Study , 59% exceeded ±15% estimation accuracy (post reconciliation)
o Of which 75% over estimate
o Of which 25% under estimate
Another serious issue that can cause difficulty is highly variable plant performance in operation. Most
process plants are designed around an average performance (throughput and recovery), based on an ore
average that usually does not actually exist. This almost always translates into poor economic performance.
This is a consequence of poor knowledge of process response of the ore and when the variable ores will
enter the plant. Figure 2 shows an example of a challenging process plant performance.
The issue with what is shown in Figure 2 is that this mine site and process plant was designed to operate at
one consistent set average, run at a steady state. That this does not happen has a direct impact on the true
metal production in comparison to the planned for and predicted metal production.

Grinding Circuit Performance

Hourly Average for 7 days
90 600
85
550
80
500
75
450
70
Mill Feed Tonnes
Mill Poer kW
65 400
60
350
55
300
50
250
45
40 200
Time (hours)
Mill Feed (tph) Mill Power (kW)
Figure 2. Mill performance at a mine somewhere in the Southern Hemisphere

(Image Copyright: Dee Bradshaw)
The examples above show challenging outcomes that have the capacity to make a mining operation fail. In
the past few decades, mining operations have become larger, with a corresponding larger CAPEX required.
This means that the financial risk is much higher. The next generation of mining operations is predicted to
be bigger again.
Alternatively, Figure 3 below shows the Batu Hijau case study, where the long term plant performance was
predicted to within 2%, and years ahead of time (Wirfiyata 2012 & Burger et al 2006).
Wirfiyata 2012 demonstrated that Batu Hijau had systematically improved long and short term mill
throughput predictions since feasibility and commissioning in 1999. This had been achieved through a
combination of ongoing improvement to the geology ore hardness and grade characterization model, drill
and blast optimization and, by application of these data to model mill throughput. Batu Hijau had
demonstrated the ability to consistently estimate long term mill throughput within ±2% accuracy. Aside from
reliably predicting expected production rates with the existing plant configuration, this predictive ability
provided a solid foundation for identifying operating and design criteria for short term grinding circuit
optimization and future expansions (Wirfiyata 2012).
The difference between Batu Hijau and the examples shown in Table 1 and Curve Types 3 and 4 in Figure 3
could be a result of a successful geometallurgical campaign (a more sophisticated feasibility study cycle
based on process orientated ore characterization).

Mill Throughput DMT/hr
Figure 3 from “Applied Geo-Metallurgical Characterisation for Life of Mine Throughput Prediction at Batu Hijau,” by
F. Wirfiyata. Reprinted with permission of the Canadian Institute of Mining, Metallurgy and Petroleum.
An evolution of the development of mining operations is required to prevent the cost overruns in CAPEX,
and the delays in production performance. There are several possible ways to do this. Geometallurgy
applied to the feasibility study cycle is a practical way to do this.
Part of the overall risk management in a geometallurgical program is to include statistics and modelling,
specifically of aspects of data which are not normally targeted for modelling per se (Suazo et al 2010 and
Senior et al 2006). A systematic and stepwise approach to running a geometallurgical program is essential.
In many cases, companies have vast amounts of stored data that few have seen, or no one has utilized,
normally because they are not sure how or what to do. Depending on the project stage a geometallurgy
programs requires value proposition creation, gathering funds, to justify the bigger picture of business
improvement with the use of those funds.

2 METHODOLOGY AND RESULTS
2.1 Campaign planning of objective goals

The question of what fundamental purpose does the geometallurgical campaign serve has to be
understood. Where in the mining life cycle is the operation in question (Figure 4)? What part of the
mining extraction process path is the problems that the campaign is to examine? How they interrelate
with other parts of the process (Figure 5)?
Exploration
Exploration &
Discovery
Scoping study
Prefeasability study Feasibility &

Construction
Feasability study
Operation and Mining &

plan execution Extraction
Site Closure & Site

rehabilitaion Rehabilitation
Figure 4. The mine operation life cycle

(Image: Simon Michaux)

Definition
• Macro scale texture Exploration
• Fragment size 1-2m
• Sample units in tonnes
• Geological fundamentals Geotechnical
Excavation
Blasting
In Pit In pit ore

Blending sorting
Crushing
Liberation through comminution

Ore blending Ore sorting
SAG Mill
Grinding
• Meso scale texture
• Fragment size -100mm + 10µm
Ore sorting
• Sample units in kilograms Ball Mill
• Physical fundamentals Grinding
Ore sorting
Isa Mill Fine

Grinding
Hydrocyclone
separation
• Micro scale texture
• Fragment size -100 + 10µm
• Sample units in grams Gravity Separation
• Chemical fundamentals Separation
Flotation Leaching
Magnetic
Separation
production
Metal
Pyrometallurgy Electrowinning
Tailings/waste
Cleanup
Environmental
plume
Rehabilitation
management
Figure 5. The mining process path

(Image: Simon Michaux, Tania Michaux, Peter Sorjonen-Ward, Pit image by Steven Harolds from Pixabay)

The decision on whether the study focus should be a valuable metal element (for example, Au; Cu or Ag), a
mineral that somehow reduces production performance, or a penalty element that reduces the value of
the final concentrate, needs to be considered. There are many kinds of process engineering that can be
modelled successfully with a geometallurgical approach. The follow list are examples of possible focus
tasks that could be integrated into a geometallurgical campaign:
Process Engineering
• Establishing the best machine efficiency process path for the flow sheet
• Choice between SAG mill or HPGR in the comminution circuit
• Recovery efficiency in flotation or leaching hydrometallurgy
• Ore type definition in context of comminution, flotation or leaching hydrometallurgy
• Choice between hydrometallurgy or pyrometallurgy in concentrate smelting and refining
• Pyrometallurgical efficiency in slag and matte management
• Penalty element impact on saleable concentrate
• Slurry rheology efficiency as a function of ore type
• Cyclone split efficiency as a function of ore type
• Impact of clays in ore processing
• Risk based process design
• Whole circuit systems-based modelling using all process areas
Economics
• Feasibility studies
• Economic viability and efficiency
• Energy efficiency
• Business model development
o Whole mine operation optimization (Enterprise Optimization)
• Metal produced per kWh for each ore type
• Developing the mine schedule
o Block model
o Cut-off grade analysis
Environmental Impact
• Mine site waste plume
o Acid mine drainage generation
o Dust generation
o Site water drainage pollution
• Mine closure and site rehabilitation
o Operation legacy impact
o Life of Mine foot print
Strategic Development
 Assessment of the true value of the resource
o Priority for which valuable metals should be targeted
 Assessing the best time for open pit expansion cutback
 Assessing the choice between an open pit and underground operation
 Long term variability
o Life of mine cycle
o Mine operation footprint modelling
 Waste dump construction in context of future re-mining
 Whether to invest in this project or not in context of market forces
 Whether to shut down an operating mine in context of a challenging market environment
 Security of corporate reputation

Geometallurgy for operations

 Show stoppers & penalties in feed stream (clay, Cl, Fl, As, etc.)
 Predicted variability in ore hardness and recovery
 Metallurgical reconciliation
 Developing the best schedule of maintenance based on risk variability
 Plant expansion design and timing
All these tasks are to be carried out in context of the whole rock. If possible, the mineral combinations which
drive and control each of these technical areas, needs to be understood and mapped out. This must be
completed in a measurable and repeatable fashion, using experimental design protocol (Napier-Munn 2014).
The final outcome can be clearly defined, success is quantified at the start, and experimental design can be
set where success is achievable.
To achieve a single geometallurgical goal, volumes of context-based data are amassed. To study another
process behavior on the same rock set may well be a comparatively small amount of test work. There is a
large amount of overlap of experimental test work scope between a numbers of process behaviour, if studied
in this way. With this in mind, it could be useful to collect samples large enough to be sampled several times
and support more than one study.
As such, a completed geometallurgy dataset could be seen as the ‘Rosetta Stone’ of the rock (Figure 6). All
parts of the mining process are working on the same rock but with different technical languages. This means
that the rock itself becomes the data transfer point to examine multiple process behaviors and ore type
mineral definitions.
Geophysics
Chemical Establishing the best
data
assays process path
Comminution Choice between SAG

data mill or HPGR
Flotation data Recovery Ore type

Geomet efficiency definition
Mineralogical dataset
data Choice between
hydromet or pyromet
Geotech XRD/XRF data
data Penalty element
impact
Figure 6. A well-constructed geometallurgy data set could become the ‘Rosetta Stone’ for analysis in that deposit
(Image: Simon Michaux, and Rosetta Stone image by Kris Åsard from Pixabay)
Conventionally, the development steps for a mine are shown in Figure 7, where these steps are done
separately and in isolation.
A geometallurgy program fits in and around the same steps and adds value by doing those same steps in a
different way. Data is collected in a fashion that can define ore types in context of a process behaviour. This
is done in a series of iterative loops, where the stability of the analytical outcome defines when data

collection should stop (like the use of variography to determine if enough drilling has been done for deposit
definition) (Figure 8).
Based on Data Collected
conventional
geology
Ore type definition
Process testing
Based on
just grade Block Model Generation
Done
Separately
Mine Engineering
Schedule (mine & process)
Done Business Model

Separately Development
Figure 7 Conventional mine development flow sheet

The tasks of mine schedule development and process engineering flowsheet development are worked in a
systems context, where everything is optimized together (instead of separately) as shown in Figure 8.
Data Collected
Geometallurgy Data Analysed Ore type definition
Process testing
Block Model Generation
Systems Optimisation
Mine Engineering
Schedule (mine & process)
Business Model
Development
Figure 8 Geometallurgically supported mine development

The outcomes of a geometallurgical study are variability, domain mapping and ore type definition in context
of process engineering performance. These outcomes may be used for economic modelling, business
development mine schedule development. All of this takes place before larger samples are taken for
conventional process tests (bankable metallurgical tests), which lead to process plant design. This means a
geometallurgical campaign is most cost effective in a pre-feasibility study, where it has the highest leverage

of influence versus cost in the whole mine design cycle, this is front-end optimization (Mackenzie 2007)
(Figures 9 to 11).
Geometallurgy
Study
Figure 9 Leverage of early work (Source: MacKenzie and Cusworth 2007)
Geometallurgy
Study
Figure 10 Impact study phases on project value (Source: MacKenzie and Cusworth 2007)
Geometallurgy
Study
Figure 11 Degree of definition in study phases on project value (Source: MacKenzie and Cusworth 2007)

2.2 Phases in a Geometallurgy campaign

There are several phases of activity in a geometallurgy campaign and are shown in Figure 12 and 13.
Depending on what the attempted goals are and scope, the following list could be a useful guide:
1. Planning and experimental design (Napier-Munn 2014)

2. Sampling from site (Gy 2004)
3. Orientation study (Section 6)
4. Geometallurgical matrix generation (Section 8)
5. Multivariate data analysis (Section 9)
6. Class definition (Section 9)
7. Process modelling by class (Section 10)
8. Mapping process domains and variability of process behavior (Section 12)
9. Hypothesis response and validation samples and analysis
10. Economic modelling and business model development
11. Assessment of study (Section 13)
12. Recommendation for a production protocol (Navarra et al 2018)
These different phases interact and relate to each other. While each is important, they vary in size and
scope. Steps within the phases will be detailed.
Figure 12 The different phases interrelate to each other.

(Image: Tania Michaux)

(0) Geometallurgical
Experimental &
Analyitcal Goals
(1) Ore Body Analysis

of Existing Knowledge
(2) Creation of foundation

geomet matrix based on
existing knowledge
Orientation Study
(4) Orientation Study

Test Work Planning
(5) Orientation
Study Analysis
(6) Update Geomet

Data Matrix
(7) Multivariate Analysis of

geometallurgical data matrix
(8) Class Definition
(9) Process Modelling

(class by class)
(10) Hypothesis Validation Study

(Reiterate Steps 4-9)
(11) Mapping Study –

domaining of process behavior
(12) Evaluate Results from Steps 4-11 in

context of original aims Step 0
(13) Geometallurgical Study Conclusion
Figure 13. Steps of the geometallurgical campaign from point of genesis to completion

2.3 Planning and experimental design

For a geometallurgical campaign to be successful, it is required to be designed to a practical experimental
objective and then carried out according to plan. Considerable time and thought should be dedicated to the
design, where the outcome is clear and consistent with its point of genesis. A baseline, fundamental project
structure is shown in Figure 14.
Fundamental hypothesis
inital point of genesis goals

Final outcome resembles
Dynamic evolution of
hypothesis possible
Highly Process
Behaviour
constrained engineering
response
minerology application
Efficiency window
Campaign final objective
Figure 14. The fundamental structure of the geometallurgical campaign

The choice of what process behavior(s) will the geometallurgy study attempt to model is the fundamental
question in experimental design (Napier-Munn 2014). Leaching for example is controlled by different
mechanisms to controlling factors in comminution (for example in this case, leaching is fundamentally a
chemical process and comminution is fundamentally physical process). This needs to be understood from
the very beginning.
Once the target of the study is understood, the methods used to determine efficiency of that process
behaviour must to be considered. What tests are used and how? What small scale tests are there and what
is considered a trusted test by the industry (or a ‘bankable’ test – results accepted for submission to the
Stock Exchange)? Then investigations into and an understanding of what mineral mapping tools are available
and the form they take is required. An understanding of the precision of these methods and how the data
is interpreted must be completed.
Part of the experimental design is a complete understanding of the options of data analyses selected for use.
All data collected experimentally must serve a coherent purpose in the overall analysis. That analysis must
be planned in the beginning of the campaign, to ensure technical limitations do not overrun the campaign
objective. It is recommended that statistical methods of experimental design are used to collect data to
support later methods of analysis, where the outcomes of that analysis can be judged in terms of statistical
significance. Figure 15 provides a structure for the initial aims and objectives of the campaign.

Settle on an objective that is Develop significant predictive relationships and true

achievable with clear metrics variability in processing performance response by
characterising intrinsic geological attributes.
of success
(0) Geometalurgical
Experimental &
Aims
Analyitcal Goals
Time
What stage of Restraints? What the
development site/deposit
Is the deposit? Budget specific issues
Restraints?
Greenfields Feasibility Brownfields Grind Recovery Oxidisation

Exploration Study Operation Sensitive? Sensitive? of Ore?
Figure 15. Step 0 - The development of the objectives and aims of the geometallurgical campaign
All concepts need to be merged together to define what samples are to be taken and what volume they
should be. Sampling is often carried out poorly due to the lack of understanding of what is to be sampled,
why and how. Ample literature is freely available to assist in defining a method (Dominy 2016,
Krishnamurthy & Sawkar 2014, Bazin et al 2013, Sampling 2014, Dominy et al 2017, and World Conference
on Blending 2017). Campaign planning of objectives to be agreed upon before sampling:
 Campaign objectives set
 Campaign mandate and scope agreed upon
 Metrics of success agreed upon
 Experimental design and analysis to be conducted agreed upon
 Test work required agreed upon (including where test work is done)
 Budget and time frame agreed upon
 Required mass of each sample understood
2.4 Sampling from site

Once the campaign objectives, experimental design, and scope are finalized, existing orebody knowledge
(OBK) needs defining. This is to ensure that previous work done is not repeated. Assessment of historical
reports, experimental test work and client personnel professional knowledge. A discussion with an
exploration geologist who worked on drill core/site and any process engineering staff that have
characterized the target ore types. Any work done on the process design flowsheet should be understood
through discussion with the engineers who developed it. Any site-specific issues are required to be
understood as early as practical.
Site geologist intuition is to be listened to and considered

 What ore types do they think are relevant?

 What are the weathering horizons and clay content understood in this deposit?
 What deposit geology structures are most significant?
 What drill holes traverse and map these structures?
Site metallurgist intuition is to be listened to and considered

 What is the process range found so far in this deposit?
 Axb & BMWi extremes
 Flotation recovery performance
 Leaching recovery
 What minerals are considered penalty elements
 Are there site specific assay models that have been developed?
 Are there any site mineralogy relationships, associations or ratios already established?
Figure 16 shows a set of ideas to start this process.
MLA/QEMSCAN
Mineralogy Thin section Comminution throughput
Mineralogy objectives
Geotechnical
UCS, PLT, Young's Mod, Chemical Batch float Plant Design
Poisson’s ratio, etc Assays test work
Liberation profiles &

Grind size targets
Geophysics Geochemistry
Site personnel Pilot plant trials
Alteration Horizons knowledge
(down hole logging)
Comminution Tests
Reports, databases, (A*b, Bond,
Lithological stratigraphic spreadsheets
systems
(down hole logging)
Geological Context Metallurgical Context
Understanding of Understanding of
Deposit system in context of Deposit system in context of
existing knowledge Geometallurgy
(1)
(1)OreOre
Body Analysis
Body Initial
of Existing Knowledge
Analysis
To be done by personnel with geometallurgical training
Figure 16 Step 1 - Assessment of existing site deposit knowledge


The data and information collected from this examination will have to be assessed whether it is useful in
context of the planned geometallurgical campaign. The quality assurance and quality control (QA/QC)
needed to build a geomet data set, is quite different to conventional methods due to the much smaller
samples masses available for testing. For example, when metallurgical test work was completed, did the
sample comprise constrained mineralogy, or was it drawn from a combination of multiple sources? How
was ore-type definition decided? Were the samples collected in a form where the spatial position was
known, and the results can be mapped back into the deposit model?
Once the existing deposit data is understood from a geometallurgy paradigm, the foundation data matrix
can be assembled. Ideally, the samples collected for a geomet campaign would be drill core alone. However,
this is not always possible and other drill products need to be utilized. Most mining operations have
extensive drill core libraries/archives either on site or in a storage facility, where half of the drill core has
already been logged lithologically and then sent for elemental assay. Core remaining at site (library core)
could be sampled from drill holes that have existing analytical interpretations, such as: geophysics and
geotechnical characterization (Point load, UCS compression test, Brazilian Disc, Fracture Frequency, RQD,
etc.) All of this is potentially useful data depending on your initial objectives.
A series of further tests could be carried out on diamond drill core that already has good context data
collected. With this in mind, a data set is to be set up with architecture to allow the import of existing data
that passes QA/QC, the results from new test work and the analytical outcomes of the campaign. This is to
be executed in a fashion where all data types can be used as model inputs, where the result is able to predict
the process behavior of constrained mineralogy in a spatially defined way. A useful approach to do this is to
limit the sample volume to the same scale as chemical assays (or a multiple of). For example, chemical assays
are samples from half drill core in 2m length intervals (or 1m interval with RC chips). Such a small volume
that is routinely tested as part of existing operational protocol is a practical method to identify specific
mineralogy and have test work consistent with the rest of the data set. In the dataset, there will be many
different kinds of data that are not conventionally reported in the same matrix due to method of collection,
and the spatial geometry associated with the result. With the understanding that this is a work in progress,
professional judgment is required from this point onwards in how to achieve this.
It is recommended that each sample in the geometallurgical campaign has a sample identification label that
can be used to manage the data set in a consistent manner throughout the whole campaign. This will be
used in data QA/QC procedures, analysis, and later the mapping of the results back into the deposit. It will
also assist in the later compositing larger samples according to the geometallurgical analytical outcomes to
provide more meaningful results from larger scale test work. Figure 17 shows a possible starting point (Stage
2).

Geotech, Geophysics, Metallurgical tests, Thin section and

Chemical Assays
Geochemistry, etc comminution & float MLA/Qemscan
mineralogy
Continuous down Spatially isolated

hole logging data data
How much of exiting data is

usable and passes QA/QC? 99% of usable data is
Understanding what has been
Chemical assays
done across the different
generations of work
(2) Creation of foundation

existing knowledge
Import & format existing

How do you site data to be compatible
Coding and labelling
incorporate to your protocols conventions
Alteration?
How does site calculate Conversion of text

mineralogy? (lith&alter) to
(appropriate?) numeric values
How do you incorporate existing
comminution & float data?
Figure 17 Step 2 - Development of a dataset which will be the foundation of the geometallurgical dataset
Once the foundational geomet data set has been established, collection of samples for further test work can
be considered. Communication with other site stakeholders is key at the point, collaboration is key to
success. A geometallurgy study is most effective when using drill core samples; planning destructive test
work procedures with limited material requires good knowledge of bankable metallurgical testing, or proxy
testing (small-scale) (Schouwstra et al 2017, Kuhar et al 2011, Chauhan 2013). Each test sample has to be of
tightly constrained mineralogy and also needs to be able to be accurately spatially positioned in the deposit.
Samples can be taken from the exploration campaign or from a mine site core yard library.

Vein
Mixed
Disseminated
Parent Rock Type

Banded Volcaniclastic Massive Porphyry
Alteration Overprint
Silica albite
sediment breccia volcanic intrusion
Transitional
Chlorite
biotite
K-spar
Figure 18. Selection and definition of end member rock texture samples
While examining drill core in the site core library, assess end member present in this deposit. Isolate ten to
twenty ore type extremes. Physically collect samples of each end member, where each sample is consistent
in lithology and rock texture (2-6m length). If veins and intrusions are in the sample, make sure they are
consistently all through the sample, not just one or two isolated places (within reason, professional judgment
required, be representative). These samples are for the Orientation Study.
Ensure there is enough sample volume for each end member, to undertake all of the planned orientation
study test work. Each end member mineralogy will need a successful (passed QA/QC) larger scale ‘bankable’
test (industry accepted in feasibility studies) as well as a number of the planned smaller scale proxy tests.
Once the end member rock texture extremes have been sampled, collect other drill core samples that
transverse all of the geological structural features of the deposit. (Texture incorporates mineral species, grain
size and shapes, weathering and alteration (Clark 1991, Lund et al 2015, Perez-Barnuevo et al 2013, and Lang
et al 2018)). Ideally, this would be continuous sections of half core from four to five drill holes. This would
represent drill core sections of the deposit that cross the different zones of the deposit, where all the end
members found so far can be put in context of how they relate to each other. This often results in something
like 1500-2000m of half drill core. Where possible, collect samples from the biggest available core size
possible to increase sample mass per meter (usually HQ or NQ core size).

Geomet Size by size

XRF/XRD Geomet Assays XRF/XRD XRF/XRD
Assays Geomet assays
FEM MLA Optical FEM MLA Optical FEM MLA Optical
Texture Ball Mill

convergence batch SMC Bond
Batch Micro- float Point Test
Grid sampling sampling grind JKCi A*b
Floats (JKMSi) Load
Texture Analysis Flotation Geotech/blasting Comminution
QA/QC protocol on Select 10-20 (?) samples with enough mass, that Destructive Tests
everything describe the end members of the texture extremes
Integration and preliminary
analysis of all outputs
(3) Geometalurgical
Orientation Study
Orientation Study
QA/QC protocol on Select 1500-2000m of continuous core after assay suite, that Non-destructive
everything describes variability and end members of whole deposit Tests
Hyperspectral EQUOTip Continuous Geological

core imaging Logging Log
mineralisation
IR/Thermal Similarity Geotech & Log Log Textural

Classification
spectrometry groupings structure lithology alteration logging
Figure 19 Step 3 - Experimental test planning ideas for the geometallurgy experimental campaign
Poor quality core is not to be rejected in sample. Traditionally metallurgical sampling (comminution in
particular) has been carried out on the good quality pieces of core, whereas the core is often broken up due
to the friable nature. This is often where the clays, heavily altered minerals and poor recovery performing
ore types are found (fines in particular in the bottom of the tray). The comminution of these samples might
be faster (often softer ore is friable) but the recovery performance is likely to be poor. As this represents
part of the deposit, it needs to be mapped and modelled if this ore type is in large quantities in the drilling
library. These samples are to be what the Mapping Hypothesis set and based on. It is with these samples
that ore variability could be assessed and mapped. The final scope of test work will probably be smaller than
noted in Figure 19 due to practical constraints. The purpose of this exercise is to consider many things, then
establish was is best for the geometallurgical campaign objectives.
2.5 Orientation study

The purpose of the Orientation Study is to investigate and plan in detail the process behavior and response
of the extreme rock textures or end members of the deposit. The full range of measurements of the process
response (for example flotation recovery, or comminution response) that might be encountered in this
deposit, and then understand the mineralogical controls which influence.

Routine QA/QC protocol on as many steps as

possible, defined by previous experience
Sample integrity and test Order of tests to be Relationship between

representatively maintained done in established bankable tests and proxy
throughout process and stable tests established
Ensure all programs interact

Maintain spatial context of samples as a Experimental design and correctly and don’t compromise
link back to site assay sample interval integration of planned tests stable each other.
(Not all at the same scale)
(4)
Test work on the
Hypothesis
Test Work Sample set
Planning
Data analysis and Test work cost and

modelling time metrics
planned out estimated
Propagation of Situational awareness of where Data matrix setup

error estimated
this experimental set sits in the established
whole geometallurgy campaign
Figure 20. Step 4 - Planning of Orientation Study testwork

3 ORE MINERALOGICAL CHARACTERISATION

Rock texture is a method of describing the structure and the style of assembly of how mineral grains are
grown together as the rock formed. Ore is a classification of a rock that has some economic value and are
defined as a mineral or aggregate of minerals that form a rock which can be mined at a profit. They can
therefore be classified, like everything else in earth sciences, into conveniently simplified categories. There
are classifications that are based on simple criteria, such as the minerals or metals contained (Pb-Zn, Cu-Au,
PGM,etc).
Rock texture has been a complicated phenomenon to measure and then classify in any consistent context.
Texture can be related to a 2-dimensional description i.e. an image, or it could relate to the physical texture.
At times, these can be miscommunicated. It is a key component to geometallurgy. Understanding rock
texture is often required to understand the ores potential process behaviour response and likely recovery.
Where the relationship between the materials of which a rock comprises is studied. The broadest and
simplified geological textural classes are but certainly not limited to (Whitten & Brooks 1972 and Craig 1994):
 Crystalline (in which the components are intergrown and interlocking crystals)
 Fragmental (in which there is an accumulation of fragments by some physical process)
 Aphanitic (in which crystals are not visible to the unaided eye)
 Glassy (in which the particles are too small to be seen and amorphously arranged)

The geometric aspects and relations amongst the component grains or crystals could be referred to as the
crystallographic texture or preferred orientation. Textures can be quantified in many ways. The most
common parameter is the grain size distribution (2D). This creates the physical appearance or character of a
rock, such as size, shape, arrangement, and other properties, at both the visible and microscopic scale.
The approach of how to characterize mineral rock texture can be complex. Measurements at different scales
will provide different kinds of results. An optical microscopy analysis on a 20mm thin section will provide
different data outcomes compared to the same sample analyzed in an automated mineralogy QEMSCAN on
100m particles. This implies that the needed answer depends on the appropriate question. The same
dilemma exists in the geotechnical mapping of the macro scale (1m to km scale). The size of the
measurement gives a different answer (Figure 21). A very small sample (say a 50mm drill core) provides a
sample of the intact rock, with no jointing, which could be used to measure intact rock strength (UCS
compressive strength or Tensile Strnegth). A larger sample of say a scan line mapping of 500mm could return
intact rock with one or two faults. A scanline mapping of 100mm could provide a rock mass with a
distribution of faults that allow the estimation of insitu block size.
Figure 21. Rock texture at macro to meso scale depends on scope of measurement
(Image. Tania Michaux)
What data needs to be collected? A geometallurgy study is not necessarily the same thing as in other studies,
thus does not have the same experimental architecture. Usually, the target is the understanding of a specific

mineral, mineral of interest (MOI), for example chalcopyrite. That being said, it is not just the target mineral
but what it is associated with and in what mineral assemblage is present. It may not be necessary to
map/analyse the entire sample. This could be relevant when planning the number of tests. A geometallurgy
campaign often requires many more such tests than a conventional study. Figure 22 illustrates different
examples of mapping the same mineral texture. The MOI is chalcopyrite, from a comminution aspect, the
host mineralogy is a priority, from a floatation aspect, the immediate mineral associations also extremely
important. Mineral association is key information when understanding floatation performance, liberation
potential and recoveries.
0.5 cm
All Minerals Chalcopyrite only Chalcopyrite neighbourhood
Figure 22. What is really needed in rock texture characterization?

(Image and copyright: Steve Walters, JKMRC, AMIRA)
Also, different process behaviors are influenced by different volumes of the rock, and are best modelled by
different scales of characterization as noted in Figure 23 (AMIRA P843 2009).
IMPACT BREAKAGE
FLOTATION
- Meso-texture
- Micro-texture (-150µm+5µm)
- (-200mm+10mm)
- Antagonistic minerals
- Rock hardness at a material level
- Particle shape
GRINDING & FINE GRINDING

LEACHING
- Micro-texture (-200µm+50µm)
- Meso-texture (-200mm+10mm)
- Mineralogy
- Mineralogy
- Grind size pf target mineral could be
smaller gangue mineral grain size - Access of target mineral to particle
surface
ENVIRONMENTAL IMPACT
- Macro-texture
- Meso-texture
- Micro-texture
- Mineralogy
Figure 23. Scales of characterization


3.1 Characterization of Meso-textures

Meso texture is scale of examination measured in centimeters (cm). A number of new technological
methods have been developed, that allow the characterization of drill core samples, where core is in the drill
tray. Before the samples are crushed and sub-sampled, the core could be imaged to characterize the meso-
scale texture.
Figure 24. Some examples of meso textures

(Source. Butcher 2019, Image: Alan R Butcher)

Examples of different meso scale rock textures can be seen in Figures 25 to 28.
Figure 25. Meso texture scale polished section sample

(Image: Marinia Sorjonen-Ward)
Figure 26. Meso texture scale drill core sample 1




3.2 Core Imaging

If the study scope and budget allows core imaging of the samples, some interesting opportunities can
develop. Historically, only micro-scale rock texture was able to be measured with relative ease. Through
technology advancements, multi-sensory capabilities, optical camera’s, x-ray imagery (CT) and gamma can
be carried out on drill core and data collected (Vatandoost et al 2008 and Shreeve 2018). The data is in the
form of imagery most recently, elemental. The images collected can be filtered in their raw form into
similarity groupings, or classified first then grouped, using specialist software (Developed at Julius
Kruttschnitt Mineral Research Centre JKMRC).
Continuous imaging
composited into 1m intervals
Classified image
Unclassified RGB image
Core imaging
classified
unclassified
FEM Modelling Modal Similarity Groupings

Process focused Mineralogy Co-occurrence matrix
Figure 29 Core imaging

(Source: AMIRA 2009, Copyright AMIRA)
The core is imaged with appropriate cameras in appropriate lighting. The images (with core depth and
intervals recorded) are then processed with appropriate software. Similar images can be grouped with other
like images (grouping lithologies in context of how they visually appear). This could be a fast and cheaper
version of hyperspectral imaging. An example of this is shown in Figure 30 to 36, which show seven of the
similarity groupings found by the software for this case study. Figure 37 shows an example of a classification
tree of images.

Figure 30. Similarity group class 1 of rock texture images from First P843 Case Study
(Source & Copyright: AMIRA P843, Ron Berry)



Class 1
Class 2
Class 3
Groups similar rock

Class 4
textures into rock types
or Image Classes
Class 5
Figure 37. Grouping of similar looking rock types from drill core imaging
(Image: Khoi Nguyen, Ron Berry, Copyright: AMIRA)

In some examples (lithology based specific cases where the signature rock looked very different), the
signatures of process behaviour can be diagnosed by proxy (with less precision). This represents a very fast
and early diagnosis of the core for previously understood signatures. This method also provides a possible
link between methods like automated mineralogy, optical microscopy and larger scale methods of
characterization like geophysics or chemical assays.
3.3 Hyperspectral characterization

Hyperspectral imaging is one of these methods and shows great potential for future work. Hyperspectral
imaging, like other spectral imaging, collects and processes information from across the electromagnetic
spectrum. The goal of hyperspectral imaging is to obtain the spectrum for each pixel in the acquisition area,
with the purpose of finding objects and identifying mineralogy, or simply elemental information. The human
eye sees colour of visible light in mostly three bands (red, green, and blue), spectral imaging divides the
spectrum into many more bands. This technique of dividing images into bands can be extended beyond the
visible.
Figure 38 Hyper spectral imaging of drill core

(Source: AMIRA 2009 & GeoTek https://www.geotek.co.uk/sensors/hyperspectral/)
Hyperspectral image characterization provides a unique opportunity, again if the budget and scope allow.
This technique method is considered to be the tool of the next generation geologist. The kind of precision
of diagnosis of minerals and rock weathering normally only was possible by an experience geologist logging
the core. Now, this method is able to provide repeatable measurement. While this method is not able to
pick up many of the subtitles of mineral diagnosis that only a geologist can observe, this form of
characterization could train an inexperienced geologist into an experienced one. Figure 39 shows an

example of this with data laid out. Each tray of core (in a continuous form in the same drill hole) is show as
a colored (characterized to mineral type) horizon line across each column. Thus, each column shows many
cores tray in a continuous form vertically. Figure 205 shows a continuous 290m length of core, in 73 trays
of core, with each tray a 4m section. The Hyperspectral instrument measured each tray (4m of core) in a
single measurement pass. Figure 49 shows 10 different spectra on the same 290m length of core (and one
column with the original grey scale image). Each tray of core (4m length) is represented by small colored
line. The lines are then stacked as a function of depth. Each column represents a different spectrum.
Figure 39. Hyperspectral imaging of kilometres of continuous drill core

Clear domains can be seen. The challenge is then to understand what the domains mean and what they
relate to. The target process behaviour may be related to only some of these signatures and perhaps not
directly. Careful and competent analysis is required to assess this sophisticated tool.
3.4 Bulk mineralogy characterization

There are many occasions in which measurement of the bulk mineralogy is useful. In particular, the
characterization (mineral phases e.g. quartz) and proportions of gangue mineralogy is useful in
understanding process behavior. In mining, gangue is the often commercially worthless material that
surrounds, or is closely mixed with, a wanted mineral in an ore deposit. It is thus distinct from overburden,
which is the waste rock or materials overlying an ore or mineral body that are displaced during mining
without being processed, and from tailings, which is rock already stripped of valuable minerals; a worthless
rock containing valuable material. There are several low cost and effective methods in achieving this.

3.4.1 Quantitative X-Ray Diffraction (QXRD)

Understanding bulk mineralogy of the sample and deposit itself is a necessary step. XRD is a versatile
analytical method to analyze material properties like phase composition and proportions of a powder
sample. Identification of the phase s is achieved by comparing the X-ray diffraction pattern obtained from
the sample being measured with a reference database. With one single measurement one can identify
mineral phases (qualitative) and their concentrations/proportions (Quantitative). X-rays are generated in a
laboratory diffractometer using x-ray tubes with a suitable anode material (Cu, Fe, etc).
For example, XRD can distinguish between a sample that contains Fe 2O3 and Fe3O4. It can also distinguish
between minerals like calcite, aragonite and vaterite (Chang-Zhong et al 2015, Ruffell & Wiltshire 2004,
Tammishetti et al 2015, Potts 1987 and Rollinson 1993).
XRD measurement requires a sample is of crystalline nature and is pulverized and homogenized, pressed in
a pellet or back loaded in a sample holder. A powder diffractogram is obtained by counting the detected
intensity as a function of the angle between incoming and diffracted x-ray beam. In the diffractogram clay
minerals have their distinguished diffracted peaks at very low 2theta angles and analysis of clays required
good low angle performance and resolution on the XRD equipment. (FWHM = Full Width Half Maximum).
Amorphous minerals, i.e. sample not having any crystals = lattice plains are particularly difficult to measure
and does require expert sample preparation, internal amorphous standards and knowledge to analyze and
process the data.
3.4.2 X-ray Fluorescence measurement (XRF)

XRF is an elemental analysis and can chemical elements are present and what are their concentrations. For
example the sample contains iron (Fe) and calcium (Ca) in a measured proportion. An X-ray fluorescence
(XRF) spectrometer is an x-ray instrument used for routine, relatively nondestructive chemical analyses of
rocks, minerals, sediments and fluids. In a laboratory X-ray fluorescence (XRF) spectrometer the emission of
characteristic "secondary" (or fluorescent) X-rays from a material that has been excited by being bombarded
with high-energy X-rays (Fitton 1997, Potts 1987 and Rollinson 1993).
It works on wavelength-dispersive spectroscopic principles that are similar to an electron microprobe
(EPMA). However, an XRF cannot generally make analyses at the small spot sizes typical of EPMA work (2-5
microns), so it is typically used for bulk analyses of larger fractions of geological materials. The relative ease
and low cost of sample preparation, and the stability and ease of use of x-ray spectrometers make this one
of the most widely used methods for analysis of major, minor and trace elements in rocks, minerals, liquids,
solids, oils and sediment.
Virtually any solid or liquid material can be analyzed with and EDXRF or WDXRF spectrometer. There are
“standardless analysis software’s” available commercially which can be used when analyzing samples. Using
of such a software give good indication of the elements and their concentrations. The best results one can
obtain with matrix specific calibrations. For rocks and minerals, typical commercial instruments require a
sample constituting at least several grams of material, pressed pellet between 15-20 grams, although the
sample collected may be much larger. For XRF chemical analyses of rocks, samples are collected that are

several times larger than the largest size grain or particle in the rock. This initial sample then suffers a series
of crushing steps to reduce it to an average grain size of a few millimeters to a centimeter, when it can be
reduced by splitting to a small representative sample of a few tens to hundreds of grams. This small sample
split is then ground into a fine powder by any of a variety of techniques to create the XRF sample. Care must
be taken particularly at this step to be aware of the composition of the crushing implements, which will
inevitably contaminate the sample to some extent. Ideally powdered sample has grain size at 10µm scale
but best results are obtained using glass beads where all grains have been melted in a flux. Sample size for
a compressed powder pellet XRF measurement is 10 to 50 grams, but less can be used depending on
mineralogical circumstance.
3.4.3 Hand Held XRF measurement

Portable XRF analyzers are used around the world for monitoring ore grade in active open pit and
underground mines. Specifically, XRF analyzers are used for direct analysis of blast hole cuttings in mines, as
a preliminary guide on direct rock faces underground, open pit faces, and within Ni mine labs after sample
preparation. Advantages of using a hand held unit could include:
 Rapid indication of potential Au mineralization through XRF analysis of pathfinder elements in soil, drill cuttings, and drill
cores
 Priority sample selection, maximization of analytical budgets, and improved drill target generation by prescreening
samples with XRF
 Improved ore deposit understanding, modelling, and vectoring with less dilution and better Au recovery via XRF mapping
of structural features and identification of mineralized and altered zones
 Inexpensive and rapid rock typing by using XRF for lithogeochemistry
Figure 40. Hand held XRF unit being used on drill core
(Image copyright: GTK)

3.5 Characterization of Micro-texture

Ore micro-texture is the mineral assembly form at the mineral grain scale (100m). Understanding what
form the ore’s micro-texture takes can be the solution to why a particular process behavior responds the
way it does. A flotation bubble will interact with a particle, which would be made up a portion of micro-
texture. Of particular interest to flotation are the following basic types of textures. Starting at the most basic
level (Figure 41 LHS), an ore texture can be evaluated as to whether it is equigranular or inequigranular. At
the most fundamental level, ores can be considered either equigranular (all grains are the same size) or
inequigranular (not all the same size). Mineral grain size is defined by the form of mineralization (Figure 41
RHS).
Figure 41. Basic ore textures, at a grain size scale (LHS), textures that influence flotation performance (RHS)
It is also important to understand the formation process of the ore texture in context of geological events
that have produced the final outcome. Figure 42 illustrates some of the geological processes that can
change rock texture. Possible changes of the three pristine ore textures is considered (one equigranular and
two inequigranular – Figure 41 LHS), which undergo progressively more aggressive modification from left to
right (oxidation, hydrothermal alteration to metamorphism).

Figure 42. Basic ore textures, at a grain size scale (LHS), textures that influence flotation performance (RHS)
To characterize a rock texture, a mineralogist would examine the following (Butcher 2019):
 complete inventory of all known mineral phases present
 detailed modal analysis for major minerals (>5%), minor phases (>1% to
 <5%) and trace phases (<%)
 textural information on both gangue and ore minerals, which includes:
o grain size estimates(used to determine optimum liberation grain sizes and predict or
prevent problems such as poor liberation, over-grinding of the valuables, production
of slimes)
o mineral associations(important for optimizing the separation process, say galena from
sphalerite, chalcopyrite from sphalerite)
o grain boundary relationships(curved, straight or irregular boundaries will affect behavior
during crushing and grinding as this imparts preferential breakage mechanisms – for
example along or across boundaries)
o elemental deportment information (important for tracking how metals behave during processing
o photomicrographs of typical textures.
Figure 43 shows examples of ore microtextures.

Figure 43. Examples of different ore micro-textures in optical photomicrographs, taken using a digital camera
attached to a reflected polarising petrographic microscope, of processed particles from an Australian lead–zinc
operation (McArthur River), which contain textures that historically would have been too fine for conventional
flotation, and now require combinations of staged and fine-grinding methods to separate the galena and sphalerite.
Figure 44 shows the mineral mapping of three particles with the chalcopyrite colored blue.
Figure 44 Chalcopyrite micro-texture of three very different forms

(Image. François Vos)
If there were three separate samples, with most of the particles in each sample were characterized by one
of the images in Figure 44, the process response would be very different for all three samples. The flotation
or leaching recovery of each of the above particles would be quite different. Measuring micro-texture in this

form allows deterministic conclusions to be drawn which could be used to guide the rest of the
geometallurgical campaign.
Figure 44 was generated using automated mineralogy (TIMA/MLA etc.). Automated mineralogy as a
characterization tool has advanced considerably. Reliable instrumentation, continuously updated software
capabilities and faster data acquisition. Samples are mounted into polished resin blocks and are mapped
using a scanning electron microscope (SEM). This method can be good for mapping micro-textures, but it is
not as effective for gangue mineralogy in some cases, or with mineralogy with very similar back scattered
electron grey scales. However, it is best practice to use complimentary analysis to establish good mineral
chemistry with probe work and LA-ICP-MS or Raman, depending on the MOI (Hrstka et al 2018, Aylmore et
al 2018, and Anderson et al 2014).
Figure 45. In the foreground is the new FEI Quanta 650 field emission scanning electron microscope, in the
background is an older MLA equipment. (Image and copyright: GTK)
As gangue mineralogy is also of interest to many geometallurgical analytical questions, automated optical
microscopy is useful. This is a cheaper and faster method compared to automated mineralogy, but does
require expert knowledge and a well-trained user. Figure 36 is an example of a desktop optical microscopy
with a resin block and a second image using cross-polarized light (thin section), clearly exhibiting gangue
mineralogy in an array of reflected colors.
Figure 46. Optical microscopy. Picture shown is an Optical Microscope Nikon MM 40

(Image: GTK)

4 CONVENTIONAL PROCESS CHARACTERIZATION TESTS FOR PROCESS BEHAVIOR

Multiple tests of the target process behavior should be carried out for each of the end member samples.
Larger scale tests that are trusted by industry in feasibility studies are referred to as ‘bankable’ tests. These
tests are considered the industry standard when characterizing ore for process engineering design. Table 2
below shows a list of these tests and reference for each for further examination.
Table 2 Larger scale ‘bankable’ ore characterization tests for process engineering
Process Behaviour
Process Engineering Field Test Reference
Characterised
Geotechnical Compressive Strength UCS Brady & Brown 2006
Tensile Strength Brazilian Disc Brady & Brown 2006
Fracture Toughness KiC Brady & Brown 2006
Comminution Impact Breakage Drop Weight Test (DWT) Napier-Munn et al 1996
Bed Breakage Lab Scale High Pressure Grinding Roll (HPGR) Wills & Napier-Munn 2005
Grinding Bond Ball Mill Work Index (BMWi) Napier-Munn et al 1996
Bond Rod Mill Work Index Napier-Munn et al 1996
Fine Grinding Laboratory Scale Isa Mill Wills & Napier-Munn 2005
Bond Abrasion Index

Abrasion Wills & Napier-Munn 2005
(Allis Chalmers Abrasion Test)
Flotation Flotation Recovery Batch Flotation Test Runge 2010
Hydrometallurgy Leaching Leaching Recovery Column Leach Test
Acid Mine Drainage

Acid Mine Drainage Static & Kinetic Tests Parbhakar-Fox 2013, 2015, 2016
(AMD)
The outcomes of tests are used as characterization inputs into process engineering. The geometallurgy
campaign objectives are to understand the mineralogy that controls process behavior response and then
map variability of that process response. This means that the feed samples mineralogy is required to be
known and consistent throughout the sample volume. Each of these larger scale bankable tests requires a
relatively large amount of mass. Many past conventional metallurgical test work campaigns have not been
that precise in ensuring the sample origins are consistent and understood in terms of source mineralogy.
Sampling large masses like what is required for the tests in Table 2 from drill core often meant that the final
sample was from mixed mineral sources.

4.1 Blasting in Geometallurgy

The prediction of blast induced fragmentation has usually been done with a combination of empirical
models. What happens in the blast defines and influences the efficiency of every downstream process.
Figure 47. A production scale blast in a mine

(Images: Codelco Flickr)
There are several blast fragmentation models, which are used as circumstance dictates, to create a
composite size distribution. The Kuz-Ram model predicts the coarse end of the size distribution of
fragmentation produced with reasonable accuracy and in a practical manner. This model is based on an
empirically derived equation that predicts mean fragment size from powder factor, explosive mass per
blasthole, relative weight strength of explosive and rock blastability constant (Kuznetsov, 1979; Cunnigham,
1983 & 1987). The fine end of the fragmentation size distribution is predicted using one of several fines (fine
fragmentation less than 1mm) estimation models, used by the JKMRC (Djordjevic 1999, Kanchibotla & Valery
1999). These models use the following inputs to predict how a volume (termed average in this section) of
rock will fragment in the applied blast:
Rock Inputs Design Parameters Explosive Inputs

•UCS strength •Burden x Spacing •Explosive Density
•Tensile Strength •Hole length •Explosive VoD
•Youngs Modulus •Charge length •Relative Weight Strength
•Rock Density •Stemming length
•P-wave Velocity •Inter hole timing
In 2010, a method was developed to predict blast fragmentation (an approximate estimate only) from drill
core (AMIRA 2010b and AMIRA 2010c). This method was developed by the W.H. Bryan Mining Centre
(University of Queensland) in to the iFrag software. A drill that has been characterized appropriately
(fracture frequency and well log geophysics parameters) that would pass through a volume of rock that later
would become a production scale blast bench is selected. The drill hole is then modelled like an explosive
borehole and fragmentation around it is estimated in a 1m interval layered fashion, where zones of
fragmentation are modelled (as in Figure 48). The explosive design and explosive products used is kept the
same to demonstrate the difference in rock type. From this model, a crushed envelope and a breakage
envelop around the drill hole is estimated (Figure 49). These results are then scaled up to a volume of a

production scale bench. It is to be remembered that this approach was designed to be done on drill core
during a pre-feasibility study, well before a ‘bench’ is assessable to characterize let alone blast.
Seismic Zone
(elastic interaction only)
Figure 48: Zones of Fragmentation around an explosive blast hole (plan view)
(Source: Michaux 2006)
Input data
Bench height
to be blasted
Predicted
fragmentation
Figure 49. Layered composite fragmentation of BH3 prediction from drill core using iFrag
(Source and copyright -AMIRA 2010b, Italo Onederra)
The model inputs into the iFrag software are:
Rock Inputs Explosive Inputs Design Parameters

•Tensile Strength (from EQUOtip) •Static •Static
•Young's Modulus
•Rock Density
•P-wave Velocity
•Fracture frequency (sliding window of 10m)

To compare the two approaches, a drill hole was selected from the client database (Case Study R) and the
relevant data was collected. A section of this drill hole was used for the comparison. The drill holes was
segmented into four 10m intervals to simulate 4 blasted benches, stacked on top of each other (where the
third bench BH3 is shown in Figure 50). The fragmentation around the drill hole in each virtual bench is then
estimated using the layered composite iFrag model (shown in Figure 51 as ‘Composite’ for each BH).
The same volume of rock is used to predict the fragmentation subject to the same explosive blast design but
using the conventional Kuz-Ram model (Cunningham 1987) and the crush zone model (Kanchibotla and
Valery 1997). This is shown in Figure 51 as ‘Average’ for each BH.
Case Study R
BH1
BH2
BH3
BH4
Figure 50. Blast induced fragmentation (Source and copyright: AMIRA 2010b, Italo Onederra)
100
90
80
70
Cumulative % passing
BH1-Average
60
BH1-Composite
50 BH2-Average
BH2-Composite
40 BH3-Average
BH3-Composite
30
BH4-Average
20 BH4-Composite
10
0
1 10 100 1000 10000
Size (mm)
Figure 51. Comparison of the layered composite fragmentation prediction from drill core to the conventional
average volume approach using Kuz-Ram and CZM (Source and copyright: AMIRA 2010c, Italo Onederra)

This section is to show that it is possible to apply geometallurgy to blasting problem solving and that it can
be used in systems based operations prediction if the input data is available. Whether this is a worthwhile
exercise has to be considered in context of the overall aims of the geometallurgical campaign. In practice,
the primary crusher can break down any larger oversize fragments reasonably easily and at relatively low
power cost. The fines produced in blasting though would heavily influence all downstream processes.
4.2 JKMRC Drop Weight Impact Breakage Test

The Drop Test was developed by the JKMRC (Julius Kruttschnitt Mineral Research Centre) to model the rock
breakage process in comminution machines. JKMRC comminution models are based on two sets of
parameters.
Classification
Recycle
Breakage
New feed Discharge
fi pi
Low
energy
High impact
energy
impact Grate
Figure 52. Schematic diagram of AG/SAG mill process mechanisms

(Source: Napier-Munn et al 1996, copyright: JKMRC)
As the JKMRC methods aspired to model the ore characteristics separately from the processing machine
characteristics, these parameters were to be measured independently. To be useful, these parameters
needed to be experimentally measured in a way that the simulation models can exploit. The Drop Test has
been the most successful way to characterize the high energy impact of mineralized ore this to date.
The principle behind the JKMRC Drop Weight Tester is simple but very effective. A known weight is dropped
from a known height onto a single particle of a known size. Thus a known energy is applied to a single
particle. This is achieved with the Drop Tester (Figure 53).
The objective of the Drop Test is to characterize the resistance to size reduction an ore has over a range of
applied energies by measuring an energy breakage curve used for comminution engineering. Individual
breakage events (points on the energy breakage curve) can be expressed as a specific comminution energy
level, Ecs (kWh/t), or as breakage or appearance functions. The principle of the drop breakage process is
well known (Napier-Munn et al, 1996). The full Drop Weight Test (DWT) procedure is described as follows.
Five different particle sizes are prepared into groups of relevant sizes to provide a representative sample

mass to be measured. These groups of particles are then broken individually at known energies (defined by
a known mass dropping from a known height onto a particle of known mass), which are then collected and
the size distribution is measured. A DWT sample has a starting mass of approximately 200kg.
Figure 53 The JKMRC Drop Weight Test

(Source and image copyright: JKTech, https://jktech.com.au/)
After breakage, each of the samples are sieved to measure the size distribution. The t 10 of each size
distribution is calculated where the t10 is the percent passing 1/10th of the initial mean particle size
(calculated with a spline function).
Product size distribution Relative size distribution showing t markers
Figure 54. Calculation of t10 from a size distribution

(Source: Napier-Munn et al 1996, Copyright: JKMRC)
The parameter of t10 is significant as it can be used to describe the whole family of t curves from t2 (1/2 of
the mean initial size) to t75 (1/75th of the mean initial size), shown in Figure 55.

Figure 55. The t family of curves and degree of breakage, t10

Figure 56 shows an example, a basalt sample of 50 particles of size fraction -11.2+9.5mm would have a
geometric mean size of 10.35mm. This in turn would give a t10 of 1.035mm. If these 50 particles were broken
in the manner shown in Figure 56, the percentage volume of the sample that is then below 1.035mm is the
t10 value.
Pure Basalt [A]

100
90 E= 2.0 (kWh/t)
E=1.0 (kWh/t)
80
E=0.5 (kWh/t)
Cumulative (%) Passing
70
E=0.2 (kWh/t)
60
50
40
30
20
10
0
0,01 0,1 1 10
Size Fraction (mm)
T10=1.03mm
Corrected E (kWh/t) 2.0 1.0 0.5 0.2

T10 (mm) 28.90 13.95 6.85 2.35
Figure 56. Size distribution of 4 basalt samples broken at 4 different energies


In the example shown in Figure 56, the sample broken at 0.2 kWh/t has 4.4% of the sample volume broken
to a degree finer than the t10 size of 1.035mm. Alternatively, the sample broken at 2 kWh/t has 38.1% of the
sample volume broken to a degree finer than the t10 size of 1.035mm. This procedure is performed for all
four breakage energies and the energy breakage curve is fitted.
Ecs (units of kWh/t) is the applied specific comminution energy, and A and b are the ore impact breakage
parameters (Napier-Munn et al, 1996). The parameters A and b are calculated by fitting the above equation
to the t10 and Ecs data, to form the energy breakage curve (Figure 57 and 58).
t10 (%)
A = 49.1 50
40
30
20 b = 0.87
10 t10 = [1-e –b.Ecs]

0
0.0 0.5 1.0 1.5 2.0 2.5
Specific Comminution Energy Ecs (kWh/t)
Figure 57. The energy breakage curve shown mathematically

T10 (%)
41.2
38.1
20.3
19.8
T10 (%)
9.9
10.4
4.9
4.4
A*bA*b
= 26=23 (hardhard
(relatively ore)ore) JKRBT
0.2 0.5 1.0 2.0

Energy Applied (Ecs)
Figure 58. The energy breakage curve shown graphically

(Image: Simon Michaux, copyright: AMIRA)

The energy breakage curve is influenced by the size of the fragments tested (Figure 48). For this reason the
full Drop Test examines five size fractions ranging from 63mm down to 13.2mm. If the objective is to
characterize the ore to design a production scale mill, then the full Drop Test would be required to capture
this spread of curves across the particle sizes.
Thus the t10 parameter can be used to describe the breakage phenomena size distribution as a single number.
The amount of breakage, or breakage index, t10 is related to the specific comminution energy with the
formula:
Equation 1
t10 = A[1-e(-b.Ecs)]
The parameter A is the t10 asymptote sill or maximum of the breakage curve (Figure 59). It indicates the
natural breakage limit of the rock that would not be exceeded no matter how much energy was used to
break it. The parameter b represents the shape of the energy breakage curve as it approaches the asymptotic
maximum (Figure 60). It represents how quickly the rock reaches this maximum.
Parameter A for Sample X
Process Path C
Sample X
Parameter A for Sample Y
Sample
Process PathY B
% TRecovery
10 (%)
Parameter A for Sample Z

Process
SamplePath
Z A
Time
Ecs
Figure 59. The A parameter in the JKMRC Axb impact breakage characterization method - conceptual example
Parameter b for Sample X
Process Path C
Sample X
Parameter b for Sample Y
Sample
Process PathY B
% TRecovery
10 (%)
Parameter b for Sample Z

Process
SamplePath
Z A
Time
Ecs
Figure 60. The b parameter in the JKMRC Axb impact breakage characterization method - conceptual example

The parameter of Axb (A times b) is used by the industry as a comminution dimensionless ranking parameter.
It is possible that both A and b parameters could be related to rock texture. This has yet to be established
and is the subject of further work.
These parameters of A and b are obtained from proprietary JKMRC Drop Test software routinely used to fit
the ore specific parameters to a set of given t10-Ecs results. A*b is used in the JKSimMet software and
accepted by mining companies as a useful rank of ore hardness. A high A*b is an indication of a soft ore to
crush, whereas a low A*b indicates a hard ore.
Hard ore Soft ore
Figure 61. A measured sample referenced against the JKTech Axb Database
(Source and copyright: JK Tech Pty. Ltd)
The objectives a geometallurgy project are however to characterize small volume samples. A full Drop Test
requires 200kg -70mm fragmented ore which is not practical for a geometallurgical campaign. Alternatively
the A, b and A*b parameters can be calculated on a single small size fraction (which could be drawn from a
much smaller volume of rock) with the understanding this is a ranking exercise and the energy breakage
curve would not have the precision required for a plant survey (see the SMC test and Rotary Breakage Test
in Section 5.3 and Section 5.4).
4.3 JKMRC Abrasion Test

The Abrasion Test is used by the JKMRC to quantify the ores resistance to low impact energy breakage
(sometimes called attrition). The test is performed in a specially manufactured tumbling mill (300 mm
diameter and 300 mm long) with four 10mm lifter bars. A representative 3kg sample of –53+37.5mm
material is selected from the ore (+/-5g). The number of particles in this sample is recorded. The sample is

then placed in the mill and run for 10 minutes at 53 rpm (70% of critical speed). The sample is removed from
the mill and sized on a 2 sieve series down to 38µm. From this size distribution, the t10 is measured.
The abrasion ore parameter, ta, used in the SAG/AG model is defined to be 1/10th of the t10 parameter.
4.4 High Pressure Grinding Roll (HPGR)

A high pressure grinding roll, (often referred to as a HPGR or roller press), consists out of two rollers with
the same dimensions, which are rotating against each other with the same circumferential speed (Figure 62
and 64). The special feeding of bulk material through a hopper leads to a material bed between the two
rollers. The bearing units of one roller can move linearly and are pressed against the material bed by springs
or hydraulic cylinders. The pressures in the material bed are greater than 50 MPa (7,000 PSI) (Figure 62). In
general they achieve 100 to 300 MPa. By this the material bed is compacted to a solid volume portion of
more than 80%.
Hydraulic Feed Stream

Pistons Floating Roll Fixed Roll
Zone of Compression
Product Stream
Figure 62: High Pressure Grinding Roll cross section schematic

The HPGR has the capacity of breaking very hard rock, where a Semi-Autogenous Mill (SAG) would struggle.
In terms of rock characterization (based on A*b impact breakage parameter and t a abrasion low energy
characterization), there is a SAG comfort zone (Figure 63). As this is the most fundamental process design
decision (SAG mill vs. HPGR) due to CAPEX cost, this means that it will be very important to understand how
hard the ore is.

Ore so
Ore too hard for SAG and

Low energy abrasion breakage (ta)
is slow to break down

soft, Ball
Mill Only
1.8
Sag Mill
Comfort Zone
0.45
Ore too hard
Ore too abrasive, slow to break with
for SAG,
attrition, will wear plant steel
HPGR only
27 83
High energy impact breakage (Axb)
Figure 63. The SAG mill comfort zone (Source: JKTech)

Extreme pressure causes the particles inside of the compacted material bed to fracture into finer particles
and also causes micro-fracturing at the grain size level. This is often referred to as compressed bed breakage
(Wills and Napier-Munn 2005). Compared to ball mills, HPGRs achieve a 30 to 50% lower specific energy
consumption, although they are not as common as ball mills since they are a newer technology.
Principle of the HPGR process
Compression
Zone
Particle bed
50-400 MPa
Working gap
Figure 64. HPGR principle of breakage

(Image: Mike Daniel)

Stress
Strain
Multi-point compression
Conventional crush Compressed bed plus shear = finer product
Figure 65 Compressed bed breakage fundamentals

(Image: Mike Daniel)
HPGR breakage is influenced by compressive rock strength (measured with the UCS compressing strength
test). This can be empirically ranked using measurement of the A, b and A*b impact breakage parameters
(Wills and Napier-Munn 2005). Bed Breakage behavior also has influence (Figure 65). This is the multi-point
loading of force on each particle under pressure, instead of the single point loading of force that happens in
impact breakage.
The accepted method of characterizing the HPGR response is to conduct a laboratory scale test run (Figure
66). Several test runs can be done at different rolls pressures. The feed and product samples are then sized
on the 2 sieve series. The feed sample for this is 25kg of -10mm crushed ore for each run.
Figure 66. Laboratory scale HPGR test work


4.5 Grinding Bond Ball Work Index

The Bond Index is used to determine the energy requirements in grinding an ore down to a target size
distribution (referred to as the closing size of the test, which is defined by a P 80). This is a comminution
design tool developed by Fred Bond (Bond 1962). Its purpose was to determine which crusher/mill is
appropriate to treat the target ore (Ballantyne & Powell 2014).
99% passing 3.35mm Ball mill ground product

Crushed sample
Figure 67. The Bond Ball work index test, crushed feed and mill product
The work index is a comminution parameter, which expresses the resistance of the ore to crushing and
grinding. The objective of the Bond ball mill test is to calculate this numerically, and is expressed as the
standard Bond Mill Work Index (BMWi). The BMWi is defined by the specific power required (kWh/t) to
reduce an ore from an infinite size to a P80 size of 100µm. This is done through a series of consecutive grinds
in the Bond mill.
According to Bond’s third theory of comminution, the work input is proportional to the new crack tip length
produced in particle breakage, and equals the work represented by the product minus that represented by
the feed. The relationship is expressed as follows:
1 1
𝑊 = 10 ∗ 𝑊𝐼 ( − ) Equation 2
√𝑃80 √𝐹80
and P = TxW Equation 3
Where W = work input (kWh/t)

WI = work index – a material specific constant (kWh/t)
P80 = size at which 80% of the product passes (m)
F80 = size at which 80% of the feed passes (m)
T = throughput of new feed (t/h)
P = power draw (kW)

To conduct a test, a prepared sample (of a specific volume of 700cc of –3.35mm material) is placed into a
mill of specific size and dimensions (known as the Bond Ball mill) with a specific ball charge (number of balls
at a range of set sizes, which all have a specific mass) and ground dry for 100 revolutions. The sample is then
removed from the mill and sized at the closing size (defined by the grain size of the ore) to remove the
undersize of the target P80.
The test is based around the constant sample volume of 700cc. When each grinding cycle is complete and
the undersize is removed, and is replenished with an equal mass of new feed. The length time of each batch
grind is adjusted until the mass of the oversize fraction is consistently 2.5 times greater than the undersize.
Under these conditions the test approximates the performance of a closed circuit of a continuous mill with
a recycle load of 250%. A full Bond test has typically 7 to 8 cycles and requires approximately 5kg of sample
as a starting mass.
The closing size is a controlling parameter for the test, and is related to the grain size of the target mineral
to be extracted. A smaller closing size will result in more energy required to grind the rock to that target size
and visa versa.
Once the standard work index (WI) measured from experimental measurement, Bond’s Third law of
comminution is used to calculate the specific power required to reduce a given F 80 to the required P80 in an
8 foot diameter wet overflow ball mill. For a given throughput (t/h) the specific power (kWh/t) is converted
to power draw (kW). Mill dimensions for a production unit are then chosen which are predicted to draw the
required power, using an appropriate mill size-power correlation.
The sample mass for a Bond Ball Work Index test is stated to be 10kg, but a sample mass of 5kg can be used
if the technicians are experienced. Coarse residues from chemical assay procedures can be used as can drill
core or coarse fragments as test feed sample. The test requires a top size of 3.35mm in particle size.
The efficiency of the ball mill is controlled by the volume of the recirculating load (250% when operating
normally). Harder minerals usually persist at the 2 to 3 mm size range and resist grinding. If a BMWi is in
the higher end of the spectrum (18-25 kWh/t), it can be useful to understand what minerals are present. To
characterize the feed and products, the following methods are useful:
 Chemical Assay
 Bulk QXRD
 XRF pellet
 Automated mineralogy in some circumstances
4.6 Fine Grinding – Laboratory scale Isa Mill

Fine grinding is increasingly required for an operation to be economically viable. As time has progressed,
many of the easy to find and easy to extract ore deposits have been mined out. In the late 1990’s a new fine
grinding comminution mill was commercialized called the Isa mill. The Isa mill made it possible to mine and
extract ore with much finer mineral grains (resulting in much higher cost of production due to cost of
grinding). The Isa mill was developed due to a very fine grained ore was discovered in an existing deposit in
Mt Isa in Australia. If the fine grained ore was to be processed, technology to fine grind had to be developed,
which would later become the Isa Mill.

Figure 68 shows the difference in complexity between coarse grained, high grade ore (LHS) and the fine
grained ore with refractory fine pyrite dilution (RHS) (Young et al 1997). Both of these samples were taken
from the No 5 orebody at Mt Isa.
Figure 68. Mineralogy and plant closing size (Source: Young et al 1997)
The IsaMill is primarily known for its ultrafine grinding applications in the mining industry, but is also being
used as a more efficient means of coarse grinding, with target product sizes ranging from 5 to 60 µm. Figure
69 shows what size fractions the different comminution devices work with.
Figure 69. Fine grinding in relation to other comminution methods

(Source: Metso blogs, https://www.metso.com/blog-hub/?pageSize=13&language=English)

A currently available method for characterization for fine grinding is a test run in a laboratory scale Isa Mill
(Figure 70). A feed sample was prepared in 15kg batches using the large rod mill. The particle size of each
of the feeds was below 100 micron. The time required to run the rod mill depended on the hardness and the
size of the initial particle and was again estimated using the ‘grind, check, predict’ method.
Figure 70. A laboratory scale Isa mill

After the correct feed size is achieved, the sample is mixed with water (45% by weight sample, 55% water)
in order to make a slurry of the correct consistency. The slurry is then transferred into a feed sump where a
stirrer keeps it in suspension.
The IsaMill was set to be running at 1605 rpm with a flow of 2.4 liters/min. A feed sample was taken from
the output hose after allowing the slurry to recirculate from the bottom of the sump, through the pump and
back into the top of the sump. A specific gravity and flow rate measurement was taken at the same time as
well.
When the IsaMill is set up according to the standard operating procedure (SOP), the slurry was directed
through the mill. As per the SOP instructions the IsaMill should not be started until 20 seconds after the
flow has been directed through the IsaMill to prevent overheating. This is the length of time needed for the
slurry to reach the mill from the sump. The first sample was taken from the output hose when approximately
half of the slurry had passed through the mill. Once the feed sump was close to empty the slurry flow was
diverted using a multi-directional valve whereby the original feed tank now serves as the product tank and
vice versa. An energy consumption reading for this cycle was recorded at the time of cycle completion. This
process was repeated many times in an attempt to reach an energy saturation point for that sample.
The size distribution of the feed and products is to be measured in an optical sizer in a slurry.

As fine grinding is often happening at a particle size smaller than mineral grains, the mineralogy of the feed
ore will control the effectiveness and efficiency of the process. To characterize the feed and products, the
following methods are useful:
 Chemical Assay
 Bulk QXRD
 XRF pellet
4.7 Selection of grind size for the whole campaign

One of the critical decisions to make that has the capacity to make the entire geomet campaign fail if done
incorrectly is the selection of sample particle grind size. This is the particle size (expressed as a P80, or particle
size that 80% of the sample is smaller than) to which all samples for process separation a milled down in size
through grinding.
If the target minerals are very small in grain size or are embedded in complex rock texture, the automated
mineralogy done on polished block mounts will be required. Figure 71 shows the possible difference in rock
texture at similar grades. Each example in Figure 71 would have a different recovery and would need to be
reduced to different grind size for full liberation.
Figure 71 Different rock textures, each would have a different process recovery
(Source and copyright: Cropp 2013)
The process of comminution is to subject the sample to size reduction. The particle size required for the
effective liberation of the target mineral is defined by the mineralogy. For each mineral in the sample, there
would be a range of mineral grain size distribution. For process separation to economically effective, enough
of those particle have to be liberated or partially liberated.

Figure 72. Particle size reduction through comminution and liberation of target minerals
(Image: Napier-Munn et al 1996, Copyright: JKMRC)
The process of size reduction through comminution is important as it has economic significance. Curry et al.
(2014) found that the mill (defined as crushing, grinding and separation) typically accounts for between 35
and 50 per cent of the total mine costs. The size reduction to energy relationship is not linear though. The
finer the grind size, the exponentially larger energy required to comminute the ore to that size (Ballantyne
& Powell 2014 and Hukki 1962) (Figure 73).
The tradeoff between the cost of comminution to the target grind size and the degree of liberation and
texture of the valuable mineral particles has to be considered carefully. Thus the full distribution of the
target mineral needs to be understood. If there is more than one target mineral associated with different
textures (with different particle sizes at liberation), then each need to be considered in context of value of
the ore.
An unconventional method of defining what the size distribution of the mineral grains are inside an ore type,
could be the use of electrodynamic fragmentation in a Selfrag unit (see Section 4.7.1). If done correctly, and
if the mineralogy is susceptible to this methodology, all minerals could be liberated intact at their true
particle size. This method has its advantages if the ore has more than one valuable mineral (possible
polymetallic process path), where each target mineral would have its own unique mineral grain size
distribution. For example, an ore with a Cu mineral with a P 80 grain size of 100 micron and a Au mineral
with a P80 grain size of 10 micron.

Figure 73. The Hukki energy size relationship shown as an imaginary example of the basic reduction characteristic
plotted on logarithmic paper. (Source: Hukki 1961)
The grade of concentrate and recovery are the most common measures of metallurgical efficiency. Batch
flotation tests are often examined together with several ores from the same deposit in a grade recovery
curve (Figure 74).
Usually this is defined through a sophisticated test procedure to determine the grade recovery curve for that
mineral in that ore type. The theoretical grade-recovery curve for an ore is a definition of the maximum
expected recovery by flotation of a mineral or element at a given grade. This is defined by the surface area
liberation of the value minerals and is consequently directly related to the grind size utilized in the process.
The theoretical grade-recovery can be readily used to quickly identify potential recovery increases that can
be gained through optimization of flotation circuits and whether the process is running efficiently.

Figure 74. Theoretical grade recovery

(Source: Reproduced from Goodall 2009, Image: Tania Michaux)
Flotation performance is most commonly represented using grade-recovery curves, with a shift upwards and
to the right indicating an improvement in performance. Figure 75 shows the effect of finer grind size on
flotation performance by virtue of more target particles being liberated.
100
Pure mineral grade
80
Grade of all mineral A bearing particles
Grade of A (%)
60
40 Pure mineral Cu grade (34.6%)

Feed grade
20 Mineral grade vs. recovery

Cu grade vs. recovery
0
0 20 40 60 80 100
Recovery of A (%)
Figure 75. A grade recovery curve for perfect separation (LHS, Cu grades based on chalcopyrite, CuFeS2)
(Source: Reproduced from Wills and Finch 2016, Image: Simon Michaux)
Figure 76 is an example that shows the key aspects of using the theoretical grade-recovery curve for a free
gold flotation circuit.

Gold Grades and Recoveries in Rougher Flotation

100
90 Test 1
Cumulative Au Recovery %
Test 2
80
70
60
50
40
30
100 120 140 160 180 200 220 240 260 280 300
Cumulative Au grade g/t
Figure 76. Experimental example of a grade-recovery curve

(Source: Tero Korhonen GTK Mintec)
To establish the theoretical grade-recovery curve for a material a mineralogical liberation study should be
undertaken using tools such as automated mineralogy (QEMSCAN or MLA).
Ideally this work should be done either before or in parallel to a geometallurgical study. Given enough
automated mineralogy data, a theoretical grade recovery curve could be estimated based on mineral grain
size across a family of ore types.
A decisions is to be made for what P80 each of the process test samples are to be reduced to through crushing
then grinding. As a laboratory scale rod mill produces a particle size distribution that better resembles a
production scale circuit compared to a laboratory scale ball mill, all sample preparation should be done with
wet grinding in a rod mill with a standard rod load. The choice of mild steel or another metal alloy for the
rods is a matter of discernment in context of the ore mineralogy.
To prevent oxidization of the sample, after rod milling, the wet sample is placed in a vacuum sealed bag
(ideally) and then stored in a freezer at a lower temperature than -180C. This should be done for any sample
that would or could be later be subject to flotation or leaching.
The ideal situation is the definition of closing samples grind size is taken from previous work done on the
same ore types.

4.7.1 Electrodynamic fragmentation

There is an unconventional opportunity to determine what the natural grain size of target minerals (possible
even several minerals at once), using electrodynamic fragmentation (EDF). Electrodynamic fragmentation
provides an alternative means to mechanical crushing when preparing rock samples for mineral separation.
Convetnional breakage fragments a rock particle by squeezing quickly between two crusher jaws, or through
impact breakage, where an external forace is applied to the particle. EDF breakage loads the particle with
electricity so quickly that the induced stress field manifests inside the particle. The particle breaks apart
from internal strain (Figure 77).
Figure 77. Conventional breakage (LHS) compared to EDF breakage (RHS)

Fragmentation by rapid powerful electrical shocks can be used for the production of clean mineral separates
for mineralogical and isotope geological purposes (Michaux 2016, Everaert et al 2019, Vaasjoki and Lindqvist
2001). Figure 78 shows a 3rd generation Selfrag unit, which can break small scale samples using EDF.
Figure 78. The Selfrag electrodynamic fragmentation instrument

(LHS University of Liege ULiege Gemme Dept Unit, Image- Eric Pirard)

EDF is a process that is fundamentally based on a High Voltage (HV) physical specificity. An electric pulse is
arced between a cathode electrode and an anode electrode over such a short time period that it will
preferentially pass through the solid sample between the electrodes than the surrounding water (Figure 79).
In such a short pulse, the solid is more conductive than the water, thus is the path of least resistance.
Figure 79. Schematic of the SELFRAG laboratory scale machine

(Source & copyright: Bru et al., 2018, Selfrag High Voltage Pulse Power Fragmentation http://www.selfrag.com/)
As the short pulse passes through the solid sample, it creates an electrostatic charged stress field with
geometry defined by material texture (Figure 80 LHS and center). Electrostatic discharge occurs through
solid, causing strong internal shockwaves across stress field geometry (Figure 80 RHS), defined by a
difference in the dielectric material properties of the minerals in rock sample. In doing so, cracks would form
along the grain boundaries using Maxwell-Wagner polarization. This results in selective breakage, where
grains are fully liberated intact (Figure 81 to 85).
~500 ns
Solid Water
Figure 80. Electrodynamic fragmentation pulse speed

(Source & copyright: Alex Weh, Selfrag High Voltage Pulse Power Fragmentation http://www.selfrag.com/)

Figure 81. Maxwell-Wagner Polarization of minerals (Kaufmann 2013)

EDF breakage happens in two basic modes. The first mode is the pulse passes through a particle and breaks
the body apart in what is termed rock body breakage original particle is broken into 5 to 6 smaller pieces of
similar mass (Wang 2012). The second mode is termed surface breakage, as the pulse passes along the
boundary between mineral grains, which is how intact mineral grains are often liberated (Figures 82 and 83).
At low voltage applications, body breakage dominates. At higher voltage applications, surface breakage
dominates and individual grains are predominantly produced.
Figure 82. EDF particle body breakage and particle surface breakage

Figure 83. Definition of selective fragmentation products

(Image & copyright. Hesse & Lieberwirth)
EDF events over time can be seen in Figure 74 and 75.
Figure 84. The electrodynamic fragmentation events over time


Figure 85. How the electrodynamic field geometry forms around a mineral inside a sample
(Source & copyright: Parvaz et al 2015, Selfrag High Voltage Pulse Power Fragmentation http://www.selfrag.com/)
This represents an unusual opportunity. EDF breakage has a very unusual signature characteristic. During
trials (Michaux 2016) it was observed that whole grains were liberated completely, with their surface
complete with their original texture (this was seen in the liberation of aggregate in concrete – Figure 86 and
the liberation of metal particles from pyrometallurgical slag – Figure 87).
Figure 86. Clean liberation from electrodynamic fragmentation of concrete. The intact 50x50x30mm block (LHS).
The fragmented sample after one EDF application (RHS).
(Images: Simon Michaux)

Figure 87. Clean liberation from electrodynamic fragmentation of steel furnace slag. The intact 80mm particle (LHS).
The fragmented sample after one EDF application (RHS).
(Images: Simon Michaux)
Figure 88 to 90 show an example of EDF breakage of a granodiorite cut sample (Michaux 2016). The sample
was broken in a single step of EDF application (Figure 88). Figure 89 shows the entire sample after being
dried in an oven, then split into two sub fractions, +850m and -850m. Figure 90 shows the same samples
as in Figure 89 but mounted on a polished block mount and then imaged under a Ziess SEM microscope.
Clear particle liberation can be seen at multiple particle sizes.
10mm
Figure 88. Granodorite sample intact, before electrodynamic fragmentation (LHS),

Same sample just after fragmentation still in the sample holder (RHS) (Michaux 2016)

+850m -850m
10mm
10mm
Figure 89. Granodorite sample after electrodynamic fragmentation (entire sample after oven drying)
(Source: Michaux 2016) (Image: Simon Michaux)
+850m -850m
Figure 90. Granodorite sample after electrodynamic fragmentation on polished block mounts
What makes this interesting is the comparison to conventional breakage. Figures 91 and 92 show a sample
that has been broken with impact breakage on a Drop Weight Tester at 1.5kWh/t (Michaux 2016). This is of
a similar particle size as Figures 89 and 90. Almost all of the particles seen in Figures 91 and 92 are binary or
tertiary (of have more different minerals again) in modal mineralogy. This is what is expected with
conventional breakage.

Figure 91. Samples from an IOCG style Cu-Au deposit -1.18+0.850mm fragments (MLA mineral map)
Figure 92. Samples from an IOCG style Cu-Au deposit -1.18+0.850mm fragments (BSE images)
(Michaux 2016) (Image: Simon Michaux)
It is postulated that EDF fragments the rock with internal stress (“shaking the rock apart from the inside”)
compared to conventional breakage which fragments the rock by external application of force on the
particle. The EDF opportunity is shown below in a conceptual example (Figure 93).

Gold mineral grain

size distribution
Cumulative (%) Passing

Copper mineral grain
size distribution
Particle size (mm)
Figure 93. Conceptual example of the grain size distribution of a polymetallic ore
(Image. Simon Michaux)
Gold is worth far more than copper, but is of a much finer grain size. The economic cost of comminution to
grind to a finer closing size would be much higher to liberate the gold compared to the copper. The only
way to resolve this interesting problem is to have accurate measurement of the liberated mineral grains for
each valuable mineral in the ore.
This could be done by breaking each end member rock texture with EDF (in the Selfrag). Then appropriately
characterize the products (in a batch flotation test or a column leach test for example). If enough sample
can be prepared, process separation tests like batch flotation could be considered.
This is a wet process that requires high purity water (with high electrical conductivity). The sample is broken
in the volume between the electrodes (an ellipsoid of 40mm in Z dimension and 50mm in X and Y dimension).
This sample would be reduced to a top particle size of approximately 2mm in a very unusual breakage
signature.
Thus this unit is able to break a relatively small sample of a few hundred grams of coarse crushed breakage
or a single particle of rock of approximately 50 x 50 x 30 mm in size. Samples of several kilograms can be
prepared in several runs that are later combined.
4.8 Flotation
Flotation (also spelled floatation), in mineral processing, method used to separate and concentrate ores by
altering their surfaces to a hydrophobic or hydrophilic condition—that is, the surfaces are either repelled or
attracted by water. The flotation process was developed on a commercial scale early in the 20th century to
remove very fine mineral particles that formerly had gone to waste in gravity concentration plants. Flotation
has now become the most widely used process for extracting many minerals from their ores.

Figure 94. Production scale flotation separation plant

(Source: Codelco Flickr, https://www.flickr.com/photos/codelco/)
(Copyright License: https://www.flickr.com/help/terms)
Figure 95. The GTK Mintec pilot scale flotation circuit

(Image: Arno-Matti Kirpala)
Most kinds of minerals require coating with a water repellent to make them float. By coating the minerals
with small amounts of chemicals or oils, finely ground particles of the minerals remain un-wetted and will
thus adhere to air bubbles. The mineral particles are coated by agitating a pulp of ore, water, and suitable
chemicals; the latter bind to the surface of the mineral particles and make them hydrophobic. The un-wetted
particles adhere to air bubbles and are carried to the upper surface of the pulp, where they enter the froth;
the froth containing these particles can then be removed (Figure 95, 96 & 97). Unwanted minerals that

naturally resist wetting may be treated so that their surfaces will be wetted and they will sink. Figure 95
shows an image of bubbles generated in a flotation experiment.
Figure 95. The flotation process photographed

(Image: Dee Bradshaw)
Figure 96. Flotation phenomena


10mm
Figure 97. Bubbles generated during flotation

(Source: Nuorivaara, T. and Serna-Guerrero 2016, Copyright: AusIMM)
Figure 98. Froth flotation separation (Image: Tero Korhonen)

Flotation is the conventional method for concentrating metal out of ore, where a feed grade of
approximately 1% could be concentrated to a product grade of 25-30%. Figure 98 shows this process at a
production scale mine. As such, the flotation response of an ore to this separation process needs to be
understood in a geometallurgical context.
There are three categories of models for the prediction of flotation (empirical models, probability models,
and kinetic models).
• Empirical models are too specific to their environment and usually involve a trial and error feedback approach to
optimization.
• The probability models basically consider the probabilities of particle-bubble collision, adhesion, froth stability, etc.
• Kinetic models are established on the basis of the analogy between a chemical reaction (collision of molecules) and an
important flotation mechanism, i.e., the collision between either hydrophilic (or hydrophobic particles) and air bubbles
in the pulp volume.
To examine and model the flotation process, the kinetic approach is certainly the oldest and most
traditionally used. It postulates that the flotation rate follows a rate equation similar to those utilized to
describe homogeneous kinetics. Its foundation is based on the well-known analogy existing between bubble-
particle and molecule-molecule collisions. Thus, there are measurable rate constants, which describe
mineral recovery in the flotation process.
The kinetic study of the flotation process includes the determination of all the factors that influence the rate
of concentrate production. This rate can be defined according to various meanings, but in the operations
carried out on ores, it is generally measured as absolute recovery (unity fraction) divided by the time (kinetic
constant units: s-1 or min-1). The overall absolute recovery is one metric of interest to measure. The time
taken to reach this recovery is another metric to measure.
4.8.1 Batch Flotation Test Characterization

Batch laboratory flotation testing is the industry accepted method for characterizing samples in context of
flotation response for modelling industrial performance. Traditionally, evaluation of batch flotation tests is
based largely on a rougher recovery value achieved at a defined time. Since the cumulative recovery of a
component in the concentrate is proportional to flotation time, the flotation process can be considered as a
time-rate recovery process. Therefore, a mathematical flotation model that incorporates both the recovery
and rate function can completely describe flotation time-recovery profiles.
The recovery in a batch flotation test is the percentage of the total target metal contained in the ore,
recovered in the test (Wills and Finch 2016). Recovery is not a straight forward concept. The flotation
separation method cannot separate out all minerals and textures.

This is done by a process that involves the imparting of a water repellent (termed hydrophobic) character to
the target mineral particles using chemicals that are called collectors or promoters. The target minerals
contain metals (like copper for example) that are naturally more hydrophobic than the surrounding rock
minerals, and the metal will preferentially stick to the surface of a bubble of air, rather than reside in water.
Under favorable conditions, these chemically coated particles become attached to the air that takes the form
of bubbles that is agitated through the pulp. The bubbles float to the surface, taking the target minerals
with it. The froth on the water surface at the top of the cell can then be wiped off, forming a concentrate
(Figures 99 & 100).
Figure 99. Batch flotation bench scale test (Image: Tero Korhonen)
Figure 100. Batch flotation bench scale test (Image: Tero Korhonen)

Batch flotation starts from three fundamental types of processes, which form the basis of flotation
experimental design.
 Bulk Flotation – the separation of like and unlike minerals (e.g. sulfides from non-sulfides).
 Differential Flotation – the concentration and removal of different minerals (with elements like Cu, Pb, Zn, Au, etc.) from
a single ore type.
 Sequential Flotation - the sequential concentration and removal of minerals with the same element in several stages from
a single ore type (the opposite to bulk flotation).
For this to be effective, the appropriate quantity of valuable target mineral needs to be liberated. To achieve
this, each sample has to be ground to the appropriate grind size (See Section 4.7).
The test objectives and resulting outcomes from batch flotation should be established in planning
beforehand. Feed size distribution, method used for chemical assays of feed and products, what the valuable
materials are, what the penalty elements are, rates of flotation and bubble agitation, etc. Once this is known,
the following can be planned:
 Cell type and size to be used in the test
 Feed sample weight required (and what mass the concentrate will probably be) for each test
 Water quality and quantity
 Slurry density
 Chemical reagents including collectors, regulators and frothers to be used
All of this is to be planned with the laboratory doing the batch flotation tests. The outcome of a batch
flotation test is a mass fraction of froth measurement (% mass that is floatable) a kinetics recovery curve
(similar to Figure 101).
Gold Flotation Kinetics in Rougher Flotation

100
90
Cumulative Au Recovery %
80
70
Test 1
60 Test 2
Test 3
Test 4
50
Test 5
Test 6
40
Test 7
30
4 9 14 19 24
Cumulative Flotation Time, Minutes
Figure 101. Flotation kinetics recovery

(Source: Tero Korhonen)

The presence of some minerals will antagonistically reduce the effectiveness of flotation for both the
absolute recovery and the time taken to reach that absolute recovery. If these minerals were subject to
different process paths, the final absolute recovery would also be different.
A useful approach is to a rougher flotation followed by a cleaner flotation step. Also, many flotation circuits
target more than one metal in separate processes. Figure 102 below is an example of an experimental design
that allows this to be modelled. Copper flotation (rougher then cleaner) could be done, and then nickel
flotation could be done on the copper rougher tails. This entails four batch flotation tests done on each ore
type. The products are all characterized with chemical assays as appropriate. The concentrate is sampled
at regular intervals and assayed to produce the recovery curve similar to Figure 101.
Rougher Cleaner Concentrate

Concentrate
% Recovery
Cu Chemical Assay
Qemscan SEM
QXRD
Cu cleaner Time
rougher flotation
flotation
Cleaner
Tails
Characterized
Feed Rougher Tails % Recovery
Chemical Assay
Qemscan SEM
QXRD
Time
Ni & Co Cleaner Concentrate

cleaner
flotation
Cleaner Tails
Ni & Co Rougher Concentrate

rougher
flotation Rougher Tails
Figure 102. Flotation separation experimental design

Sample mass requirements for a batch flotation test is about 1kg for each actual batch flotation run. As each
ore type may need three to five separate tests, and sometimes sequential testing is done, thus needs larger
sample feed mass, it then it is appropriate to prepare about five (5) kilogram´s for each ore type tested. This
sample would then be prepared with wet grinding in a rod mill to the selected grind size (see Section 4.7).
To reduce the influence of ore oxidization (which would coat each particle with a layer that would inhibit
flotation), the ground sample (with some water decanted) would be vacuum packed in nitrogen, and then
stored in buckets in a freezer at -180C.

4.9 Hydrometallurgy Leaching

Hydrometallurgy is a method for obtaining metals from their ores. It is a technique within the field of
extractive metallurgy involving the use of aqueous chemistry for the recovery of metals from ores,
concentrates, and recycled or residual materials. Metal chemical processing techniques that complement
hydrometallurgy are pyrometallurgy, vapour metallurgy and molten salt electrometallurgy. Hydrometallurgy
is typically divided into three general areas:
 Leaching
 Solution concentration and purification
 Metal or metal compound recovery
Leaching involves the use of aqueous solutions to extract metal from metal bearing materials which is
brought into contact with a material containing a valuable metal. The lixiviant solution conditions vary in
terms of pH, oxidation-reduction potential, presence of chelating agents and temperature, to optimize the
rate, extent and selectivity of dissolution of the desired metal component into the aqueous phase. Through
the use of chelating agents, certain metals can selectively extracted. Such chelating agents are typically
amines of schiff bases.
The five basic leaching reactor configurations are in-situ, heap, vat, tank and autoclave.
1- In-situ leaching
In-situ leaching is also called "solution mining." The process initially involves drilling of holes into the ore
deposit. Explosives or hydraulic fracturing are used to create open pathways within the deposit for solution
to penetrate into. Leaching solution is pumped into the deposit where it makes contact with the ore. The
solution is then collected and processed.
2- Heap leaching
In heap leaching processes, crushed (and sometimes agglomerated) ore is piled in a heap which is lined with
an impervious layer. Leach solution is sprayed over the top of the heap, and allowed to percolate downward
through the heap. The heap design usually incorporates collection sumps, which allow the "pregnant" leach
solution (i.e. solution with dissolved valuable metals) to be pumped for further processing. An example is
gold cyanidation, where pulverized ores are extracted with a solution of sodium cyanide, which, in the
presence of air, dissolves the gold, leaving behind the nonprecious residue.
3- Vat leaching
Vat leaching, also called agitation leaching, and involves contacting material ore, which has usually
undergone size reduction and classification, with a leach solution in large vats.
4- Tank leaching

Stirred tank, also called agitation leaching, involves contacting material, which has usually undergone size
reduction and classification, with leach solution in agitated tanks. The agitation can enhance reaction kinetics
by enhancing mass transfer. Tanks are often configured as reactors in series.
5- Autoclave leaching
Autoclave reactors are used for reactions at higher temperatures and pressures, which can enhance the rate
of the reaction. Similarly, autoclaved enable the use gaseous reagents in the system.
After leaching, the leach liquor must normally undergo concentration of the metal ions that are to be
recovered. Additionally, undesirable metal ions sometimes require removal.
 Precipitation is the selective removal of a compound of the targeted metal or removal of a major impurity by precipitation
of one of its compounds. Copper is precipitated as its sulfide as a means to purify nickel leachates.
 Cementation is the conversion of the metal ion to the metal by a redox reaction. A typical application involves addition
of scrap iron to a solution of copper ions. Iron dissolves and copper metal is deposited.
 Solvent Extraction is a mixture of an extractant in a diluent is used to extract a metal from one phase to another. In
solvent extraction this mixture is often referred to as the "organic" because the main constituent (diluent) is some type
of oil.
 Ion Exchange is where chelating agents, natural zeolite, activated carbon, resins, and liquid organics impregnated with
chelating agents are all used to exchange cations or anions with the solution. Selectivity and recovery are a function of
the reagents used and the contaminants present.
 Gas reduction. Treating a solution of nickel and ammonia with hydrogen affords nickel metal as its powder.
 Electrowinning is a particularly selective if expensive electrolysis process applied to the isolation of precious metals. Gold
can be electroplated from its solutions.
Metal recovery is the final step in a hydrometallurgical process. Metals suitable for sale as raw materials are
often directly produced in the metal recovery step. Sometimes, however, further refining is required if ultra-
high purity metals are to be produced. The primary types of metal recovery processes are electrolysis,
gaseous reduction, and precipitation. For example, a major target of hydrometallurgy is copper, which is
conveniently obtained by electrolysis. Cu2+ ions reduce at mild potentials, leaving behind other
contaminating metals such as Fe2+ and Zn2+.
4.9.1 Column Leaching Characterization

Leaching hydrometallurgical response can be characterized by using a Column Leach test. This test requires
20-30kg of ore crushed to a top size of approximately 15-20mm.

Leach parameters such as, amount of sulfuric acid in the cure solution, cure technique, particle size of the
material, leach solution’s flowrate application, leach solution’s acid strength and height of the lifts, can be
evaluated using column leach tests results. This can be done by putting prepared ore aggregate in a column
vessel with 12″ inches leach diameter.
Figure 103. The Column Leach Test

Figure 104. Laboratory scale column leach experimental setup

(Source: CSIRO Mineral Resources, Copyright: CSIRO)
The column is then placed under a drip of cyanide solution made to concentration (of around 0.1% or
1g/litre), where the flow rate from the top should be adjusted to give a constant drip approximately 2ml/min.
This example is for gold leaching.

After 24 hours, and every 24 hours thereafter, the pH of the underflow solution should be taken and
adjustments to the solution flow can be made if necessary. After the approximate number of days (this can
be anything from 10 days to 70 days), stop the test and flood the column with hot water to clean out any
residual solution. The ore in the column is laid out and dried, then sent for chemical assay characterization.
The pregnant effluent collected from under the column an also sent for chemical assay. A final cyanide
sample is taken to determined residual cyanide by titration as a QA/QC measure.
Deposit E
Ore A
Ore B
Ore C
X Ore D
Figure 105 Leaching recovery over time for different ore types (Chart: Simon Michaux)
Sample mass required for a column leach is 20 to 30kg of -30mm crushed aggregate.
4.10 Gravity Separation

Gravity separation is an industrial method of separating two components, either a suspension, or dry
granular mixture where separating the components with gravity is sufficiently practical: i.e. the components
of the mixture have different specific weight. All of the gravitational methods are common in the sense that
they all use gravity as the dominant force.
One type of gravity separator lifts the material by vacuum over an inclined vibrating screen covered deck.
This results in the material being suspended in air while the heavier impurities are left behind on the screen
and are discharged from the stone outlet. Gravity separation is used in a wide variety of industries, and can
be most simply differentiated by the characteristics of the mixture to be separated - principally that of 'wet'
i.e. - a suspension versus 'dry' -a mixture of granular product.

Table 3 Characteristics of metals and their minerals (Reproduced from: Klein 1999, and Gaines et al 1997)
Metal Composition Mineral Formula Specific Gravity Hardness Colour & Characteristics
Stibnite (antimony
Antimony Sulphide Sb2S3 4.5 to 4.6 2.0 lead-gray, metallic luster, slightly sectile
glance)
Cobalt Arsenide Smaltite CoAs2 6.4 to 6.6 5.5 to 6 Tin-white, metallic, opaque
Sulpharsenide Cobaltite CoAsS 6.0 to 6.3 5.5 Silver-white to reddish, cubic, metallic
Copper Metallic Native Cu 8.8 to 8.9 2.5 to 3 Copper-red, malleable and ductile
Melaconite (black
Oxide CuO 5.8 to 6.2 3 Black
oxide)
Cuprite (red
Oxide Cu2O 5.8 to 6.1 3.5 to 4 Red, brittle
oxide)
Carbonate Malachite CuCO3Cu(OH)2 3.9 to 4.0 3.5 to 4 Green, brittle
Carbonate Azurite 2CuCO3Cu(OH)2 3.7 to 3.8 3.5 to 4 Azure-blue
Sulphide Chalcopyrite CuFeS2 4.1 to 4.3 3.5 to 4 Brass-yellow, brittle
Sulphide Bornite Cu3FeS3 4.9 to 5.4 3.0 Copper-red to brown, brittle
Sulphide Covellite CuS 4.59 Indigo-massive
Chalcocite
Copper Sulphide Cu2S 5.5 to 5.8 2.5 to 3 Blackish lead-gray, often tarnished, sectile
(copper glance)
Tetrahedrite
Antimonide Cu8Sb2S7 4.4 to 5.1 3 to 4 Flint-gray to iron-black
(gray-copper)
Sulpharsenate Enargite Cu3AsS4 4.4 3.0 Grayish-black, brittle
Silicate Chrysocolla CuSiO3 + 2H2O 2 to 2.2 2 to 4 Turquoise-blue, translucent, vitreous, rather sectile
Oxychloride Atacamite Cu2ClH3O3 3.75 3 to 3.5 Bright bottle green, brittle
Gold Metalic Native Au 15.6 to 19.3 2.5 to 3 Gold-yellow, malleable and ductile
Telluride Sylvanite (Au.Ag)Te2[Au:Ag = 1:1] 7.9 to 8.3 1.5 to 2 Steel-gray to silver-white to yellow, brittle
Telluride Calaverite (Au.Ag)Te2[Au:Ag = 6:1] 9.0 2.5 Pale bronze-yellow, massive
Telluride Petzite (Au.Ag)2Te[Au:Ag = 1:3] 8.7 to 9.0 2.5 to 3 Steel-gray to iron-black, brittle
Graphite Carbon Plumbago C 2.0 to 2.23 1 to 2 Iron-black, dark-steel gray, greasy, flexible
Iron Metallic Native Fe 7.3 to 7.8 4 to 5 Steel-gray to iron black malleable

Oxide Hematite Fe2O3 4.9 to 5.3 5.5 to 6.5 Steel-gray, red
Oxide Limonite 2Fe2O3, 3H2O 3.6 to 4.0 5 to 5.5 Brown, opaque
Oxide Magnetite FeO, Fe2O3 5.16 5.5 to 6.5 Iron-black, magnetic
Carbonate Siderite FeCO3 3.8 3.5 to 4 Gray-brown, reddish, brittle
Sulphide Pyrite FeS2 4.9 to 5.1 6 to 6.5 Brass-yellow, brittle
Sulphide Marcasite FeS2 4.85 to 4.9 6 to 6.5 Pale bronze-yellow, brittle
Sulphide Pyrrhotite Fe5S6 to Fe16S17 4.58 to 4.64 3.5 to 4.5 Bronze-yellow to copper-red, easily tarnished
Sulpharsenide Mispickel FeAsS 5.9 to 6.2 5.5 to 6 Silver-white, brittle
Titanate Ilmenite FeTiO3 4.5 to 5 5 to 6 Iron-black, slightly magnetic
Tungstate Wolframite (Fe, Mn)WO4 7.2 to 7.5 5 to 5.5 Brownish-black, brittle
Chromate Chromite FeCr2O4 4.3 to 4.57 5.5 Iron and brown black, brittle
Manganate Franklinite (FeZnMn)O(FeMn)2O3 5.1 to 5.22 5.5 to 6.5 Iron-black
Gravity dressing is done by means of water and air. Gravity separation is one of the efficient methods of
separating minerals from barren rocks by using the difference in densities.
The most notable advantages of the gravitational methods are their cost effectiveness and in some cases
excellent reduction. Gravity separation is an attractive unit operation as it generally has low capital and
operating costs, uses few if any chemicals that might cause environmental concerns and the recent
development of new equipment enhances the range of separations possible. Types of gravity separators
are:
 Conventional jigs  Centrifugal jigs
 Pinched sluices  Shaking tables
 Reichert cones  Dense medium separation
 Spirals  Chute separation

Gravity separation is the most well-proven and accepted technique of concentrating minerals and has been
used as a primary form of mineral concentration for centuries. Due to its high efficiency and low cost, gravity
separation is always the first consideration in any mineral processing program. In the case of gold and PGE,
gravity separation can quickly generate a precious metal concentrate that can be sold direct to refineries,
which gives immediate payment for the work carried out.
Possible reasons to consider using Gravity Separation:
 To reject barren waste as an initial pre-concentration step
 To recover malleable and/or friable coarse heavy minerals from grinding circuit circulating loads. Such minerals are
otherwise hard to recover after regrinding
 To pre-concentrate heavy minerals to minimize downstream processing costs
 To concentrate heavy minerals
 To clean low weight yield bulk concentrates
 To scavenging plant tailings
 To generate a precious metal concentrate that can go direct to a refinery rather than a smelter
4.10.1 Strong Gravity Separation

To characterize the ore to strong gravity separation response, centrifugal gravity concentrators used for gold
such as the Knelson separator can be considered (IM 2014). The purpose of this is to examine the possibility
of the extraction of precious metals (gold or PGE platinum group elements). The Knelson Concentrator relies
on an enhanced gravitational force together with fluidization process to recover even very micron sized
particles (Wills & Finch 2015). First water is injected into the rotating concentrating cone through series of
fluidization holes. Then feed slurry is introduced through a stationary feed tube. Once the slurry is filled in
each cone, create a concentrating bed and high specific gravity particles are retained in the cone and then
flushed into the concentrate launders. This procedure can be completed in less than one minute.
Figure 106 below shows a possible experimental program for strong gravity separation using a Knelson unit.
The heavy fraction of a Knelson separator would be quite small in volume compared to the lighter fraction
(2-3% of the feed mass). This needs to be considered when planning possible further testwork of the heavy
fraction. The sample products are then to be characterized appropriately.
Final
Concentrate
Could be
cleaned by a
shaking table
Low mass
Tails
pull
Gravity Heavy
Characterized
Separation
Sample
(Knelson) Light
Flotation
and/or
Leaching
Figure 106. Strong gravity separation experimental design (Simon Michuax)

The light fraction could then be sent to flotation and/or leaching recovery. Sample mass requirements for
Knelson gravity testing can be as little as 200g but should be at least 2kg. The particle size can vary but
should be a sample feed F80 of approximately of 100m.
4.10.2 Weak Gravity Separation

To characterize the ore to weak gravity separation response, a gravity spiral concentrator separator and/or
shaking tables can be used for fine grained minerals. A gravity jig can be used for recovery of coarse grained
minerals. The purpose of this is to examine the possibility of the extraction of gangue minerals that could
negatively impact recovery.
Spiral concentration separators of the wet type, also called spiral concentrators, are devices to separate solid
components in a slurry, based upon a combination of the solid particle density as well as the particle's
hydrodynamic properties (e.g. drag). The device consists of a tower, around which is wound a sluice, from
which slots or channels are placed in the base of the sluice to extract solid particles that have come out of
suspension.
As larger and heavier particles sink to the bottom of the sluice faster and experience more drag from the
bottom, they travel slower, and so move towards the center of the spiral. Conversely, light particles stay
towards the outside of the spiral, with the water, and quickly reach the bottom. At the bottom, a "cut" is
made with a set of adjustable bars, channels, or slots, separating the low and high density parts. Many things
can be done to improve the separation efficiency, including:
 changing the rate of material feed

 changing the grain size of the material
 changing the slurry mass percentage
 adjusting the cutter bar positions
 running the output of one spiral separator (often, a third, intermediate, cut) through a second.
 adding washwater inlets along the length of the spiral, to aid in separating light minerals
 adding multiple outlets along the length, to improve the ability of the spiral to remove heavy contaminants
 adding ridges on the sluice at an angle to the direction of flow.
Typical spiral concentrators will use a slurry from about 20% - 40% solids by weight, with a particle size
somewhere between 1.5-0.075 mm (17-340 mesh), though somewhat larger particle sizes are sometimes
used. For good separation, the density difference between the heavy minerals and the light minerals in the
feedstock should be at least 1 g/cm3; and because the separation is dependent upon size and density, spiral
separators are most effective at purifying ore if its particles are of uniform size and shape. A spiral separator
may process a couple tons per hour of ore, per flight, and multiple flights may be stacked in the same space
as one, to improve capacity.
A shaking gravity table is a mechanized gold pan, that operates with a high degree of efficiency and
continuously. The table is comprised of a deck, in somewhat of a rectangular shape, covered with riffles
(raised bars running perpendicular to the feed side of the table), mounted in a near flat position, on a
supporting frame that allows the table to slide along the long axis of the table.

Figure 107 shows a laboratory (or bench) scale gravity shaking table. Figure 108 shows a pilot scale (5 tonne
per hour) gravity shaking table.
Figure 107. Bench scale shaking table gravity separation

(Image: David Bastin & Walter Rivera, ULiege University of Liege)
The mechanism is attached to the table, and it moves the table along the long axis a distance adjustable
between 10mm and 30mm and then back to the starting position between 200 and 300 times per minute.
This reciprocal movement is faster on the reverse stroke than it is on the forward stroke. This shaking
movement helps transport the concentrates or heavy material to the concentrate end of the table. A very
important operating variable of a shaking table is the tilt adjustment. Normally, the feed side is lower, and
the concentrate end is higher on the table, which creates an upward slope where the heavy material will
ascend, while the light density material will not, and consequently, flow over the riffles. The tailing (low
density) side is usually near level to lower than the feed side.
The gravity concentrating tables remove the high density material from the low density material, since the
high density material will reside behind the riffles and allow the low density material to flow over the top of
the riffles with the wash water, to the tailings discharge. As a general rule, the frequency and stroke
relationship are similar to screens, short stroke, high frequency is better for fines (-180µm), while a shorter
stroke and lower frequency is better for coarse material (-3+0.180mm).

Figure 108. The gravity circuit section of the GTK Mintec process pilot scale plant, gravity spirals and shaking table
(Image: Arno-Matti Kirpala)

Mineral jigs are a type of mining equipment, also referred to as gravity concentrators or jig concentrators
that are used in operations to separate different ore materials based on their densities. Usually they will
process material that is a similar size after the ore has passed through a crusher or over a screening plant.
The particles are introduced to the jig bed (usually a screen) where they are thrust upward by a pulsing water
column or body, resulting in the particles being suspended within the water. As the pulse dissipates, the
water level returns to its lower starting position and the particles once again settle on the jig bed. As the
particles are exposed to gravitational energy whilst in suspension within the water, those with a higher
specific gravity (density) settle faster than those with a lower count, resulting in a concentration of material
with higher density at the bottom, on the jig bed. The particles are now concentrated according to density
and can be extracted from the jig bed separately.
Figure 109 below shows a possible experimental program for weak gravity separation using a gravity spiral
unit. The lighter fraction could potentially contain minerals like quartz, clays or talc, all of which negatively
impact flotation.
High mass
pull Flotation
and/or
Weak Gravity Heavy Leaching
Characterized
Separation
Sample
(Spiral) Light
Removes
light weight
gangue
Figure 109. Weak gravity separation experimental design
The heavy fraction could be sent flotation and/or leaching recovery. In doing so, the recovery could be
greatly improved once the ore had been cleaned of minerals that could reduce efficiency.
Sample mass requirements for gravity spiral testing should be at least 20 to 30kg (needed to fill the spiral).
It is possible to use a micro spiral which can take a much smaller feed sample. The particle size can vary but
should be a sample feed F80 of approximately of -1mm + 75m.
Figure 110 below is an idealized experimental design, combining Figure 106 and 109. The strong gravity
separation could remove a portion of the precious metals (Au & PGE) and/or any elements that are too heavy
to separate with flotation. Then, weak gravity separation could remove the lighter gangue minerals which
could reduce recovery.

Final gravity
concentrate
(Au, PGE, Ag, REE)
Characterized
Sample Could be cleaned
Low mass by a shaking table
pull
Tails
Stong Gravity Heavy
Separation
(Knelson) Light
High mass
pull
Weak Gravity Heavy Flotation

and/or
Separation Leaching
(Spiral) Light
Remove
light
minerals
gangue
Figure 110. Ideal gravity separation experimental design

One issue to consider with this design is the difference in sample feed between the two units, the Knelson
separator and the gravity spiral. They may require different feed sizes and different sample masses to be
efficient.
4.11 Magnetic Separation

Magnetic separators exploit the difference in magnetic properties between the minerals in a deposit and are
used to concentrate a valuable mineral that is magnetic (e.g., magnetite from quartz), to remove magnetic
contaminants, or to separate mixtures of magnetic and nonmagnetic valuable minerals (Wills 2015).
It is not practical to apply electromagnetic separation to an ore stream before flotation, due to the volumes
required. What usually is feasible is to use electromagnetic separation to clean impurities from the flotation
concentrate.

All materials are affected in some way when placed in a magnetic field, although with many substances the
effect is too slight to be easily detected. For the purposes of mineral processing, materials may be classified
into two broad groups, according to whether they are attracted or repelled by a magnet:
1. Diamagnetic materials are repelled along the lines of magnetic force to a point where the field
intensity is smaller. The forces involved here are very small and diamagnetic substances are often
referred to as “nonmagnetic”, although this is not strictly correct. Diamagnetic minerals will report
to the nonmagnetic product (“non-mags”) of a magnetic separator as they do not experience a
magnetic attractive force.
2. Paramagnetic materials are attracted along the lines of magnetic force to points of greater field
intensity. Paramagnetic materials report to the “magnetic” product (“mags”) of a magnetic separator
due to attractive magnetic forces. Examples of paramagnetic minerals which are separated in
commercial magnetic separators are ilmenite (FeTiO3), rutile (TiO2), wolframite ((Fe, Mn)WO4),
monazite ((Ce, La, Nd, Th)PO4), xenotime (YPO4), siderite (FeCO3), chromite (FeCr2O4), and
manganese minerals.
Figure 111 shows a laboratory scale (also called bench scale) magnetic separation unit in operation.
Figure 111. Bench scale magnetic separation process testing

(Image: David Bastin & Walter Rivera, ULiege University of Liege)

Table 4. Magnetic separation characteristics of minerals (Source: reproduced from 911Metallurgist.com)
Non-Conductors (High Tension Pinned) Conductors (High Tension Thrown)

Specific Weakly Highly Weakly Non
Magnetic Non Magnetic Magnetic
Gravity Magnetic Magnetic Magnetic Magnetic
Over 8.0 Gold
Copper
8.0
Galena
Ferberite Wolframite
7.5 Cassiterite
7.0
6.5
6.0 Scheelite Columbite

Tantalite
5.5
5.0 Monazite Magnetite Samarskite Pyrite

Bastnasite Ilmenite Ilmenite Euxenite
Zircon Hematite
(high iron) Davidite Molybenite
Barite Chromite
4.5
Xenotime
Rutile
Chalcopyrite
4.0 Corundum Brookite
Garnet Celestite Limonite
Siderite Perovskite
3.5 Staurolite Kyanite
Diamond
Epidote
Olivine Topaz
Sphene
Apatite Sillimanite
3.0 Flourite
Hornblende Anhydrite
Tourmaline Mica (Muscovite)
Mica (Biotite) Beryl
2.5 Feldspars
Calcite Graphite
Quartz
Gypsum
2.0 Chrysotile
Sulphur
Under 2.0

4.11.1 Low-intensity magnetic separation - ferrite magnet LIMS

Low-intensity separators are used to treat ferromagnetic materials and some highly paramagnetic minerals.
Minerals with ferromagnetic properties have high susceptibility at low applied field strengths and can
therefore be concentrated in low intensity (<~0.3T) magnetic separators. This is usually a dry process but
also can be a wet process. This often depends on the particle size of the feed.
4.11.2 Medium intensity magnetic separation - permanent magnet MIMS

Permanent magnetic drum separators combine the attributes of a high-strength permanent magnetic field
and a self-cleaning feature. These separators are effective in treating process streams containing a high
percentage of magnetics and can produce a “clean” magnetic or non-magnetic product. The magnetic drum
separator consists of a stationary, shaft-mounted magnetic circuit completely enclosed by a rotating drum.
The magnetic circuit is typically comprised of several magnetic poles that span an arc of 120 degrees. When
material is introduced to the revolving drum shell (concurrent at the 12 o’clock position), the non-magnetic
material discharges in a natural trajectory. The magnetic material is attracted to the drum shell by the
magnetic circuit and is rotated out of the non-magnetic particle stream. The magnetic material discharges
from the drum shell when it is rotated out of the magnetic field.
4.11.3 High intensity magnetic separation – electromagnetic HIMS

High-intensity magnetic separators are effective in collecting paramagnetic particles. These focus on the
separation of very fine particles that are paramagnetic. This is a wet process.
The current is passed through the coil, which creates a magnetic field, which magnetizes the expanded steel
matrix ring. The paramagnetic matrix material behaves like a magnet in the magnetic field and thereby
attracts the fines. The ring is rinsed when it is in the magnetic field and all the non-magnetic particles are
carried with the rinse water. Next as the ring leaves the magnetic zone the ring is flushed and a vacuum of
about – 0.3 bars is applied to remove the magnetic particles attached to the matrix ring.
In a geometallurgical experimental campaign, there would most probably not be enough mass in any given
flotation concentrate to test directly with magnetic separation. What can be done though is the magnetic
separation of a given ore type in a series of increasing magnetic intensities. After each separation step, the
product concentrate could be characterized appropriately. The comparatively non-magnetic tails of each
separation could then be subject to a stronger intensity magnetic separation. Once the last separation step
had been completed, the tails could be characterized appropriately, and then a mineral reconciliation could
be done across the whole electromagnetic process.
In doing so, each ore type could be characterized in context of what minerals are present that would be
susceptible to magnetic separation. This has several possible useful outcomes. Some minerals that are
magnetic (see Table 4) are also antagonistic to pyrometallurgical refinement. Thus removing them would
increase the value of ore concentrate. Also, some magnetic minerals can also be closely associated with
valuable metals (e.g. platinum embedded in magnetic pyrrhotite).

Figure 112 below shows an experimental design to determine the ore response to magnetic separation, using
a laboratory scale electromagnetic unit. Each separation step was to be done at a different applied current.
The ferrite LIMS could be estimated at a 1 Ampere current. A permanent magnet MIMS estimated with a 2
Ampere current. An electromagnetic HIMS could be estimated with a 3 Ampere current.
Characterized
Sample
Electromagnetic
Separation – Concentrate 1
1 Ampere
(LIMS)
Electromagnetic
Ferrite Separation – Concentrate 2
equivalent 2 Ampere
(MIMS)
NdFeB permanent Electromagnetic Concentrate 3
magnet equivalent Separation –
3 Ampere
(HIMS)
Electro-magnet
equivalent Final
Tails
Figure 112. Electromagnetic separation experimental design

Sample mass requirements for magnetic testing should be at least 2 to 5kg. The particle size can vary but
should be a sample feed F80 similar (the same) as what is feed to flotation and leaching. This would be
defined by the selected grind size (see Section 4.7).
4.12 Ore Sorting

Modern, automated sorting applies optical sensors (visible spectrum, near infrared, X-ray, ultraviolet), that
can be coupled with electrical conductivity and magnetic susceptibility sensors, to control the mechanical
separation of ore into two or more categories on an individual rock by rock basis. Also new sensors have
been developed which exploit material properties such as electrical conductivity, magnetization, molecular
structure and thermal conductivity. Sensor based sorting has found application in the processing of nickel,
gold, copper, coal and diamonds.
There are many ways of improving recovery processes for a wide variety of ores using sorting equipment.
Sensor-based ore particle sorting has the potential for a powerful method for pre-concentration or waste
rejection early in the comminution process.
This process has the potential to upgrade the value of the ore by increasing the grade, and/or remove a
penalty mineral that would negatively impact the recovery of the valuable target metal (Figure 113). This is
usually done dry, with crushed ore feed of size -100mm. The viability of sorting is ore specific and is not
traditionally used. That being stated, technology is evolving quickly and at some point sorting will become
viable. When this happens sorting technology has the potential to make whole mining operations viable
where before they were not.

Figure 113. Sorting of ore using an XRF characterization technique (Source and Images: IMA Engineering Finland)
There are several ways for ore to be sorted. One method uses air jets to separate out the target sub-fraction.
The ore characterization technology (based around the principles of XRF) analyzes each and every rock
particle on-line for unique physical and chemical properties and separates the particles by high pressure air
jets. This maximizes recovery or upgrade and is suitable for various minerals and applications. A shortcoming
of this method is the use of air jets consumes a considerable amount of power (therefore is expensive). This
often is enough to make this method uneconomical.
To determine whether sorting would be effective, an assessment study on the range of end member rock
textures within the ore being extracted needs to be done. As each rock texture would respond differently

and to an efficiency curve, this fits the geometallurgy paradigm well. If sorting is viable, it should be
considered as a separation process in the same context as flotation or hydrometallurgy leaching. As such
ore should be characterized so all effective processes can be optimized together.
4.13 Pyrometallurgy
Following separation and concentration by mineral processing, metallic minerals are subjected to extractive
metallurgy, in which their metallic elements are extracted from chemical compound form and refined of
impurities. This is done with the products from flotation and leaching hydrometallurgy.
Metallic compounds are frequently rather complex mixtures (those treated commercially are for the most
part sulfides, oxides, carbonates, arsenides, or silicates), and they are not often types that permit extraction
of the metal by simple, economical processes. Consequently, before extractive metallurgy can affect the
separation of metallic elements from the other constituents of a compound, it must often convert the
compound into a type that can be more readily treated.
Pyrometallurgy is a branch of extractive metallurgy. It consists of the thermal treatment of minerals and
metallurgical ores and concentrates to bring about physical and chemical transformations in the materials
to enable recovery of valuable metals. Pyrometallurgical treatment may produce products able to be sold
such as pure metals, or intermediate compounds or alloys, suitable as feed for further processing. Examples
of elements extracted by pyrometallurgical processes include the oxides of less reactive elements like iron,
copper, zinc, chromium, tin, and manganese.
Pyrometallurgical processes are generally grouped into one or more of the following categories:
 calcining,
 roasting,
 smelting,
 refining.
Most pyrometallurgical processes require energy input to sustain the temperature at which the process
takes place. The energy is usually provided in the form of combustion or from electrical heat. When sufficient
material is present in the feed to sustain the process temperature solely by exothermic reaction (i.e. without
the addition of fuel or electrical heat), the process is said to be "autogenous". Processing of some sulfide
ores exploit the exothermicity of their combustion.
The efficiency and effectiveness of any given pyrometallurgical process is often related to the character of
the feed material. Impurities and toxins in the feed material are subject to penalty rates and directly affect
the value of an ore concentrate. To characterize the feed, the following methods are useful:
 Chemical Assay
 Bulk QXRD
 XRF pellet

5 SMALL SCALE PROXY TESTS

Also, a geometallurgy campaign will require multiple measurements of different kinds of the same small
volume. This means that small scale tests are required. The AMIRA P843 and P843A projects were successful
in developing many such tests (AMIRA P843 2008a & 2008b, AMIRA P843 2009, AMIRA P843A 2010a, 2010b
& 2010c, AMIRA P843A 2011a & 2011b, AMIRA P843A 2012a & 2012b and AMIRA P843A 2013). To reach
the geometallurgy campaign objectives, a large number of successful tests are required (where the larger
scale bankable tests usually are done in much smaller numbers for each campaign). A series of tests that
require only very small sample masses, are inexpensive and have a fast turnaround times. Table 5 below
shows a list of some of these comparative tests.
Table 5. Small scale comparative proxy tests for process engineering ore characterization
Process Engineering
Test Reference
Field
Geotechnical Point Load Index (PLT) Brady & Brown 2006 [33]
Comminution RBT Wills & Napier-Munn 2005 [39]
Wills & Napier-Munn 2005 [39]

SMC
JKTech [42]

JKCi
Kojovic et al 2011 [48]

SPi
SGS 2018 [44]
Vos & Bradshaw 2014 [41]
Flotation Separability Indicator (JKMSI) Chauhan 2013 [20]
Morgan et al 2012 [46]
Hydrometallurgy Kuhar et al 2011 [19]

CSIRO Diagnostic Leach Test
Leaching Li et al 2016
Acid Mine Drainage Geo -environmental modelling Parbhakar-Fox 2015 [36]
Rinse Rinse /paste pH Parbhakar-Fox 2016 [34]
Carbonate staining Parbhakar-Fox 2013 [35]
Meso scale mineralogy Hyperspectral image analysis Schodlok et al 2016
Rock Hardness EQUOtip hardness Keeney & Nguyen 2014

It is to be remembered that a proxy test is an estimate of a correlation between a

physical/chemical/mineralogical phenomenon and a test response. This correlation can be related to the
larger scale tests as shown in Section 4. Each ore type family will respond differently to this relationship.
Also, there is the problem of what the test results actually mean and the same phenomena behavior at a
production scale. There are all sorts of difficulties in scaling up tests between laboratory scale and
production scale. It is for this reason to be mindful of what tests are being used where and for what purpose.
JKMSI
Batch
Pilot plant
Full Scale
Drill
Core
Figure 114. Proxy tests and scale of precision

(Image Copyright: Dee Bradshaw)
One of the purposes of the Orientation Study is to examine and establish the best combination of small-scale
tests that describe the target process behavior and develop a working relationship to the relevant bankable
tests. The approach should be that every deposit is different and unique. Some tests will work well, whereas
others will not, in a deposit specific fashion.
5.1 EQUOtip test

This is a hardness tester developed for steel liners, which is based on the concept of the Schmitt hammer.
Now used on drill core as a non-destructive test (Keeney 2014 and Verwaal 1993) that is to be done on intact
drill core while it still sits in the core tray. Can be done on lots of drill core easily and can be used as a useful
domaining tool. The EQUOtip output is also useful as a Geometallurgy model input.
While the core was still in the tray, the EQUOtip test was done across all core at 2cm intervals (Keeney 2014
and Verwaal 1993). The EQUOtip is a non-destructive test that was originally designed to measure the
hardness of steel (Figure 115). It operates very similar to a Schmitt Hammer (Figure 116). While it does not
directly correlate with rock hardness as it is empirically measured, it is an excellent model input for
comminution.

Figure 115. EQUOtip Impact Device

(Source: AMIRA P843, AMIRA 2008b, Copyright AMIRA)
Figure 116. (a) Cross section of EQUOtip impact device and (b) voltage signal generated by movement of the impact
body (Source: AMIRA P843, AMIRA 2008b, Copyright AMIRA)

EQUOtip is a measurement that is different to chemical assay analysis, or visual measurement like
hyperspectral imaging. As such it often provides a useful contrast in domaining drill core. Figure 117 shows
an example of continuous EQUOtip measurements on core. The raw data is variable, and it is required to
smooth the data to the point where it is possible to discern domains.
850 850
800 800
750 750
700 700
650 650
600 600
550 Direct proxies? 550
500 500
150 200 250 300 350 400 450
850 Depth (m)
800
750
700
650
600
550
500
1100 1150 1200 1250 1300 1350 1400 1450 1500
Figure 117. EQUOtip hardness tester used as a domaining tool

(Source: Amira 2008b, Image: Steve Walters, Copyright: AMIRA)
Figure 118 shows a probability plot of all EQUOtip data for 7 different deposits, each with very different
geological and lithological characteristics. This is shows an interesting opportunity. What can be seen is the
deposits all have a very different deposit scale signatures. If EQUOtip was done as standard practice as a
test done in exploration campaigns, it could be a very fast assessment of what signature each new deposit
could be against mature deposits that have been measured in the same fashion. This could be a very early
indicator to some of the challenges a deposit might present if it was mined and then processed.

Sediment hosted Pb-Zn

Porphyry/Skarn Cu-Au
Archean Bedrock Au-
Cu
Porphyry Cu-Au
Iron Oxide Cu-Au
Ni-Cu-PGE
Figure 118. Deposit scale EQUOtip comparison

(Source: AMIRA 2008b, Image: Steve Walters, Copyright: AMIRA)
5.2 Point Load Index Measurement

The Point Load Index is a geomechanical test used to measure rock fragment strength (Broch and Franklin
1972) and has been correlated with rock mass parameters like UCS compressive strength and tensile strength
(ISRM 1985 and Butenuth 1997). It was originally designed to test 50mm diameter cores, but there is an
empirically derived correction model that allows the testing of irregular shaped fragments and ½ drill core
(Brook 1985 and Brook 1980).
The device comprises of two point-shaped "platens" between which the specimen rock is placed. One of the
platens is effectively stationary (though its initial starting position can be adjusted) whilst the other is free
to move through the application of pressure, delivered via a hand pump and piston arrangement. As the
hand pump is activated the pressure and hence force applied to the rock is increased and eventually causes
the rock to fail. The peak pressure applied is indicated on a pressure gauge.

A successful/appropriate measurement for the Point Load Index is based on a criterion for fragment failure.
Did the fragment fail completely between the two platens? If the answer to this is no, then the result must
be discarded.
The standard formula for Point Load Index calculation is as follows:
Is50 = F P/(De)2 Equation 6
Where:
F = size correction factor (used for irregular particles)
= (De/50)0.45 Equation 7
P = force at failure and is calculated from the pressure and geometry of the hydraulic system.
De = (4A/)0.5 Equation 8
A = minimum cross-sectional area of specimen
There are several methods of sample selection and manner of conducting the Point Load Index (Figure
119).
L > 0.5D 0.3W<D<W
Diametrical Test Axial Test
0.3W<D<W 0.3W<D<W
Block Test Irregular Lump Test
Figure 119: Sample shape requirements for the Point Load Index (Source: ISRM, 1985)

5.2.1 Diametral test

The Point Load Index was originally developed to test core rock samples with a length (L)/diameter (D)
ratio greater than 1 (Figure 119).
 The core is placed into the tester so that the platens catch the sample at a point where the full diameter of the core is
being loaded between them.
 The distance between the platen contact points (D) is to be recorded along with the specimen width (W) at an accuracy
of +/- 2%.
 The sample is loaded over a time period between 10 and 60 seconds.
 The test is rejected if the sample did not fail through a plane passing through the two platens where the rock is held.
The full procedure for the diametrical point load test is shown in ISRM 1985.
5.2.2 Axial test

Core specimens have a ratio of length (L)/diameter (D) between 0.3 and 1.0 for axial testing.
 The sample is inserted into the tester as per shown in Figure 119 until the platens have contacted the sample.
 The distance between the platens (D) is measured along with the specimen width (W) to an accuracy of +/-2 to 5%.
5.2.3 Block and irregular lump tests

Irregular lumps of rock (for example from the muckpile) can be also tested with a core correction factor
(Equation 7). This is very useful, as the samples no longer have to be from drilled core. The rock particles
should come from the size fraction of –75+35mm. Shape is important, as the rock has to be held in the tester
where the smallest dimension is between the platens. The Diameter (D)/width (W) ratio should be between
0.3 and 1.0, but as close to 1.0 as practical.
 The sample is inserted into the tester in a manner where the rock has to be held in the tester where the smallest
dimension is between the platens (Figure 119).
 The distance between the platens (D) is measured along with the specimen width (W) to an accuracy of +/-2 to 5%.
Tests with irregular particle samples produce much more variation in the results than core samples.
5.2.4 Anisotropic rock samples

Rock types like shale or schist have structures at the meso-scale, which can influence th3e measurement of
strength with the point load tester. Case studies have shown that strength measurements parallel to the
foliation’s are mush different in magnitude to strength measurements perpendicular to the foliation’s in the
same rock type. Table 6 shows a summary of the Point Load Index data taken from Mt Coot-tha quarry
(Australia). The rock tested was a highly foliated phyllite.

Table 6: Summary of Mt. Coot-tha Point Load Index data (Source: Michaux 2006)
Test Orientation to foliations Mean Is50 Standard Deviation No of successful measurements

75*63mm perpendicular 5.37 1.49 32
75*63mm parallel 2.94 1.06 25

63*53mm parallel 3.61 1.44 20
53*45mm parallel 3.33 1.42 19
For cored tests, a number of diametrical and axial tests should be done as two sub sets to quantify the
influence of the meso structure of the rock (to ensure that both orientations are measured). For irregular
particle samples an attempt to orient the samples parallel and perpendicular to the bedding structures.
Depending on how the rock is oriented in the bench will decide how to use this data. An average of the two
sub sets is a practical solution.
5.2.5 Collecting samples

The Point Load Index is an approximation of strength as opposed to a direct measurement. As a consequence
of this, there is a significant amount of variation in the measurements. This is not such a problem as the rock
strength characterisation is also highly variable (even over a few metres). To deal with this problem, a large
number of tests are required. How many depends on the variability of the rock type. After each test, some
statistics analysis should carried out to determine the variability of the rock type. The t-test could be used
to compare point load tests done on the same rock type. Rock density and particle mass before testing could
be used to normalise the data. As a practical start, 30 successful measurements (where the force has been
measured after a sample failure through a plane passing through the two platens where the rock is held)
would provide enough data to approximate strength. It should also be remembered where the samples
were collected. Correct sampling procedure is required, especially when spatially defining rock strength
across a bench to be blasted.
In terms of sampling points on the bench or muckpile, a spatially defined grid (resolution of 1 to 2m) could
be used as a sampling arrangement to collect particles (remembering shape is important and not all
fragments collected will produce a successful measurement).
Sample mass requirements for a point load index is 30 to 40 drill core (NQ or HQ half core) pieces that have
a length (X dimension) 2.5 times the width (Y dimension). This can be taken from 4m of drill core depending
on the quality of the core (how broken up it is). The Point Load Index can also be done on fragments of rock
-90+50mm fraction. The test needs 30 successful measurements that pass QA/QC requirements. This can
be taken from 20kg of -90+50mm fraction rocks.
5.2.6 Use of the Point Load Index

The unit of the point load index is Mpa and is considered to be a measure of tensile failure and fracture
toughness (Kic). There is a crude relationship between the Point Load Index and unconfined compressive
strength (UCS), which has been used in mining research in the past. This is simply multiplying the Point Load
Index by 24 to obtain the UCS. This however is not appropriate as this relationship change with rock type.
The Point Load multiplication factor to estimate UCS can range from 10 to 50, depending on rock type (Thuro
and Plinninger 2001). To estimate UCS from Point Load Index data, it is necessary to conduct a series of UCS
tests in the same rock type to calibrate the multiplication factor.

5.3 Impact Breakage SMC test

The SMC Test is has the same objectives at the Drop Test but requires much less sample. The SMC was
originally developed to measure the energy breakage characteristics of core samples and has been used as
a comparison test to the Drop Test. A comparatively small amount of core is cut to a very specific size and
shape (volume).
To achieve this shape and size, the mass of each fragment is measured and with the measured ore density
(wet & dry method), the volume is calculated. For each fragment to pass, it must be within a statistical
tolerance of an average volume. Five sets of twenty of these wedge shapes are taken for the SMC test.
Each set of twenty is broken at a specified energy in the Drop Test machine as per Section 4.2. These 5
energy breakage points are then used to estimate the energy breakage curve and calculate A and b
parameters.
Figure 120. Sampling strategy to ensure the Comminution Index and the SMC tests were as representative as possible
The SMC Test was developed as a cost-effective means of profiling an ore body to predict its comminution
circuit throughput as well as its rock mass characteristics and blasting properties. Energy requirements for
crushing and HPGR options can now be determined in addition to AG/SAG mills, and with Mic, ball mills
(Morrell 2004 & 2011, JKTech 2009). The results from conducting the SMC Test are used to determine the
so-called drop-weight index (DWi) which is a measure of the strength of the rock” (JKTech 2009). Figure 121
shows the results of SMC test with operation data.
Table 7. SMC comminution test length of core required for the two methods (Source: JKTech SMC Test)
For larger diameter core, greater lengths are required for the cut core method. Half core, quarter core and
crushed rock can also be used. Lengths in Table 7 must be multiplied by a factor of two for half core and four

for quarter core, respectively. Predicted Comminution Circuit Specific Energy Using the SMC Test Compared
to Values Measured from Over 120 Operating Plants from Around the World.
Figure 121 The SMC impact breakage test outcomes vs. operation data
(Source & Image copyright: Morrell 2011 and SMC)
The SMC Test requires significantly less sample than the JK Drop Weight Test or JK Rotary Breakage Test.
Two sample preparation methods are available- the cut core method where the ore sample is competent
and restricted in quantity, and the crush and select method where the ore is available in larger quantity or
is friable.

5.4 RBT Impact Breakage test

The industry accepted JK impact breakage parameters, A and b, can now reliably be determined using the
new generation high throughput JK Rotary Breakage Tester (JKRBT). It does this by projecting a particle (from
a rotating rotor arm inside the machine) against metal anvil around the edge of the interior of the machine.
The breakage products are treated the same way as in the DWT test.
Stator/Anvil
Rotor
Vacuum
Unit
Base Frame
Recovery
Assembly
Bin
Figure 122. The JKMRC RBT impact breakage characterization test unit
(Source and Copyright: JK Tech Pty. Ltd, JK Rotary Breakage Tester)
The A and b impact parameters can be determined using the high throughput JK Rotary Breakage Tester
(JKRBT). The machine is very different to the JK Drop weight tester, but outcomes are the same (See Section
4.2). Samples required are the following: 100 kg of crushed rock, -53+12.5 mm, or 100 kg of drill core”.
(JKTech Rotary Breakage Tester 2018). The JK Rotary Breakage Test provides these breakage parameters for
use in the JKSimMet software to analyze and predict AG/SAG mill performance.
The typical minimum quantity of sample required to provide sufficient particles for testing is 100 kg of
crushed rock in the -53+12.5 mm size range, or 100 kg of drill core. For diamond drill core samples, the core
must be un-split and have a diameter of at least 50 mm.
5.5 Piston and Die Compress Bed Breakage (CBB)

The Compressed Bed Breakage (CBB) Piston and Die test is a small scale breakage test that can relate ore
breakage to laboratory scale HPGR (Benzer et al 2017, Kumar 2014 and Daniel 2007). This test used a piston
and die assembling in a hydraulic press (Figure 123). Approximately 300g of crushed ore (known size
distribution) was put in the piston which was subject to a known pressure from the hydraulic press, thus
breaking the sample in compressed bed breakage. The piston was instrumented to measure displacement

over time in conjunction with applied pressure (Figures 124 and 125). Thus a known applied energy could
be applied. The sample was recovered and the product sized.
Avery Hydraulic Piston & Die Equipment

Press
Figure 123. Conventional piston and die unit in a hydraulic press

(Image: Steve Walters)
Figure 124. Instrumented piston and die test unit

(Image: Steve Walters LHS, and Rajiv Chandramohan RHS)
Piston and die HPGR
Figure 125. How the piston and die relates to the mechanism of breakage in a HPGR
(Image: Mike Daniels)

This approach is an interesting opportunity. The normalized feed size distribution of the laboratory scale
HPGR is very similar in shape and form as the size distribution of the -3.35mm crushed ore of the same ore
(Figure 126).
100
Ave
LabBodd
HPGRHPGR
FeedFeed
-3.35mm
Ave Bodd crushed
-3.35mmfeed
Feed
80
Cum % Passing
60
40
20
0
0.00 0.01 0.10 1.00
Normalized Size (X/Xmax)
Figure 126. Lab Scale HPGR feed is very similar in character to the -3.35mm crushed ore (Image: Simon Michaux)
The products of the piston and die assembly also have a very similar size distribution compared the
laboratory scale HPGR samples of the same ore (Figure 127).
Figure 127. Lab HPGR and CBB tests can produce similar products (Daniel 2007 – P9N samples)
(Image: Mike Daniels)

The CBB test can be used for the comparative ranking of different ore types possible in context of HPGR
response. The feed size distribution F50 is compared to the piston and die product P50 (where P80 not as
useful product size marker as P50) using an algorithm to output an index that ranks the process performance
of a HPGR crushing the ore (Figure 128).
CB50i = Ecs/[10* (1/√P50 – 1/√F50)] Equation 9
SampleCE002
C at 121-123
Ecs = 1.0 kWh/t Comparitative HPGR Index
100
-3.35mm Feed 16
1.0 kWh/t Product
Cumulative % Passing
80 14
12
CB50i Index (kWh/t)

60
10
40 8
6
20
4
0 P50 = 0.55mm F50 = 1.22mm 2
0.01 0.1 1 10 0
Size Fraction (mm) Sample A
CE002 108-111 Sample
CE002 B
121-123 Sample
CE002 C
126-128 Sample
CE002 D
130-132
Figure 128. CBB small scale test for bed breakage (AMIRA 2009)
Figure 128 LHS shows an example of a CBB test applied to Sample C, which produces a CB50i of 7.2 kWh/t.
Another interesting opportunity involves transforming the CBB product mass size distributions into surface
area distributions using the method developed in Michaux 2010. There is a fundamental relationship
between surface area generated and energy applied in the breakage event (Michaux 2006). So several CBB
tests could be done on the same ore, but at different applied Ecs energies (Figure 129).
New Surface Area Generated vs Specific Energy (CBB)

10
9
New Surface Area Generated
8
7
6
(m /kg)
5
2
4
NSAG = 8.77(1-e(-Ecs*Sb))
3
AS = 8.77
2
bS = 4.12
1 AS*bS = 36.2
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Specific Energy (kWh/t)
Figure 129. A series of CBB tests done at a number of applied energies (Source: AMIRA 2009)
(Image and Chart: Simon Michaux)

An exponential model could be fitted for new surface area generated (accounting for surface are in the feed
product) as a function of energy is shown in Figure 117. This method has not been developed but shows
promise in that it is analogous to Drop Weight A and b parameters for t10 vs Ecs model. This may provide
more robust index relevant to HPGR, in the form:
CBBi=1000/AS*bS (kWh/t required per m2/t of new surface area generated) Equation 10
The shape of this surface area curve would be different for different rock textures.
Each CBB test run would require 200-300g of ore sample. The new surface are generated would require 4
CBB runs requiring 800-1200g of crushed -3.35mm ore sample.
5.6 Grinding - Batch Operating Work Index

A comparative grinding test was developed to rank the grinding behaviour of the small volume samples
(AMIRA 2009). This test is known as the Batch Operating Work Index. In contrast to a full Bond Ball test which
requires about 5kg of starting material the Batch Operating Work Index test requires only 700cc of material
(crushed to 99% passing -3.35mm) as a feed sample. The procedure for Batch Operating Work Index test
was as follows:
1. Select a representative sample 700cc in volume. Record its mass.
2. Size this feed sample on the √2 sieve series down to 38µm. The sample was then subject to a mass balance QA/QC
step.
3. Place the sample in a standard Bond Ball mill (which has been cleaned out with compressed air and has the
appropriate measured ball charge)
4. Run the mill with the sample secured in it for 358 revolutions or 5 minutes at 71.6 rpm.
5. Extract the sample from the mill and measure the mass.
6. Wet screen the sample on a 38µm sieve (with a scalping screen to protect the 38 µm screen)
7. Dry the +38 µm sample in an oven over night at 1050C.
8. Size the dried sample on the √2 sieve series down to 38 µm. The sample was then subject to a mass balance QA/QC
step.

9. Calculate the F80 from the feed size distribution using a spline function.
o Each feed sample had its size distribution measured on the √2 sieve series using a mechanical shaker (Ro-tap).
Each distribution was measured down to 38 µm. Each sample was subject to a mass balance QA/QC step.
o The measured size distribution of the feed was used to calculate the F80 for each sample. The F80 represents the
volume of sample that passes 80% of the top particle size, and was calculated from the cumulative percent
passing plot of the size distribution.
10. Calculate the P80 from the product size distribution using a spline function. The P80 represents the volume of sample
that passes 80% of the top particle size, and was calculated from the cumulative percent passing plot of the size
distribution.
Figure 130. Feed and product samples for the Batch Operating Work Index
11. Using the sample mass, revolutions of the Bond Ball mill, F 80 and P80, calculate the Batch Operating Work Index,
using the equation below. This equation accounts for the mass of the sample in the test and the energy applied to
that sample by the Bond Mill (number of revolutions).
Equation 11
Where:
W = 0.02*(Number of revolutions of the Bond mill) Equation 12
Mass of sample (kg)
P = The P80 of the product (mm)
F = The F80 of the feed (mm)
The sample mass and energy applied are normalized for each test. Thus the Batch Operating Work Index
examines the grind behavior of the ore and different samples can be compared in a ranking fashion.

5.7 SGS SPi test

The SGS SPI test measures the energy usage for a standard size reduction. It is interpreted into an index. It
measures the hardness of the ore from a SAG or AG milling perspective. It is a batch test and requires 2kg
of ore material and runs in a 10cm x 30 cm mill. Time is measured (minutes) for the sample to grind 80%
passing 12.5mm to 8-% passing1.70mm. (SGS 2018)
5.8 Mergan Work Index

The Mergan work index is a small scale wet grinding test developed by Niitti (1970), where the net energy
consumption of the mill is continuously determined by a simple pendulum torque measurement from the
mill axis. Nitty concluded that to accurately determine the grinding signature of an ore, a wet grinding test
was required where a dry grinding test would be inaccurate (Heiskari et al 2019). Also, for a test to be
reliable, it must be carried out to the same fineness as the actual grinding is done (Niitti, 1970).
The Mergan index gives good correlation to the Bond work index (the industry accepted standard for the
measurement of energy required to grind to a given closing size) (Figure 131). The Bond Ball work index is a
dry grinding test, but grinding done at an industrial scale is a wet process, which could make the Mergan
Index more appropriate for mill design.
Figure 131. Comparison of the normalized Mergan work index values and the Bond work index values from the test
work. (Source and Image: Heiskari et al 2019)
This test is small enough to enable taking mineralogical variability into account, and the ground product from
the Mergan grindability test can be used for mineralogical characterization studies and beneficiation tests.

The selected primary sample size and the size of sub-samples are located left of Gy’s Safety line and their
integrity will not be compromised (Lotter et al 2014).
The required amount of ore, about 5 kg or 1500 cm3 of packed material. This material is crushed ore (that
has been dried) where 99% of the sample volume passes a 3.35mm sieve. The sample is then homogenized.
The dimensions of the Mergan mill are 268 × 268 mm, with a smooth inner lining (no lifters). The speed of
the mill is kept at a constant 60 rpm throughout the test. The mill was equipped with a torque transducer to
measure the torque of the mill. The ball charge used in the Mergan tests consisted of 318 steel balls. The
ball charge was an “expanded” Bond ball mill charge. This means that the total weight of each ball size class
was expanded until the total mass of the ball charge reached at least 22 kg, in this case of Heiskari (2019),
22.23 kg.
To conduct the Mergan test, the 1500cm3 sample of packed crushed -3.35mm ore is added to the mill with
enough water, so that the pulp density of the slurry was 55% solids by weight. The material is ground for an
arbitrary amount of time, while the torque transducer is used to record the mill torque at 1-second intervals.
After grinding, the mill is emptied, the sample is collected, the slurry was separated from the ball charge,
and particle size analysis is conducted for the sample (particle size analysis was done by using a wet sieving
method ultrasonic bath). After particle size analysis, the slurry was weighed and water was removed from
the slurry or added to the slurry to achieve the desired slurry density of 55% solids by weight. The slurry was
then returned to the mill, and grinding time was estimated based on the results of the particle size analysis.
This was continued until the desired P80 value was reached (Heiskari 2019). The grindability may be
expressed as kWh/t of the new 74-µm material produced, or as kWh/t per 1000 cm2/cm3 of the new specific
surface area produced (Niitti 1970).
The Mergan Index is calculated as follows.
Equation 13
The calculation of (E0), the grindability of the tested material can be calculated directly as energy
consumption in kWh/t when ground from a given initial fineness to a desired final fineness. This is based on
empirical data from the development of the test.
The mill power draw was obtained from the torque measurements, and the momentary power draw can be
calculated from:

Equation 14
Momentary power draw could be summed over the grinding time to obtain the energy consumption (E 0) of
the grinding according to Equation 15. Once P80 and E0 were calculated, the Mergan work index, M-Wi, could
be calculated according to Equation 8.
Equation 15
5.9 Comminution Index (JKCi)

The Comminution Index (Ci) is a small-scale crushing test that was developed as part of the AMIRA P843
GEMIII project (Kojovic et al 2010, AMIRA P843 2008a & 2008b and AMIRA P843 2009). It is based on the
concept of crushing drill core in very specific experimental conditions. This test can be done during sample
preparation for other tests (for example chemical assay, or batch flotation). The test has a quick turnaround
(~10min/test) and has proven to be repeatable and robust across a range of core sizes and geometries.
Natural rock fragments can also be used as feed. The Ci has undergone preliminary trials within a major
commercial laboratory involving 13,000 samples as part of routine assay sample preparation (Yildirim 2016).
The resulting size distribution from the Ci test is used to derive two parameters: Ci (CRU) which is related to
the JKMRC Drop Weight Test A*b impact breakage parameter - range 0.0-6.0; and the Ci (GRD) which has a
good relationship with the Bond Ball Mill Work Index (BMWi) – range 0.0 to 1.0. An orientation study is
recommended to determine the optimal application of the Ci testing protocol for individual sites and sample
mediums. The indices of Ci-CRU and Ci GRD are also excellent comminution based domaining tools.

Selected core intervals e.g. 2m

Domianing of AG/SAG mill behaviour 25
bond)
BMWI (kWh/t)
20
(Ci, modified
Crush core sample
@ 2.5:1 reduction ratio
15
Then measure Size Distribution
100.0
EH512_236 CI=1.61 Ci CRU Crushing behaviour
Percent Retained on Sieve (%)
EH512_290 CI=3.17
ProxyPredicted
10.0
CE107_364 CI=4.29
Comminution footprint 10
CE107_176 CI=5.88
domaining at similar scale

1.0
to assays
Grinding behaviour 5
0.1
0.01 0.1 1 10 100
Ci GRD
Size Fraction (mm)
Comparative comminution testing of individual 0

rock type textures. 0
Can provide estimates of key ore parameters Meas

A*b and BMWi used to forecast plant throughput.
Key inputs into defining comminution domains. Domaining of Ball mill behaviour
Figure. 132 The Comminution Index to domain comminution behavior

(Source: AMIRA 2008b, Copyright: AMIRA)
The circumstancesA $50thetest

where Ci testthat canhasbe
protocol done
proven on limitations include:
to show
 2m of half drill core by an assay lab in
Friable core (pieces not the same size in each sample)
 numbers of 1000’s
High clay content in rock (plasticity and elasticity considerations)
 Variable feeding (choke vs single piece at a time)
 Crusher gap CSS is left unchecked to drift
 Natural fragments that have large variation in shape
What is an interesting opportunity is that it is able to test poor quality and friable core. This is something
conventional impact breakage tests like the Drop Weight Test or the SMC test cannot do. Conventional
comminution usually is done to determine the hardness of the rock, so a mill can be built that is strong
enough to cope with the required throughput. This usually does not need to measure the softest parts of
the deposit (usually the friable and poor-quality core). This is useful the softer core will have a higher
throughput through the SAG mill but not necessarily through the ball mile or fine grinding mill. A full domain
can now be done that can related to impact breakage and grinding behavior.
Figure 133 shows an example of the Comminution Index done down hole then compared to site lithology
classification of the drill core section. The Comminution Index in conjunction with chemical assays and
lithology logs can be an effective way domaining ore in context of comminution response as a function of
rock type.

Drill Hole UE 022

(Cadia East Underground)
Brecciated Mixed
Chlorite/ Magnetite/
4.5 Zone Alteration
Sericite zone
highly
(Ci-GRD).
brecciated
Crush
4.0
Index
JKCi CRD
Cominution
3.5
Propylitic Magnetite
alteration zone
3.0
0 20 40 60 80 100 120 140 160 180 200
Down Hole
Down HoleDepth
Depth(m)
(m)
Figure 133. Comminution Index (Ci) as a domaining tool

The outcome of a domaining study is an understanding of the domaining structure and variability of the
deposit in context of the target process behavior. While this outcome is useful, further work is to be done
in a non-spatial fashion to really secure understanding of what mineralogy controls the target behavior.
Figure 134 shows a 360m section of core, tested for the Comminution Index in a continuous fashion (each
test is of a 2m length of half core, some HQ, some NQ in size).
5.0
4.5
Crushing indexIndex
4.0
3.5
Comminution
3.0
2.5
2.0
EH635
Drillhole ABC
1.5
Calibration
GEM Calibration
1.0
795 845 895 945 995 1045 1095 1145
Down Hole Depth (m)
Figure 134. Comminution Index in a continuous 360m downhole section of core

(Simon Michaux)

The indices that come from the Comminution Index are used to predict comminution process parameters.
The CiCRU index is used to empirically predict the Axb impact breakage parameter (Figure 135). The CiGRD
index is used to empirically predict the Bond Ball Work Index (BMWi), shown in Figure 136. These are a
deposit specific relationship that requires calibration (usually using the SMC tests and BMWi tests) and is not
meant to replace the full Drop Weight Test for ore characterization in context of mill design.
300
A*b prediction requires an ore-
specific calibration using an
250 orientation set of RBT/SMC
and JKCi data from at least 12
samples in a given ore system
200 Development
Resolution
Boddington
Pred A*b
Ernest Henry
150 Cadia East
` Aqqaluk
ideal
100
50 68 Samples
6 Ore Systems
9.2% relative error
0
0 50 100 150 200 250 300
Meas A*b
Figure 135 Comminution Index (Ci-CRU) prediction of impact breakage (Axb)

(Source: Kojovic et al 2010, Copyright: AusIMM)
The universal correlation of Ci-CRU with A*b shows lower level of correlation compared to BMWi estimates
(Figure 136). It is also apparent that for some rock types with low A*b (i.e. hard crush) the Ci shows a greater
variation than A*b. For example, in Figure 121 at a Ci value greater than 3.5, the A*b values have a limited
range between approximately 20 and 50, however the Ci shows a much greater range. The significance of
this relationship is still under investigation.

30.0
Model appears universal
across different closing sizes
25.0
20.0
Pred BMWi (kWh/t)
15.0
10.0 Cadia East

Ernest Henry
Aqqaluk
87 Samples
Resolution
5.0 10.8% relative error Develop1
212,150,106 75mm Develop2
Closing size on Bond tests ideal
0.0
0 5 10 15 20 25 30
Meas BMWI (kWh/t)
Figure 136 Quality of the prediction of bond mill work index BMWi from Ci-GRD
(Source: Kojovic et al 2010, Copyright: AusIMM)
Results indicate that the Ci-GRD offers a reliable index for estimating BMWi, with 8.7% average relative error
across 190 samples from 7 mine sites and 10 development rock types (Figure 136). The universal correlation
with BMWi is very promising, but integration with supporting information is also important at some sites for
example where the mineralogy range is extremely varied, or the core is heavily broken.
To illustrate the application of the Ci in geometallurgical mapping of throughput a large-scale comparative
study was carried out at an AMIRA P843 sponsor site. This involved comparing the results of a large
underground bulk sample extracted from a drive and tested using conventional methods for A*b and BMWi
with the results of three drill holes drilled along the drive. 443 Ci tests were conducted at two metre assay
intervals along the drill core and used to estimate BMWi and A*b for each sample.
The results for the bulk ore sample treated through the existing SABC plant are compared with the Ci
estimates in Table 8. The Ci based throughput estimates were based on the estimated ore hardness
parameters (A*b and BMWi), actual mill operating powers during the plant trial, and surveyed grind size (P80)
and SAG feed size (F80). As can be seen (Table 8) the mill throughput estimates are consistent with the plant
trial on the bulk ore sample.
Table 8 Comparison of actual and forecast parameters and plant throughput (Kojovic et al 2010)
BMWi Bulk Ore Plant JKCi Plant TPH

Sample A*b
(kWh/t) Throughput Forecast
Bulk Ore Sample (average) 28.9 20.8 - 1417
JKCi tested Hole 1 -average 29.7 20.2 - 1458

2000
RR222=0.3557
=0.6473
=0.5149
1900
1800
1700
..
1600
TPH
2 TPH
Ultimate
1500
Level 1
1400
1300
1200
Relative
1100
Error= 3.16%
Relative
1000
1000 1100 1200 1300 1400
4.49%
Error= 4.03%
1500 1600 1700 1800 1900 2000
LevelPredicted
P843 Analysis 3 TPH Tph
Figure 137. AMIRA P843 sponsor site Throughput Predictions vs. Underground production trial
(Source: AMIRA P843 2009, Copyright AMIRA)
As can be observed, the mill throughput estimates consistent with production trial on underground ore
samples (to support a mine expansion). This suggests at least in this case that the geometallugical approach
can provide process attributes fit for purpose to model the mining process in a sophisticated manner.
Figure 138 shows the same data as Figure 134, but with an estimated mill throughput based on a simplified
comminution circuit from the client process plant.

4.0 0
Crushing Index
3.5 500
Tonnes per hour

Comminution Index
1000
3.0
1500
2.5
2000
2.0
2500
1.5 Throughput 3000
1.0 3500
800 840 880 920 960 1000 1040 1080 1120
FROM
Down Hole Depth (m)
Figure 138. Estimated

Mill mill throughput
Dimensions 36’x18’ using theLoad
Ball Comminution
12% Index
78% inofaCritical
continuous 360m
Speed downhole
25% section of core.
Mill Load
Mill Dimensions 11x5.5m, Ball Load 12%, 78% of Critical Speed, 25% Mill Load
(Image: Toni Kojovic, Copyright: AMIRA)
The Ci had proven to be a fast and reproducible comparative comminution test designed specifically to
support large-scale geometallurgical mapping and modelling. As such it is not designed as a replacement for
existing higher precision tests used for ‘bankable’ engineering design. Insertion of the Ci test into routine
assay preparation as shown through a large-scale laboratory trial provides an opportunity to collect the data
during feasibility studies. This represents significant value adding to routine drill core testing. Estimates of
BMWi from extensive Ci development and validation studies show a high level of correlation suitable for
geometallurgical domaining.
Within the emerging field of geometallurgy rapid ‘front line’ testing methods such as the Ci provide the
opportunity to support a new approach to definition of inherent geological variability in terms of processing
performance parameters.
The sample mass required for JKTech or ALS to conduct the Ci Comminution Index is circumstance
dependant. This is a destructive test that allows for the whole sample mass to be used for further testing.
Drill core requirements define how the test is done. The test can be done on samples as small as 1m of drill
core (PQ/HQ/NQ, whole/half/quarter core, natural fragments -40+20mm).
5.10 Comparative Comminution Hardness Test Geopöyrä

Another small scale proxy test to measure the A*b impact breakage parameter is the Geopyörä device
(Matus et al 2020, Torvela 2020, Torvela et al 2020). This device is a new testing device is a variation of an
instrumented roll crusher with an adjustable gap to measure breakage forces and energy applied to rock
particles during the breakage process. This experimental setup can rapidly test rocks on a wide range of

sizes and energy levels. The device was capable of measuring energy with great accuracy and a standard
deviation of 1.07 Joules.
The device was developed at the University of Oulu and consists of two instrumented wheels of 600
millimeters diameter each, with an adjustable gap as shown in Figure 139. The steel wheels are powered by
integrated electric motors placed in the frame; the available energy is range between 100 to 250 Joules. The
device provides data of the impact force and measure the loss of rotational moment to determine the energy
required for each particle breakage event (Torvela 2020 and Matus et al 2020).
Figure 139. (LHS) 3D model of the double wheel crusher frame with wheels and wheel covers (Torvela et al 2020),
(RHS) the constructed Geopyörä device (Matus et al 2020)
This device can test a variety of small discrete samples, either bulk or drill cores on a range of particle sizes
and can conduct the test very quickly. The short testing time results in low operating costs, which allows a
large number of samples to be tested for a better understanding of the ore variability.
Figure 140 shows a series of high-speed photographs of a Geopyörä crushing event were taken with a Sony
Cyber-shot DSC-RX100 IV camera (lent by the Oulu Mining School) as part of the device development (Torvela
2020). The individual frames were cropped and combined into a single image seen in Figure 140. The series
shows the particle breaking into two distinct parts at the middle while passing through the gap with a spray
of dust and fragments ejected below.
Sample preparation for the standard test requires two sets of 15 particles of equal size, where each set
represents an energy level, and the energy of the breakage is adjusted by modifying the gap between the
two wheels. The device has been tested with particles from 31.5 to 9.5 millimeters. Particles are placed on
an automatic feeder, which drops it from the top of the device to the crushing zone between the two rollers
one at a time. The instrumentation and CPU record the force and energy data for each breakage event. After
the crushing, the product is collected from a tray in the bottom of the equipment. This approach was
developed to establish a test procedure that can be replicated and minimize operator-induced bias.

Figure 140. A series of high-speed photographs of a typical crushing event.

(Source: Matus et al 2020)
The resulting samples were sized and the t10 for each sample was calculated, and then the A*b impact
breakage parameter was calculated (see Section 4.2) at the Oulu Mining School (done by Tabatha Chavez
Matus as part of her thesis work). The measured t10 of each sample at the measured energy level of that
sample was compared against a t10 curve predicted by a commercial Drop Weight Test on a control sample
taken out of the same parent sample of ore for multiple mine sites. Figure 141 below shows how the
Geopöyä device A*b measurement directly compares to a commercial Drop Weight Test result on parallel
samples.
A further opportunity was explored by the Geopöyrä development team, where correlation with the Bond
Ball Work Index (see Section 4.3) was examined. Using a series of parallel samples taken from each ore type
(from serveal different mine sites), standard Bond ball mill tests were performed to determine the Bond ball
mill work index (BBMWi at a closing size of 150m) for each sample, to compare to existing Geopöyrä A*b
measurements.
Figure 142 shows a comparison between the BBMWi measured using the standard procedure and the
BBWMi estimated using the Geopyörä data. A satisfactory correlation can be observed with four out of the
five results within a relative error range of ± 20% (dashed lines).

50
40
GRBT t10, [%]
30
20
10
0
0 10 20 30 40 50
DWT t10, [%]
Kittila Kevitsa Terrafame Kylylahti Yara

Figure 141. Geopöyrä Double Wheel Crusher versus DWT measured versus predicted t10
35
Geopyörä estimated BBMWi
30
25
[kWh/t]
20
15
10
5
0
0 5 10 15 20 25 30 35
BBMWi [kWh/t]
Mine A Mine B Mine C Mine D Mine E
Figure 142. Comparison of BBMWi and the Geopyörä estimated BBMWi


This represents a unique opportunity. The way this data would be used is not to design a ball mill, but to
map variability of ore hardness. If the same test can measure directly the A*b impact breakage parameter,
and approximately estimate the BMWi, and do so quickly and cheaply, then this test would have potential
to be a useful geometallurgical tool.
What could be an interesting application would be the use of the Geopöyrä device installed in a process
plant to measure A*b real time. This would require a periodic sample taken from the SAG mill feed, then
that sample broken in the device, then the products sized quickly. If this was done by an optical sizing device,
then the sample could be sized as it falls out of the Geopöyrä unit. The A*b parameter and the BMWi could
be estimated. This data could be used to predict the SAG mill specific energy and the Ball mill product P 80.
This information could be used to optimize the SAG mill and ball mill together in real time during operation.
This concept is shown in Figure 143.
Automatic sampling
device to feed unit Optimize together
Geopöyrä A*b Prediction of SAG mill

real time Prediction of ball mill
P80 and Specific Energy
measurement product P80
applied (kWh/t)
t10 (%)
Broken product real time
PSD sizing measurement Calculation of A*b
A = 49.1 50
40
(optical system) parameter 30
20 b = 0.87
10 t10 = [1-e –b.Ecs]

0
0.0 0.5 1.0 1.5 2.0 2.5
Specific Comminution Energy Ecs (kWh/t)
Figure 143. Possible use of Geopöyrä to measure A*b real time in an operating plant
The opportunity is to use this device to match ore signatures found in a geometallurgy study with real time
operation. This concept has been discussed in Section 15.
5.11 Comparative Comminution Hardness Index Tester HIT

Another proxy test for the A*b impact breakage parameter (see Section 4.2) is the Hardness Index Tester
(HIT), developed by SimSAGe Pty Ltd (https://simsage.com.au/). This is a relatively simple device (Figure
144) that is a low cost device for estimating the comminution parameter A*b and the Bond Ball Mill
Work Index at any mine site, with potential applications on fragments up to 25 mm from drill core.

Figure 144. Image of the HIT device (Prototype V2; Patent Pending 62/241,852, PCT/I B2016/001591)
(Source: Kojovic et al 2019, Copyright: Toni Kojovic)
The application of the HIT was not to replace the standard comminution tests like the Drop Weight Test (see
Section 4.2) or the Bond Mill Work Index (see Section 4.5), but rather to generate a high number of
comminution and geomechanical parameters for the rock mass. Ultimately, the objective would be to
determine the variability of the rock mass and, therefore, easily distinguish any potentially problematic
zones for the mill.
The principle phenomena behind the HIT methodology is that A*b is the slope of the curve of ‘zero’ input
energy (Napier-Munn 2005) of the ore being tested. Mathematically, this means the slope of the curve for
that ore as it connects to the origin. This could be used as a fundamental rock breakage characteristic in
some applications of geometallurgy modelling. The slope at a relatively low Ecs (~0.2 kWh/t) is a reasonably
effective estimate of the slope at zero for that ore (i.e. the true A*b for the fitted curve).
So the HIT test will break particles one at a time at an applied energy of 0.2 kWh/t according to the machine
settings and procedure. The HIT test requires a minimum of ten fragments in a narrow size fraction (for
example -22.4 + 19 mm or -16 + 13.2 mm). Fragments may be sorted to ensure all fragments are within a
set mass tolerance around the bulk sample mean (for example ±20 %). This will ensure the mean mass of

any set of ten fragments will be within five per cent of the population mean. As such the process was in
theory in line with the particle sorting adopted by the SMC Test® protocol (see Section 5.3). The specific
energy (Ecs) calculated from input energy (Ei) and average particle mass (m) as follows:
𝐸𝑖 𝑀𝑔𝐻
𝐸𝑐𝑠 = = (J/kg) Equation 16
𝑚 𝑚
where:
M = a mass of the crusher assembly (crusher hammer, crusher shaft and crusher weight) (kg)
g = the gravitational constant (9.8 m/s2)
H = a height of the crusher hammer (m)
m = a weight of the fragment of granular material in the sample cup
Ecs can be converted to the units of kWh/t by dividing by 3600
The broken product from all ten fragments is dry sized for one minute using a single sieve, representing the
t10 size (for example, the t10 would be 2 mm for -22.4 + 19 mm size fraction fragments). The mass per cent
of undersize is referred to as t10. The slope, or t10/Ecs, is the raw hardness index (HDI):
𝑡10 %
𝐻𝐷𝐼 = ( ) Equation 17
𝐸𝑐𝑠 𝑘𝑊ℎ/𝑡
HDI needs to be corrected for the slope offset and particle size to estimate the JKDWT A*b parameter. This
is done in consultation of the manufacturer (SimSAGe Pty Ltd). Calibration against the standard SMC or DWT
parameters is possible, taking into account the offset for the slope and effect of particle size. Table 9 shows
the sample preparation and experimental procedure to run this test.

Table 9. Standard Operating Procedure for HIT Testing (Source: Kojovic 2016, Copyright: Toni Kojovic)
Figure 145 shows a comparison between the HIT measured A*b and the Full JKDWT Drop Weight Test
measured A*b on parallel samples. Figure 146 shows a comparison between the HIT measured A*b and the
SMC test measured A*b, again on parallel samples. As can be shown, correlation between both conventional
tests is reasonably good. It is to be remembered that the purpose of this test is not to replace the JKDWT in

the characterization of ore to design a mill, but to domain map a deposit in context of comminution process
response.
Figure 145. Comparison of JKDWT A*b Values and Corresponding A*b Estimates Using T10 for 31.5×26.5 mm Fraction
(93 samples, 32 ore deposits) (Source: Bergeron et al 2017, Copyright: Toni Kojovic)
Figure 146. Comparison of SMC Axb Values and Corresponding HIT Axb Proxy Estimates (62 drill core samples)

Figure 147. Comparison of Estimated and Measured Bond BMWi Results (62 drill core samples)
Figure 147 shows the HIT proxy estimated Bond Ball Work Index, BMWi (see Section 4.5) and the actual
measured BMWi on parallel samples. This result is interesting. One test can estimate the comminution
response of both impact breakage and grinding. So one test on one sample group (10 fragments) can be
used to estimate and comparatively rank the comminution footprint of that ore.
The opportunity could be to predict the performance of the comminution circuit. Figure 148 and Figure 149
shows an example of this, compared against actual plant performance. This ore was processed in an
established mining operation, where the SAG mill specific energy, and the Ball Mill product output was
measured for a trail period of 53 days. The SAG mill specific energy and the Ball Mill P80 was calculated in
simulation based on a HIT test. Figure 148 shows a comparison between simulated SAG specific energy and
measured SAG mill specific energy. Figure 149 shows comparison between simulated and measured Ball
Mill P80.

Figure 148. Trend Showing Actual and Calculated SAG Specific Energy for the Initial Trial Period
Figure 149. Trend Showing Actual and Calculated Ball Mill Cyclone Overflow P80 for the Initial Trial Period
The HIT evaluation program, running for 53 days was extended shown in Figure # and # (Kojovic et al 2019)
was extended, generating a further 107 days of sampling. This additional period provided an ideal data set
for model validation. To this end, the model developed from initial 53 days of monitoring was used to
forecast performance for the remaining 107 days not used in model calibration. This approach was
considered a textbook example of empirical model validation (Kojovic and Whiten, 1994). The calculated
and actual SAG Specific energy and ball mill P80 trends over the periods covered in the regression analysis
and validation are shown in Figure 150 and Figure 151.

Figure 150. Trend showing Actual and Calculated SAG Specific Energy for the Initial Trial and Forecast Periods
Figure 151. Trend showing Actual and Calculated BM Cyclone Overflow P80 for Initial Trial and Validation Periods
This result shows the potential value of this basic approach. There are many ways this testing approach
could be used, in geometallurgy and operational process prediction.

5.12 Comparative flotation test JKMSi

The JK mineral separation index (JKMSi) is a small-scale flotation test that requires feed material of 10 g of
sample (Vos & Bradshaw 2014, and Morgan et al 2012). Figure 152 illustrates the comparison between the
small-scale float test and a batch float response. Figure 153 is an example of a plant flotation response
compared to the JKMSi.
Figure 152. Comparative flotation response test JKMSi compared to conventional batch flotation tests
(Source: Vos & Bradshaw 2014, Image: François Vos, Copyright: AusIMM)
100
Versus plant data
80
Cu Recovery %
60
40
20
0
Particle Size (um)
Figure 153. Comparative flotation response test JKMSi compared to production plant recovery
(Source: Vos & Bradshaw 2014, Image: François Vos, Copyright: AusIMM)

5.13 CSIRO proxy diagnostic leaching test

The focus of the small-scale leach tests has been to determine the leach index (relative indication of leach
performance), recovery, impurity deportment, reagent consumption, effect of size/degree of liberation and
mineralogy of samples for their relative ranking (Figure 154). Predictive tests for heap leach hydrology
(related to agglomerate characteristics) and slurry rheological behavior are also being developed (Kuhar et
al 2011).
Figure 154. Correlation between nickel leached in the CSIRO proxy test and that from conventional column least tests
(Source: Kuhar et al 2011, https://www.csiro.au/en/Research/MRF and https://people.csiro.au/K/L/Laura-Kuhar
Image: Laura Kuhar, Copyright: CSIRO)
To use this test, a sample of 5 to 10 kg of each end member ore type is to be sent to CSIRO for
characterization. This would result in the diagnostic leach test being refined to target minerals of interest in
context of the geometallurgical campaign. Sample mass requirements for the diagnostic test itself is 15 to
20 grams but can be less depending on circumstances.
5.14 Establishing what comparative tests are appropriate for the target deposit
For each end member ore rock texture, conduct a bankable large-scale test (for example if the process
behaviour of interest was leaching, do a column leach test), characterize the rock in multiple scales and
forms (SEM mineralogy, optical mineralogy, QXRD, chemical assays, etc.), and conduct a series of small-scale
comparative tests. Assemble the result together for all of the samples end member extreme rock textures.
Develop a deposit specific relationship between one (or a combination of) comparatives tests and the
bankable test. Develop understanding of why these samples might rank in context of mineralogy. Ensure
sample representivity has been maintained across all tests. Design a propagation of analysis for all analytical
outcomes.

‘Comparative’ ‘Quantitative’
• Smaller-scale lower-cost tests that • Larger-scale tests for use in ‘bankable’
Same Ore type

generally produce qualitative outputs. design parameters kWh/t, A*b,
Grade/Recovery, etc….
• Results define relative similarity and
variability. • Recognised methodologies
with constrained protocols.
• Influences sample selection and
composites for more quantitative testing • Samples composited and blended to
represent actual operational feed
scenarios.
Deposit specific
proxy relationship
Figure 155. Development of the relationship between comparative proxy tests and larger scale bankable tests
Comparative tests are selected by past demonstrated usefulness and intuition-based selection of new
technology. As each deposit is unique, the data outcomes will vary in usefulness. A good metric for
usefulness is the quality of the relationship between the comparative test and the bankable test. Strength
of correlation between tests is perhaps the most practical measure. Another good metric is the range of
results returned. Does the full range of experimental results of the comparative tests exceed the set error
bars of the same comparative test? The Orientation Study is an excellent time to test assumptions and
question new ideas. The final Orientation Study conclusions need to be measurement based in a practical
fashion that can demonstrate propagation of error, sample representivity and has a clear data QA/QC
procedure.

6 ORIENTATION STUDY DESIGN

The tested sample interval is selected based on the number of tests (and what sample mass they require).
If possible select the same intervals of chemical assay measurement, or a multiple. This is usually a 2m
interval of core (sum several intervals together?). By the time the tests being selected for this part of the
study is done, a great deal of work should be done to consider the context of these tests and what the data
is needed for. It is simple to try and solve all problems in one experiment campaign. Budgets and time frame
to delivery agreements rarely allow for this.
Figure 156 represents a starting point, from which a cut down test schedule, fit for purpose to the original
geometallurgical objective, can be designed. A clear understanding of what the data is needed for and how
it will be analyzed needs to be in place before this is done. Some tests are destructive, which disqualifies
some kinds of tests after they are done (e.g. most metallurgical tests). Some do not (e.g. hyperspectral
analysis or EQUOTip). As such, the experimental test work is to be planned together, then done in a planned
order.
Non–Destructive
Chemical Hyperspectral XRF / XRD
Assay Imaging
Polished blocks &

Geophysics thin sections (optical
EQUOtip & SEM)
Lithology
Acid Mine
Destructive Test
Crushing Flotation
Blasting Drainage
Response
Grinding
Propensity
Leaching for dust
Geotechnical Liberation Response generation
Figure. 156 Experimental test work done on each sample core length interval
Below is a case study, where the objectives were to understand comminution and flotation process behavior.
The questions the experimental design and analysis were developed to answer were:
 What is the mineralogy that controls impact breakage?
 What is the mineralogy that controls grinding?
 What is the mineralogy that controls flotation recovery?
 Can these phenomena be related to a small-scale proxy test?
 Is it possible to map these into the deposit?

Figure 157 shows the Orientation Study and Figure 158 shows the Hypothesis Sample study.
Site data
Non–Destructive
Chemical Hyperspectral
Assay Imaging
Geophysics EQUOTip
Lithology
Destructive Test
Crushing Flotation
SMC
JKCi Response
QXRD MLA
Grinding
BMWi Batch Flotation
OWi Tests
Recommendations
Analysis
Figure 157. Case Study P, Orientation Study 16 samples

At the end of the Orientation study, a working relationship between the JKCi, an operating work index (OWi)
and the BMWi had been established. It was decided that the SMC test would be used to estimate the Axb
impact breakage parameter. A small-scale comparative test for flotation was not able to be done in this
study due to logistical constraints. A full batch flotation test was done for each sample in the Hypothesis
set. An understanding of what mineralogy might control comminution and flotation was determined with
the Hyperspectral imaging, QXRD and MLA results. EQUOTip was and Point Load Index was to be used as
model inputs with both JKCi-CRU and JKCi-GRD outputs of the Comminution Index to model BMWi domains.
Existing Site Data

Chemical Lithology Non-destructive Destructive Tests
Assays Logging Tests
Well Log Batch
Chemical
Crush to 99% passing 3.35mm
Rotary divide in 4 equal parts
Geophysics Flotation
Assays
EQUOTip Comminution Batch Chemical

Index JKCi Grind OWi
Sample Assays
2m length of HQ Point Load
Index
half drill core
QXRD
Hyperspectral SMC Impact
Imaging Breakage
Reserve sample
Figure 158. Order of test work in Hypothesis Sample set in Case Study P

Figure 158 shows the established order of test work. Data QA/QC was based on a mass balance reconciliation
between tests against the original sample mass. The Point Load Index breaks intact core. The Comminution
Index and the SMC test can be run on the Point Load Index products. The products of the Comminution
Index and the SMC tests were re-combined and then crushed to 99% passing 3.35mm and then was rotary
divided into four equal sub samples. One sub-sample went into the freezer as a flotation feed sample (to
reduce the risk of sample oxidization). One sub-sample was tested with the Batch Grind Operating Work
Index (OWi) procedure. The products were sampled again at target size fractions, which were sent to
chemical assays. One sub-sample was sent to QXRD analysis. One sub-sample was kept in reserve if needed
for future work (in the freezer at -200C in case flotation work was needed). Figure 159 shows the planned
Hypothesis Sample analysis and test work flow sheet.
Orientation Study Hypothesis on Comminution & Flotation
Non–Destructive
Chemical
Assay Hyperspectral
Imaging
Geophysics
EQUOtip
Lithology
Batch
Destructive Test
Point Load SMC
5 min Batch Flotation
Index (PLT) Impact
Grind (OWi)
Breakage
JKCi
Comminution Index
Chemical Chemical
Assay Assay
Data base
assembly
Data Analysis Domaining
Comminution & Flotation Ore Type Definition
Recomendations to Propagate Back into the Deposit
Figure 159. Case Study P, Hypothesis Sample set 218 samples


7 ORIENTATION STUDY DATA ANALYSIS

Data analysis of orientation study samples needs to address the following questions achieve the following
tasks:
 What tests showed the greatest spread in variability?
 What tests are useful here?
 Are some end member ore types very similar in process response space?
 Should they be merged into one ore type in this process?
 Do the tests interrelate?
 Is a good relationship between bankable tests and proxy tests viable for this deposit?
 What are mineralogical influences that can be seen as patterns across the samples?
 Set hypothesis spectrum for next experimental phase
 Define what tests are to be done in Hypothesis Sample Study and what QA/QC metrics they are subject to
Selection of process path Selection of polymetallic

for each target mineral process path for all ore types
What minerals What minerals

control separation? control recovery?
Gravity Magnetic
Leaching Sorting Flotation
Separation Separation
Grind Size Selection
Ranking of economic
value of target minerals

Analysis
Tools for domaining have been tested An understanding of what works and
and experimental protocols developed what does not for this deposit
What works and
what does not?
Useable relationship between Representatively of each test across
bankable tests and their proxies the sample maintained
Figure 160. Orientation Study analysis


7.1 Experimental design of the Orientation Study

The following is an example of an Orientation Study experimental design.
7.1.1 Step 1 – Sample selection.

Selection of all of the rock texture end members (in this case 10), together forming the samples for the
Orientation Study. Each sample would be approximately 20-30kg. It is of paramount importance that the
mineralogy and lithology rock texture is consistent throughout this sample. It is preferable to have less
sample mass and more consistent mineral rock texture, than a larger sample mass. This sampling is to be
done with the consultation of a geologist that has experience in working with this deposit. Usually, this takes
the form of drill core, and/or can be coarse residue from chemical assay sampling campaigns.
Figure 161 and Table 10 shows an example from the literature. This was a case study from Nickel Rim South
(NRS), which was a high grade copper–nickel–PGE deposit (Lotter et al 2011). In this case study,a total of six
distinct geometallurgical populations were defined within the deposit based on their mineralogical
characteristics. These include four ores within the primarily Cu bearing, PGM.
Footwall:
 Main (core).
 Upper.
 Fringe.
 Low Sulphur PGM.
Along with two ores from the Ni bearing, higher pyrrhotite contact zone:
 Sublayer Breccia.
 Footwall Breccia
Table 10 shows a summary of the key mineralogical features within each ore type. Figure 161 shows typical
textures from each of the geometallurgical units, as imaged by QEMSCAN. The scale of each image is
approximately 3.5 mm x 3.5 mm. Images show that the Main Footwall and Upper Footwall textures are
coarsest. Sulphide textures in all other geometallurgical units are finer grained and more commonly
associated with silicate gangue (Lotter et al 2011).
Table 10. Key mineralogical features within each ore type of the NHS case study (Source: Lotter et al 2011)

Figure 161. Geometallurgical unit textures as measured by QEMSCAN for the ore type of the NHS case study (Lotter
et al 2011)
For the purpose of this report, Figure 161 will be broken up and used as a theoretical example of pictures of
different samples for the orientation study. The QEMSCAN mapping has really highlighted the different
mineral signatures with the use of color. This is shown in Figure 162.
Orientation Orientation Orientation Orientation Orientation Orientation

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6
Figure 162. Theoretical example of six end member texture samples selected for the Orientation Study
(Image: Lotter et al 2011)

7.1.2 Step 2 – Sample preparation.

A number of experimental tests are to be run of each sample, where a number of very different process
separation methods are examined. This has to be done representatively, where all tests done represent
particles of rock taken from all parts of the original sample mass. As a general rule of thumb, each
Orientation Sample would be approximately 20 to 30kg in mass.
Crush the sample to a form where it can be rotary divided or riffled into subsections. If sorting response is
being examined, the sample is to be crushed to a top size of approximately 10-12mm, and a sorting sample
riffled out (in this case 1/6th). The remaining 5/6th of sample mass is then crushed to a top size of 3.35mm
(99% of sample passing a 3.35mm sieve). The top size of 3.35mm is selected to ensure that there are enough
particles to allow for particle statistics to be applied to each sub-sample.
The remaining 5/6th sample is rotary divided into 5 sub-products using a rotary divider. So of the six sub-
products, one is -12mm crushed material (for sorting) and the other five are -3.35mm crushed material.
Flotation
(5kg)
Gravity
Sorting
Separation
(5kg)
(5kg)
Orientation Sample φ
(25kg)
Magnetic
Leaching
Separation
(5kg)
(5kg)
Characterization
(5kg)
Figure 163. Primary crushing preparation for each Orientation Study sample
7.1.3 Step 3 – Characterization of each Orientation Study sample.

Before any other testwork (if possible), each of the Orientation Study samples need to be comprehensively
characterized. A number of methods are used in parallel for a variety of reasons.

Orientation Orientation Orientation Orientation Orientation Orientation Orientation

Sample Preparation
•
Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 7 Sample 8
Orientation
Sample 1
Orientation
Sample 2
Orientation
Sample 3
Orientation
Sample 4
Orientation
Sample 5
Orientation
Sample 6 2 stages of crushing
• PSD -12mm+38µm
For each • Size by Size XRF
Orientation • 1 stage of rod mill grinding
Sample
• Bulk Chemical Assay

o Lead collection Fire Assay (to measure for Pt, Pa and Au)
o 4 acid digest (to measure for 60 elements)
o Ammonium Citrate leach analysis (to measure supplied nickel minerals)
o LECO/ELTRA (Suplhur combustion test for high sulphur content)
o XRF pellet
• Bulk QXRD
• -12mm PSD + size by size XRF
• MLA polished block analysis after grinding
Characterization Point
o -150+75µm x 3 blocks
• Qemscan
o -75+38µm x 2 blocks • XRD/XRF
o -38µm x 1 block • Chemical Assay
Figure 164. Primary sample characterization for each Orientation Study sample
Chemical assay measurement is the basic standard for the whole campaign and for further routine work,
due to the comparatively low cost and small mass requirements. The following suite have shown to be
useful:
 Lead collection Fire Assay (to measure for Pt, Pa and Au), (requires 50g sample)
 4 acid digest (to measure for 60 elements)
 Ammonium Citrate leach analysis (to measure supplied nickel minerals)
 LECO/ELTRA (Suplhur combustion test for high sulphur content)
If there is the presence of precious metals like gold (Au), platinum (Pt) or palladium (Pd), then one of the fire
assay tests are required. Due to the very fine grain size, and very low grades mostly encountered, the sample
mass required is often much larger than other assay methods. A minimum of 50g is recommended.
While most mining operations will only assay for their target metal, process response is usually controlled
by gangue mineralogy in some form. The 4 acid digest chemical assay measurement method is
recommended.

Often, the target element is associated with some kind of sulfide mineral. These can be difficult to quantify
with conventional assay measurement methods. The Ammonium Citrate leach analysis and the ELTRA (or
LECO) tests are recommended.
In addition to this, further characterization can be done if the rock texture (and campaign objectives) has a
signature that could influence process response. An XRF pellet (requires 10g sample) test will give the
element distribution in the sample. A QXRD (requires 50g sample) test will give mineralogy present in a bulk
context.
There are three systems of automated mineralogy. QEMSCAN, MLA and Ziess Mineralogic. All three have
strengths and weaknesses. Selection of what method to use is a judgment call to be done by the
geometallurgy analyst team.
The limitation of this kind of analysis is the cost of test work and the long turnaround time. As such,
automated mineralogy is done at the start of the process (characterization of sample before process
testwork) and at the very end of the best process path (as shown by chemical assays and XRD/XRF analysis).
If this is done for each orientation study sample, this can create an expensive experimental program budget,
which may need to be reviewed at each stage of planning for this reason.
 -250+150µm x 3 blocks
 -150+75µm x 2 blocks These three size fractions are recomended. The coarser fractions will
 -75µm x 1 block need more poliched blocks measured to have a number of particles
measured relevant in context of sampling particle statistics.
7.1.4 Step 4 – Selection of grind size for the whole campaign.

One of the critical decisions to make that has the capacity to make the entire geomet campaign fail if done
incorrectly is the selection of sample particle grind size. This is the particle size (expressed as a P 80, or particle
size that 80% of the sample is smaller than) to which all samples for process separation a milled down in size
through grinding (See Section 4.7).
The process of comminution is to subject the sample to size reduction. The particle size required for the
effective liberation of the target mineral is defined by the mineralogy. For each mineral in the sample, there
would be a range of mineral grain size distribution. For process separation to economically effective, enough
of those particle have to be liberated or partially liberated.
An unconventional method of defining what the size distribution of the mineral grains are inside an ore type,
could be the use of electrodynamic fragmentation in a Selfrag unit (see Section 4.7.1). If done correctly, and
if the mineralogy is susceptible to this methodology, all minerals could be liberated intact at their true
particle size. This method has its advantages if the ore has more than one valuable mineral (possible
polymetallic process path), where each target mineral would have its own unique mineral grain size
distribution. For example, an ore with a Cu mineral with a P 80 grain size of 100 micron and a Au mineral
with a P80 grain size of 10 micron.
Usually this is defined through a sophisticated test procedure to determine the grade recovery curve for that
mineral in that ore type. The theoretical grade-recovery curve for an ore is a definition of the maximum
expected recovery by flotation of a mineral or element at a given grade. This is defined by the surface area
liberation of the value minerals and is consequently directly related to the grind size utilized in the process.
The theoretical grade-recovery can be readily used to quickly identify potential recovery increases that can
be gained through optimization of flotation circuits and whether the process is running efficiently.
To establish the theoretical grade-recovery curve for a material a mineralogical liberation study should be
undertaken using tools such as automated mineralogy (QEMSCAN or MLA).

Ideally this work should be done either before or in parallel to a geometallurgical study. Given enough
automated mineralogy data, a theoretical grade recovery curve could be estimated based on mineral grain
size across a family of ore types.
A decisions is to be made for what P80 each of the process test samples are to be reduced to through crushing
then grinding. As a laboratory scale rod mill produces a particle size distribution that better resembles a
production scale circuit compared to a laboratory scale ball mill, all sample preparation should be done with
wet grinding in a rod mill with a standard rod load. The choice of mild steel or another metal alloy for the
rods is a matter of discernment in context of the ore mineralogy.
To prevent oxidization of the sample, after rod milling, the wet sample is placed in a vacuum sealed bag
(ideally) and then stored in a freezer at a lower temperature than -180C. This should be done for any sample
that would or could be later be subject to flotation or leaching.
The ideal situation is the definition of closing samples grind size is taken from previous work done on the
same ore types.
7.1.5 Step 5 – 1st pass process separation for each Orientation Sample
Once the samples have been characterized and then prepared to an appropriate grind size in a rod mill, the
different process separation methods can be trialed. Any engineering process could be studying in this
context. The well understood methods are:
 Flotation
 Hydrometallurgical leaching
 Gravity separation
 Electromagnetic separation
 Sorting technology
The first step is to trial each of these process methods on the sample directly
Examine(Figure
Mawson 166). This gives a base
data G
recovery performance for flotation and leaching in each ore type. It also examines the effectiveness of what
gravity separation, electromagnetic separation and sorting has in context of what minerals report to what
Select 4-10 samples
process sub-product. The sub-products of each of these process tests are to be characterized
based on extremeappropriately.
data signatures ba
This could be termed as Orientation Step 1, where each of these methods Sample couldβHF
be consideredRotary a process
divide
Heavy fraction
path. Examine Mawson
Gravity A
each sample
data Sample βLF into 4 parts
Light fraction F
Select 4-10 samples B (flow) • Qemscan
• XRD/XRF
based on extreme Leach
CSIRO Examine Mawson
data signatures background bi (components) data
• Chemical Assay
RG
Rotary divide Separation Sample δConc
each sample Process Concentrate
Feed
into 4 parts Select 4-10 samples
Sample Flotation A Sample δTail based on extreme
Characterization Point Tails
Product
data signatures ba
• Qemscan Samples
•
•
XRD/XRF
Chemical Assay
A (flow) Rotary divide
each sample
Reserve 1 Size distribution
ai (components) measurement
Size by size hand
into 4 parts
F
held XRF &
C (flow)
Representitively sample across
• Qemscan
• Chemical
XRD/XRF Assay
• Chemical Assay
c
whole sample size
Sample prep in
distribution.
i 4(components)
size fractions
R
Sample α
Figure 165. Mineral reconciliation for each process separation test done

Leaching Leaching
(5-6kg) (5-6kg)
Examine what works and where

Examine what works and where
Orientation Orientation Orientation Orientation Orientation Orientation
Orientation Orientation Orientation Orientation Orientation Orientation Orientation
Sample Orientation
Sample 1 1Sample 2 Orientation
SampleOrientation
Sample 3 2Sample 41 Sample 3Orientation
Sample 25 Sample
Sample 37
Orientation
Sample 4Sample 5
Sample
Sample48 Sample 5Sample 8
Sample 7 Sample 6
Sorting Sorting
(5-6kg) (5-6kg)
For each For each Flotation

GrindFlotation
size
Characterization
Characterization
Orientation Orientation (5-6kg) (5-6kg)
Sample Sample
(5-6kg) Selection
(5-6kg)
Magnetic Magnetic
(5-6kg) (5-6kg)
Gravity Gravity
(5-6kg) (5-6kg)
Figure 166. First pass process separation for each sample in Orientation Study Step 1
Each process separation method should conducted in a way where the feed sample was characterized
appropriately and the process products are also characterized in the same way. The separation process
could be examined in context of what minerals where separated, by reconciling the mineral content of all
the products in reference to the feed (Figure 165).
All data is collected for each orientation study sample and examined in context of what was successful in
context of recovery and what possible pre recovery steps may be effective.

7.1.6 Step 6 – Data Analysis of the 1st pass process separation for each Orientation Sample
Each of the process separation tests done (4 or 5 different test methods for each Orientation Sample) would
be examined in context of what minerals separated into what products. The characterized products would
be compared to the characterized feed sample (Figure 136 and Section 3) to do a mineral mass balance
reconciliation. The separation performance of valuable minerals as well as gangue minerals would be
considered.
All separation methods are then considered together within each Orientation Sample and an understanding
of what minerals were separated out into products and in what circumstances. Knowing what minerals are
associated with the valuable target minerals is required. For example:
 Platinum PGE metals could be in grains 5µm in size, embedded into pyrrhotite and that pyrrhotite could be magnetic.
Thus a magnetic separation has the capacity to isolate the valuable platinum group metals away from the sulfides.
 Gravity separation could extract out all the heavy minerals containing Au, PGE and heavy REE away from the sulfides.
 Gravity separation could extract out gangue mineralogy that may negatively influence flotation recovery.
 Leaching could be effective in extracting cobalt.
 Flotation response is poor and is being hampered by gangue minerals of a particular character
The products of each of these process separation tests could be subject to a proxy test like a diagnostic leach
(done on small volume samples).
Consider how process combinations might work together. Using these insights, make recommendations for
Step 7.
7.1.7 Step 7 – 2nd pass process separation for each Orientation Sample
Possible cleaning of the ore by removing a gangue mineral could be considered to improve later recovery.
Assemble each test series into a proposed process path. This could be considered as Orientation Step 2. For
example:
 Gravity separation could possibly remove quartz, clays and biotite
 Flotation could be used to remove sheet silicates and clays
 Magnetic separation could have practical difficulties to consider before flotation due to the required throughput volume
in a production scale plant. That being stated, the possibility that gangue removal could improve recovery substantially
could be considered for future developments.
So, for example gravity separation could be done and one of the products could be subject to a further
process separation test (like flotation and/or leaching for example).

Diagnostic
Leach
% Recovery
Chemical Assay
Qemscan SEM
QXRD
Time
Gravity Conc
Heavy Flotation
Separation Tails
(5kg) Light -150+75µm x 2 blocks
-75+38µm x 2 blocks
-38µm x 2 blocks
Diagnostic
Diagnostic
• Qemscan Leach
Leach
• XRD/XRF
• Chemical Assay
Figure 167. Flotation response after gravity separation

The products of each of these process separation tests could be subject to a proxy test like a diagnostic leach
(done on small volume samples).
7.1.8 Step 8 – Data Analysis of the 2nd pass process separation for each Orientation Sample
As in Step 6, a detailed examination at the performance of each separation method will be done by
conducting a mineral reconciliation by comparing the characterized products against the characterized feed.
What minerals are separated for each method and where do these minerals report to? Examples of this can
be seen in Sections 7.3, 7.4, and 7.5.
A method of how to compare and rank the recoveries from flotation and leaching is shown in Section 7.2.
In the examination of what worked and why, metrics of how success is to be judged need to be established.
Include several metrics of success
 Ultimate recovery
 Time taken to reach ultimate recovery
 Recovery compared to grade to show the actual metal produced
 Allow for economic metrics to be calculated from these results
Good recovery of the target mineral is the ultimate goal, but there are tradeoffs to be considered. For
example, is the recovery of one mineral better than others (for example Au vs. Cu) where one is more
valuable than the other? Consider how process combinations have worked together.
Is there a connection between recoveries and mineral content? The flotation recoveries of the base line
response (1st pass) could be compared directly to the flotation recoveries from the same Orientation Sample
that has been subject to gravity separation beforehand (2nd pass). Is there an improvement? If so, what
difference in mineral quantities in the feed to flotation correlate with this improvement? This may be done
by comparing all Orientation Samples together in the same fashion to give a reasonable number of data
points.

7.1.9 Step 9 – Comparison process paths for each Orientation Sample

Compare the process paths developed in Steps 5 to 9 for each Orientation Sample. For each ore type, 8 to
10 different process paths can now be considered. The feed and products of each step within each of these
process paths has been appropriately characterized, which means a mineral mass balance reconciliation can
be conducted across each process path, and then process paths can be compared directly.
Process Path 1
Gravity
SGA
Orientation Step 1
Leach
Process Path 2
SLA
Representitive sample of Starting
Flotation Characterization end member orientation sample.
Process Path 3
SFA (in 4 size fractions)
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting • Which process path is more effective in the
Process Path 5 recovery of each target metal?
SOSA
• Which process path is most effective in
Analysis on what works and what does not
recovery of the 2-3 most valuable metals?
Flotation
Process Path 6 Gravity • What is the mineralogical signature that
SGFB
controls that process path?
Leach
Orientation Step 2
Process Path 7 Flotation

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Separation SGFDLB
Process Path 10 Leach

Gravity Flotation
SGFDLB
Orientation Gravity Flotation Leach
Process Path 11 Ore Sorting
Step 3 SOSBGC SOSGFC SOSGFDLC
Figure 168 Comparison of different process paths for each Orientation Study sample
A method of how to compare and rank the recoveries from flotation and leaching is shown in Section 7.2.
Questions to answer in this section could be (allow for economic metrics to be calculated from these results):
 What is the most effective process path for the recovery each target metal? (Au, PGE, Cu, Ni, Co,
etc.)
o Consider the tradeoff between grade or ore vs. the efficiency of recovery of that ore in a monometallic circuit
o Rank the process paths in order of preference (to an agreed upon metric)
 What is the most effective process path for the recovery of several mineral targets in a polymetallic
context?
o Consider the tradeoff between grade or ore vs. the efficiency of recovery of that ore in a ploymetallic circuit
o What are the tradeoff decisions (say PGE grade + recovery vs. Cu grade + recovery)?
o Rank the process paths in order of preference (to an agreed upon metric)

 What is the mineral signatures (usually gangue minerals) that control each process path?
o What separation steps are most useful in managing these minerals?
 What is the mineral signatures in context of what reports where for potential penalty elements in
each process path?
7.1.10 Step 10 – Selection of the most effective process path for each Orientation Sample
For each Orientation Sample, select one process path in context of the original geometallurgical
experimental aims and the requirements set by the client (owner of the deposit).
The most effective outcome will be a combination of the separation methods, done in a specific order. This
order would be defined by the ore, where if the orientation samples have been selected appropriately, a
range of recoveries and efficiencies would be observed.
Make the choice in a manner that is data supported and defendable in an audit.
7.1.11 Step 11 – Comparison of Process Paths Between Orientation Samples

Take the experimental results that have been analyzed in Step 9 for all Orientation Samples to compare them
together. If the work has been done consistently, each process path can be directly compared at each point
in the flow sheet between all Orientation Samples, where like is directly compared to like.
 Re-examine how the metrics of will define the best outcome in context of the original experimental aims of the
geometallurgy campaign.
 The characterized products can be compared in context of effective separation, both in starting feed samples and in final
products. Consider how separation process steps could work together in combination.
 For each end member ore type, examine the recoveries vs. grade for all the economic metals present and make a
judgment call on how the valuable target minerals could be ranked in economic importance.
o Does the high grade ore samples recover more effectively than the lower grade ores in some process paths
compared to others?
o For example Au may be worth more than Cu per unit mass, but if the Au is very low grade and very fine in grain
size (e.g. Au in ore 0.02g/t at 5µm grain size) compared to Cu (e.g. Cu in ore 1% at 100µm), the cost of producing
Cu will be much lower and may be more economic.
 Once the valuable minerals have been ranked, make a judgment call on what could be the overall most economically
effective process path. Rank the process paths in this context.
 Gangue minerals and penalty elements can be examined in context of where they report to in a process path

Orientation Orientation
Process Path 1
Gravity
Sample 8
SGA
Orientation Step 1
Sample 7 Orientation
Leach
Process Path 2
SLA
Gravity Representitive sample of Starting
Flotation Process Path 1 Characterization end member orientation sample.
Sample 8
Process Path 3 SGA
SFA
Orientation Step 1
(in 4 size fractions)
Magnetic Process Path 2
Leach Sample SC1-4
Process Path 4 SLA
Separation
Ore Sorting Process Path 3 Flotation Characterization end member orientation sample.
Process Path 5
SOSA SFA (in 4 size fractions)
Analysis on what works and what does not Sample SC1-4
Magnetic
Process Path 4
Separation
Flotation
Process Path 6 Gravity
SGFB Ore Sorting
Process Path 5
Leach SOSA
Orientation Step 2
SFADLA
Magnetic Flotation
Process Path 8
Separation Flotation
Magnetic Leach SGFB
Separation SGFDLB
Leach
Orientation Step 2
Process Path 10 Leach SFADLA
Gravity Flotation
SGFDLB
Magnetic Flotation
Orientation Process PathGravity
8
Process Path 11 Ore Sorting Separation Flotation Leach
Magnetic Leach
Separation SGFDLB

Gravity Flotation
SGFDLB
Process Path 1
Gravity
SGA
Orientation Step 1
Leach
Process Path 2
SLA
Flotation Characterization
Process Path 3 end member orientation sample.
Sample SC1-4
Magnetic
Process Path 4
Sample 7
Separation
Gravity
Ore Sorting Process Path 1 SGA
Process Path 5
Orientation Step 1
SOSA
Leach
Process
Analysis on what works and what doesPath
not 2 SLA
Flotation Representitive sample of Starting
Process Path 6 Gravity Flotation Characterization end member orientation sample.
Process Path SGFB
3
Orientation Step 2 Leach Sample SC1-4
SFADLA Magnetic
Process Path 4
Separation
Magnetic Flotation
Process Path 8
Separation Ore Sorting
Process Path 5
Magnetic SOSA Leach
Separation SGFDLB
Gravity Flotation
SGFDLB Flotation
SGFB
Step 3 SOSBGC SOSGFC Leach SOSGFDLC
Orientation Step 2

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Orientation
Gravity
Process Path 1 Separation SGFDLB
SGA
Study
Orientation Step 1
Leach Process Path 10 Leach

Process Path 2 Gravity Flotation
SLA SGFDLB
Flotation
Orientation Characterization
Process Path 3 Gravityorientation sample.
end member Flotation Leach
StepSFA
3 (in 4 size fractions)
SOSBGC SOSGFC SOSGFDLC
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Orientation Study
Process Path 5
SOSA

SampleOrientation
Gravity
•
Process Path 1
Flotation
Process Path 6 Gravity SGA
Which process path is more effective in the

Study
SGFB
Orientation Step 1
Leach
Process PathLeach
2
Orientation Step 2
Sample 5
Process Path 7 Flotation SLA

SFADLA
Magnetic Process Path
Flotation Flotation Characterization end member orientation sample.
Process Path 8 3
Separation SFA (in 4 size fractions)
Magnetic Leach Sample SC1-4
Process Path 9 Flotation Magnetic
Separation Process Path 4 SGFDLB
Separation
recovery of each target metal?

Gravity FlotationOre Sorting
Process Path 5 SGFDLB
SOSA
Step 3 Analysis onSOSBGC
what works and what does not
SOSGFC SOSGFDLC
•
Flotation

SGFB
Leach
Orientation Step 2

SFADLA
Magnetic Flotation
Orientation
Process Path 8
Separation
Process Path 1
Gravity
SGA Magnetic Leach
Orientation Step 1

Leach Separation SGFDLB
Process Path 2
4
• recovery ofiseach
mosttarget metal?
SLA
Leach
Which process path effective in

Process Path 10 Gravity Representitive
Flotationsample of Starting
Flotation Characterization end member orientation sample.SGFDLB
Process Path 3
Orientation Gravity
Sample SC1-4 Flotation Leach
Magnetic Process Path 11 Ore Sorting
Process Path 4 Step 3 SOSBGC SOSGFC SOSGFDLC
Separation
Ore Sorting
Process Path 5
SOSA

Flotation
SGFB

Orientation
Leach
Orientation Step 2

SFADLA
Orientation
Gravity
Flotation SGA
Process Path 8
Orientation Step 1
Separation
Leach
•
Sample 4
Leach
Process Path 9
Separation
Flotation SLA SGFDLB
Which process path is most effective in

Process Path 10 Process Path 3 Leach
Gravity Flotation SFA SGFDLB (in 4 size fractions)
Orientation
Sample SC1-4
Gravity Magnetic Flotation Leach
Process Path 11 Ore Sorting Process Path 4
Step 3 SOSBGCSeparation SOSGFC SOSGFDLC
•
Ore Sorting
Process Path 5
What is recovery
the mineralogical signature
most that
SOSA
Process Path 1
Gravity
SGA
Flotation
SGFB
of the 2-3 valuable metals?

Orientation Step 1
Leach
Orientation Step 2
Leach Process Path 7 Flotation

3
Process Path 2 SFADLA

SLA
Flotation Magnetic
Characterization Flotation
Process Path 3 Process Path 8 end member orientation sample.
SFA Separation (in 4 size fractions)
Sample
Sample SC1-4
Magnetic Magnetic Leach
Process Path 4
Separation
Separation SGFDLB
Ore Sorting Leach
Process Path 5 Process Path 10
SOSA Gravity Flotation
SGFDLB
Process Path 6
Step 3
Gravity
Flotation SOSBGC SOSGFC SOSGFDLC
SGFB
•
Leach
Orientation Step 2
What is the mineralogical signature that

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Orientation
Separation SGFDLB
Process Path 1
Gravity
Process Path 10 SGA Leach
Gravity Flotation
Orientation Step 1
SGFDLB
Leach
Sample 3
Process Path 2

Process Path 11 Ore Sorting SLA
Step 3 SOSBGC SOSGFC SOSGFDLC Representitive sample of Starting
Process Path 3
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Process Path 5
SOSA
Orientation
Process Path 1
Gravity Analysis on what works and what does not
SGA
Orientation Step 1
Flotation
Leach Process Path 6 Gravity
2
Process Path 2 SGFB

SLA
Flotation Characterization Leach orientation sample.
Orientation Step 2
Process Path 3 Process Path 7 Flotation end member

SFA SFADLA
Sample
Sample SC1-4
Magnetic Magnetic Flotation
Process Path 4
Separation Process Path 8 Separation
Ore Sorting
Process Path 5 Magnetic Leach
SOSA Process Path 9 Flotation
Separation SGFDLB
Mapping Study
Flotation Gravity Flotation
SGFDLB
SGFB
11 Ore Sorting
Process PathLeach
Orientation Step 2

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Separation SGFDLB

Gravity Flotation
• Process domains
SGFDLB
Orientation
Step 3 SOSBGC Gravity SOSGFC SOSGFDLC

Process Path 1
SGA
Orientation Step 1
Leach
Sample 2
Process Path 2
SLA
Process Path 3
Process Path 1
Gravity SFA (in 4 size fractions)
SGA
Sample SC1-4
Orientation Step 1
Magnetic
Orientation
Leach Process Path 4

Process Path 2
SLA
Separation
• Variability
Flotation Ore Sorting
Characterization end member orientation sample.
Process Path 5
1
Process Path 3
SFA SOSA (in 4 size fractions)
Sample SC1-4
Magnetic Analysis on what works and what does not
Process Path 4
Separation
Mapping Study
Flotation
Sample
Ore Sorting Process Path 6 Gravity

Process Path 5 SGFB
SOSA
Leach
Orientation Step 2

Process not 7 Flotation
SFADLA
Flotation
SGFB Magnetic Flotation
Process Path 8
Separation
Leach
Orientation Step 2

SFADLA Magnetic Leach
Magnetic FlotationSeparation SGFDLB
Process Path 8
Separation
Gravity Flotation
• Process domains
Magnetic Leach SGFDLB
Separation SGFDLB
Process Path 10 Process Path 11 Ore Sorting Leach
Step 3
Gravity Flotation SOSBGC SOSGFC SOSGFDLC
SGFDLB
Process Path 1
Gravity
SGA
Orientation Step 1
Orientation
Leach
Process Path 2
SLA
• Variability
Sample 1
Process Path 3
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Process Path 5
SOSA
Figure 169. Comparison of different process paths for all Orientation Study samples
Flotation
SGFB
Leach
Orientation Step 2

SFADLA
Magnetic Flotation
(Image. Simon Michaux)

Process Path 8
Separation
Magnetic Leach
Separation SGFDLB

Gravity Flotation
SGFDLB

7.1.12 Step 12 – Trial of best Combination Process Separation for the Deposit
Based on what has been learned in Step 11, consider what might be the best overall combination of
separation processes for the family of ore types in this deposit. If this process path has not been already
tested, assemble the ideal process path and test all Orientation Samples. Once the experimental results
have been completed, add the data to what has already been done, and repeat Step 11.
Magnetic
Separation Diagnostic
Fraction Leach
% Recovery
Chemical Assay
Qemscan SEM
QXRD
Time
Conc
Fraction A Flotation
Sorting or
Method X Tails
Fraction B -150+75µm x 2 blocks
-75+38µm x 2 blocks
-38µm x 2 blocks
Gravity Diagnostic
Separation Leach
Fraction Diagnostic
Leach
• Qemscan
• XRD/XRF
• Chemical Assay
Figure 170. Development of the ideal process path based on previous testwork done and analyzed in Orientation
Study (Image: Simon Michaux)
7.1.13 Step 13 – Selection of the most effective process path for the Deposit
Based on what was learned in Step 11, select for the whole deposit, one preferred process path in context
of the original geometallurgical experimental aims and the requirements set by the client (owner of the
deposit).
The most effective outcome will be a combination of the separation methods, done in a specific order. This
order would be defined by the ore, where if the orientation samples have been selected appropriately, a
range of recoveries and efficiencies would be observed.
 What is the most effective process path for the recovery of several mineral targets in a polymetallic context?
 What tradeoff decisions were made to make this the best choice?
 What is the mineral signatures (usually gangue minerals) that control each process path?
 What is the mineral signatures in context of what reports where for potential penalty elements in each process path?
Make the choice in a manner that is data supported and defendable in an audit.

7.1.14 Step 14 – Selection of the mineralogical mapping signatures of the most effective process path
Based on the choice made in Step 13 and what was learned in Step 11, assemble the recommendations of
what mineralogy controls the chosen process path for each of its separation steps. Usually gangue
mineralogy controls process behavior. Also, valuable minerals could be associated with other minerals.
These minerals will form the focus of analysis for the Geometallurgical Mapping Study (Section 12). Each
mineral (estimated with chemical assay’s if possible) would be examined in context of domains and
variability along continuous sections of drill core.
This would form the list of minerals that the Mapping Study would examine.
7.2 Method to Mathematically rank and compare process recoveries

The purpose of the geometallurgical Orientation Study is the assessment of each of the end member rock
textures, studied in context of multiple process paths, where the final absolute recoveries can be examined
and ranked. In this section, a method of ranking recoveries is presented as an option to consider.
Consider the leaching recovery over time from four column leach tests shown in Figure 147. In terms of
ultimate recovery, clearly Ore Type D is the most preferable and the four ore responses can be ranked as
follows:
 Ore Type D
 Ore Type A
 Ore Type B
 Ore Type C
100%
90%
80%
Ore Type A
70%
60%
Cu Recovery
Ore Type B
50%
40% Ore Type C
30%
Ore Type D
20%
10%
0%
0 5 10 15 20 25 30
Time Leached (h)
Figure 171. Leaching recovery of four different ore types in the same deposit

How could this be done mathematically consistently in a form that could be graphically plotted against other
parameters?
A model is fitted to this data using the same basic form as the A*b curve shown in Section 4.2 to rank and
model comminution. This mathematical tool can be used in a similar fashion for leaching recovery. The
equation below was fitted to the data in Figure 141, with the results shown in Table 9 and Figure 172.
]
Equation 18
Where:
Q = Cu Output %
t = Time leached
Table 11. Fitted model parameters
Sample R s R*s
Ore Type A 76.9% 0.45 34.6
Ore Type B 54.3% 0.16 8.7
Ore Type C 49.2% 0.17 8.2
Ore Type D 81.8% 1.21 98.9
100%
90%
80%
Ore Type A
70% Ore Type B
60% Ore Type C

Cu Recovery
Ore Type D
50%
Ore Type A Fitted
40%
Ore Type B Fitted
30% Ore Type C Fitted
20% Ore Type D Fitted
10%
0%
0 5 10 15 20 25 30 35
Time Leached (h)
Figure 172. Fitted models to leaching recovery curves


According to Table 11 the ultimate recovery could be ranked by using the ‘R’ parameter. The time taken to
reach the ultimate recovery could be ranked by using the ‘s’ parameter. R*s could be used as a ranking
parameter accounting for both ultimate recovery and the time taken to reach that recovery (in a similar way
to how the A*b parameter is used), resulting in:
 Ore Type D R*s = 98.9
 Ore Type A R*s = 34.6
 Ore Type B R*s = 8.7
 Ore Type C R*s = 8.2
The R*s recovery ranking parameter is dimensionless and can be used to mathematically correlate with other
attributes or parameters. Figure 149 is an example where the data in Figure 148, and Table 9 is plotted
against mica clay content for each sample.
2,5
1,5
Mica Clays (%)
0,5
0
0,00 20,00 40,00 60,00 80,00 100,00 120,00
R*s recovery ranking parameter
Figure 173. An example how R*s ranking parameter could

The presence of clays or sheet silicates often negatively impact process performance and recovery. Figure
173 shows how the method of quantifying each sample by mathematically comparing in a consistent form,
using the form described in Table 11.
7.3 Example A1 – What mineralogy controls grinding?

Figure 144 results show an example of assessing the variability of grinding as a function of end member rock
type. Identical procedures has been run on all of these samples. A feed sample of 700cc in volume was
prepared by crushing the sample where 99% of the sample mass passes the 3.35mm sieve screen (this is the
same as the sample feed for the Bond Ball Index test). Each feed sample first had its size distribution
measured on the √2 sieve series down to a size of 38m. Then each sample was placed in a standard bond
ball mill with a standard ball charge for the BMWi test. The mill was run (sample without water) for exactly
358 revolutions at a speed of 71.5rpm (5 minutes). The dry ground sample was collected, its mass measured
and then wet screened on a 38m screen. The resulting +38m sample was dried overnight in an oven at

1050C and its size distribution determined on the √2 sieve series down to a size of 38m (1.18mm, 850µm,
600µm, 425µm, 300µm, 212µm, 150µm, 106µm, 75µm, 53µm, +38µm, -38µm and cyclosize).
Figure 174 shows the +38m size distribution in the form of percent mass in each sieve (the calculation step
before calculating the conventional cumulative percent passing size distribution). This method of data
presentation is useful in that it returns peaks and troughs that are signatures of the grinding process and the
mineralogy of the sample. The peak at 100m is often seen in grinding tests and is thought to be related to
the geometry of the spaces between the grinding balls when the empty mill is at rest. The peak at 2.36mm
is also seen in grinding tests and is related to the size fraction that the mill has the most difficulty in grinding.
The volume recirculating load in a standard bond ball index test is influenced by the volume in this size
fraction. The harder the rock being ground, the more particles that are left at this coarse size fraction. The
magnitude of these peaks is related to the number of particles left on the sieve at that size fraction (or the
volume of the original sample at that particle size).
End Member Samples Batch Grind Product

25
20
15
(%)
10
0
0,01 0,10 1,00 10,00
Size Fraction (mm)
Figure 174. Grinding behavior of several tests from the same deposit
It is noted that there is quite a range across both peaks. A high peak at 100m corresponds to a low peak at
2.36mm; a sample that is soft to grind. Alternatively, a low peak at 100m corresponds to a high peak at
2.36mm; a sample that is hard to grind.
The next step would be to examine and assess the mineralogy of the particles in both of those size fractions
(-150+106m and -3.35+2.36mm) by taking a sample of each and testing them. Gangue mineralogy usually
controls grinding behavior (Figure 175). A useful approach could be to measure the QXRD of each peak size
fraction for each of the end member rock textures, then compare these results to the mineral mapping done
of the intact sample. The objective is to understand what mineralogy the rock breaks into and how what is
observed in the peak size fractions might relate to the original rock.
An examination of how many particles (peak magnitude or % of original sample at that size fraction) in
context of which minerals are dominant would provide an understanding of what minerals control grinding
in the deposit. In this example, the dominant minerals found in the -3.35+2.36mm size fraction were

potassium feldspar. The dominate minerals found at the -150+106m size fraction were magnetite,
carbonate and pyrite. If budget allows the size fractions could be characterized with automated mineralogy
methods (which is often useful in studying other process behaviours like flotation or leaching).
Intact Mesotexture Dry grinding
5mm
Sample Q112 Batch Grind Product

18
Of original samples
16
14
12
10
(%)
8
6
4
2
0
0,01 0,10 1,00 10,00
Size Fraction (mm)
-150+106 microns -3.35+2.36mm
Figure 175. Study of mineralogy that dominates grinding


7.4 Example A2 – What mineralogy controls comminution footprint

The comminution footprint refers to the deposit specific relationship between impact breakage (measured
with the Axb parameter from the DWT) and grinding behaviour (measured with the Bond Ball Work index
BMWi). An RBT test (to estimate the Axb Impact Breakage Parameter) and a Bond Ball Work Index (BMWi
at a closing size of 150m) was done for each end member sample. Care was taken to ensure these tests
were representative samples from all parts of each end member sample. The curve shown in Figure 176 is
referred to as the comminution footprint.
Carbonate-magnetite
End Memember Orientation Samples
300
Axb Impact Breakage parameter
A
KSpar dominant
250
M F
K 200
G
150
100
50
0
6 7 8 9 10 11 12 13 14
P
D
BMWi @ 150 micron cls
Q
Figure 176. An example of a mineral hypothesis of the impact breakage vs. grinding behaviour curve
(Image: Steve Walters, Copyright: AMIRA)
At the top left end of the curve (where the ore is softest), sample rock texture was dominated with carbonate
and magnetite mineralogy. At the bottom right end of the curve, sample rock texture was potassium feldspar
and quartz dominant.
Examples A1 and A2 were two different parts of the orientation study for the same deposit. The conclusions
were that the minerals that controlled comminution in this deposit magnetite, carbonate, quartz and pyrite.
This would form a hypothesis to guide the next step in the geometallurgical campaign where a smaller
number of tests would be done on a much larger number of input samples. The outcomes of which would
be propagated into the deposit for spatial domaining.

7.5 Example A3 – Process Circuit Performance Estimation Economic Modelling of each End Member
Sample
The measured Axb and BMWi characterization parameters from Example A2 were then used to estimate the
performance of a proposed comminution circuit (Figure 153). The objective would be to rank the ranked
throughput and energy power draw consumption of each of the end members (if the feed was 100% one
end member only). This is a crude estimate only to assist in the domaining of the deposit in context of
process behavior. This is not a substitute for larger scale ore characterization and engineering mill design.
 For each end member sample, the Axb impact breakage parameter was measured with an SMC test and the BMWi
was measured with a Bond Ball test. These results were then used as input into the procedure to estimate
throughput and estimation of performance of a conceptual comminution process plant.
 For each end member sample, the comminution throughput, measured recovery (batch flotation study – not shown
here) and the geological characteristics of the original rock texture were all used to estimate the economic
signatures.
The mill throughput was estimated using a set of engineering equations developed from 30 operating
AG/SAG mills (closed and open circuit), covering a range of diameters, lengths, speeds, ball loads, grate
designs, ore characteristics and feed size distributions. These models are similar to the models developed by
Morrell (2004) for scale-up, design and optimization. The equations have the form:
SP(kWh/t) = SAG(kW)/TPH = f´(F80,T80,SG,DWAB,BMWi,BL,CS,D,D/L) Equation 19

where:
 SP = AG/SAG specific power required for given set of operating conditions
 F80 = SAG mill feed 80 per cent passing size
 T80 = SAG Mill product 80 per cent passing size (or transfer size)
 SG = ore specific gravity
 DWAB = Drop Weight specific power required to break feed to transfer size, which is dependent on
 ore A and b parameters, feed size distribution and trommel size.
 BMWi = Bond ball mill grindability index
 BL = ball charge (per cent)
 CS = mill speed (per cent of critical)
 D = mill diameter
 D/L = aspect ratio = mill diameter/mill length (EGL)
The above equations are combined with the Bond ball mill design equation to complete the circuit
throughput model, providing a link between the SAG and ball milling stages. That is, the throughput capacity
is dependent on both the SAG and ball mill performance as they interact in practice.

Equation 20
where:
 TPH = expected feed capacity (tph)
 PBM = installed ball mill power (kW)
 EF1, EF2 … = Bond’s efficiency factors
 T80 = SAG circuit transfer size
 P80 = final ball mill circuit product size.
The maximum throughput may be SAG or Ball Mill limited, depending on ore conditions. The sensitivity of
the Excel model to A*b was fine-tuned using JKSimMet software (JKTech 1991, and Wiseman 1991) to better
capture the effect of recycle load and pebble crushing on transfer size to the ball mill.
This modification provides a correction to the SAG transfer size T80 (from the baseline) when A*b changes
at constant BMWi and increasing recycle fraction.
The effect of A*b and feed size F80 is included in the Drop Weight Axb (DWAB) term. However, ideally the
complete feed size distribution is required to determine DWAB. The simplification presented here assumes
the shape of the distribution remains constant when the F80 changes as part of a comparative
geometallurgical mapping type approach.
The estimated throughput is based on the existing SABC circuit, comprising an open circuit SAG followed by
two ball mills in closed-circuit with hydrocyclones (Figure 177).
Figure 177. SABC flowsheet used in CI throughput modelling (Kojovic et al 2010)


Figure 177 can be simulated more effectively using JKSimmet (JKTech 1991 and Wiseman 1991). The same
approach can be achieved with flotation, hydrometallurgy and pyrometallurgy using the HSC
thermochemical simulation software.
This approach was comminution based. A more comprehensive approach would be to include mineralogy,
recovery estimation and possibly a penalty element estimate (Figure 178). In doing so, an estimate of
throughtput, cost of production and metal recovered as a function of time unit (per hour) could developed
for each rock type (or in this case orientation sample).
Axb
Simulations
BMWi
Input Data Desired

Outcome
Recovery Estimate
Throughput
Model recoverable
Grind Size metal per hour
Mineralogy
Recovery Estimate cost
of production
Penalty
Elements
Figure 178. Process simulation for each end member sample

7.6 Economic modelling and business model development

 Economic modelling and optimization
 Original objectives reassessed, but refitted in context of the business model
 Scope and mandate reassessed based on analytical outcomes
 Metrics of success agreed upon to optimize towards
 Links established in how this could be related to operation protocols
 Links to reconciliation protocols
For each end member (the rock texture extremes of the deposit being studied), an estimate of the metal
production throughput and cost of production in a proposed process plant. In addition to this, an estimate
of the economic net revenue value was determined. This was based on the formula shown in Figure 155.

Figure 179. Economic Net Revenue value of a parcel of ore

This was to provide the possible extremes of revenue generated by the operation. At this stage, the
proposed circuit is a crude and simplistic model. This procedure was to be carried out on all of the samples
planned in the geometallurgical campaign. If successful, the results would be useful in making ore blending
recommendations.
7.7 Conclusions of orientation study

At the conclusion of the orientation study, the following should be understood:
 An understanding of what works and what does not for this deposit
 Useable relationship between bankable tests and their proxies
 A series of hypotheses of what characterizes the target process behaviour, to be used in the Hypothesis Sample set
 Experimental design for what tests are to be determined on the Hypothesis Sample set (the remaining samples of
the 1500-2000m drill core)
 Representatively of each test across the sample maintained
 The order of tests to be done, where some of them will destroy the core
 Data QA/QC protocols established
 Data analysis & modelling planned out to milestone conclusion
 Sample size and drill core depth interval for each Hypothesis sample
 Tools for domaining have been tested and experimental protocols developed

8 BUILDING THE GEOMETALLURGY DATA SET

As each test in the Orientation Sample set is completed, the resulting data is to be assessed in context of
QA/QC protocols then entered in the geometallurgy database (see Figure 180-184). This step is often
misunderstood or not done very well. All analysis going forward will be using this data set in the assumption
the data is reliable and in an appropriate form.
The nature of a geometallurgical study is that data is collected in various forms that is conventionally not put
together for analysis (metallurgical test work in context of spatially defined meso texture mapping). Also,
many of the tests done do not have the same precision and reliability of outcomes as the larger scale
bankable tests.
Core imaging
Optical microscopy
MLA microscopy
€ and Resolution
Figure 180. Different kinds of data results from different measurement platforms
(Image: Steve Walters, Copyright: AMIRA)
Much of the data collected in a geometallurgical campaign may come from different measurement platforms
that are not traditionally used together. Some thought is required in how to appropriately integrate these
concepts in a useful manner.
A very clear understanding of what each data outcome is and what form it takes needs to be developed,
across the whole data set. This is important to understand and manage the analysis outcomes. The purpose
of this analysis is to provide signatures for process behaviour that can be domained into the deposit.

Indirect Direct Direct Linking

proxies Proxies Measures Models
Assays Rock type EQUOTip PLT Index Plant performance

Alteration hardness A*b values prediction
Hyperspectral… Mineralogy BMWi Additivity
texture….. Flotation Economic net
Recovery… value
Trends Rankings Calibrations Behaviors
Geometallurgical dataset building
Engineering design & optimum schedule

Corporate decision making & business development
Figure 181. Integration of the different kinds of data

The structure of the data set needs to be thought through in a fashion where all the relevant information
can be collected as each test is done. Figure 182 shows some of the things to consider while attending to
this step.
Make recommendations to do all

tests across the remaining 1500- Only the orientation
2000m of drill core samples that passed
QA/QC
Flag where samples are

Merge orientation study output compromised (and why) and
into Geomet matrix delete from results set
(6) UpdateGeomet
(5) Update Geomet
Data
DataMatrix
Matrix
QA/QC on this
process of
Define the architecture of the
integration geometallurgical data matrix
Figure 182. Update the geometallurgical database and develop its final architectural structure

While establishing the structure of the data matrix with the planned tests, also consider how to link to
documentation and procedures. Consider what level of error would be acceptable to allow a test result to
be included into the matrix (for example, a maximum mass loss of 1%). The data from the Orientation Study
should be included in this matrix.
Work completed previously at site will provide excellent data in terms of spatial domaining. The
geometallurgy analytical team is to examine this work and collect the raw data used in these client outcomes.
The lithology domains as defined by site geologists have good data backing their analytical decisions. The
raw data behind the lithology domains should be incorporated into the geometallurgical data matrix if
possible.
The final data set (Figure 183) is an integration of very diverse kinds of information, all linked back to a single
sample identification (usually an alphanumeric related to drill hole number and sample depth section). Once
the data set has been completed with test data that has been assessed in terms of its suitableness, the
analysis steps can begin.
Existing data from site

Small scale comminution testing
Chemical Lithology
PLT CrushingAssays
Index A*b fromLogs
RBT Batch Grind
New data from geometallurgical campaign

New data from geometallurgical campaign
Batch Batch
Grind and Flotation Testing
Grind OWi Hyperspectral
Flotation Characterization &
Core Imaging
Comminution Hyperspectral
Texture-based
Index Ci Classified
Imaging
recoveries images
Point Load
Texture-based EQUOTip
Index Texture
liberation groupings
SMC Impact
Sample ID data
Breakage
Intact texture Geophysical

Classified Geotech
Mineralogical Texture-based
libraries Logs
textures Logs
definition FEM models
Existing data from site
MLA and optical microscopy
Figure 183. Case Study P: Building the geometallurgical dataset structure

In the final matrix, each test entry should also be accompanied by each number entry used in its final
calculation. This may seem excessive, but it will be easier to track QA/QC issues if this is done as standard
practice from the beginning. The number of Hypothesis Set samples usually is between 150 and 350. The

easiest way to manage this is to have each sample and everything associate with that sample on a single
row. Thus, each sample has its own unique row in the data base with many columns. For example, in Case
Study P, each sample studied had (not all tests were done for all samples). Sample identification (Sample ID,
drill hole, core type, core depth, original mass, etc.) 6 columns
• Lithology 3 columns
• Chemical Assays 21 columns
• Geotechnical logging 7 columns
• Downhole Well Log Geophysics 25 columns
• EQUOTip measurements 5 columns
• Hyperspectral Imaging 17 columns
• Point Load Index test 8 columns
• Comminution Index (Ci) 16 columns
• XRD 23 columns
• SMC test 8 columns
• Comminution Batch Grind Operating Work Index (OWi) 36 columns
• Batch Flotation tests (including chemical assays) 38 columns
Only some of these columns are to be used in analysis. The rest are there to support data QA/QC. Figure
184 shows a starting point in the relationship between experimental data acquired, the selection of what
data is selected for modelling and then the data outcomes, all in the data matrix structure. Also, links of
documented procedures on how these data analyses was done also needs to be present and linked to the
experimental design. The ultimate form of a geometallurgical matrix is a cross-correlation table populated
with selected attributes integrated into defined sample intervals.
The ultimate form of a Geometallurgical Matrix is a linked cross-correlation table
populated with selected attributes integrated into defined sample intervals
Acquisition
Data Processing
and reduction
QA/QC
Selection or
modelling of
parameters
Reference
Technical against
documentation documented
protocols
Parameters
accepted into matrix
Figure 184. Data acquisition, QA/QC, documentation and building the data matrix

9 MULTIVARIATE ANALYSIS OF ORE DEPOSIT DATA

Once the experimental data has been collected and assessed for its QA/QC reliability, further analysis can
be done. There are several stages.
• Statistical analysis of all parameters
• Cross correlation of all parameters
• Statistical data domain definition with Principle Component Analysis (PCA)
• Mineralogical signatures and patterns of each ore type understood in context of target process behaviour
• Ore type definition
• Propagation of engineering simulation (HSC) outcomes from Orientation Study (if possible)
• Propagation of error analysis across all stages of study
• Assessment of analytical outcomes in context of original geometallurgical objectives
Some of these procedures are substantial and complex works in their own right. As the data is processed,
the outcomes need to be added to the geometallurgy data matrix. This is a step that also needs to be thought
through and planned. What gets used and what for needs to be documented and a protocol developed
before starting.
Each of the measured parameters and trends should be subject to statistical analysis. Each parameter should
be subject to the following procedures (across the whole study, across domains, across ore types after they
are defined, across individual tests with multiple readings):
• Mean
• Median
• Mode
• Range
• Maximum
• Minimum
• Number of data points
Different
Sub-populations Mean
Population
spread
around mean
Skewness
tail
Figure 185. Box and Whisker Plot Figure 186. Probability Plot Figure 187. Histogram
(Image: Simon Michaux) (Image: Simon Michaux) (Image: Simon Michaux)

Each parameter should be plotted in

Compare each
In context of outcomes of the Domain Study

parameter against
target process
behaviour
• Mean Compare each T-test

• Median parameter against
Calculate the statistical • Mode parameters shown ANOVA tables
• Maximum to be relevant in
profile of each parameter
• Minium Domaining Study
in the data set • Standard Deviation Repeatability
• Variance with replica
• Number of points Compare each tests
parameter against
parameters shown
to be relevant in
Orientation Study
(7) Multivariate Analysis

of geometallurgical data
matrix
An iterative process to understand the
mathematical variability scope and
controls of the data set Plot each parameter against
Patterns of correlation,
every other parameter in a
Different sub-populations
cross correlation graph matrix
Determining what is relevant and
what is not in matrix data set Examine the probability plot of Differences in gradient,
each parameter Different sub-populations
Examine outcomes in context of

Examine the histogram plot of Shape of distribution,
Domaining Study & Orientation Study
each parameter Tight or spread
Mirrored patterns and

Determine interaction terms correlations, leading
for parameters if possible and lagging indicators
Figure 188. Multivariate Analysis of Geometallurgical Matrix data set

Each parameter should be plotted in cumulative plots, box and whisker plots and histograms. This provides
a strong foundation to base conclusions of the study. Figures 189 to 192 show data from Case Study P. In
this case study, impact breakage (Axb) was estimated with SMC tests and grinding was measured with a
Bond Ball work index (BMWi). In addition to this, the Comminution Index was measured on a much wider
sample range. In Figure 189, all parameters were compared to all others, (where the two algorithm outputs
of the Comminution Index were CRU.INDEX or Ci-CRU (related to crushing) and GRD.INDEX or Ci-GRD (related
to grinding)). The percentage mass of the -4.75mm and the -106 µm fractions are data routinely measured
as part of the Ci data QA/QC checks and are not used in the calculation of the Comminution Index. These
were included in the data analysis as they are also appropriately measured crushing responses of the
representative sample.

Figure 189. Compare all parameters against all others (Case Study P)
Figure 190. Probability plots of variables studied (Case Study P)


Figure 191. Histogram plots of variables studied (Case Study P)

The outcomes of the Domain Study and the Orientation Study are to be understood when conducting this
multivariate analysis. T-tests should be applied for each parameter between domains and ore types once
they are defined. If the experimental design has been developed appropriately, parameters and trends can
be placed in a series of ANOVA tables to test statistical significance and interaction terms. All of this is to be
added to the geometallurgy data matrix in the same form that the experimental data was structured.
Figure 192. Correlation between parameters in the geometallurgy data matrix (Case Study P)

9.1 Statistical data domain definition with Principle Component Analysis (PCA)
One of the methods used to partition a marge and varied data matrix into meaningful groups is the
application of a statistical technique called Principle Component Analysis (PCA).
PCA is a mature statistical technique that is widely used for finding patterns in data of multiple dimensions.
PCA finds a set of orthogonal dimensions, which account for all the variance in a particular dataset, by
reducing the dimensionality of a complex system of correlations into a smaller number of dimensions. The
first principal component (PC1) accounts for as much data variance as possible and each subsequent
principal component accounts for remaining data variance (PC2, PC3, PC4…).
This approach is to spatially model classes created during principal component analysis and attempt to
partition the data set into groups that describe the most variability (PC1, PC2, etc.). Each class has a given
distribution for each parameter, where the target processing response and variability is examined in each
block by association.
Principal Component Analysis (PCA) is a mature statistical technique that is widely used
dimensions. PCA finds
Principal Component Analysis of orthogonal
a set(PCA) is a maturedimensions, which account
statistical technique for all
that is widely thefor
used variance
finding
9.1.1 Principal component Method:
dimensionality of a complex system of correlations into a smaller number of dimensions.
dimensions. PCA finds a set of orthogonal dimensions, which account for all the variance in a parti
1. dimensionality
much of the
Subtract thedata
mean complex
of avariance
from eachas the of
system
ofpossible
data and eachinto
correlations
dimensions inaorder
subsequent to number
smaller a of
principal
create dimensions.
datacomponent The
set that has first p
accoun
much
PCA of the
expressesdata variance as possible and each subsequent principal component accounts for as m
a mean of zero. the data in such a way as to highlight similarities and differences within th
PCA expresses the data in such a way as to highlight similarities and differences within the data se
Principal Component
Principal Method:
Component Method:
2. Calculate the covariance matrix for the dataset. The covariance is a measure of how much
1.Subtract
1.each
Subtract the
dimension mean from each
fromeach
variesfrom
the mean the meanthethe
of of data
with data dimensions
dimensions
respect in order
in order
to each other create
to(i.e. the create
to relative
a data set data
aspread
that has
of meah
set athat
2.data in a dataset).
2.Calculate
Calculate the The formula
covariance
the covariance forfor
matrix
matrix theforcovariance
the the is: covariance
dataset.
dataset. The The covariance is a measure
is a measure of how
of how much eachm
mean respect totoeach
with respect
meanwith eachother
other(i.e.(i.e. relative
the the spread
relative spread a dataset).
of dataofindata in a dataset). Thefor
The formula th
form

n
(1) (nX (  )(Y  Y )
XXi  X )(Yi  Y )
(1)
cov( X , Y )  i 1 i i
Equation 21
cov( X , Y )  i 1
(n  1)
(n  1)
The value obtained from the covariance calculation is not as important as if the sign is positive o
fromvalue
The value obtainedThe the covariance from the covariance
obtained calculation calculation
is not as important is not
as if they asisimportant
sign positive or as if the sign is
negative.
- If the result is positive, it implies that a direct relationship exists between the two datasets.
• If the result
If the result is-positive, is positive,
it implies it implies
that a direct that
relationship a direct
exists relationship
between exists between the two dataset
the two datasets.
- If the sign is negative; the data sets are inversely related to each other.
• If the sign is negative: the data sets are inversely related to each other.
- -IfIfthe sign is negative; the data sets are inversely to each
relateddata other.are independent of e
• If the result of the thecovariance the covariance
result ofcalculation equals zero, the twoequals
calculation zero, the two
data dimensions dimensions
are independent of each other.
- If the result of the covariance calculation equals zero, the two data dimensions are indep
As there are multidimensional datasets, a covariance matrix is used. The covariance matrix is d
As there are multidimensional datasets, a covariance matric is used. The covariance matrix is defined as:
As there are (2)multidimensional covDimi , a
C nn  ci , j , ci , j datasets, j 
covariance
Dim matrix is used. The covariance m
(2) C nn  ci , j , ci , j  covDimi , Dim j  Equation 22
This equation states that if you have an n-dimensional dataset, then the matrix has n rows and c
matrix is the result of calculating the covariance between two separating dimensions.
This equation states thatequation
This states
if you have that if you have
an n-dimensional dataset, n-dimensional
an then the matrix has n rowsthen
dataset, and columns andhas n r
the matrix
matrix is the result of calculating the covariance between two separating dimensions.
each entry in the matrix is the result of calculating the covariance between two separating dimensions.
3. Calculate the eigenvectors and eigenvalues for the matrix:

Let A be a square matrix. A non-zero vector C is called an eigenvector of A if and only if there exists a number
(real or complex) λ such that AC = λC. If such a number λ exists, it is called a eigenvalue of A. The vector C
is called the eigenvector associated to the eigenvalue associated to the eigenvalue λ.
4. Select components and form a feature vector. A feature vector is an n-dimensional vector
of numerical features representing an object.
5. Derive new datasets by condensing multidimensional data into orthogonal dimensions (i.e.
Principal Components)
Describes the proportion of variance for each eigenvector

in this case the first two components account for around
60% of total variance.
List the variables used in the principal component analysis

in this case only six components represent a relatively
simple system to analyze
Shows the relative scores for each component, high

values have the strongest effect and positive versus
negative values indicate opposing discriminant trends
Actual weighting values used to calculate the principal

components. These values are based on training sets but
can be transferred into Excel and used to calculate
principal components for new ‘unknown’ samples
V1 -0.13685955*'%py'-0.25199805*'%cp'+0.087340186*'%bn'+0.58723749*'%cc'-0.0014709545*'U3O8 (ppm)'-0.63977939*'Au (ppm)'+1.2815472

V2 0.22465695*'%py'+0.039256248*'%cp'-0.36381083*'%bn'-0.51623515*'%cc'-0.0011000002*'U3O8 (ppm)'-1.0202435*'Au (ppm)'+0.74344882
Figure 193. Principle component methodology
PCA expresses the data in such a way as to highlight similarities and differences within the data set. It is well
suited to analysis of rock-based attributes using large datasets in geometallurgy.
It is to be remembered that this PCA based characterisation has been set up to group based on fundamental
multivariate controls of the whole data set, not on constraining the processing response. This can potentially
result in large variability within classes that may cover the entire distribution of processing performance
results.
It is for this reason that the PCA procedure is not the final analytical goal of a geometallurgical study. The
decision of what a Class (or process defined ore type) could or should be is to be guided by patterns found
in the target process behavior data with a secondary consideration for what the PCA tool returned and what
was learned in the domaining study. Figure 194 to 196 shows several PCA plot examples from Case Study P.

Primary Influence
• Ca, S, %-106, Cu/S
• Ci-GRD, Cu, K, Fe/S
• Au, Ci-GRD_Norm
• %-4.75, %+8mm
Secondary Influence
• Ag, Zn, Al, Fe, Pb
• Ci-CRU, Ci-CRU_Norm
• Spec_Power
Reduced Influence
• Mn, Sb, Mg/Al
• As, Hg
Figure 194. Principle Component Analysis of Geometallurgical Matrix data set

PCR3
PCR1
PCR2
PCR2
PCR1
PCR3
Figure 195. PCR1 vs. PCR2 vs. PCR3 (Case Study P)


Figure 196. Principle Component Analysis (PCA) plots (Case Study P)

10 PROCESS BEHAVIOR ORE TYPE CLASS DEFINITION

Geometallurgy is an iterative process where the experimental plan and the data analysis is constantly
reexamined to assess to ensure that the overall campaign objectives are achieved.
A new data set is outputted that is principle

component based. PC1 describes most variability.
PC2 describes most of the variability that is left, etc. Create cluster windows to
describe similarity in PCA
context Assess the
PCA Variability PCA Eigen Vector statistical profile
PCA Data matrix of each Class of
Matrix Significance Matrix
each parameter
Groupings and patterns
(mean, variability,
found in domaining study
etc.)
Query diagram plot against Reduce /change population of cluster Groupings and patterns
inputs in model windows to describe discrete found in multivariate
(PC1=Ax +By+Cz+...) boundaries to form classes of similarity analysis
Process behaviour
defined Class
Previous Guide final Class definition
work definition in context of
Plot PC1 vs PC2, PC1
Campaign target process behaviour
vs PC3, (etc)
goal patterns in data
Are the classes

appropriate?
Validate cluster windows against external If validation fails, repeat domain
data. (Chemical assays not in set, XRD, MLA, analysis, multivariate analysis, PCA
important to understand context of data) analysis (steps 6-8)
This is a QA/QC step.
Figure 197. Class Definition of Geometallurgical Matrix data set


Geometallurgy Data Matrix
Domain Multivariate &

Study
PCA Statistical Analysis
Process behaviour ore type Class definition

Done with the professional judgement of the
geometallurgical analyst team
Class 1 Class 2 Class 3 Class 4 Class 5

Mineral Profile Mineral Profile Mineral Profile Mineral Profile Mineral Profile
& Process & Process & Process & Process & Process
Behavior 1 Behavior 2 Behavior 3 Behavior 4 Behavior 5
Class 1 Class 2 Class 3 Class 4 Class 5

Process Behavior Process Behavior Process Behavior Process Behavior Process Behavior
Signatures Signatures Signatures Signatures Signatures
Characterisation Characterisation Characterisation Characterisation Characterisation

Signatures Signatures Signatures Signatures Signatures
Domain Domain Domain Domain Domain

boundary rules boundary rules boundary rules boundary rules boundary rules
Figure 198. Process behavior guided ore type class definition


Figure 199. Box and whisker plots of the Comminution Index CRU (crushing), GRD (grinding), crushing QA/QC steps
(Case Study P) (Image: Simon Michaux)
Figure 200. Box and whisker plots (Case Study P)





Figure 203. Scatter plots between parameters: Comminution Index Ci-CRU (Case Study P)
Figure 204. Scatter plots between parameters: Cu/S (Case Study P)


Figure 205. Scatter plots between parameters: Fe/S (Case Study P)

Figure 206. Probability plots for parameters (Case Study P)


Mg_ppm
Ca_ppm
10 90
10 90
20 80
20 80
30 70
30 70
40 60
40 60
50 50
50 50
60 40
60 40
70 30 70 30
80 20 80 20
90 10 90 10
10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90
Al_ppm Ca_ppm K_ppm Fe_ppm
CRU.INDX
CRU.INDX
10 90
10 90
20 80
20 80
30 70
30 70
40 60
40 60
50 50 50 50
60 40 60 40
70 30 70 30
80 20 80 20
90 10 90 10
10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90
S_pct Fe_pct Cu/S Mg/Al
Figure 207. Ternary plots for parameters (Case Study P)

Figure 208. Histograms for parameters (Case Study P)


Figure 209. Histograms for parameters (Case Study P)


Figure 210. Principle Components (PCA) plotted against parameters (Case Study P)

When the established classes are stable enough under scrutiny, the study can move to the next step. The
outcomes should be cross examined again in context of what was learned in the Domaining Study. Figure
211 below shows an example of Class domaining down hole according the boundary conditions decided
upon by the analyst team. This example only has two chemical assay profiles, when to assess the Class
boundary domains, eight chemical assays are needed.
60000 Blue Class

Black Class
50000 K Mg
Pink Class
40000
(ppm)
30000
20000 Green Class
10000
0
480 530 580 630 680 730 780
Down Hole Depth (m)
Figure 211. An example of Class domaining down hole in a single drill hole (Case Study P)
The process behaviour ore type definition step was done using chemical assays (in this Case Study P), not
using the collected hyperspectral characterization data. The hyperspectral characterization data was used
in the domaining study and also as a validation step in the Assessment of the Study step (Section 13).

10.1 Process modelling Class by Class

The comminution process behaviour of each class was studied, and a predictive model was developed.
Figure 212 shows the starting point for the planning of this step.
METAL PRODUCED
Installed PER kWh
Blast Power draw
Whole (kW)
Circuit
METAL PRODUCED
Primary
PER HOUR
Crusher
tph
SAG Mill
Metal Metal recovery rate
Ball Mill content in
ore
Gravity
Flotation Hydrometallurgy
Throughput modelling based separation
around a simplistic
concentrator design
For each class, predict
recovery for each target metal

(class by class)
Propagation
of error
Process In context or original
behavior geometallurgy expt goals
Are the classes Validation of models against measured
different from parameters
each other? Geological
behavior Predict process results of
samples not part of data used to
develop models (Hypothesis
Statistical Analysis Response Sample Set)
Figure 212. Step 9 - Process modelling class by class (Case Study P)

For each defined mineral association class, a model is developed to predict the target process. Class-based
modelling in is based on a combination of principal component analysis (PCA), exploratory data analysis and
class-based multiple regression.
Develop a working relationship between the target measured parameter (bankable test) and a proxy
estimates (proxy test measurement(s)), specific to that family of rock types in the target deposit. This would
become a proxy process model for that mineral association class. To make a proxy model, enough data
points are needed to provide data support for further analysis (Kojovic 1988).
 Approximately ratio of 4:1 data points to variables in model
 Acceptable level of R2 defined by where you are in the mining process
 Acceptable average residual error defined by where you are in the mining process
Individual class-based regression is designed to produce predictive process models (for example impact
breakage comminution). This is model would be based on at data parameter inputs considered from the
geometallurgical database used in PCA definition of that class. It would be very useful if the final class based

process model be based on multivariate mineralogy from assay. This would mean that acceptable models
could be extrapolated to many tens of thousands of assay data points. The ability to have this type of
predictive data at the same scale of assays is profound. This would be a major step-forward in supporting
spatial modelling and process domaining linked to resource optimization.
Define Problem Establish goals
Stage 1: Correlation Matrix
Select Response i.e. std error, R2,
Planning First Regressions
Suggest Variables model variability
Plot new variables

Check quality of data
Stage 2: Survey Experts Create New Models
Plot Data
Group Evaluation
Development Try Models
Are goals met?
Are Coefficients reasonable?

Are parameters Has model been
Stage 3: stable over Is Equation plausible?
Validated?
Verification sample space? Is Equation Usable?
Figure 213. The Model Building Process

Mathematical modelling is a process designed to achieve a method for predicting a response given a set of
input variables. In the context of mineral processing this usually involves adjusting model parameters or
coefficients to fit measured data.
The methods used in model building could be (but are not limited to) multiple linear regression analysis:
y = a1x1+a2x2+……anxn Equation 23
Where: y is the dependent variable

x1,x2,…,xn are the independent variables
a1,a2,…,an are the coefficients to be determine
Models have been fitted using nonlinear least squares method. A least squares solution (SSQ) is used to
obtain the initial approximation of the coefficients for the linear model.
 Minimize over a1,…,an:
 This equation gives the maximum likelihood estimates of the fitted parameters if the measurement errors are
independent and normally distributed with constant standard deviation and mean.
 The above quantity in equation 1 is referred to as X2 (chi-squared).
 The parameter can be used to obtain a quantitative measure for the goodness of fit for a model.
After a model has been determined which provides an acceptable fit to the data, before it is used some form
of validity should be made:

 Model validation is primarily concerned with determining that the models output behaviour has the accuracy required
for the models intended purpose.
 As a model is only an approximation of the real process, a difference between the model and system results (model error)
must be anticipated.
 Validation process deals with the evaluation of the magnitude of this model error.
 Validation Techniques.
 Face validity, comparison with other models, traces of model terms, historical data validation, predictive validation,
parameter variability, operational graphics.
A comment on model building:

“It is important to realize that model building is a creative process, where many factors such as
the model builder’s experience, preferences and ideas as well as the purpose for which the model
is to be used enter into selection of the form of the model.
Statistical procedures for assessing a model exist, but the model building process cannot
ultimately be completely reduced to a recipe based on several statistical tests.”
Toni Kojovic, 1988 (Kojovic 1988)
Shown below are examples of class based process models (crushing and grinding).
23
Hematite-pyrite Class BMWI Model
21
Model:
19
17
BMWI= -66.65 +1.24Flu0.61 -0.27Hem -1.49Py +19.23SG
Predicted
15
13
+8.39(Py/Sul)0.84 +0.72QHard1.22 +0.72(Chl/Sul)1.31
11
9
S.E.=0.90 R2=0.84 Model prediction relative error =7.4%
7
N=32
7 9 11 13 15 17 19 21 23
23 Measured
Feldspar Class BMWI Model
21
Model:
BMWI= -104.4 -0.17Ksp -0.88Qtz -0.12Sid -0.47Flu -0.43Sul
Predicted
19
-0.54(Ser/Ksp) +0.88(Sul/Hem) +8.56QHard

17

15
N=47
15 17 19 21 23
23
Measured
21 Hematite Class BMWI Model

19
Model:
Predicted
17
BMWI= -44.99 +0.67Ser0.56 +0.19Sid +0.13Flu 1.64 +0.63Bar0.81
15
13
+6.80SG +1.90QHard
11
N=36
9
9 11 13 15 17 19 21 23
Measured

70
Hematite-pyrite Class A*b Model Model:

60
A*b=177.1 +0.5Ser -1.05Sul +0.87Hem +0.05Py2.75 -44.66SG

Predicted
50
-2(Qtz/Sul) +12.36(Py/Qtz)1.65 -17.3(Py/Sul)1.55
40

30
N=32
30 40 50 60 70
70
Measured
Feldspar Class A*b Model
60
Model:
A*b= 24.99 +1.29Qtz0.8 -1.58Flu0.65 15.36Sul0.49 +4.06Hem0.61
Predicted
50
+2.74Cpy1.42 +7.89Bn0.49
40
N=47
30
30 40 50 60 70
90 Measured
Hematite Class A*b Model

80 Model:
70
A*b= -282-1.93Sid +0.01Bar1.69 +1.65Hem +194.3(1/Ser) -

Predicted
60
50
41.01SG +383.1(Qtz/Hem)0.09 +5124(Sul/Hem)3.62
40
+41.75((Sid+Ser)/Qtz)0.82 -82.97((Sul/Hem)/((Sid+Ser)/Qtz))
N=36
30
30 40 50 60 70 80 90
Measured
The viability of the proxy model has to the bankable test is then established. In an ideal practice, many
different proxy models developed in parallel to predict each measured/bankable process target. Once the
form and structure of the proxy model is established, the next step is to constrain it using the mill circuit to
identify the sensitivity required on measurements like BMWi and A*b. From this, identify the critical
parameters relating to that target process that should be collected for that ore deposit system.
The quality of the outcome has to be balance against the cost of tests required vs accuracy of model,
available budget and time pressure to complete the study.
Once the most efficient proxy model is established, it can then be used in conjunction with other process
target proxy models. Figure 190 below shows the outcome of two proxy models (Impact breakage for
crushing and BMWi for grinding) to study the comminution footprint in context or ore characterization.
A useful data analysis to conduct is determining what mineralogy controls comminution footprint. It is
understood that mineralogy controls most other mining processes. In comminution, it is the gangue
mineralogy that dominates the outcome. As such QXRD is the preferred mineralogical measurement tool.
Figure 191 below shows an example of this being done successfully.
10.1.1 Example Case Study P

Specifically, the impact breakage parameter Axb and grinding Bond Ball Work Index (BMWi). After the
process behaviour ore type Classes were defined, the boundary conditions were used to select samples to
be sent off for larger scale tests. The SMC test was selected to measure the Axb parameter. The Bond Ball
Index test was selected at the defined closing size (defined by automated mineralogy assessing target metal
mineral grain).

The drill core to source these samples was selected by the Class definitions boundary conditions. A series of
composite samples were selected from characterized drill core, where different tests were taken from the
same core intervals in a representative manner. The tests were completed and passed QA/QC test
standards. The data for each Class was isolated and modelling was able to be conducted. The parameters
looked at first were defined by the PCA and multivariate studies (similar to Figure 194). Influencers are
ranked in Table 12.
Table 12. Known influences from the completed tests in Case Study P (Simon Michaux)
PRIMARY INFLUENCE SECONDARY INFLUENCE

Potassium (K) Calcium (Ca)
Comminution Index impact breakage (Ci-CRU Sulfur (S)
Comminution Index impact breakage (Ci-GRD) Fe/S ratio
Magnesium (Mg) Lead (Pb)
Aluminium (Al) %4.75_PRE CRUSH
Mg/Al ratio Specific_power
Silver (Ag) %-106_CiCRUSH
Gold (Au)
Copper (Cu)
Cu/S ratio
All of these parameters were studied for correlation in a multivariate and non-linear regression specialist
modelling software. Through an iterative process, the most robust model was selected for each Class to
predict both Axb and BMWi.
Each Class was controlled and influenced by different parameters, or the same parameters but with different
model inputs.

Green Class - Axb Prediction Model

200
180
160
140
Predicted Axb
120
100 y = 0,978x
R² = 0,9082
80
60
40
20
0
0 50 100 150 200
Measured Axb
Figure 214. BMWi for Green class prediction, Case Study P (Simon Michaux)
Black Class - BMWi Prediction Model
20
Predicted BMWi (kwh/t)
18
16
y = 0,9966x
14 R² = 0,935
12
10
8
8 10 12 14 16 18 20
Measured BMWi (kwh/t)
Figure 215. BMWi for Black class prediction, Case Study P (Simon Michaux)
Figures 214 and 215 show class based BMWi grinding prediction models. These examples were very
successful with a high R2 (the best examples for Axb and BMWi from across the five classes). The full outcome
of this example is not shown in this report.
Normally, the correlation is not so strong. It is to be remembered that the number of parameters in the
developed model has to be limited by the number of degrees of freedom.

10.2 Engineering Simulation

Once the relevant parameters have been measured or estimated, they should be used to estimate process
performance. This can be done by using a simulation software (Figure 216 and 217) to model an existing
plant, or develop a model of a conceptual plant. In doing so, plant throughput, recovery and power draw
estimation can be modelled. This should be done for each Orientation sample if possible.
Figure 216. The HSC Thermochemical process simulation tool developed by Outotec
(Image and copyright. Outotec)
Figure 216 shows a model developed in the Outotec HSC software. This software conducts thermochemical
simulations of process behavior. It works well with flotation, hydrometallurgy and pyrometallurgy
applications. It can be used to estimate outcomes from bench scale test work inputs.
Figure 217 shows a model developed in JKSimMet, which is a, general-purpose computer software package
for the analysis and simulation of comminution and classification circuits in mineral processing operations.
This software was developed from the research outcomes at JKMRC from University of Queensland. This
software has been designed to work with a plant survey done on an existing process plant running at steady
state. The prediction models of this software are empirical and have the capacity to optimize plant
performance over a wide range of operation.

Figure 217. The JkSIMmet software

(Image and copyright. JKTech)
The outcomes of these kinds of simulations can be used to estimate the performance of the target process
operation. If the model inputs of these simulations was based on the Orientation Samples discussed
previously in this report, then it could be possible to predict the operational extremes the target operation
may undergo.
In doing so, a series of process paths around each orientation sample could be developed and the outcomes
ranked in context of effectiveness.
11 HYPOTHESIS SAMPLE STUDY

Once the Orientation Study is finished, a clear understanding of what small-scale tests, work in the target
deposit is developed and planning of the experimental design of Hypothesis samples is able to be done. The
Orientation Study examined a target process behavior in the 10 to 20 end member rock texture extreme
samples (tightly constrained mineralogy across the whole sample of drill core). The purpose of the Mapping
Hypothesis study is to map out the variability of the process behavior being studied.
A series of hypotheses can developed as a result of the work done previously. Where previous work was
conducted in an exploratory fashion, this work would be done in a validation of hypothesis fashion. For
example:

 This ore deposit responds best to a combination of leaching and flotation
 Gravity separation will remove gangue minerals X, Y and Z, thus improve recovery
 The prediction model where minerals A, B and C control comminution can be validated with testing samples of known
compositions of A, B and C, where the comminution result is compared to the predicted value.
Further sampling and test work (Hypothesis response and validation samples) is a smaller sample set 500-
1000m of drill core.
Hypothesis
Validation Study
Evaluation and
assessment
Orginal
Experimental Goal
Orientation Further
implications
Study
Domain
Study
Conclusions &
Hypothesis Recomendations
Study
Figure 218. Evaluation of hypothesis response against orientation and mapping studies
11.1 Hypothesis Sample experimental test work

A reduced number of tests are done on each Mapping Hypothesis sample, based on what was learned in the
Orientation Study. By the very nature of what is required in a geometallurgy mapping campaign, many data
points (sample points) are needed.
The Hypothesis Sample set is taken from the original 1500-2000m of drill core (representing 4-5 continuous
sections across the target deposit). Ideally, the Orientation Study sample set was also taken from these drill
core sections. The Hypothesis Sample study should be thought through, where the test work has been
statistically designed to allow the various stages of analysis the best possible chance for clear outcomes. The
best way to ensure this, is to select the input samples appropriately and to manage the experimental test

work to a high level of precision. This would allow for a lower value for propagation of error when examining
the final results. Things to consider:
 For each sample, an intelligent selection of tests that interrelate with final objectives
 Order of tests understood (non-destructive then destructive)
 Competent technicians who understand methodology and objectives
 Sample preservation for future work
Once the analysis of these samples is completed, examine the analytical outcomes against previous work to
determine if the geometallurgical campaign has been appropriately designed.
Predict process response Validate the developed Does the predicted

based on models developed process behaviour Ore type
in Step 9 Classes previously defined process results correlate
with measured values?
Return to mine site. Select a 2nd Phase of

samples from the core yard using the
guidelines developed so far. Run developed
experimental protocol
(10) Hypothesis Validation

Study (Reiterate Steps 4-9)
Use Hypothesis Response set to

answer any new queries raised in the
Evaluation and Assessment (Step 10)
Incorporate new domains of the Query why a class is Fill in gaps Finish experimental
deposit into the geomet analysis different to other classes in objectives that were
that were previously unavailable unusual cases
in sampling unable to be completed in
Hypothesis Sample Study
Figure 219. Hypothesis Validation of geometallurgical experimental aims

12 MAPPING STUDY PROCESS DOMAINING ACROSS DRILL CORE SECTIONS

One of the main objectives of a geometallurgy study is to determine zones of behavior for the target process
in context of how it appears across the deposit (Liipo et al 2019). Ideally, the process signatures developed
in Section 4 would then be projected onto continuous drill core is some form.
Once the Orientation Study is finished, a clear understanding of what small-scale tests, work in the target
deposit is developed and planning of the experimental design of Mapping Study samples is able to be done.

The Orientation Study examined a target process behavior(s) in the 10 to 20 end member rock texture
extreme samples (tightly constrained mineralogy across the whole sample of drill core). The purpose of the
Mapping Study is to map out the variability of the process behavior being studied.
Correlations and
associations of each Variability of gangue Process Domains
mineral minerals that control along continuous
(Multivariate Analysis) process separation drill core sections
What combination of
minerals control the
preferred process path?
(Orientation Study)
Small scale Diagnostic Comminution

flotation (JKMSI) Leach Index
CuSUM
Done on regular Hyperspectral Chemical

EQUOTip
intervals Image Analysis Assays
Continuous drill
core data
(11) Mapping Study –

domaining of process behavior
Establish a data collection procedure

to be used to map process domains
into the deposit to be done as a
routine production procedure
Figure 220. Mapping Study experimental design


12.1 Geometallurgy Mapping Study Experimental Design

The basic steps to be considered for a Geometallurgy Mapping Study could be as follows:
1. Use the minerals that control and influence the chosen process path from the Orientation Study, as
discussed in Section 7.
2. Examine each of the minerals in the chemical assay data base using statistical analysis. Determine the
correlations and associations of each mineral. Use Principle Component Analysis.
3. Examine domains and variability of each of the minerals along sections of continuous drill core.
4. If needed, conduct extra experimental tests using the proxy tests discussed in Section 5.
5. Examine the combination of minerals that control the preferred process path together in context of domains
and variability.
6. Establish a data collection procedure to be used to map process domains into the deposit to be done as a
routine production procedure.
12.2 Mineral signatures that control process response

A successful outcome of the Orientation Study is an understanding of the mineral signatures that control a
process response. All orientation samples can be examined to determine what mineralogy controls a given
process response. For all orientation samples, the very good recoveries of flotation (for example) ranging to
the very poor flotation recoveries can be ranked. Within this ranking, trends of what minerals (or
combination of minerals) also trend. In doing so the mineralogy that controls flotation can be determined.
As such, a combination of minerals (usually both penalty and gangue) can now be used to guide the mapping
study. In some cases, this has been done quite successfully, by using chemical assays done on drill core. It
can even be possible that the client is already collecting the relevant data, which could then be used in a
different context.
So the relationship between process path and minerals present could be mapped and modelled using
chemical assays. This is a case by case possibility of viability. As rock texture often controls recovery, this
method would not work in all situations.
12.3 Proxy test approach to mapping process domains

Establish proxy signatures that can be mapped with cheaper tests. Geometallurgy is the mapping of process
behaviour into the deposit in a spatially defined domain fashion. The Orientation Study was to establish
what mineral signatures controlled process behavior (Section 5). Often to achieve this, complex and
expensive test work on a small number of samples is involved.
Firstly, establish which minerals control the preferred process path. This should be an outcome of the
Orientation Study. Then establish if that process behavior can be mapped by a proxy test, or a proxy test in
conjunction with something else.

Process Path 1
Gravity
Sample 8
SGA
Orientation Step 1
Leach
Process Path 2
SLA
Gravity Representitive sample of Starting
Flotation Process Path 1 Characterization end member orientation sample.
Sample 8
Process Path 3 SGA
SFA
Orientation Step 1
Leach Sample SC1-4
Process Path 4 SLA
Separation
Ore Sorting Process Path 3 Flotation Characterization end member orientation sample.
Process Path 5
SOSA SFA (in 4 size fractions)
Analysis on what works and what does not Sample SC1-4
Magnetic
Process Path 4
Separation
Flotation
SGFB Ore Sorting
Process Path 5
Leach SOSA
Orientation Step 2
SFADLA
Magnetic Flotation
Process Path 8
Separation Flotation
Magnetic Leach SGFB
Separation SGFDLB
Leach
Orientation Step 2
Process Path 10 Leach SFADLA
Gravity Flotation
SGFDLB
Magnetic Flotation
Orientation Process PathGravity
8
Process Path 11 Ore Sorting Separation Flotation Leach
Magnetic Leach
Separation SGFDLB

Gravity Flotation
SGFDLB
Process Path 1
Gravity
SGA
Orientation Step 1
Leach
Process Path 2
SLA
Flotation Characterization
Process Path 3 end member orientation sample.
Sample SC1-4
Magnetic
Process Path 4
Sample 7
Separation
Gravity
Ore Sorting Process Path 1 SGA
Process Path 5
Orientation Step 1
SOSA
Leach
Process
not 2 SLA
Flotation Representitive sample of Starting
Process Path 6 Gravity Flotation Characterization end member orientation sample.
Process Path SGFB
3
Leach
Orientation Step 2
Process Path 7 Flotation Sample SC1-4
SFADLA Magnetic
Process Path 4
Separation
Magnetic Flotation
Process Path 8
Separation Ore Sorting
Process Path 5
Magnetic SOSA Leach
Separation SGFDLB
Gravity Flotation
SGFDLB Flotation
SGFB
Step 3 SOSBGC SOSGFC Leach SOSGFDLC
Orientation Step 2
SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Orientation
Gravity
Process Path 1 Separation SGFDLB
SGA
Study
Orientation Step 1
Leach Process Path 10 Leach

SLA SGFDLB
Flotation
Orientation Characterization
Process Path 3 Gravityorientation sample.
end member Flotation Leach
StepSFA
3 (in 4 size fractions)
SOSBGC SOSGFC SOSGFDLC
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Orientation Study
Process Path 5
SOSA

SampleOrientation
Gravity
•
Process Path 1
Flotation
Process Path 6 Gravity SGA

Study
SGFB
Orientation Step 1
Leach
Process PathLeach
2
Orientation Step 2
Sample 5
Process Path 7 Flotation SLA

SFADLA
Magnetic Process Path
Flotation Flotation Characterization end member orientation sample.
Process Path 8 3
Separation SFA (in 4 size fractions)
Magnetic Leach Sample SC1-4
Process Path 9 Flotation Magnetic
Separation Process Path 4 SGFDLB
Separation
recovery of each target metal?

Gravity FlotationOre Sorting
Process Path 5 SGFDLB
SOSA
Step 3 Analysis onSOSBGC
what works and what does not
SOSGFC SOSGFDLC
•
Flotation

SGFB
Leach
Orientation Step 2

SFADLA
Magnetic Flotation
Orientation
Process Path 8
Separation
Process Path 1
Gravity
SGA Magnetic Leach
Orientation Step 1

Leach Separation SGFDLB
Process Path 2
4
• recovery ofiseach
mosttarget metal?
SLA
Leach
Which process path effective in

Process Path 10 Gravity Representitive
Flotationsample of Starting
Flotation Characterization end member orientation sample.SGFDLB
Process Path 3
Orientation Gravity
Sample SC1-4 Flotation Leach
Magnetic Process Path 11 Ore Sorting
Process Path 4 Step 3 SOSBGC SOSGFC SOSGFDLC
Separation
Ore Sorting
Process Path 5
SOSA

Flotation
SGFB

Orientation
Leach
Orientation Step 2

SFADLA
Orientation
Gravity
Flotation SGA
Process Path 8
Orientation Step 1
Separation
Leach
•
Sample 4
Leach
Process Path 9
Separation
Flotation SLA SGFDLB
Which process path is most effective in

Process Path 10 Process Path 3 Leach
Gravity Flotation SFA SGFDLB (in 4 size fractions)
Orientation
Sample SC1-4
Gravity Magnetic Flotation Leach
Process Path 11 Ore Sorting Process Path 4
Step 3 SOSBGCSeparation SOSGFC SOSGFDLC
•
Ore Sorting
Process Path 5
What is recovery
the mineralogical signature
most that
SOSA
Process Path 1
Gravity
SGA
Flotation
SGFB
of the 2-3 valuable metals?

Orientation Step 1
Leach
Orientation Step 2
Leach Process Path 7 Flotation

3
Process Path 2 SFADLA

SLA
Flotation Magnetic
Characterization Flotation
Process Path 3 Process Path 8 end member orientation sample.
SFA Separation (in 4 size fractions)
Sample
Sample SC1-4
Magnetic Magnetic Leach
Process Path 4
Separation
Separation SGFDLB
Ore Sorting Leach
Process Path 5 Process Path 10
SOSA Gravity Flotation
SGFDLB
Process Path 6
Step 3
Gravity
Flotation SOSBGC SOSGFC SOSGFDLC
SGFB
•
Leach
Orientation Step 2
What is the mineralogical signature that

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Orientation
Separation SGFDLB
Process Path 1
Gravity
Process Path 10 SGA Leach
Gravity Flotation
Orientation Step 1
SGFDLB
Leach
Sample 3
Process Path 2

Process Path 11 Ore Sorting SLA
Step 3 SOSBGC SOSGFC SOSGFDLC Representitive sample of Starting
Process Path 3
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Process Path 5
SOSA
Orientation
Process Path 1
Gravity Analysis on what works and what does not
SGA
Orientation Step 1
Flotation
Leach Process Path 6 Gravity
2
Process Path 2 SGFB

SLA
Flotation Characterization Leach orientation sample.
Orientation Step 2
Process Path 3 Process Path 7 Flotation end member

SFA SFADLA
Sample
Sample SC1-4
Magnetic Magnetic Flotation
Process Path 4
Separation Process Path 8 Separation
Ore Sorting
Process Path 5 Magnetic Leach
SOSA Process Path 9 Flotation
Separation SGFDLB
Is thereMapping
a useful proxy Study
Flotation Gravity Flotation
test that relates

Process Path 6 Gravity SGFDLB
SGFB
11 Ore Sorting
Process PathLeach
Orientation Step 2

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Separation SGFDLB
to the process behaviour of the

Gravity Flotation
• Is
SGFDLB
Process a domains
Orientation
Step 3 SOSBGC Gravity SOSGFC SOSGFDLC
thereprocessuseful
Process Path 1
SGA
Orientation Step 1
Leach
Sample 2
Process Path 2
SLA
preferred path?
Process Path 3
Process Path 1
Gravity SFA (in 4 size fractions)
SGA
Sample SC1-4
Orientation Step 1
Magnetic
Orientation
Leach Process Path 4

Process Path 2
SLA
Separation
• Variability
Flotation Ore Sorting
Characterization
end member orientation sample.
Process Path 5
1
Process Path 3
SFA SOSA (in 4 size fractions)
Sample SC1-4
Magnetic Analysis on what works and what does not
Process Path 4
Separation
Mapping Study
Flotation
Sample
Ore Sorting Process Path 6 Gravity

Process Path 5 SGFB
SOSA
Leach
Orientation Step 2

Process not 7 Flotation
SFADLA
Flotation
SGFB Magnetic Flotation
Process Path 8
Separation
Leach
Orientation Step 2

SFADLA Magnetic Leach
Magnetic FlotationSeparation SGFDLB
Process Path 8
Separation
Gravity Flotation
• Process domains
Magnetic Leach SGFDLB
Separation SGFDLB
Process Path 10 Process Path 11 Ore Sorting Leach
Step 3
Gravity Flotation SOSBGC SOSGFC SOSGFDLC
SGFDLB
Process Path 1
Gravity
SGA
Orientation Step 1
Orientation
Leach
Process Path 2
SLA
• Variability
Sample 1
Process Path 3
Sample SC1-4
Magnetic
Process Path 4
Separation
Ore Sorting
Process Path 5
SOSA
Flotation
SGFB
Leach
Orientation Step 2

SFADLA
Magnetic Flotation
Process Path 8
Separation
Magnetic Leach
Separation SGFDLB
Can a combination (e.g.

Leach
Can the proxy test results be

SGFDLB
Can these minerals Orientation

Step 3
Gravity
SOSBGC
Flotation
SOSGFC
Leach
SOSGFDLC
be tracked directly, diagnosed by a combination Chemical Assays + QXRD) be

or indirectly? of routine measurements used or a statistical method
(like chemical assays) be used like PCA, CuSUM,
Chosen combination of
proxy tests & methods for
Mapping Study
Figure 221. Choice of which small scale proxy tests and methods to be used in the Mapping Study
Another approach is to develop a working relationship between an expensive and complex test and a proxy
test combination. In doing so, the signatures found in the more complex characterization tests could be
mapped in greater numbers, enough to conduct a domaining study.

To develop for example, a relationship between the rock texture mineralogical signatures observed in the
automated mineralogy tests and the less precise tests (e.g. QXRD and chemical assays) (Figure 222). In doing
so a larger number of cheaper and less complex tests can be applied in the Mapping Study.
Orientation Signatures OrientationMapping

Signatures
Signatures Mapping Signatures
• Process separation behavior • Process•separation
Chemicalbehavior
Assay • Chemical Assay
(all flotation tests done on a single

Signatures XRD + Chemical Assays
+ • Hyperspectral
+ • Hyperspectral
• Automated mineralogy • Automated• XRD
mineralogy • XRD
Orientation Sample)
• Chemical Assay • Chemical Assay
• Hyperspectral • Hyperspectral
• XRD • XRD
Define what mineralology controls Define what
Apply
mineralology
proxy tests controls
in 2m intervals Apply proxy tests in 2m intervals
process behavior, for each end processacross
behavior,
all geology
for eachstructures
end to across all geology structures to
memember rock texture memember define
rock texture
domains define domains
(Example, not real data)
Establish a link between complex EstablishApply
a linkproxy
between
testscomplex
in 2m intervals Apply proxy tests in 2m intervals
Signatures observed from QEMSCAN
tests and simpler proxy tests, for tests across
and simpler
all geology
proxy structures
tests, for to define across all geology structures to define
(all flotation tests done on a single
each end memember rock texture each end memember
variability ofrock
process
texture
behavior
Orientation Sample)
variability of process behavior
Figure 222. Develop a relationship between the mineralogical signatures observed in the expensive and time
consuming tests and the less precise tests that could be used as a proxy for this ore type family
This method may be required if metal recovery is heavily influenced by the rock texture, and that texture is
complex. It is to be considered if the process response can be modelled in this fashion at all. If the rock
texture defining process response can only be measured by automated mineralogy, then the concept of a
mapping study may not be viable.
The viability of the proxy model has to the bankable test is then established. In an ideal practice, many
different proxy models developed in parallel to predict each measured/bankable process target. Once the
form and structure of the proxy model is established, the next step is to constrain it using the mill circuit to
identify the sensitivity required on measurements like BMWi and A*b. From this, identify the critical
parameters relating to that target process that should be collected for that ore deposit system.
The quality of the outcome has to be balance against the cost of tests required vs accuracy of model,
available budget and time pressure to complete the study.

12.4 Use of Proxy Test Outcomes to Map Process Domains

Once the most efficient proxy model is established, it can then be used in conjunction with other process
target proxy models. Once those signatures have been measured and mapped in large enough numbers,
domains of process behavior can be examined. What is to be attempted is the grouping of different process
behaviors to a resolution that would change process engineering design, coupled with mineral character
content.
Figure 223 below shows the outcome of two proxy models (Impact breakage for crushing and BMWi for
grinding) to study the comminution footprint in context or ore characterization.
A*b and BMWi calculated using regressions

A*b from regression
Mineralogical Discriminant Diagram

PCA plot – PC1 vs. PC2
BMWI from regression
How does this change as What is the mineral character of

the samples get harder? these two sample cluster groups?
Figure 223. Class by class linked to comminution footprint (Aqqaluk ore deposit)
(Source: AMIRA P843 2009, Copyright: AMIRA)
A useful data analysis to conduct is determining what mineralogy controls comminution footprint. It is
understood that mineralogy controls most other mining processes. In comminution, it is the gangue
mineralogy that dominates the outcome. As such QXRD is the preferred bulk mineralogical measurement
tool. Figure 224 below shows an example of this being done successfully. The data can then be domained
into a comminution footprint (two separate examples in Figure 225 and 226).

Figure 224. QXRD Mineralogy Linked to Comminution Performance

Figure 225. Comminution footprint in throughput space at a mine site in the Southern Hemisphere

Figure 226. Comminution footprint domaining in conjunction with ore hardness derived throughput bottlenecks for a
mine site in the Northern Hemisphere (Source: AMIRA P843 2011, Copyright: AMIRA)
12.5 Domaining behaviour and quantifying variability

Domaining behaviour and quantifying variability is now examined. If the 1500-2000m of core that form the
source of Hypothesis Set were sampled correctly, not only the extremes should be mapped but several
continuous sections across all major structures should appear in this data matrix.
The sorts of parameters that can be used in continuous measure that is useful in domaining is:
• Core imaging to group similar rock textures
• Chemical Assays done as a routine by the client already
 valuable metals
 gangue mineralogy
 penalty elements
• Chemical Assays done as a new request as outcomes of the Orientation Study
 to track minerals that control process response
• Lithology domains and rock alteration domains from site geologist
• CuSUM statistical method applied to chemical assays and other parameters
• Imaging of core (if possible, and if can be analyzed by the client)
• QXRD to give mineralogical distribution
• XRF pellet to give elemental distribution
• Hyperspectral drill core characterization
• EQUOTip
• SMC (if done in enough samples)
• Small scale flotation test
• Small scale diagnostic leaching tesst

Once a decision has been made in how to approach the mapping in context of useful proxy tests, develop a
smaller scope testing regime to be done consistently on each Mapping Study sample. Run this set of small
scale cheaper tests on the rest of the samples (of the order of 100 to 500 measurements) to map variability
and domains. This is often called the Mapping Study (Figure 227).
Another critical decision to be made after careful consideration is the core length interval to be used for
each sample. Each Mapping Study test regime will require a sample mass to be effective. Ideally this should
be kept as small as possible to allow the targeting of genuine variability. Chemical assays are often done at
2m intervals (sometimes 4m). Usually the mass in these intervals is more than enough.
The mapping study can be approached in context of continuous measurement. That is, all chemical assay
intervals over a continuous section of core. Ideally this section of core would transverse major geological
boundaries and/or geological structures. Selection of these drill core sections should be done in context of
collaboration with a geologist that has worked with the deposit and has excellent knowledge of the deposits
unique characteristics.
It is also possible to have a selection of samples that are not continuous but are consistent in spatial
positions. For example 1 chemical assay interval in 10 intervals are sampled. This is a decision to be made
by the geometallurgical team in context of campaign objectives and project budget.
Examine Orientation
Study results Design final
Define required sample
experimental Sample preparation
mass and corresponding
flowsheet for (crushing & grinding)
Examine
ExamineMawson
existing data orginal core length
company
Mapping Study
in light data in light of
of Orientation
Orientation
StudyStudy results
results
Data QA/QC
Testwork done on each
protocols
Mapping Study sample
Sample representivity
maintained
Sample labelling and

data base structure
Figure 227. How the Orientation Study relates to the Mapping Study

Collect the data from the Mapping Study. Put the results with the context of, where each test is defined
spatially (drill hole + depth). Assemble all data in a downhole context for each sample. Enter the data into
the developed geometallurgical matrix (see Section 7 & 8).
Use cross correlation of different data types to define domains of process behavior. In doing so, process
defined domains can be diagnosed and process response variability can be quantified.
The objective here is to put all available data in a continuous downhole fashion in a cross-correlation form.
In doing so, domains of behaviour can be examined, although in this case, the target context for domaining
is the process behaviours. Comminution and flotation often have very different domains, because one being
physical in nature (comminution is driven by stress and strain applied to meso texture scale rock) and the
other is chemical in nature (flotation is driven by surface chemistry of micro scale particles). Thus, it is usual
to domain these process behaviors separately and accept that they may have different spatial signatures.
12.6 Chemical Assays and CuSUM statistical method

In terms of examining continuous downhole data, chemical assays are an excellent start as this data is
collected as routine across the entire deposit and will continue to be the case during operation (AMIRA P843
2009, AMIRA P843 2011). It is to be remembered though that chemical assays are not the same as
minerology, but a proxy for mineral content. Figure 228 shows an example of a continuous section of drill
core from Case Study P with several assays in a downhole format.
Figure 229 shows the same data as Figure 228 but with the CuSUM statistical tool applied to the same
chemical assays. As can be seen, the CuSUM tool is excellent and highlighting different domains from data
that seemingly has no structure. It is to be remembered, it is the gradient of the CuSUM line, where other
parts of the curve of the same gradient are very similar are considered to be the same statistical domain for
that parameter. This would be a parameter to study in context of the target process behavior. Different
minerals (and their proxy chemical assays) will influence different process behaviors. Comminution for
example is influenced by gangue mineralogy, where magnesium, aluminum, potassium and calcium are all
of note. Flotation is generally influenced by different minerals and at trace element concentrations.

Parts per million (ppm) Cu %, Cu/S
Fe %, S %
0
0,1
0,2
0,3
0,4
0,5
0,7
0,8
0,6
0
1
2
3
4
5
6
7
8
9
5000
0
10000
15000
20000
25000
30000
35000
40000
128 - 130 128 - 130
128 - 130
130 - 132 130 - 132
132 - 134 132 - 134
132 - 134
134 - 136 134 - 136
136 - 138 136 - 138
136 - 138
138 - 140 138 - 140
Cu/S
140 - 142 140 - 142
140 - 142
Cu_pct
142 - 144 142 - 144
K_ppm
Al_ppm
Ca_ppm
144 - 146
Mg_ppm
144 - 146 144 - 146
146 - 148
146 - 148
148 - 150 148 - 150 148 - 150
150 - 152 150 - 152
152 - 154 152 - 154 152 - 154
154 - 156 154 - 156
156 - 158 156 - 158 156 - 158
158 - 160 158 - 160
160 - 162 160 - 162 160 - 162
162 - 164 162 - 164
164 - 166 164 - 166 164 - 166
166 - 168 166 - 168
168 - 170 168 - 170 168 - 170
170 - 172 170 - 172
S_pct
172 - 174 172 - 174 172 - 174

Fe_pct
174 - 176 174 - 176

176 - 178 176 - 178 176 - 178
Depth (m)
Depth (m)
178 - 180 178 - 180
Depth (m)
180 - 182 180 - 182 180 - 182
182 - 184 182 - 184
184 - 186 184 - 186 184 - 186
186 - 188 186 - 188
188 - 190 188 - 190 188 - 190
190 - 192 190 - 192
192 - 194 192 - 194 192 - 194
194 - 196 194 - 196
196 - 198 196 - 198 196 - 198
198 - 200 198 - 200

200 - 202 200 - 202 200 - 202
202 - 204 202 - 204
204 - 206 204 - 206 204 - 206
206 - 208 206 - 208
Figure 228. Chemical Assay’s (Case Study P) (Simon Michaux)

208 - 210 208 - 210 208 - 210
210 - 212 210 - 212
212 - 214 212 - 214 212 - 214
214 - 216 214 - 216
216 - 218 216 - 218 216 - 218
218 - 220 218 - 220
220 - 222 220 - 222 220 - 222
222 - 224 222 - 224
224 - 226 224 - 226 224 - 226
226 - 228
200/237
226 - 228
228 - 230 228 - 230 228 - 230
230 - 232 230 - 232
232 - 234 232 - 234 232 - 234
-1,8
-1,6
-1,4
-1,2
-0,8
-0,6
-0,4
-0,2
0
0,2
-2
-1
-4,0
-3,0
-2,0
-1,0
0,0
2,0
3,0
4,0
5,0
1,0
-35000
-30000
-25000
-20000
-15000
-10000
5000
0
-5000
10000
15000
128 - 130 128 - 130 128 - 130
130 - 132 130 - 132
132 - 134 132 - 134 132 - 134
134 - 136 134 - 136
136 - 138 136 - 138 136 - 138
138 - 140 138 - 140
140 - 142 140 - 142 140 - 142
142 - 144 142 - 144
144 - 146 144 - 146 144 - 146
146 - 148 146 - 148
148 - 150 148 - 150 148 - 150
150 - 152 150 - 152
152 - 154 152 - 154 152 - 154
154 - 156 154 - 156

156 - 158 156 - 158 156 - 158
cusum Cu
158 - 160
cusum Cu/S
158 - 160
160 - 162 160 - 162 160 - 162
162 - 164 162 - 164
164 - 166 164 - 166 164 - 166
166 - 168 166 - 168
168 - 170 168 - 170 168 - 170
170 - 172 170 - 172
172 - 174 172 - 174 172 - 174
174 - 176 174 - 176
176 - 178 176 - 178 176 - 178
178 - 180 178 - 180
180 - 182 180 - 182 180 - 182
182 - 184 182 - 184
184 - 186 184 - 186 184 - 186
Depth (m)
Depth (m)
Depth (m)
186 - 188
cusum S
186 - 188
cusum Fe
188 - 190 188 - 190 188 - 190

190 - 192 190 - 192
cusum K
cusum Al
192 - 194
cusum Ca
192 - 194
cusum Mg
192 - 194
194 - 196 194 - 196
196 - 198 196 - 198 196 - 198
198 - 200 198 - 200
200 - 202 200 - 202 200 - 202
202 - 204 202 - 204
204 - 206 204 - 206 204 - 206
206 - 208 206 - 208

208 - 210 208 - 210 208 - 210
210 - 212 210 - 212
212 - 214 212 - 214 212 - 214
214 - 216 214 - 216
216 - 218 216 - 218 216 - 218
218 - 220 218 - 220
Figure 229. CuSum on Chemical Assay’s (Case Study P) (Simon Michaux)

220 - 222 220 - 222 220 - 222
222 - 224 222 - 224
224 - 226 224 - 226 224 - 226
226 - 228 226 - 228
228 - 230 228 - 230 228 - 230
230 - 232 230 - 232
232 - 234 232 - 234 232 - 234
201/237
0
-1
0,2
0,0
2,0
4,0
6,0
8,0
-1,4
-1,2
-0,8
-0,6
-0,4
-0,2
-8,0
-6,0
-4,0
-2,0
10000
20000
30000
40000
-20000
-10000
The objective here is to put all available data in a continuous downhole fashion in a cross-correlation form
(Figure 230). In doing so, domains of behavior can be examined, although in this case, the target context for
domaining is the process behaviors. Comminution and flotation often have very different domains, because
one being physical in nature (comminution is driven by stress and strain applied to meso texture scale rock)
and the other is chemical in nature (flotation is driven by surface chemistry of micro scale particles). Thus,
it is usual to domain these process behaviors separately and accept that they may have different spatial
signatures.
Recovery
Geological Chemical Assays Petrophysical EQUOTtip Comminution Flotation
Figure 230. Cross correlation of different data types to define domains of behavior
(Source: Amira 2008b, Copyright: AMIRA)

13 ASSESSMENT OF GEOMETALLURGY CAMPAIGN
“A good reality check can not only save your career, it can elevate it in the face of adversity”
S. Michaux – some time ago
Are the selected Ore Classes genuinely Were the samples selected to become
different from each other in context of original the Orientation Study samples, the true
experimental objectives stated in (0)? End Member rock textures?
Does data collected in this geomet study Should a different process behaviour
provide the relationships and true variability be considered to build upon what has
of key process parameters in context of been already done?
intrinsic geological attributes?
(12) Evaluate Results from

Steps 4-11 in context of
original aims Step 0
Are the relevant questions

being asked in this study in Is the methodology in the current
context of existing objectives study able to determine when different
as stated in Step (0)? geomet characteristics are being
observed for the first time?
Figure 231. Assessment of study

13.1 Assessment for potential sampling gaps in the study

Once the domaining Mapping Study has matured to the point where statistical structures in the deposit are
understood spatially, examine whether the process test samples are truly representative. The multivariate
statistical analysis, the PCA analysis and the domaining study should give a good understanding of the data
set. The Orientation Study also should show what mineralogy is important in controlling mineral separation
through the preferred process path.
This combination of data and knowledge has the capacity to assess if:
 Whether the samples taken for process separation test work are representative in the range for the relevant minerals.
 It is at this point where it can be asked whether the original end member texture Orientation Samples selected, genuinely
were the end member textures of the deposit.

Figures 232 and 233 show an example of this with Case Study P. As can be seen, samples collected for the
comminution test SMC do not have a representative range in the gangue mineralogy that later was shown
to influence breakage. The Case Study P sampling campaign was based on the requirement to collect
samples across the full range of valuable target metal (in this case Au and Cu). Gold at a grade in the range
of parts per million will have little influence on the breakage of meso-scale particles. In Case Study P, it was
later shown that the calcium content and the Fe/S ratio were important in defining different ore classes.
Full data set Samples with SMC
Figure 232. Gaps in sampling for comminution SMC tests in the calcium range (Case Study P)
Full data set Samples with SMC
Figure 233. Gaps in sampling for comminution SMC tests in the Fe/S (Case Study P)
If there are gaps in the sampling (only look at the controlling minerals that control/influence the preferred
process path), then draft up a mineralogical signature profile of what characteristics of the missing samples
might have. Then go back to the client and together with their geological staff, use this signature profile to
find a section of drill core and collect new samples. Make the judgment call whether only a small amount of
testwork is required, or would the new samples require the full spectrum of tests as done on the Orientation
Study.

13.2 Assessment on whether the study addressed the original experimental aims and hypotheses
At regular points during the geometallurgy campaign, do a reality check on what the data outcomes are
reporting in context of the original campaign objectives (see Section 2). The approach of honour what the
mineralogy is telling you is very important. A geomet study is often a trail blazing exercise into the unknown.
• Did it work?
• Was the original Orientation

Study samples really the process
extremes observed?
• Was the grind size used in the

study appropriate?
• What is missing when looking at

all data, across the whole range
of each parameter, in context of
original objectives?
• Does more sampling need to be

done?
• How do the different campaign

phases interrelate?
• Is the link between this

geometallurgy campaign and
conventional process modelling
viable and useful?
• Is the geomet matrix dataset able

to be audited in context QA/QC?
Figure 234. Finish of the geometallurgical campaign


14 POSSIBLE USES OF THE GEOMETALLURGY OUTCOMES

Domianing of AG/SAG mill behaviour
A geometallurgical campaign can be expensive and take time. Once this is done, however, the data and
Domaining of Ball mill behaviour

knowledge collected can be used as support for further tasks with great effect.
domaining at similar scale

14.1 Use of Geometallurgical Analytical Outcomes to Define Sampling Strategies for Full Scale Crushing behaviour
Grinding behaviour
Comminution footprint
Laboratory Testing for Process Design
The conventional approach to ore characterization was to collect a number of samples and send them off
to assays
for metallurgical testing. These samples were sometimes collected using the advice from the site geologists
and sometimes not. Most of the time, little or no mineralogy characterization was done on those exact
samples. Knowledge of how those samples can be related back to the ore deposit was often vague. Often
samples were a blend of several ore types to make up the required mass.
Part of the geometallurgy paradigm approach is the identification and selection of what is termed end
member orientation samples, which are to represent the rock texture extremes of the deposit in context of
a process response behaviour. If this is done correctly, they should represent the extremes to be observed
for a target process test. If the orientation study is done successfully, the mineralogy that controls process
performance would be understood.
Also part of the geometallurgy paradigm approach is the domaining of process behaviour. This, if done
successfully should show what parts of the deposit will respond to a given process separation or liberation.
The combination of these two concepts could be used to select larger scale samples for conventional
A*b and BMWi used to forecast plant throughput.
bankable tests (tests that are accepted by the mining industry in a due diligence context).
Ci GRD
Ci CRU
Key inputs into defining comminution domains.

Comparative comminution testing of individual
Can provide estimates of key ore parameters
Conventional
column leach Feed
tests material
Exploration
Geometallurgy much more
drill core
modelling reliable
Then measure Size Distribution
100
Selected core intervals e.g. 2m
ratio
Crush core sample
10
% recovery
@ 2.5:1 reduction
Size Fraction (mm)
rock type textures.

1
Time
CI=1.61
CI=3.17
CI=4.29
CI=5.88
Proxy
diagnostic
EH512_236
EH512_290
CE107_364
CE107_176
0.1
leach tests
Engineering
design
0.01
1.0
0.1
100.0
10.0
Percent Retained on Sieve (%)
Figure 235. Using a geometallurgy study to select feed samples for a bankable larger scale test campaign

The mineral signatures that control the process test in question could be used as a guide in sample selection
to ensure that the ore collect is of the same and only the same rock type texture. Sample dilution could
introduce variability without being quantified. In doing so, samples for bankable tests could be meaningfully
selected to establish what the true range of process performance would be.
The same ore type signatures could be used to create blends of samples. In doing so, all the test extremes
could be quantified as well as a series of blends could be tested in a controlled fashion.
14.2 Mine and Process Plant Design

Mining and extraction of ore deposits is becoming more complex with higher risk over time. The choice of
what process path is not as straight forward as it has been in the past. In some cases, there is multiple open
pits, and multiple process paths in the same operation, resulting in much higher CAPEX and OPEX. How to
manage the extraction of ore in a planned fashion for a complex operation like this requires a more
sophisticated feasibility study process.
To design and then operate such an operation, disciplined and sophisticated ore characterization is needed.
Much more has to be put into initial ore body knowledge and risk mitigation in design due to scale and
complexity. Measurement based efficiency of each individual process in context of a family of ore types is
required to support decision like which process path is more appropriate. Table 13 and Figure 236 show a
theoretical example in how this could be used to design parallel process paths and decide what parts of the
deposit to send to each plant. This is could be required in future, as opposed to a vague one recovery target
for the whole process circuit for all ore types that is rarely observed in practice.
ORE DOMAINS TO PROCESS PLANT ECONIMICS

To flotation
(NPV $φ, CAPEX $ϕ, OPEX $β)

Ore Domain 3 • Net revenue per quarter
time period
Ore Domain 4 • Metal produced total
• Metal produced per hour
over schedule
(NPV $A, CAPEX $B, OPEX $C)

To hydromet
Ore Domain 1 • Net revenue per quarter
leaching
time period
Ore Domain 2 • Metal produced total
• Metal produced per hour
over schedule
Figure 236. The choice of possible process path for different parts of the deposit – theoretical example linked to
Table 11. (Image: Simon Michaux)

Table 13. A geometallurgy study can domain the deposit in context of process performance
Process Attribute Ore Domain 1 Ore Domain 2 Ore Domain 3 Ore Domain 4
Ore Value
Valuable metal 1 (Au) grade 0.06g/t 1.1g/t
Valuable metal 2 (Cu) grade 1.20% 0.02%
Valuable metal 3 (Ag) grade 2.2g/t 1.3g/t
Valuable metal 4 (Mo) grade 0.96% 0.47%
Valuable metal X (?) grade
Penalty elements Yes (High As content) No No Yes (Low F content)
Process Attribute Ore Domain 1

Ore Characterisation
Ore Domain 2 Ore DomainOre4DomainOre
2 Ore Domain 3
Domain 7Ore Domain 4
Mineral Liberation Size 75 micron 5 micron 150 micron 212 micron
Clay and sheet silicate content High Medium Low Low
Ore sorting feasible? Yes No No No
Ore conditioning feasible? No No No Yes
Impact breakage (Axb) ranking 21 19 45 75
Low energy abrasion energy (ta) 0.35 0.41 1.6 2.1
Bed breakage propensity (outside SAG comfort zone) Yes Yes No No
Grinding energy (BMWi) ranking 22 20 13 11
Fine grinding needed? No Yes No No

Separation Process
Ore Domain 2 Ore Domain Ore
4 Domain
Ore 2 Ore Domain 3
Domain 7 Ore Domain 4
Flotation recovery Poor Average Good Good

Flotation kinetics Poor Average Good Good
Leaching recovery Average Poor Good Average
Leaching kinetics Average Poor Good Average
Gravity Au recovery Yes No
Pressure oxidisation required for Au recovery No Yes

Engineering Design Request
Ore Domain 4
Ore Domain 1
Ore Domain
Ore Domain 2 Ore Domain 3
7 Ore Domain 4
SAG Mill Yes Yes

HPGR Yes Yes
Ball Mill Yes Yes Yes Yes
Fine Grinding Mill Yes
Flotation Circuit Yes
Yes Yes Yes
Leach Dump Yes
Leach Heap
Leach Tank Yes
CiL Carbon Leaching circuit Yes Yes
Process Attribute Ore Domain 2 OreOreDomain

Domain 1 4Ore Domain
Ore 2 Domain
Ore Domain 3 7 Ore Domain 4
Environmental impact
Acid Mine Drainage Potential High Medium Low High
Dust Generation Potential Low High Low Medium
Water Pollution Potential High Medium Medium High
Impact of Flora & Fauna Potential Low High Low Low
Mine Site Rehabilitation Requirements High Medium Low High
Choices like what are shown in Table 13 and Figure 236 can be used for design. There are several options.
Ore can be mined into multiple stockpiles, each feeding an optimized process path with a unique
comminution closing size and cutoff grade (Figure 237).

Flotation
Geomet
Block Model
Sorting
Blast Flash
Flotation
Tank
Leach
Dump
Leach
Pad
Flotation
Figure
Each 237. Multiple
stream with itsstock piles,closing
own feeding multiple
size process plants grade
and cutoff
Conventional process plants are designed to operate at a steady state throughput to a predicted average
recovery. Technology has advanced greatly in the last few decades, which now allows a more sophisticated
complexity in engineering. It is now conceptually possible to engineer a process path that is designed to
operate to a dynamic throughput conditions. Variable throughput that has the capacity to keep ore for a
longer residence time in the process plant, or for a shorter time (high grade softer ore should be kept in the
process plant for as long as possible to increase recovery, low grade harder ore should be pushed through
as fast as possible). This is only possible with more sophisticated ore knowledge, where the process response
of different ores is known. What is also required to be known is what ore is entering the plant from the
stockpile (using ore tracking technology). Operational protocol needs to be developed accordingly.
Figure 238. One process path that is heavily instrumented and optimized for a dynamic feed
(Image Copyright: JKMRC)
Sorting technology can be engineered but is difficult to justify economically. A technological breakthrough
is considered to be likely. If this does happen, then sorting could be used to greatly reduce what volume of

ore needs to be sent to grinding (the highest mining cost), for a similar metal production to all the ore being
processed (Figure 239). This will only be effective if ore sorting response is understood.
Waste dump
Blast
Flotation
Dump
Leach
Pad
Sorting Sorting
Sorting
Working Ore
Geomet
Block Model
Future Ore Problem ore with

‘show stoppers’
Flotation
Figure 239. Sorting technology being integrated into process design

Only a fraction of the ore volume goes
(Image: Simon to ball mill for similar recovery
Michaux)
14.3 Geometallurgical Block Model Generation

The development of a block model with geometallurgical attributes has been the ultimate objective to many
geometallurgical research and developments. Some of the outcomes of the Geometallurgical Mapping Study
could be incorporated into a block model (Figure 240).
Figure 240. Classes and domains integrated into a block model


There is a very difficult challenge to this idea however. For the geostatistical kriging to be successful, the
attributes modelled are required to be ‘additive’. This means that experimental measurements of an
attribute could be averaged in a spatially appropriate fashion to predict what that attribute might be at a
theoretical point between them.
Most metallurgical attributes are not additive, due to how they are calculated. For example the BMWi and
the A*b impact breakage parameter have been shown to be probably not additive. Figure 241 shows an
example where the Comminution Index done continuously on a 161m section of drill core. This example is
done where the measured particle size fractions are summed together for several interval sections (2m, 4m
and 8m).
3,4
2,9
Comminution Index Ci - CRU
2,4
2m intervals
4m intervals
1,9
8m intervals
6 per. Mov. Avg. (2m intervals)
1,4
1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153 161 169
Continuous Down Hole Depth (m)
Figure 241. Comminution Index done continuously on a drill core section, calculated on several interval lengths, from
an IOCG style Au and Cu deposit in the Southern Hemisphere
This suggests data used to calculate the Comminution Index may well be additive when using size fraction
data summed together. This implies that while process test results like BMWi and A*b are not additive, the
supporting size fractions could be.

Current state of knowledge of how to build a genuinely additive block model with process attributes is not
developed enough to be considered reliable enough to use. Future research and developed may resolve
these difficulties.
14.4 Economic Modelling and Business Model Development

The true value of this kind of integrated data is how to develop the business model. To do this, the
integration of these diverse attributes into more effective mine planning and optimization models is
embodied in the cross-discipline area of geometallurgy. Figure 242 shows the tasks that are needed to be
done to develop an economically viable operation.
Open Pit
Optimisation
Mining Phase Required Capital
Optimisation CAPEX
Integrated &
Mine Schedule Simultaneous
Optimisation
Cut Grade(s) Process Plant

Analysis Calibration
Stockpile Infrastructure
Management Logistics
Ore Blending
Figure 242. Enterprise Optimization (Whittle Consulting)

In the past these tasks were done separately and insolation. As technology has improved, it has become
possible to optimize all these tasks together at the same time (Whittle Consulting
https://www.whittleconsulting.com.au/). The tasks in Figure 225 are all able to happen with data inputs. A
geometallurgy campaign would be able to support each task with more sophisticated information. Just so,
the operation planning would be more sophisticated, where the plan would resemble reality more closely
(avoiding some of the problems discussed in Section 1).
The difference between a conventional assessment and a geometallurgical styled assessment can be seen in
Figure 226. As can be seen, the optimized case has variability embedded in it as a function of time. This
means that tasks like plant expansions, pit cutbacks and maintenance can be planned, with the foresight of
when performance could be poor and when it could be productive.

14.5 Corporate problem solving with the Theory of Constraints

There are operational opportunities using geomet data. With the capability to map this ore type process
behavior into the deposit (here previously, only grade and geology domains were mapped). As extraction
progresses through the deposit, it is possible to examine operational challenges. Data can be used to support
the study of production bottlenecks across the operation. These bottlenecks will move along the logistics
chain over the life of the mine. The methodology to do this would be to use the options analysis and theory
of operational constraints (TOC) (Rochman 2003).
The Theory of Constraints is a methodology for identifying the most important limiting factor (i.e. constraint)
that stands in the way of achieving a goal (usually generating maximum revenue) and then systematically
improving that constraint until it is no longer the limiting factor (Goldratt & Cox 1993). The goal for mining
may be maximum metal produced, or maximum recovery.
In manufacturing, the constraint is often referred to as a bottleneck in production parts. In a mining, the
constraint could be something like ball mill throughput, or rate of tonnes produced ROM. The Theory of
Constraints (TOC) takes a scientific approach to improvement. It hypothesizes that every complex system,
including manufacturing processes, consists of multiple linked activities, one of which acts as a constraint
upon the entire system (i.e. the constraint activity is the “weakest link in the chain”).
The Theory of Constraints provides a specific methodology for identifying and eliminating constraints,
referred to as the Five Focusing Steps. As shown in the following in Figure 243 and Table 14, it is considered
a cyclical process.
5 - Repeat the 1 - Identify the

process Constraint
What is the
goal?
4 - Elevate the
2 - Exploit the
Performance of the
Constraint
Constraint
s
3 - Subordinate &
Synchronize to Constraint
Figure 243. The Five Focusing Steps (a methodology for identifying and eliminating constraints) (Goldratt & Cox
1993) (Image: Simon Michaux)

Table 14: The Five Focusing Steps (a methodology for identifying and eliminating constraints) (Goldratt & Cox 1993)
Step Objective
Identify the current constraint (the single part of the process that limits the
1 Identify
rate at which the goal is achieved).
Make quick improvements to the throughput of the constraint using existing

2 Exploit
resources (i.e. make the most of what you have).
Review all other activities in the process to ensure that they are aligned with
3 Subordinate
and truly support the needs of the constraint.
If the constraint still exists (i.e. it has not moved), consider what further
actions can be taken to eliminate it from being the constraint. Normally,
4 Elevate actions are continued at this step until the constraint has been “broken”
(until it has moved somewhere else). In some cases, capital investment may
be required.
The Five Focusing Steps are a continuous improvement cycle. Therefore,

once a constraint is resolved the next constraint should immediately be
addressed. This step is a reminder to never become complacent –
5 Repeat aggressively improve the current constraint…and then immediately move on
to the next constraint. If, in the previous steps, a constraint has been
broken, go back to Step 1. But do not allow inertia to cause a system
constraint
The TOC methodology (Goldratt & Cox 1993) introduced a staged logical thinking process to be used in
conjunction with the five focusing steps. The thinking process assists with working through the change
process by identifying the following:
• What to change
• What to change to, and
• How to effect that change
The thinking processes consist of logic tools used to identify problems, then the development and
implementation of solutions. These tools include effect-cause-effect (ECE) diagramming (also known as
Cause and Effect Diagrams, Fishbone Diagrams, Ishikawa Diagrams, Herringbone Diagrams, and Fishikawa
Diagrams) and its components:
 negative branch reservations
 the current reality tree
 the future reality tree
 the prerequisite tree
 the transition tree
 the evaporating cloud
 the negative branch reservation, and
 the ECE audit process.

These tools allow any organization to analyze and to verbalize cause and effect around any issue. In doing
so, a complex and/or serious problem can be unravelled and understood at its fundamentals. This
methodology assists in the search is for the possible cause / causes or basic reason with regard to a specific
process, outcome, problem or situation, and makes relationships and interactions between different causes
visible with regard to specific process. It also allows for the process of analyzing existing problems with the
aim of initiating corrective measures and identifying areas where additional activities (improvement
processes) can be started.
To conduct this procedure, the target industrial operation needs to be mapped out in a systems context,
where all inputs, throughputs, outputs, actions and decisions are accounted for to produce a saleable
product. Usually, a combination of different systems architecture are used and merged, similar to what is
shown in Figure 244.
LINEAR CHAIN PYRAMID HERINGBONE
V SYSTEM T SYSTEM
SYSTEM SYSTEM SYSTEM
Output- Output- Output- Output- Output-
Product Outcome Products Outcome Product Outcome Products Outcome Product Outcome
Input- Start Input- Start Input- Start Input- Start Input- Start
Figure 244. Different kinds of systems architecture

This will help diagnose bottlenecks and strategically important parts of the process. Then each bottle neck
can be examined where different choices can be made. Each different choice will have a different outcome
in context of the goal (whether it be revenue, or metal produced). The Net Present Value (NPV) tool can be
used to account for the time needed to make this change.
Net Present Value is the difference amount between discounted sums: cash inflows and cash outflows. It
compares the present value of money today to the present value of money in the future at a discounted rate
to account for inflation and returns. This is a fundamental tool in Discounted Cash Flow analysis (DCF).
Formula 24 below shows how NPV is calculated.
𝑇
𝐸{𝐶𝐹𝑡 }
𝑁𝑃𝑉 = ∑ 𝑡
− 𝐶𝐴𝑃𝐸𝑋 Equation 24
𝑡=0 (1+𝑅)

The NPV calculation will increasingly discount the value contribution of costs/incomes with each addition
time period from the present. This means that the NPV calculation will only account for a time period of 5-
8 years, after which cash flows are disregarded. Formula 25 below shows a general rule of thumb in how to
quickly estimate NPV (only to be used appropriately).
NPV = (1/5 CAPEX) + OPEX for 1 year Equation 25
All changes to the system can then be ranked in terms of effectiveness. A theoretical example is shown in
Figure 245.
Option 1A - ###
Cash
Option 1 - ###
TOC Flow Operation
Option 1B - ###
Option Required Upfront Year 1 NPV @ 20%
Option 2 - ### new CAPEX Revenue Rank
The Goal = Option 1A 104 180 46M 1
maximise the Option 1B 150 160 43M 2
Option 3 - ###
production of metal
Option 2 24 45 13M 5
per hour Option 4A - ###
Option 3 11 29 6.6M 5
Option 4 - ###
Option 4A 35 50 14M 4
Option 4B - ###
Option 4B 12 45 15M 3
Option 5 - ### Option 5 19 33 6.5M 7
Figure 245. Options analysis & theory of constraints decision tree – theoretical example
An example of TOC could be the choice for a mine operation to transport concentrate in trucks or build a
pipeline to the port. The development of decision making could resemble:
 For the Life of Mine (LOM):

o Pipe is cheaper but is a bottle neck
o Trucking is more costly, may not be a bottle neck
o Pipeline and trucking is an option
 Trucks are very flexible (an optimization value add) allowing cash to be moved up-front, improving NPV
 Optimize whole project for 3 options: pipe, truck, pipe and truck
 Choose the option with the best NPV, progress outcome to next study phase
Ore mineral characteristics data would form only part of the input data required to do a TOC analysis.
Traditionally, a very simplistic input was used in conjunction with data inputs from other parts of the
operation like power consumption, transport time and handling logistical metrics.
The value of the concentrate (market price * grade * recovery – cost of production – penalty element) has
the capacity to change this sequence. This value would be highly variable and change for different parts of
the deposit. A geometallurgy data set used effectively in TOC, could change the outcome of the analysis.

14.6 Corporate decision making for the executive board

The executive board often make decisively important decisions with poor data quality available. Have to rely
on past experience and intuition. There is now a need to make defendable decisions now that the risk is so
much higher and the time frame for things to happen is much longer. The market is now much more volatile
and unforgiving (in 2019). Also, due to the time frame projects now have, there are multiple generations of
executive board members, all of which need better data support at handover. Having quality data in place to
do this requires the work to be done ahead of time. Questions an executive board may need to answer and
then later defend their decisions could be:
 Should we authorize the start of this project?
 Or should we delay it until market conditions are better?
 And considering debt obligations of company
 Can we afford site rehabilitation?
 When will mine start producing a saleable product?
 Should this project have an open pit or should it be an underground block cave?
 Should this project use:
o a HPGR or a SAG mill?
o Should recovery be flotation circuit or a leaching circuit, or a hybrid of both?
o Should final metal production use pyrometallurgy or hydrometallurgy?
 When should we do the plant expansions and pit cutbacks?
Risk could be mitigated by:
 At the feasibility stage(s), CAPEX closely resembles the final commissioning reality.
 Once operating, the deposit is comprehensively characterized which allows flexible changes in the mine schedule and
process response can be accurately predicted.
 Where decisions on process plant expansion, open pit cutback expansions and maintenance shutdowns can all be planned
in context of risk uncertainty with more precision.
14.7 Geometallurgy to support due diligence

The financing of mining operations is evolving quickly. The CAPEX for a typical copper mine is now of the
order of 4 to 5 billion $USD. Shareholders and venture capitalist often put up approximately 30% of the
required capital. The remaining 70% often is supplied by a banking hedge fund. Once a mining corporation
proposes a project, due diligence is required to audit the feasibility study before capital is available. The
same due diligence procedures would be needed state governments to authorize a mining license.
Geometallurgy can facilitate more sophisticated Due Diligence methodology
 What is the real NPV?
 What is the real CAPEX?
 How long will commissioning take?
 When will CAPEX be paid off?
 What is the real OPEX?
 What is the real rate of metal production?
 What is the real IRR for revenue flow?
 What are the macro scale steps of the mine Life of Mine?
 What is the probably mine operation waste plume?
 What is the mine operation environmental impact?
 What are the requirements for mine site environmental rehabilitation?
 Should the investment hedge fund support this project?

Investors &
Business Model & Revenue Flow
State Government & Regulation

Exploration
Engineering financial
& Science backers Conceptual
Study
Geometallurgy Pre-Feasibility
Study
Design & Mining
Business Model Corporation Feasibility
Study
Design &
Construction
Mine Site Site

Operation Rehabilitation
Figure 246. Geometallurgy supported due diligence

15 DEVELOPMENT OF PRODUCTION PROTOCOL

Traditionally, once a mine operation has been designed, constructed and then commissioned, there is a
handover to the operating staff. Often, very little information on why choices in engineering design were
made is transferred to the operating staff. This handover point is often the subject of corporate amnesia.
This means that the implications of the process engineering design in the plant interacting with the ore
deposit have to be discovered the hard way.
Advances in engineering (information technology, instrumentation and automation in particular) offer a
unique opportunity to resolve this. The comprehensive information and knowledge a geometallurgical
campaign that has been effectively used in design can be used in an operational context.
As ore is excavated, it can be tracked (using ore tracking technology) as it progresses to a stock pile, then
later into the process concentrator plant.
For example, a mine site will need to operate multiple working faces, each of which will have logistical
challenges on a day to day basis. If multiple stockpiles have been developed as part of the mine plan, then
different ore types can be put into appropriate stockpiles.

All brought onto a level playing field, managed

with the same metrics of efficiency Refined Process Path
Stockpile A Optimised for A
Working Face 1
Working Face 2 Refined Process Path

Geometallurgy Stockpile B Optimised for B
Characterisation
Working Face 3
Working Face 4 Refined Process Path
Stockpile C Optimised for C
Blending Recommendations
Disposal Process Path
Penalty Elements X Optimised for X
Logistical Recommendations
Paradigm: Each stockpile is of ore types that will be most

efficiently/effectively processed by the process plant
Figure 247. Multiple stockpiles

For this to work, a coordinated and heavily interconnected optimized mining and processing operation has
to be designed and then operated, where ore body knowledge and a real time operations center all need to
be integrated. This can only work if ore body knowledge is at a degree of sophistication, where each parcel
of ore is known in context of its process performance and true value.
The outcomes of a geometallurgy study can be used to develop a routine protocol at a production site. The
Orientation Study, if done well, will show process behavior for the rock texture extremes of the deposit being
mined. Ideally the process path being used in the operating mine is similar to one of the process paths
examined in the Orientation Study, and the mineralogy that controls the effectiveness of that process path
has been diagnosed.
Those mineralogy groupings that control process behavior across all the Orientation Samples will have
measurement signatures that could be captured with online instrumentation in the process plant.
Something as simple as an online XRF measurement tool or a more sophisticated Raman Spectroscopy
instrument placed at the right place in the plant has the capacity to diagnose which combinations of the rock
textures so far examined is passing into the process circuit. This in turn could predict ahead of time what
the process response could be, through understanding the outcomes of the geometallurgy Orientation
Study.

Geometallurgy
Orientation Study
Production Scale Process
Orientation
Sample 1
Circuit Operation
Orientation Behavior 1 Behavior A

Sample 2
Behavior 2 Behavior B
Orientation
Sample 3
Behavior 3 Behavior C
Orientation
Sample 4
Behavior 4 Behavior D
Orientation
Sample 5
Behavior 5 Behavior E
Geometallurgical
Orientation operational protocol
Sample 6
Mineral signatures Instrumentation measurement
that control process in plant that shows different
behavior process behavior metrics
Figure 248. Geometallurgy during process operation

To do this, a link between the process behavior of the Orientation Samples and some kind of online mineral
measurement in the process plant to diagnose which combination of rock textures are being processed,
needs to be established. This would be part of a geometallurgical operational protocol that is required to be
developed, based on the outcomes of work done.
Regardless of the level of development a mine site has, a routine operating protocol is needed to guide data
collection and reconciliation. Based on what has been learned from the Orientation Study and the Mapping
Study, the following could be considered:
 What mineral signatures control process behavior?

 What data should be collected routinely (that are practical in a production environment)?
 How is that data to be used by site personnel?
 How should the process plant operation evolve over time to keep pace with the mine schedule?

16 CONCLUSION AND SUMMARY

The geometallurgy process can be complex but it has 6 broad brush steps. A more complex demonstration
of this can be seen in Figures 249-251.
1. The geologically informed selection of a number of ore samples
2. Laboratory-scale test work to determine the ore's response to mineral processing unit operations
3. Determination of what mineralogy parameters control each process path, then a judgment on what is the most effective
process path
4. Domaining and mapping of variability of the chosen process path
5. The propagation and distribution of these parameters throughout the orebody using an accepted geostatistical technique
(or in more practical terms a chemical assay signature).
6. The application of a mining sequence plan and mineral processing models to generate a prediction of the process plant
behavior.
The purpose of this report was to show the paradigm behind geometallurgy. In practice, only a part of this
procedure would be done, as defined by the original goals of the campaign.

MLA/QEMSCAN
Mineralogy Thin section Comminution throughput
Mineralogy objectives
Geotechnical
UCS, PLT, Young's Mod, Chemical Batch float Plant Design
Poisson’s ratio, etc Assays test work
Liberation profiles &

Grind size targets
Geophysics Geochemistry
Site personnel Pilot plant trials
Alteration Horizons knowledge
(down hole logging)
Comminution Tests
Reports, databases, (A*b, Bond,
Lithological stratigraphic spreadsheets
systems
(down hole logging)
Settle on an objective that is Develop significant predictive relationships and true Geological Context Metallurgical Context
achievable with clear metrics variability in processing performance response by
characterising intrinsic geological attributes.
of success
Understanding of Understanding of
Deposit system in context of Deposit system in context of
existing knowledge Geometallurgy
(0) Geometallurgical (1) Ore(1)(1)Body Analysis of
Experimental & (0) Geometalurgical
Analyitcal
Experimental &
Aims Goals OreOre
Body
of Existing
Analysis
Body Initial
Knowledge
Analysis
Analyitcal Goals
Existing Knowledge
Time To be done by personnel with geometallurgical training
What stage of Restraints? What the
development site/deposit
Is the deposit? Budget specific issues
Restraints?
Greenfields Feasibility Brownfields Grind Recovery Oxidisation

Exploration Study Operation Sensitive? Sensitive? of Ore?
Geomet Size by size

XRF/XRD Geomet Assays XRF/XRD XRF/XRD
Geotech, Geophysics, Metallurgical tests, Thin section and Assays Geomet assays
Chemical Assays
Geochemistry, etc comminution & float MLA/Qemscan
mineralogy FEM MLA Optical FEM MLA Optical FEM MLA Optical
Continuous down Spatially isolated Texture Ball Mill

convergence batch SMC Bond
hole logging data data Batch Micro- float Point
sampling grind JKCi A*b Test
Grid sampling Floats (JKMSi) Load
How much of exiting data is

usable and passes QA/QC? 99% of usable data is Texture Analysis Flotation Geotech/blasting Comminution
Understanding what has been
Chemical assays
done across the different
generations of work
QA/QC protocol on Select 10-20 (?) samples with enough mass, that Destructive Tests
everything describe the end members of the texture extremes
Integration and preliminary
(2) Creation(2) Creation
of foundation
of foundation
geomet (3) (3)(3)
Geometallurgical
Geometallurgical
Geometalurgical analysis of all outputs
matrix based on existing

existing knowledge knowledge Orientation Study Study
Orientation Study
Orientation
QA/QC protocol on Select 1500-2000m of continuous core after assay suite, that Non-destructive
Import & format existing everything describes variability and end members of whole deposit Tests
How do you site data to be compatible
Coding and labelling
incorporate to your protocols conventions Continuous Geological
Alteration? Hyperspectral EQUOTip
core imaging Logging Log
mineralisation
How does site calculate Conversion of text
mineralogy? (lith&alter) to
(appropriate?) numeric values IR/Thermal Similarity Geotech & Log Log Textural
How do you incorporate existing Classification
spectrometry groupings structure lithology alteration logging
comminution & float data?
Selection of process path Selection of polymetallic

Routine QA/QC protocol on as many steps as
for each target mineral process path for all ore types
possible, defined by previous experience
What minerals What minerals

Sample integrity and test Order of tests to be Relationship between control separation? control recovery?
representatively maintained done in established bankable tests and proxy
throughout process and stable tests established
Gravity Magnetic
Leaching Sorting Flotation
Ensure all programs interact
Maintain spatial context of samples as a Experimental design and correctly and don’t compromise
link back to site assay sample interval integration of planned tests stable each other. Grind Size Selection
(Not all at the same scale)
Ranking of economic
(4) (4)Orientation
(4) Test
Orientation Study Study
work on the value of target minerals
Test Work
Hypothesis Planning
Sample set
Test Work Planning
Data analysis and
modelling
Test work cost and
time metrics
(5) Orientation
(5) Orientation StudyStudy
Analysis
planned out estimated Analysis
Tools for domaining have been tested An understanding of what works and
and experimental protocols developed what does not for this deposit
Propagation of Situational awareness of where Data matrix setup What works and
error estimated
this experimental set sits in the established
Useable relationship between
what does not?
Representatively of each test across
whole geometallurgy campaign bankable tests and their proxies the sample maintained
Figure 249. The geometallurgy steps 1 of 3


Compare each
parameter against
Make recommendations to do all target process
In context of outcomes of the Domain Study

tests across the remaining 1500- Only the orientation behaviour
2000m of drill core samples that passed
QA/QC
• Mean Compare each T-test
• Median parameter against
Calculate the statistical • Mode parameters shown
• Maximum
ANOVA tables
profile of each parameter to be relevant in
Flag where samples are • Minium
in the data set Domaining Study Repeatability
Merge orientation study output compromised (and why) and • Standard Deviation
into Geomet matrix • Variance with replica
delete from results set
• Number of points Compare each tests
parameter against
parameters shown
(6) Update Geomet to be relevant in
(5)(6) Update
Update Geomet
Geomet Orientation Study
DataMatrix
Data Matrix
Data Matrix (7) Multivariate Analysis
(7) Multivariate
of geometallurgical data Analysis of
matrix
An iterative process to understand the geometallurgical data matrix
mathematical variability scope and
QA/QC on this
process of
Define the architecture of the controls of the data set Plot each parameter against
Patterns of correlation,
every other parameter in a
integration geometallurgical data matrix cross correlation graph matrix
Different sub-populations
Determining what is relevant and
what is not in matrix data set Examine the probability plot of Differences in gradient,
each parameter Different sub-populations
Examine outcomes in context of

Examine the histogram plot of Shape of distribution,
Domaining Study & Orientation Study
each parameter Tight or spread
Mirrored patterns and

Determine interaction terms correlations, leading and
for parameters if possible lagging indicators
A new data set is outputted that is principle

component based. PC1 describes most variability.
PC2 describes most of the variability that is left, etc. Create cluster windows to
describe similarity in PCA
context Assess the
PCA Variability PCA Eigen Vector statistical profile
PCA Data matrix of each Class of METAL PRODUCED
Matrix Significance Matrix
each parameter Installed PER kWh
Groupings and patterns
(mean, variability, Blast Power draw
found in domaining study
etc.) Whole (kW)
Circuit
Query diagram plot against Reduce /change population of cluster METAL PRODUCED
Groupings and patterns Primary
inputs in model windows to describe discrete PER HOUR
found in multivariate Crusher
(PC1=Ax +By+Cz+...) boundaries to form classes of similarity analysis
Process behaviour tph
SAG Mill
defined Class
Metal Metal recovery rate
Previous Guide final Class definition
Ball Mill content in
work definition in context of
Plot PC1 vs PC2, PC1 ore
Campaign target process behaviour Gravity
vs PC3, (etc) Flotation Hydrometallurgy
goal patterns in data separation
Throughput modelling based
around a simplistic
concentrator design
For each class, predict
(8) Class Definition recovery for each target metal

(class by class)
Are the classes Propagation
of error
(class by class)
appropriate?
Validate cluster windows against external If validation fails, repeat domain Process In context or original
data. (Chemical assays not in set, XRD, MLA, behavior geometallurgy expt goals
analysis, multivariate analysis, PCA Are the classes Validation of models against measured
important to understand context of data) analysis (steps 6-8) different from
This is a QA/QC step. Geological parameters
each other?
behavior Predict process results of
samples not part of data used to
develop models (Hypothesis
Statistical Analysis Response Sample Set)
Correlations and
associations of each Variability of gangue Process Domains
mineral minerals that control along continuous
(Multivariate Analysis) process separation drill core sections
What combination of
minerals control the
preferred process path?
(Orientation Study)
Predict process response Validate the developed Does the predicted

based on models developed process behaviour Ore type Small scale Diagnostic Comminution
in Step 9 Classes previously defined process results correlate
flotation (JKMSI) Leach Index
with measured values? CuSUM
Return to mine site. Select a 2nd Phase of Done on regular Hyperspectral Chemical
EQUOTip
samples from the core yard using the intervals Image Analysis Assays
guidelines developed so far. Run developed
experimental protocol
Continuous drill
(10)(10) Hypothesis Validation
Hypothesis Validation core data
Study
Study(Reiterate Steps 4-9)
(Reiterate Steps 4-9)
(11) (11)
Mapping Study
Mapping Study – –
Use Hypothesis Response set to domaining of ofprocess
domaining behavior
process behavior
answer any new queries raised in the
Evaluation and Assessment (Step 10)
Establish a data collection procedure

to be used to map process domains
into the deposit to be done as a
Incorporate new domains of the Query why a class is Fill in gaps Finish experimental
routine production procedure
deposit into the geomet analysis different to other classes in objectives that were
that were previously unavailable unusual cases
in sampling unable to be completed in
Hypothesis Sample Study


Are the selected Ore Classes genuinely Were the samples selected to become
different from each other in context of original the Orientation Study samples, the true
experimental objectives stated in (0)? End Member rock textures?
• Did it work?
Does data collected in this geomet study Should a different process behaviour • Was the original Orientation
provide the relationships and true variability be considered to build upon what has Study samples really the process
of key process parameters in context of been already done?
extremes observed?
intrinsic geological attributes?
• Was the grind size used in the
study appropriate?
• What is missing when looking at

(12) Evaluate Results
(12) Evaluate Results from
Steps 4-11 in context of
from Steps 4-11 all data, across the whole range
of each parameter, in context of
in context ofSteporiginal
original aims 0 aims Step 0 original objectives?
• Does more sampling need to be

done?
Are the relevant questions
being asked in this study in • How do the different campaign
Is the methodology in the current phases interrelate?
context of existing objectives study able to determine when different
as stated in Step (0)? geomet characteristics are being
• Is the link between this
observed for the first time?
geometallurgy campaign and
conventional process modelling
viable and useful?
• Is the geomet matrix dataset able

to be audited in context QA/QC?


17 REFERENCES
1. AMIRA Project D483 (1997a): Optimisation of Mine Fragmentation for Downstream Processing:
First Progress Report. November 1996-April 1997 AMIRA Project D483, JKMRC Internal
Report (Mine to Mill 1)
2. AMIRA Project D483 (1997b): Optimisation of Mine Fragmentation for Downstream Processing:
Second Progress Report. May 1997 – November 1997 AMIRA Project D483, JKMRC Internal
Report (Mine to Mill 1)
3. AMIRA P843 (2008a) Geometallurgical Mapping and Mine Modelling - Technical Report 1,
February, 2008 – [confidential]
4. AMIRA P843 (2008b) Geometallurgical Mapping and Mine Modelling - Technical Report 2,
November, 2008 – [confidential]
5. AMIRA P843 (2009) Geometallurgical Mapping and Mine Modelling - Technical Report 3, June,
2009 – [confidential]
6. AMIRA P843A (2010a) Geometallurgical Mapping and Mine Modelling - Technical Report 4,
February 2010 – [confidential]
7. AMIRA P843A (2010b) Geometallurgical Mapping and Mine Modelling - Technical Report 5, June
8. AMIRA P843A (2010c) Geometallurgical Mapping and Mine Modelling - Technical Report 6,
November 2010 – [confidential]
September 2011 – [confidential]
10. AMIRA P843A (2011b) Geometallurgical Mapping and Mine Modelling - Technical Report 8,
December 2011 – [confidential]
September 2012 – [confidential]
12. AMIRA P843A (2012b) Geometallurgical Mapping and Mine Modelling - Technical Report 10,
December 2012 – [confidential]
13. AMIRA P843A (2013) Geometallurgical Mapping and Mine Modelling - Technical Report 11, May

14. Anderson, K.F.E., et al., Quantitative mineralogical and chemical assessment of the Nkout iron ore
deposit, Southern Cameroon. Ore Geology Reviews, 2014. 62: p. 25-39.
15. Aylmore, M.G., et al., Applications of advanced analytical and mass spectrometry techniques to the
characterisation of micaceous lithium-bearing ores. Minerals Engineering, 2018. 116: p. 182-195.
16. Ballantyne, G.R., Powell, M.S., (2014). Benchmarking comminution energy consumption for the
processing of copper and gold ores. Minerals Engineering 65, 109-114.
17. Benzer, H. & Tavares, L. & Tavares, M. & Powell, M. & Kruttschnitt, J. & Batista, D. &
Francisco, J. & Russo, C. & Mazzinghy, D. (2017). Hakan Dündar Okay Altun Chair in
Comminution HPGR simulation from piston-die tests with an itabirite ore, Mining Mineração. Rem
Revista Escola de Minas. 70. 99-107. 10.1590/0370-44672015700058.
18. Bond, F.C., (1961): Crushing and grinding calculations, part 1. Br. Chem. Eng. 6 (6),
378–385.
19. Bond, F.C., (1962). Crushing and Grinding Calculations - April 1962 Additions and Revisions.
Allis-Chalmers Manufacturing Co., Milwaukee, Wisconsin.
20. Brady, B.H.G., Rock Mechanics for underground mining : Third edition / by B.H.G. Brady, E.T.
Brown, ed. E.T. Brown and SpringerLink. 2006, Dordrecht: Dordrecht : Springer Netherlands.
21. Broch, E. and J. Franklin (1972). "The Point Load Strength Test." Int. J. Rock Mech. & Min. Sci.
9: 669-697.
22. Brook, N. (1980). "Size Correction for Point Load Testing." Int. J. Rock Mech. Min. Sci. &
Geomech 17: 231-235.
23. Brook, N. (1985). "The Equivalent Core Diameter Method of Size and Shape Correction in Point Load
Testing." Int. J. Rock Mech. Min. Sci. & Geomech 22(2): 61-70.
24. Bazin, C., D. Hodouin, and R.M. Blondin, Estimation of the variance of the fundamental error
associated to the sampling of low grade ores. International Journal of Mineral Processing, 2013.
124: p. 117-123.
25. Bergeron, Y., Kojovic, T., Gagnon, M-d-N. and Okono, P., (2017): Applicability of the HIT for
Evaluating Comminution and Geomechanical Parameters from Drill Core Samples – The Odyssey
Project Case Study, Proceedings from COM2017, Vancouver, August.

26. Burger, B & McCaffery, K & Jankovic, A & Valery, W & Mcgaffin, I. (2006). Batu Hijau Model
for Throughput Forecast, Mining and Milling Optimisation and Expansion Studies Newmont
Mining Corporation. Advances in Comminution. pp. 461-479.
27. Butcher, AR (2019). A Practical Guide to Some Aspects of Mineralogy that Affect Flotation. Chapter
4 in: Methodology for Identifying and Solving Problems with Base Metal Sulphide Flotation Plants.
Editor C Greet. Published by AUSIMM, Spectrum Series 25, pp 137 – 156.
28. Butenuth, C. (1997). "Comparison of Tensile Strength Values of Rocks Determined by Point Load
and Direct Tension Tests." Rock Mechanics and Rock Engineering 30(1): 65-72.
29. Castel, G. (1985). Feasibility studies and other pre-project estimates. How reliable are they?
30. Chang-Zhong L, Lingmin Z, Kaimin S, (2015). Quantitative X-ray Diffraction (QXRD) analysis
for revealing thermal transformations of red mud, Chemosphere, Volume 131, Pages 171-177,
ISSN 0045-6535
31. Chauhan, M., Investigation of a mineral flotation separability test for ore characterisation in
geometallurgy, in CRC ORE. 2013, crcore.org.au: Brisbane Australia.
32. Chibaya, A., Geometallurgical Analysis - Implications on Operating Flexibility. (A Case for
Geometallurgy for Orapa A/K1 deposit). in Faculty of Engineering and the Built Environment. 2013,
University of the Witwatersrand: South Africa. p. 136.
33. Clark, A.M., Mineral textures. (1991), London, United Kingdom: Mineralogical Society:
London. p. 301-486.
34. Craig, J.R., (1994) Ore microscopy and ore petrography / James R. Craig, David J. Vaughan. 2nd ed..
ed, ed. D.J. Vaughan. 1994, New York: New York : Wiley.
35. Cropp, A., (2013), LIBERATION AND FREE SURFACE AREA IN THE FLOAT FEED, Min
Assist Blog, Process Mineralogy Today, http://www.minassist.com.au/blog/liberation-and-
free-surface-area-in-the-float-feed/
36. Cunningham, C. V. B. (1983). "The Kuz-Ram Model for Prediction of Fragmentation from Blasting."
First International Symposium on Rock Fragmentation by Blasting 2: 439-454.
37. Cunningham, C. V. B. (1987). Fragmentation Estimators and the Kuz-Ram Model - Four Years On.
Proc. Second Int. Symposium on Rock Fragmentation by Blasting, Colorado.

38. Curry, J.A., Ismay, M.J.L., Jameson, G.J., (2014), Mine operating costs and the potential impacts
of energy and grinding. Minerals Engineering, 2014, 56(0), 70-80.
39. Daniel, M.J., 2007. Energy efficient mineral liberation using HPGR technology, Julius Kruttschnitt
Mineral Research Centre. The University of Queensland, (PhD Thesis).
40. Djordjevic, N. (1999). Two Component Model of Blast Fragmentation. Fragblast 1999,
Johannesburg, South African Institute of Mining and Metallurgy.
41. Dominy, S.C., (2016) Importance of good sampling practice throughout the gold mine value chain.
Mining Technology, 2016. 125(3): p. 129-141.
42. Dominy, S., and O'Connor, L., (2016) Geometallurgy - beyond conception, in Geometallurgy 2016,
T. AusIMM, Editor. 2016, The Australasian Instititute of Mining and Metallurgy: Perth. p. 3-
10.
43. Dominy, S., (2017) Geometallurgical sampling protocol validation by bulk sampling in a sheeted vein
gold deposit. in Eighth World Conference on Sampling and Blending 2017. 2017. Perth: AusIMM.
44. Everaert, M., & Stein, R., & Michaux, S., & Goovaerts, V., & Groffils, C., & Delvoie, S., & Zhao,
Z., & Snellings, R., & Nielsen, P., & Broos, K. (2019). Microwave Radiation as a Pre-Treatment
for Standard and Innovative Fragmentation Techniques in Concrete Recycling. Materials. 12.
10.3390/ma12030488.
45. Fitton, G., (1997), X-Ray fluorescence spectrometry, in Gill, R. (ed.), Modern Analytical
Geochemistry: An Introduction to Quantitative Chemical Analysis for Earth, Environmental and
Material Scientists: Addison Wesley Longman, UK.
46. Flyvbjerg, B. (2002). Understanding costs in public works projects. Errors or Lies?
47. Gaines, R. V., Skinner, C. W., Foord, E.E, Mason, B., and Rosenzweig, A. (1997): Dana's New
Mineralogy: The System of Mineralogy of James Dwight Dana and Edward Salisbury Dana 8th
Edition, Wiley-Interscience; 8th edition (November 10, 1997), ISBN-10: 0471193100
48. GOLDRATT, E.M., COX, J. The goal: a process of ongoing improvement. 2nd edition. London:
Gower, 1993.
49. Goodall, W. (2009), WHAT IS A THEORETICAL GRADE-RECOVERY CURVE? AN

EXAMPLE, Min Assist Blog, PROCESS MINERALOGY TODAY
https://www.minassist.com.au/blog/what-is-a-theoretical-grade-recovery-curve-an-example/

50. Gy, P., (2004). Sampling of discrete materials—a new introduction to the theory of sampling.
Chemometrics and Intelligent Laboratory Systems, 2004. 74(1): p. 7-24.
51. Gypton, C. (2002). How have we done? Engineering and Mining Jounral Jan 2002; 203, 1;
ProQuest pg. 40
52. Hamlyn, P, (2011). Sampling, analysis and quality control, in Field Geologists’ Manual, fifth
edition, pp 311-334 (The Australasian Institute of Mining and Metallurgy: Melbourne).
53. Heiskari, H., Kurki, P., Luukkanen, S., Sinche Gonzalez, M., Lehto, H., Liipo, J. (2019):
Development of a comminution test method for small ore samples, Minerals Engineering 130, 5–11
54. Holliday,J R and Cooke, D R, (2007). Advances in geological models and exploration methods for
copper ± gold porphyry deposits, in Proceedings Exploration 07: Fifth Decennial International
Conference on Mineral Exploration (ed:B Milkereit), pp 791-809.
55. Hrstka, T., et al., (2018), Automated mineralogy and petrology – applications of TESCAN Integrated
Mineral Analyzer (TIMA). Journal of Geosciences, 2018. 63(1): p. 47-63.
56. Hukki, R.T., (1961), Proposal for a solomonic settlement between the theories of von Rittinger, Kick
and Bond. AIME Trans. (mining), 1961, 220, 403-408.
57. Hunt, J & Berry, Ron & Walters, Steve. (2010). Using mineral maps to rank potential processing
behaviour. XXV INTERNATIONAL MINERAL PROCESSING CONGRESS (IMPC) 2010
PROCEEDINGS / BRISBANE, QLD, AUSTRALIA / 6 - 10 SEPTEMBER 2010
58. IM (2014 7th Oct): Byron Knelson (Knelson Concentrator) to be inducted into the International
Mining Technology Hall of Fame, IM International Mining Newsletter, https://im-
mining.com/2014/10/07/byron-knelson-knelson-concentrator-to-be-inducted-into-the-
international-mining-technology-hall-of-fame/
59. ISRM (1978). "Suggested Methods for Determining Tensile Strength of Rock Materials." Int. J. Rock
Mech. Min. Sci. & Geomech 15(3): 99-103.
60. ISRM (1979). "Suggested Methods for Determining the Uniaxial Compressive Strength and
Deformability of Rock Materials." Int. J. Rock Mech. Min. Sci. & Geomech 16(2): 135-140.
61. ISRM (1985). "Suggested Method for Determining Point Load Strength." Int. J. Rock Mech. Min.
Sci. & Geomech 22(2): 51-60.

62. JK Tech Pty. Ltd., SMC Test, in JK Tech, U.o.Q. Julius Kruttschnitt Mineral Research Centre,
Editor., University of Queensland: Brisbane.
63. JK Tech Pty. Ltd., J.T.P., JK Rotary Breakage Tester, U.o. Queensland, Editor. 2018, University
of Queensland: Brisbane.
64. Johnson,N W and Munro, P D, (2008). Methods for assigning domains in the primary sulfide zone
of a sulfide orebody, in Proceedings Ninth International Congress for Applied Mineralogy
(ICAM), pp 597-603 (The Australasian Institute of Mining and Metallurgy: Melbourne).
65. Jones, K., and Morgan, D., On-Site Process Mineralogy - Linking Departments and Disciplines
with the Common Goal of Increasing Plant Performance, in 13th AusIMM Mill Operators
Conference, AusIMM, Editor. 2016, The AusIMM: Perth. p. 149-155.
66. Kumar, A. (2014). Predicting HPGR performance and understanding rock particle behavior through
DEM modelling (T). University of British Columbia. Retrieved from
https://open.library.ubc.ca/collections/ubctheses/24/items/1.0103429
67. Kanchibotla, S. and Valery, W (1999). Modelling Fines in Blast Fragmentation and its impact on
Crushing and Grinding. Proc Explo '99 - A Conference on Rock Breakage, Kalgoorlie Australia,
AusIMM.
68. Kaufmann, A., (2013). INNOVATIVE BUILDING MATERIAL RECYCLING, Fraunhofer

Institute IBP
69. Keeney,L, Walters, S G and Kojovic, T, (2011). Geometallurgical mapping and modelling of
comminution performance at the Cadia East porphyry deposit, in Proceedings First AusIMM
International Geometallurgy Conference 2011, pp 73-83 (The Australasian Institute of
Mining and Metallurgy: Melbourne).
70. Keeney, Luke & Nguyen, K. (2014). The use of EQUOtip as a hardness domaining tool. IMPC
2014 - 27th International Mineral Processing Congress.
71. Klein, Cornelis, (1999) Manual Of Mineralogy : (after James D. Dana 1937). New York :Wiley.
72. Krishnamurthy, P. and R. Sawkar (2014), Handbook on sampling of ores and minerals. India. p.
466-466.
73. Kojovic, T., (1988): Mathematical Model Building, JKMRC University of Queensland
professional training course

74. Kojovic, T. & Whiten, W.J. (1994): Evaluating the quality of simulation models. Journal
Innovations in Mineral Processing (Editor: Yalcin, Laurentian University, Sudbury), Sudbury,
p. 437-446.
75. Kojovic,T, Michaux, S and Walters, S, (2010), Development of new comminution testing
methodologies for geometallurgical mapping of ore hardness and throughput, 2011. in Proceedings
XXV International Mineral Processing Congress (IMPC) 2010, pp 891-899 (The
Australasian Institute of Mining and Metallurgy: Melbourne)
76. Kojovic, T., (2016): HIT - A Portable Field Device for Rapid Hardness Index Testing at Site.
Proceedings AusIMM Mill Operators’ Conference 2016, October, Perth WA, pp9-16.
77. Kojovic, T., Bergeron, Y. and Leetmaa, K., (2019): The value of daily HIT ore hardness testing of
the SAG feed at the Meadowbank Gold Mine. Proceedings from SAG2019, Vancouver,
September.
78. Kuhar L.L., J., M.I., McFarlane, A.J., Benvie, B., Botsis, N.m., Turner, N. and Robinson, D.J.
(2011), The Development of Small-Scale Tests to Determine Hydrometallurgical Indicies for Orebody
Mapping and Domaining. in First AusIMM International Geometallurgy Conference (GeoMet) 2011.
2011. Perth: AusIMM.
79. Kuznetsov, V. M. (1973). The mean diameter of fragments formed by blasting rock. Soviet Min
Science 9(2): 144-148.
80. Lang, A.M., K. Aasly, and S.L. Ellefmo, (2018), Mineral characterization as a tool in the
implementation of geometallurgy into industrial mineral mining. Minerals Engineering, 2018. 116:
p. 114-122.
81. Li, J., Kuhar, L., Meakin, R., Chapman, N., McFarlane, A., and Yu, J. (2016) A Geometallurgical
Proxy for Nickel Laterite Heap Leaching, CSIRO publication
82. Liipo, J., Hicks, M., Takalo, V-P., Remes, A., Talikka, M., Khizanishvili, S., and Natsvlishvili,
M. (2019): Geometallurgical characterization of South Georgian complex copper-gold ores. The
Southern African Institute of Mining and Metallurgy
83. Lotter, N.O., Kormos, L.J., Oliveira, J., Fragomeni, D., and Whiteman, E. (2011). Modern
Process Mineralogy: Two case studies. Minerals Engineering 24 (2011) 638–650
84. Lotter, N.O., Whiteman, E., Bradshaw, D.J., (2014): Modern practice of laboratory flotation

testing flowsheet development–a review. Miner. Eng. 66–68, 2–12
85. Lund, C., P. Lamberg, and T. Lindberg, (2015), Development of a geometallurgical framework to
quantify mineral textures for process prediction. Minerals Engineering, 2015. 82: p. 61-77.
86. Matus, C., Bueno, M., Powell, M., Michaux, S. and Luukkanen L. (2020): The Geopyörä brekage
test for geometallurgy, IMPC 2020: XXX International Mineral Processing Congress, Cape
Town, South Africa, 18-22 October 2020
87. Mackenzie, W and Cusworth, N, (2007). The use and abuse of feasibility studies, in Proceedings
Project Evaluation 2007, pp 65-76 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
88. McNulty, T. (2014). Plant Ramp-Up Profiles an Update with Emphasis on Process Development.
Proceedings of the 2014 Conference of Metallurgists, Canadian Institute of Mining, Metallurgy and
Petroleum.
89. Michaux, S. (2006). Analysis of fines generation in blasting, PhD Thesis JKMRC University of
Queensland
90. Michaux, S. & Djordjevic, N. (2010). Approximation of surface area of fines in blast induced
fragmentation. XXV International Mineral Processing Congress – IMPC 2010
91. Michaux, S., (2016). Electrodynamic Fragmentation of Stainless Steel Slag and Concrete Materials,
Universite de Liege, Department of GeMMe (ULg) internal research development
presentation
92. Montoya Lopera, P. (2014) Geometallurgical Mapping and Mine Modelling - Comminution
Studies: La Colosa Case Study, AMIRA P843A, Master’s Thesis, CODES, University of
Tasmania.
93. Morgan, S., Bradshaw, D.J. and Schwarz, S. (2012), Julius Kruttschnitt Mineral Separation Index
– Small-Scale Test to Characterise the Flotation Response of an Ore Type. in 11th AusIMM Mill
Operators’ Conference 2012. 2012. Brisbane: AusIMM.
94. Morrell S, and Morrison R. (1996), AG and SAG mill circuit selection and design by simulation.
SAG 1996: Mining and Mineral Process Engineering Vancouver, Canada, pp 769-791.
95. Morrell, S, (2004). Predicting the specific energy of autogenous and semi-autogenous mills from
small diameter drill core samples, Minerals Engineering,17(3):447-451.

96. Morrell,S, (2011). Mapping orebody hardness variability for AG/SAG/crushing and HPGR circuits,
presented to Fifth International Conference on Autogenous and Semi-autogenous Grinding
Technology, Vancouver, 25 - 28 September.
97. Napier-Munn, T., Morrell, S., Morrison R. and Kojovic T. (1996) JKMRC Monograph Mineral
Comminution Circuits, Their Operation and Optimisation. The University of Queensland, Julius
Kruttschnitt Mineral Research Centre JKMRC Brisbane Austalia.
98. Napier-Munn, T. and B.A. Wills, Wills' Mineral Processing Technology. (2005).
99. Napier-Munn, T. J., Morrell, S., Morrison, R. D.,& Kojovic, T. (2005): Mineral Comminution
Circuits: Their Operation and Optimization, Julius Kruttschnitt Mineral Research Centre, The
University of Queensland, Brisbane.
100. Napier-Munn, T.J.a., Statistical methods for mineral engineers : how to design experiments
and analyse data / T.J. Napier-Munn, ed. C. Julius Kruttschnitt Mineral Research. 2014:
Indooroopilly, Queensland : Julius Kruttschnitt Mineral Research Centre.
101. Navarra, A., T. Grammatikopoulos, and K. Waters, (2018), Incorporation of

geometallurgical modelling into long-term production planning. Minerals Engineering, 2018. 120:
p. 118-126.
102. Niitti, T., (1970): Rapid evaluation of grindability by a simple batch test. In: International
Mineral Processing Congress Proceedings, Prague, pp. 41–46
103. Noort, D & Adams, C. (2006). Effective Mining Project Management Systems. Australasian
Institute of Mining and Metallurgy Publication Series.
104. Nuorivaara, T. and Serna-Guerrero, R. (2016), A STUDY ON THE PROPERTIES OF

FOAMS PRODUCED WITH CELLULOSE-BASED MOLECULES WITH AN OUTLOOK ON
FLOTATION, IMPC 2016: XXVIII International Mineral Processing Congress Proceedings -
ISBN: 978-1-926872-29-2
105. O'Connor, L. (2016), The important role of geometallurgy in today’s mining industry. The
AusIMM Bulletin 2016 [cited April 4]; 1]. Available from:
https://www.ausimmbulletin.com/feature/the-important-role-of-geometallurgy-in-todays-
mining-industry/.
106. Outotec Oy, (2017): HSC Chemistry 9, Thermochemical and Process Simulation
Calculation
Software. Outotec Research Center. www.outotec.com

107. Parbhakar-Fox, A., (2016), Geoenvironmental characterisation of heap leach materials at

abandoned mines: Croydon au-mines, QLD, Australia. Minerals, 2016. 6(2): p. <xocs:firstpage
xmlns:xocs=""/>.
108. Parbhakar-Fox, A., B. Lottermoser, and D. Bradshaw, (2013), Evaluating waste rock
mineralogy and microtexture during kinetic testing for improved acid rock drainage prediction.
Minerals Engineering, 2013. 52(C).
109. Parbhakar-Fox, Anita & Aalders, John & Lottermoser, Bernd. (2015). Effective Field-Based
Testing Tools for Rapid ARD Prediction. 10th International Conference on Acid Rock Drainage
& IMWA Annual Conference, At Santiago, Chile, Volume: 1
110. Parbhakar-Fox, A. and B.G. Lottermoser, (2015), A critical review of acid rock drainage
prediction methods and practices. Minerals Engineering, 2015. 82(C): p. 107-124.
111. Parbhakar-Fox, A.a.L., B.G., (2013), Establishing environmental geometallurgy indicators for
predicting acid rock drainage, Copper Mines Tasmania. 2013, CODES TMVC: University of
Tasmania.
112. Parian, M. (2017) Development of a geometallurgical framework for iron ores - A

mineralogical approach to particle-based modeling, Doctor of Philosophy PhD, University of
Lulea
113. Parvaz, P.B., Mosaddeghi, A., and Weh, A. (2015). Pre-concentration Attributes of
SELFRAG High Voltage Treatment, Conference Proceedings: European Symposium for
Comminution and Classification At: Chalmers University of Technology; Gothenburg,
Sweden, DOI: 10.13140/RG.2.1.1813.8006
114. Pérez-Barnuevo, L., E. Pirard, and R. Castroviejo, (2013), Automated characterisation of

intergrowth textures in mineral particles. A case study. Minerals Engineering, 2013. 52(C) p. 136-
142.
115. Potts, P.J., (1987), A Handbook of Silicate Rock Analysis: Chapman and Hall.
116. Rochman, R. (2003). Real Options Valuation of Companies Run by Theory of Constraints.
117. Rollinson, H., (1993), Using Geochemical Data: Evaluation, Presentation, Interpretation:
John Wiley, NY.

118. Ruffell A., Wiltshire P., (2004). Conjunctive use of quantitative and qualitative X-ray
diffraction analysis of soils and rocks for forensic analysis. Forensic Sci Int. 2004 Oct 4;145(1):13-
23.
119. Runge, K., (2010) Laboratory Flotation Testing – An Essential Tool for Ore Characterisation,
Ch 9 of Flotation Plant Optimisation: A Metallurgical Guide to Identifying and Solving Problems in
Flotation Plants, AusIMM, Spectrum Series. Melbourne ISBN: 978 1 921522 14 7
120. Sampling, C., (2014). Sampling 2014 : where it all begins, Sampling : 29-30 July 2014,
Perth, Western Australia / AusIMM. ed. M. Australasian Institute of, et al. 2014: Carlton,
Victoria Australasian Institute of Mining and Metallurgy.
121. Schodlok, M., Green, A. & Huntington, J. (2016) A reference library of thermal infrared
mineral reflectance spectra for the HyLogger-3 drill core logging system, Australian Journal of Earth
Sciences, 63:8, 941-949, DOI: 10.1080/08120099.2016.1234508
122. Schouwstra, R.P., D.V. De Vaux, and Q. Snyman, (2017), Further development of a
chemistry proxy for geometallurgical modelling at the Mogalakwena mine. Journal of the Southern
African Institute of Mining and Metallurgy, 2017. 117(7): p. 719-726.
123. Senior, G.D., P.J. Guy, and W.J. Bruckard, (2006), The selective flotation of enargite from
other copper minerals — a single mineral study in relation to beneficiation of the Tampakan deposit
in the Philippines. International Journal of Mineral Processing, 2006. 81(1): p. 15-26.
124. SGS, SPI Testing. 2018, SGs A.A.
125. Shreeve, J., Multi sensory mapping. (2018), GeoTek Limited: United Kingdom.
126. Suazo, C.J., W. Kracht, and O.M. Alruiz, (2010), Geometallurgical modelling of the
Collahuasi flotation circuit. Minerals Engineering, 2010. 23(2): p. 137-142.
127. Tammishetti, V. & Rai, B., & Kumar, B., & Kumar, R., & Pradip, P., (2015). Quantitative
Estimation of Mineral Phases from Chemical Assays and Powder X-Ray Diffraction Rietveld
Analysis: A Case Study on Selective Flocculation of Iron Ore Slimes. Transactions of the Indian
Institute of Metals. 69. 10.1007/s12666-015-0721-7.
128. Torvela, J. (2020): DOUBLE WHEEL CRUSHER PROTOTYPE, Univeristy of Oulu MSc
Thesis
129. Torvela, J., Bueno, M., Liedes, T. and Luukkanen, S. (2020): The Geopyörä Breakage Test,
University of Oulu, Minerals Engineering journal “in press”

130. Vaasjoki M., and Lindqvist, K., (2001) ELECTRODYNAMIC FRAGMENTATION TESTS
AIMED AT SEPARATION OF MINERAL SPECIES FOR MINERALOGICAL AND ISOTOPIC
STUDIES, Geologian tutkimuskeskus Geological Survey of Finland, Arkistoraportti - Open
File Report M41.1/2001/2
131. Vann, J., Bertoli, O., and Jackson, S., (2002), An Overview of Geostatistical Simulation for
Quantifying Risk in Quantifying Risk and Error”. 2002, Geostatistical Association of Australasia
symposium.
132. Vatandoost, A., P. Fullagar, and M. Roach, (2008), Automated multi-sensor petrophysical
core logging. Exploration Geophysics, 2008. 39(3): p. 181-188.
133. Verwaal, willem & MULDER, A. (1993). Estimating rock strength with the Equotip hardness
tester. International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts. 30. 659-662. 10.1016/0148-9062(93)91226-9.
134. Vos, F. and Bradshaw, D., (2014), Geological Engineering Investigators from Julius
Kruttschnitt Mineral Research Center Target Geological Engineering (A new small-scale test to
determine flotation performance - Part 1: Overall performance). Journal of Engineering, 2014: p.
919.
135. Yildirim, B.G., (2016), Development of a correlation between mineralogy, rock strength
measures, and breakage of Copper Porphyries, in SMI JKMRC. 2016, University of Queensland:
Brisbane. p. 293.
136. Young, M.F., Pease, J.D., Johnson, N.W., and Munro, P.D., (1997). Developments in
Milling Practice at the Lead/Zinc Concentrator of Mount Isa Mines Limited from 1990, AusIMM
Sixth Mill Operators Conference, 6-8 October 1997
137. Wang, Hao-liang (Eric) (2012). Use of high voltage pulse technology to investigate the
mechanism of mineral breakage and selective liberation. PhD Thesis, Sustainable Minerals
Institute, The University of Queensland.
138. Wills, B.A. and Finch, J., (2015), Wills' mineral processing technology : an introduction to the
practical aspects of ore treatment and mineral recovery / Barry A. Wills, James A. Finch. Eighth
edition.. ed. Mineral processing technology, ed. J.A.a. Finch. 2015: Oxford : Butterworth-
Heinemann is an imprint of Elsevier.
139. Wirfiyata, F. (2012). Applied Geo-Metallurgical Characterisation for Life of Mine Throughput
Prediction at Batu Hijau.

140. Wirfiyata, F., (Jan 1, 2012),"Applied Geo-Metallurgical Characterisation for Life of Mine
Throughput Prediction at Batu Hijau""Canadian Institute of Mining, Metallurgy and Petroleum,
214408, CIM
141. Wiseman, D.M.a.R., (1991) J.M. JKSimMet - the mineral processing simulator. in 2nd
Canadian Conference on Computer Applications in the Minerals Industry. Univeristy of British
Columbia:
142. Whitten, D. G. A. and Brooks J. R. V., (1972). The Penguin Dictionary of Geology. 495 pp.,
161 figs, tables. Penguin Books Ltd., Harmondsworth, Middx. Price 75p, in Geol. Mag. 1972. p. 198.
143. World Conference On, S. and Blending, (2017), Eighth world conference on sampling and
blending 2107 : 9–11 May, in Perth, Western Australia, 2017 edited by S. C. Dominy and K. H.
Esbensen, ed. K.H. Esbensen, et al. 2017: Carlton (Victoria) : AusIMM.

How To Set Up and Develop A Geometallurgical Program

Uploaded by

Copyright:

Available Formats

How To Set Up and Develop A Geometallurgical Program

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

How To Set Up and Develop A Geometallurgical Program

Uploaded by

Copyright:

Available Formats

Geological Survey of Finland

Economic Minerals Unit

How to Set Up and Develop a

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

GEOLOGICAL SURVEY OF FINLAND DOCUMENTATION PAGE

BATCircle Project – Business Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Associate Professor/ Apulaisprofessori GTK BATCircle Project Manager

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Bon Scott ACDC 1975

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

4.11.3 High intensity magnetic separation – electromagnetic HIMS 90

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Figure 1 Start-up curve classification showing % of design capacity vs. time

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

The conclusions were as follows:

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Grinding Circuit Performance

Mill Feed (tph) Mill Power (kW)

Figure 2. Mill performance at a mine somewhere in the Southern Hemisphere

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Mill Throughput DMT/hr

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

2 METHODOLOGY AND RESULTS

2.1 Campaign planning of objective goals

Prefeasability study Feasibility &

Operation and Mining &

Site Closure & Site

Figure 4. The mine operation life cycle

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

In Pit In pit ore

Liberation through comminution

Isa Mill Fine

Figure 5. The mining process path

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Geometallurgy for operations

Comminution Choice between SAG

Flotation data Recovery Ore type

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Done Business Model

Figure 7 Conventional mine development flow sheet

Geometallurgy Data Analysed Ore type definition

Block Model Generation

Figure 8 Geometallurgically supported mine development

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

Figure 9 Leverage of early work (Source: MacKenzie and Cusworth 2007)

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

2.2 Phases in a Geometallurgy campaign

1. Planning and experimental design (Napier-Munn 2014)

Figure 12 The different phases interrelate to each other.

Geologian tutkimuskeskus | Geologiska forskningscentralen | Geological Survey of Finland

(1) Ore Body Analysis

(2) Creation of foundation

(4) Orientation Study