1

HOW TO SET UP AND DEVELOP A

GEOMETALLURGICAL PROGRAM

Simon Michaux
28/05/2020
IT’S HARDER THAN IT LOOKS…
2

”It’s a long way to the top if you want to rock ‘n’ roll”

Bon Scott ACDC 1975

 3

Sound experimental design Multi-tiered data analysis

done up front tied to strategic objectives

Strategic objectives set with How the final product is to

tactical milestones established be used is planned out

Victory loves preparation

30.5.2020
OBJECTIVE 1: DEFINE MINERALOGICAL BASED 4

PROCESS BEHAVIOUR

Measurement based efficiency of

each individual process in context Crushers
of a family of ore types
Flotation
As opposed to a vague one
recovery target for the whole SAG/AG Mill
process circuit for all ore types
C C
C C Dump Leach
With the capability to map this V V
V V
ore type process behavior into HPGR
the deposit
Where previously, only grade and
CIL Leach Tank
geology domains were mapped Ball Mill
OBJECTIVE 2: DEFINE MOST EFFECTIVE PROCESS
5

PLANT DESIGN
Increases precision Exploration
Geometallurgy
Large scale characterization
Process design Business model
Go/no go decision
Mining sequence to
stockpile
Construction
De-risked
Commissioning
 Decommissioning & 6
Exploration Site Rehab
program
What minerals will generate Operation
environmental hazard?
Conceptual
Study
What is the best process path
for this deposit? Construction

Geometallurgy What minerals control the

Program best process path? Raising Capital
(Prefeasibility Study) Investment
Process domains and process
behaviour response spatially
mapped into the deposit Final Business
Design

Larger scale process

separation Process Design Pilot Scale
characterization program (Feasibility Study) Trial
(Hypothesis Study)
30.5.2020
OBJECTIVE 3 - IDEAL OPTIMISATION CYCLE 7

Open Pit
Optimisation
Mining Phase Required Capital
Optimisation CAPEX

Integrated &
Mine Schedule Simultaneous
Optimisation

Cut Grade(s) Process Plant

Analysis Calibration

Stockpile Infrastructure Geometallurgy is

Management Logistics sophisticated data support

Ore Blending
 OBJECTIVE 4 – MITIGATE RISK
8

• At the feasibility stage(s), CAPEX closely resembles the final commissioning reality

• Once operating, the deposit is comprehensively characterised 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

• The corporate executive board benefit the most from this outcome
SETTING OBJECTIVES
9

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

EVALUATION OF EXPERIMENTAL OBJECTIVE TARGETS 10

AGAINST ORIENTATION AND MAPPING STUDIES

Hypothesis
Validation Study
Evaluation and
assessment
Orginal
Experimental Goal

Orientation Further
implications
Study

Domain Domain
Mapping
Study
Study
Conclusions &
Hypothesis Recomendations
Study
PHASES IN A GEOMETALLURGY CAMPAIGN 11

• Did
What
Orderitdomain
Original
Ore work?
Objectives
Contact ofwith set
signatures exploration
objectives
testsdefinition
understood
control geologist
reassessed,
(PCA)
process but
1. Campaign planning of objectives
who worked
behaviour?
refitted on drill
in context
o Recovery? ofcore/site
the business
2. Sample selection • model
o Penalty
What
Competent
Mineralogical element
is missing management?
when
technicians
signatures looking
who
of eachat all
• o Economic
data, across
Examination
understand
What
ore typedata performance?
the whole
of existing
methodology
should
understood rangeandof data
deposit
be collected
3. Experimental test work • each parameter?
objectives
routinely
Scope and(that are practical
mandate reassessed in a
•• Mandate
production
based
Assessment and
Processonbehaviour ofscope
environment)?
analytical
the in agreed
outcomes
number
each ore upon
of end
4. Analysis of data outcomes
• In context
member
Sample
type of original objectives
lithology's
perseveration
understood for future
5. Assessment of analytical outcomes •• Metrics
work
How is that
Metrics of
of success
data toagreed
success be usedupon
agreed upon
by site
to
• personnel?
optimise
Sampling
Engineering towards
of each end member
simulation (HSC)
6. Further sampling and test work •• Budget
lithology,
Data and
matrix time
where
QA/QC framesample
each agreedisupon
• consistent.
How
Links
Spatial should
established
mapping the process
in how
into plant
this could
deposit
7. Economic modelling and optimisation
• Experimental
operation
be related evolve campaign
over time
to operation planned
to keep
protocols out
8. Development of production protocol •• pace
Geomet withblock
Collection the
of allmine schedule?
existing
model context data
generated
• for
Masseach
Links sample takenunderstood
of reconciliation
to each sample protocols

30.5.2020
AND NOW A WARNING…
12

Also known as the

‘seagull’ problem

Be very clear in your mind what you are about and what success looks like

Other departments may try and ‘influence’ what you spend your budget on
THE FULL PROCEDURE
13

(0) Geometallurgical
Experimental &
Analyitcal Goals

(1) Ore Body Analysis

of Existing Knowledge

(2) Creation of foundation

geomet matrix based on
existing knowledge

(3) Geometallurgical
Orientation Study

(4) Orientation Study

Test Work Planning

30.5.2020 (5) Orientation

Study Analysis
 14

(0) Geometallurgical
Experimental &
Analyitcal Goals
CAMPAIGN PLANNING AND SETTING OF 15

OBJECTIVES
• What is the fundamental question?
• Establishing the best process path
• Choice between SAG mill or HPGR
• Recovery efficiency, if so, which metal(s)?
• Ore type definition
• Choice between hydromet or pyromet
• Penalty element impact

• What are the related questions the same samples could be use for at a later
date?
• Map out existing site knowledge
• Reports, feasibility studies, data sets
• Talk to site personnel of all stripes
FOR A GIVEN JOB, WHERE ARE YOU IN THE MINING
16

CYCLE? • Commissioning
• Bringing plant production up to design expectations
• Feasibility study
• True value of the resource
• Economic viability
• Geometallurgy for operations
• Most effective process design • Show stoppers & penalties in feed stream (clay, Cl, Fl, As,
etc)
• Predicted variability in ore hardness
• Developing the mine schedule
• Metallurgical reconciliation
• Block model
• Maintenance schedule
• Cut off grade analysis
• Plant expansion

• Geomet for long term variability

• Life of Mine cycle
• Environmental Impact
• Operation total footprint • Mine site waste plume
• Legacy impact
• Mine closure and site rehabilitation
30.5.2020
 MLA/QEMSCAN 17
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

 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

geomet matrix based on
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?
EXPERIMENTAL TEST PHASES OF A GEOMET 19

CAMPAIGN
Orientation Study Hypothesis Response samples
• 10-20 samples of end member samples • 500-1000m of drill core
• The Race • To fill in and test missing sections shown up in
Mapping Hypothesis Sample set
• Deposit specific hypothesis formation

Mapping Hypothesis samples

To be collected only once you
• 1500-2000m of drill core
know what is missing, not before
• 4-5 continuous sections across deposit major
structures If you are able to, collect these
• Intelligent selection of tests that interrelate samples at the same time
with final objectives
30.5.2020
 SAMPLE SELECTION FOR THE GEOMETALLURGY 20

STUDY
Site geologist intuition is to be listened to and considered
• What ore types do they think are relevant
Settle on the number of extreme
• What deposit geology structures are most significant textures (end members)
• What drill holes traverse and map these structures • Spend time in the core yard and
help do the sampling and packing
Site metallurgist intuition is to be listened to and considered
• What is the process range found so far in this deposit?
Ideal: 1500-2000m of
• Axb & BMWi extremes
half drill core
• Flotation recovery performance so far (biggest size possible)
• Leaching recovery so far
• What minerals are causing them grief so far?
• Are there site specific assay models that have been developed?
30.5.2020
POOR QUALITY CORE IS NOT TO BE REJECTED
21

• Traditionally metallurgical sampling (comminution in particular) has been

done one only the good quality bits of core

• Whereas the core is often broken up due to the friable nature of the rock

• This is often where we find the clays and poor recovery performing ore types
• The 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 could be poor
• Needs to be mapped and modelled if this ore type is in large quantities in the drilling library

30.5.2020
 Geomet Size by size
XRF/XRD Geomet Assays XRF/XRD XRF/XRD 22
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)
(3) Geometallurgical
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
SAMPLE SELECTION FOR THE ORIENTATION STUDY
23

• Using the 1500-2000m of half core you sampled from site previously

• Isolate the end member textures into assay interval sections

• We want to target the extremes of rock textures in the deposit
• Variability mapping will come later

• Select for each end member a section (2-6m length HQ half core / 20-25kg)
• Consistent lithology across whole section (do your best with what you have)
• If veins and intrusions are in the sample, make sure they are all though the sample, not just one or two isolated places

• So all tests (bankable and proxy) will be done on as close to the same lithology and
mineralogy as practically possible

• A good number of end member samples is 10-15

• Deposit specific, let the rock speak Keep in mind possible future
work on these samples

30.5.2020
SELECT THE ORIENTATION SAMPLES
24

Orientation Orientation Orientation Orientation Orientation Orientation Orientation

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 7 Sample 8

Selection of all of the rock texture end members, 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.
 25
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)
(4) Orientation Study
Test work on the
Test Work Planning
Hypothesis Sample set

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

Robust data QA/QC in conjunction with competent experimental test
work is the best way to manage propagation of error

(Plan the flight, then fly the plan)

PROCESS RESPONSE
What minerals control
flotation separation?
Flotation
(5kg)

Gravity
What minerals control Sorting What minerals control
Separation
sorting separation? (5kg)
(5kg) gravity separation?
OrientationSample
Orientation Sampleφ
(25-30kg)
(25kg)
Magnetic
What minerals control Leaching
Separation
What minerals control
leaching? (5kg) magnetic separation?
(5kg)

Characterization
How do those minerals interact (5kg) Are the controlling minerals at
with the process separation? each expt stream different?
How do I characterize this ore?
If so, what is really an ore type?
CHARACTERIZATION ANALYSIS OF EACH
ORIENTATION SAMPLE
4 Acid Digest Multi- Bulk Element
element analysis by ICP-MS Analysis

Fire Assay, Au, Ag, Pd, Pt Samples

determination by ICP-OES X-Ray
Chemical SKC-PM1
Fluorescence Lead Collection Fire Assay (50-100g)
Assay
Determination of Sulphur XRF
SKC-PM2 4 acid digest (to measure for 60
by sulphur S analyzer (Eltra) elements) (1g)
SKC-PX1
Ammonium Citrate leach analysis (to
Determination of carbon by SKC Konttijärvi SKC-PX2 measure supplied nickel minerals) (1g)
carbon C analyzer (Eltra) Orientation
LECO/ELTRA (Suplhur combustion
Characterization Sample SKC-MS1 test for high sulphur content) (1g)

SKC-MS2 XRF pellet (1g)

SEM X-Ray SKC-BAS1 Bulk QXRD (50-100g)

Particle Mineral Texture,
Automated Diffraction SKC-BAS2
Content & Association
Mineralogy XRD

Bulk Mineral
Analysis
 PX1 4Average
Acid Digest (306P
Grain Size306M),
(mm)
Fire Assay (711P), ELTRA (810L, 811L) (%)
100
SKC-PX1 AveragePrecious
5,0
ModalMetals
Mineralogy
Pt (mg/kg); 0,57
Chem Assay XRD/XRF MLA – gangue

100
150
200
250
300
350
50
4,0

0
Precious Metals 80 Chlorite2,22
Pd (mg/kg);
3,0 Tremolite
(mg/kg);
Diopside3,91 150,13 60
2,0 Au (mg/kg); 0,15
Ferrosilite
Enstatite

Make a rock type mineral

88,20 40
Base Metals (mg/kg); 1,0 Enstatite
Ag (mg/kg); 0,97
4789,73
Ferrosilite 77,50 20 0,0 Diopside
SKC-PX1
0

content profile
Tremolite
-75µm -150+75µm -250+150µm
Na (mg/kg); 167,67 169,80
Biotite 27,63
Base Metals
PX1 - (-250+150mm) Average Particle Surface
Ti Pectolite
(mg/kg); 826,33
9,08 6000
B Area (mm)
(flow)
Characterization Point
• Qemscan
Diopside • XRD/XRF
Enstatite
Chlorite
P (mg/kg); 57,67
121,80
5000
bi
(components)
• CoChemical
(mg/kg);Assay
108,33

Pyrophyllite 9,38 Separation 12,3 %

31,00Feed
Process 28,2 % 1,2 %
Ferrosilite Tremolite
Cr (mg/kg); 604,00 MLA – Smelter
Plagioclase
Sample
0,9 %
4000
Product
MLA – Value 1 MLA – Value 2 Penalty 1
Characterization Point 56,2 % Chlorite Plagioclase
K_feldspar 26,66 Samples Cu (mg/kg); 1190,00
A (flow)
• Mg (mg/kg);
Qemscan
• XRD/XRF
135000,00
Chemical Assay 9,99
• Allanite 3000 Li (mg/kg); 8,00
ai (components)
Quartz Pyrrhotite
Quartz 19,31 Mn (mg/kg); 1350,00
Characterization Point
Wollastonite 8,72
2000
CAverage Particle• Surface
(flow)
PX1- (-150+75mm) Qemscan
• XRD/XRF
Area (mm)
ci (components)
•Ni (mg/kg);
Chemical1380,00
Assay
Titanite 7,32
1000 Diopside Enstatite
Epidote 6,18 V (mg/kg); 63,73
Fe (mg/kg); 11,1 %
Ferrosilite Zn Tremolite
(mg/kg); 85,67
79166,67
Calcite 61,08 1,2 % 0
31,6 % 1,3 %
SKC-PX1
Chlorite Plagioclase
Magnesite 2,02 53,7 %
Quartz Ankerite
Ankerite
Ca (mg/kg); 52,75 Penalty Elements
52700,00 3,0 Pyrrhotite Chalcopyrite
Apatite 67,96

Magnetite
Al (mg/kg); 58,48 2,5
PX1- (-75mm) Average Particle Surface Area
33033,33 (mm)
Ilmenite 60,39
2,0
ALL PROCESS SEPARATION METHODS

B (flow) •
•
Characterization Point
Qemscan
XRD/XRF
bi (components) • Chemical Assay

Separation
Feed Process
Sample Product
Characterization Point
Samples
A (flow)
• Qemscan
• XRD/XRF
• Chemical Assay

ai (components)
Characterization Point
C (flow) •
•
Qemscan
XRD/XRF

ci (components)
• Chemical Assay

Mineralogical mass balance

ORIENTATION STUDY - THE RACE…
31

• This is a race.
• Racers are selected by past demonstrated usefulness and intuition based selection of new
technology
• Stragglers will be shot
• Not everyone will make it across the finish line

• Nothing is sacred. Sacred cows get eaten first unless they continue to prove their
usefulness
• Steak is good but maybe we are now vegetarians?
• And can meatarians and vegetarians eat at the same table, or should they be separated?

This is why sound experimental design is

• Measurement is the key to demonstrating usefulness required at the very beginning, coupled with
competent data analysis & statistical
• Don’t tell me, come and show me
comparison
• Reason, consequence and accountability

30.5.2020
 Process Behaviour 32
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)
 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]

Wills & Napier-Munn 2005 [39]

JKCi
Kojovic et al 2011 [48]

Wills & Napier-Munn 2005 [39]

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

FLOTATION FOR
B (flow) Characterization Point
• Qemscan
• XRD/XRF
bi (components) • Chemical Assay

Separation

RAJAPALOT
Feed Process

Free Au Cleaner Conc Characterization Point

• Qemscan
• XRD/XRF
Sample

A (flow)
Product
Samples

cleaner
• Chemical Assay

ai (components)
Characterization Point

flotation C (flow)
ci (components)
• Qemscan
• XRD/XRF
• Chemical Assay

Prepared Rougher Cleaner

Sample Conc Tails
Free Au

% Recovery
rougher
Chemical Assay
Qemscan SEM
QXRD

Time

Characterization flotation
Cleaner
Data Rougher
Conc
Tails
Co
cleaner Cleaner Lead Collection Fire Assay (50-100g)

flotation Tails 4 acid digest (to measure for 60

Gold elements) (1g)
Cobaltite Ammonium Citrate leach analysis (to
Pentlandite Rougher measure supplied nickel minerals) (1g)
Co
Linarite Conc

% Recovery
Scheelite rougher Chemical Assay
Qemscan SEM
LECO/ELTRA (Suplhur combustion
test for high sulphur content) (1g)
flotation
QXRD

Uraninite Rougher
Time

XRF pellet (1g)

Tails
Bulk QXRD (50-100g)
GRAVITY SEPARATION KONTTIJÄRVI Final
conc

Could be cleaned by
Low mass a shaking table
pull
Tails
Gravity Heavy
Separation
(Knelson) Light

High mass
Flotation
Prepared pull Stage 2
Sample Gravity Heavy
Separation Lead Collection Fire Assay (50-100g)
(Spiral) Light 4 acid digest (to measure for 60
Remove elements) (1g)
light
Ammonium Citrate leach analysis (to
B (flow) Characterization Point minerals
• Qemscan measure supplied nickel minerals) (1g)
gangue
Characterization
• XRD/XRF
bi (components) • Chemical Assay

Separation
LECO/ELTRA (Suplhur combustion
Data Characterization Point
Feed
Sample
Process
Product test for high sulphur content) (1g)
Samples
A (flow)
• Qemscan
• XRD/XRF
• Chemical Assay

ai (components) XRF pellet (1g)

Characterization Point
C (flow) • Qemscan
• XRD/XRF
Bulk QXRD (50-100g)
ci (components)
• Chemical Assay
MAGNETIC SEPARATION KONTTIJÄRVI
Lead Collection Fire Assay (50-100g)

4 acid digest (to measure for 60

Prepared elements) (1g)

Sample Ammonium Citrate leach analysis (to

measure supplied nickel minerals) (1g)
Electromagnetic
Separation – Conc 1 LECO/ELTRA (Suplhur combustion
test for high sulphur content) (1g)
1 Ampere
Characterization XRF pellet (1g)
(LIMS)
Data Electromagnetic Bulk QXRD (50-100g)
Ferrite Conc 2
Separation –
equivalent
2 Ampere
(MIMS)
NdFeB permanent ElectromagneticS Conc 3
magnet equivalent eparation –
B (flow) Characterization Point

3 Ampere
• Qemscan
• XRD/XRF
bi (components) • Chemical Assay

Feed
Separation
Process (HIMS)
Sample Product

Electro-magnet
Characterization Point
Samples
A (flow)
• Qemscan
• XRD/XRF
• Chemical Assay

ai (components)
C (flow) •
Characterization Point
Qemscan
equivalent Final
Tails
• XRD/XRF

ci (components)
• Chemical Assay
 37
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

(5) Orientation Study

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
 38

DATA ANALYSIS OF ORIENTATION STUDY SAMPLES

• 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?

• What is the process engineering simulation outcomes using orientation samples as input
TO DIFFERENTIATE AND RANK THESE THREE
SAMPLES IN CONTEXT OF LEACHING RECOVERY

% recovery

Needs to be done on small sample

masses and still be relevant

Time
Which of these results would I prefer to have in my plant?
Rank them and compare against mineral content
THE AXB BREAKAGE CURVE

T10 (%)
38.1

19.8

10.4
t10 = A[1-e(-b.Ecs)]

4.4
A*b =23 (hard ore)

0.2 0.5 1.0 2.0

THE RANKING OF RECOVERIES
]
100%
100%
90%
90%
80%
80% Ore Type A Where:
Ore Type A
70%
70% Ore Type B
Ore Type C
Q = Cu Output %
60%
60%
Recovery
Cu Recovery

Ore Type B
50%
50%
Ore Type D t = Time leached
Ore Type A Fitted
40%
40% Ore Type C
Cu

30%
30%
Ore Type B Fitted
Ore Type C Fitted
Sample R s R*s
20%
20%
Ore Type D
Ore Type D Fitted Ore Type A 76.9% 0.45 34.6
10%
10%

0%
0%
00 55 10 10 15 15 20 2025 30
25 35 30
Ore Type B 54.3% 0.16 8.7
Time
Time Leached(h)
Leached (h)

Ore Type C 49.2% 0.17 8.2

Ore Type D 81.8% 1.21 98.9

 This is what is studied in the
Mapping Study to look at

Talc/chlorite mineral content (mg/kg)

process domains & variability

Develop a relationship between XRD/XRF

and chemical assays for each rock type
Using data you have already collected

This is modelled in a
geometallurgical block model
Leaching/Flotation Recovery ranking R*s

(Theoretical example – not data)

CONCLUSIONS OF ORIENTATION STUDY
43

• An understanding of what works and what does not for this deposit

• Useable relationship between bankable tests and their proxies

• Experimental design for what tests are to be done on the Mapping Study 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
• QA/QC 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

30.5.2020
 44

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
THE GEOMETALLURGICAL DATA MATRIX IS READY!!! 45

Small scale comminution testing

PLT Crushing Index A*b from RBT Batch Grind
Grind and Flotation Hyperspectral Characterization
Testing & Core Imaging
Texture-based Classified
recoveries images
Texture-based Texture
liberation groupings

Intact texture Classified Mineralogic Texture-based

libraries textures al definition FEM models

MLA and optical microscopy

30.5.2020
 Compare each
46
parameter against
target process

In context of outcomes of the Domain Study

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
in the data set Domaining Study Repeatability
• Standard Deviation
• 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 Plot each parameter against
Patterns of correlation,
controls of the data set every other parameter in a
Different sub-populations
cross correlation graph matrix

Determining what is relevant and Examine the probability plot of Differences in gradient,
what is not in matrix data set each parameter Different sub-populations

Examine outcomes in context of Examine the histogram plot of Shape of distribution,

each parameter Tight or spread
Domaining Study & Orientation Study
Mirrored patterns and
Determine interaction terms correlations, leading and
for parameters if possible lagging indicators
EXAMINE THE CORRELATION BETWEEN ALL 47

PARAMETERS AGAINST ALL OTHER PARAMATERS

EXAMINE ALL PARAMETERS IN PROABLITY PLOTS 48

TO DIAGNOSE SUB-POPULATIONS
EXAMINE ALL PARAMETERS IN HISTOGRAMS TO 49

DIAGNOSE SKEWNESS SHAPE OF POPULATIONS

PRINCIPLE COMPONENT ANALYSIS (PCA) 50

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
PRINCIPLE COMPONENT ANALYSIS (PCA) – CASE STUDY P
51

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
 52
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

(8) Class Definition

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.
 Geometallurgy Data Matrix 53

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
PCA – CASE STUDY P 54
CASE STUDY P 55

Comminution
CASE STUDY P 56
CASE STUDY P 57

Valuable target metals

CASE STUDY P 58

Comminution
CASE STUDY P 59

Gangue minerals
CASE STUDY P 60

Penalty elements
CASE STUDY P 61
CASE STUDY P 62
CASE STUDY P 63
CASE STUDY P – CLASSES PROJECTED BACK INTO 64

DRILL CORE

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)
 65
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

(9) Process Modelling

(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)
MODELING EXAMPLE R
66

23
Hematite-pyrite Class BMWI Model
21
Model:
19

17
BMWI= -66.65 +1.24Flu0.61 -0.27Hem -1.49Py +19.23SG
Predicted

+8.39(Py/Sul)0.84 +0.72QHard1.22 +0.72(Chl/Sul)1.31

15

13

S.E.=0.90 R2=0.84 Model prediction relative error =7.4%

11

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

S.E.=0.41 R2=0.83 Model prediction relative error =2.1%

15
N=47
15 17 19 21 23
23
Measured

21 Hematite Class BMWI Model

19
Model:
BMWI= -44.99 +0.67Ser0.56 +0.19Sid +0.13Flu1.64 +0.63Bar0.81
Predicted

17

+6.80SG +1.90QHard
15

13

11

N=36
S.E.=0.99 R2=0.66 Model prediction relative error =6.2%
9
9 11 13 15 17 19 21 23
Measured
 67

PROCESS SIMULATION FOR EACH END MEMBER 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
ECONOMIC SIGNATURE MODELLING
COMPARISON OF DIFFERENT PROCESS PATHS FOR EACH 69

ORIENTATION STUDY SAMPLE

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
Process Path 9 Flotation
Separation SGFDLB

Process Path 10 Leach

Gravity Flotation
SGFDLB
Orientation Gravity Flotation Leach
Process Path 11 Ore Sorting
Step 3 SOSBGC SOSGFC SOSGFDLC
 70

EXPERIMENTAL DESIGN TO STATISTICAL ANALYSIS

• How can we know if which process separation method is more

effective?
• Flotation/Hydrometallurgy/Gravity/Magnetic

• How can we know that the process response is really statistically

different between rock types?
• T-test
How do we draw conclusions
• ANOVA if you can do it that are defendable?

30.5.2020
 71

COMPARISON OF
DIFFERENT

Orientation Study
• Which process path is more effective in the

PROCESS PATHS
recovery of each target metal?

• Which process path is most effective in

FOR ALL
recovery of the 2-3 most valuable metals?

• What is the mineralogical signature that

ORIENTATION
controls that process path?

STUDY SAMPLES Mapping Study

• Process domains
• Variability
 72

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
 Correlations and 73
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
CUMULATIVE SUMMATION (CUSUM) ANALYSIS
74

Time Series Recovery Chart

96
Change in Circuit
94

A change was made to a flotation plant circuit

92
at day 85 and the data was analyzed to
determine if there was a change in recovery
Recovery (%)

90
performance of the circuit.
88
The time series plot does not provide any
86 visible indication of any change in the day to
day recovery data.
84

82
0 20 40 60 80 100 120 140 160
Day

Example Source: T. Napier-Munn

CUMULATIVE SUMMATION (CUSUM) ANALYSIS
75

5
Change in Circuit
0
The cusum plot identifies four periods:
-5 • two –ve gradients
-10 • one horizontal gradient
CUSUM

-15 • one positive gradient

-20
Difference between lowest and
-25 highest recoveries are only 1%
-30
0 20 40 60 80 100 120 140 160
Day

μ=88.87% (overall mean of dataset)

Example Source: T. Napier-Munn
 Cu %, Cu/S
Fe %, S %

0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0
1
2
3
4
5
6
7
8
9
128 - 130 128 - 130
130 - 132
132 - 134 132 - 134
134 - 136
136 - 138 136 - 138
138 - 140

Cu/S
140 - 142 140 - 142

Cu_pct
142 - 144
144 - 146 144 - 146
146 - 148
148 - 150 148 - 150
150 - 152
152 - 154 152 - 154
154 - 156
156 - 158 156 - 158
158 - 160
160 - 162 160 - 162
162 - 164
164 - 166 164 - 166
166 - 168
168 - 170 168 - 170
170 - 172
S_pct

172 - 174 172 - 174

Fe_pct

174 - 176
176 - 178 176 - 178

Depth (m)

Depth (m)
178 - 180
180 - 182 180 - 182
182 - 184
184 - 186 184 - 186
186 - 188
188 - 190 188 - 190
190 - 192
192 - 194 192 - 194
194 - 196
196 - 198 196 - 198
198 - 200
200 - 202 200 - 202
202 - 204
204 - 206 204 - 206
206 - 208
208 - 210 208 - 210
210 - 212
212 - 214 212 - 214
214 - 216
216 - 218 216 - 218
218 - 220
220 - 222 220 - 222
222 - 224
224 - 226 224 - 226
226 - 228
228 - 230 228 - 230
230 - 232
232 - 234 232 - 234

Parts per million (ppm)

0
5000
10000
15000
20000
25000
30000
35000
40000

128 - 130
130 - 132
132 - 134
134 - 136
136 - 138
138 - 140
140 - 142
K_ppm

142 - 144
Al_ppm
Ca_ppm
Mg_ppm

144 - 146
146 - 148
148 - 150
150 - 152
152 - 154
154 - 156
156 - 158
158 - 160
160 - 162
162 - 164
164 - 166
166 - 168
168 - 170
170 - 172
172 - 174
174 - 176
176 - 178
Depth (m)

178 - 180
180 - 182
182 - 184
184 - 186
186 - 188
Case Study P

188 - 190
190 - 192
192 - 194
194 - 196
HOW DO WE LOOK AT DOWN HOLE DATA?

196 - 198
198 - 200
200 - 202
202 - 204
that can filter data

204 - 206
206 - 208
208 - 210
210 - 212
212 - 214
214 - 216
216 - 218
218 - 220
Need a statistically valid method

220 - 222
222 - 224
224 - 226
226 - 228
228 - 230
230 - 232
76

232 - 234
 -1,8
-1,6
-1,4
-1,2
-0,8
-0,6
-0,4
-0,2
0

-2
-1
0,2
-4,0
-3,0
-2,0
-1,0
0,0
1,0
2,0
3,0
4,0
5,0
128 - 130 128 - 130
130 - 132
132 - 134 132 - 134
134 - 136
136 - 138 136 - 138
138 - 140
140 - 142 140 - 142
142 - 144
144 - 146 144 - 146
146 - 148
148 - 150 148 - 150
150 - 152
152 - 154 152 - 154
154 - 156
156 - 158 156 - 158

cusum Cu
158 - 160

cusum Cu/S
160 - 162 160 - 162
162 - 164
164 - 166 164 - 166
166 - 168
168 - 170 168 - 170
170 - 172
172 - 174 172 - 174
174 - 176
176 - 178 176 - 178
178 - 180
180 - 182 180 - 182
182 - 184
184 - 186 184 - 186

Depth (m)
Depth (m)
186 - 188
cusum S

cusum Fe

188 - 190 188 - 190

190 - 192
192 - 194 192 - 194
194 - 196
196 - 198 196 - 198
198 - 200
200 - 202 200 - 202
202 - 204
204 - 206 204 - 206
206 - 208
208 - 210 208 - 210
210 - 212
212 - 214 212 - 214
214 - 216
216 - 218 216 - 218
218 - 220
220 - 222 220 - 222
222 - 224
224 - 226 224 - 226
226 - 228
228 - 230 228 - 230
230 - 232
232 - 234 232 - 234
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

-35000
-30000
-25000
-20000
-15000
-10000
0
5000

-5000
10000
15000

•
•

128 - 130
130 - 132
132 - 134
134 - 136
136 - 138
138 - 140
140 - 142
142 - 144
144 - 146
146 - 148
148 - 150
THE CUSUM TOOL

150 - 152
152 - 154
154 - 156
156 - 158
158 - 160
160 - 162
162 - 164
164 - 166
166 - 168
168 - 170
170 - 172
172 - 174
prevailing mean.

174 - 176
176 - 178
178 - 180
180 - 182
182 - 184
184 - 186
Depth (m)

186 - 188
188 - 190
190 - 192
cusum K
cusum Al
cusum Ca
cusum Mg

192 - 194
Case Study P

194 - 196
any point is not important

196 - 198
198 - 200
200 - 202
202 - 204
204 - 206
The gradient of the line over a

206 - 208
208 - 210
210 - 212
212 - 214
characteristic period indicates the

214 - 216
The absolute value of the cusum at

216 - 218
218 - 220
220 - 222
222 - 224
224 - 226
226 - 228
228 - 230
230 - 232
232 - 234
0
10000
20000
30000
40000

-20000
-10000
77
HYPERSPECTRAL IMAGING OF DRILL CORE 78

Depth
Each line is 4m of drill core characterised
79

WRD 432 Calcium

10000
9000
8000
7000
(ppm)

6000
5000
4000
3000
2000
1000
0
250 300 350 400 450 500 550 600 650 700 750
Down Hole Depth (m)
WRD 432 Cu/S
1
0.9
0.8
0.7
(ppm)

0.6
0.5
0.4
0.3
0.2
0.1
0
250 300 350 400 450 500 550 600 650 700 750
Down Hole Depth (m)
WRD 432 Magesium
50000
45000
40000
35000
(ppm)

30000
25000
20000
15000
10000
5000
0
250 300 350 400 450 500 550 600 650 700 750
Down Hole Depth (m)
 Conventional
column leach Feed
tests material
Exploration
Geometallurgy much more
drill core
modelling reliable

Proxy
diagnostic
leach tests
Engineering
design
 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
leaching Ore Domain 1 • Net revenue per quarter
time period
Ore Domain 2 • Metal produced total
• Metal produced per hour
over schedule
 83
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?
 84
• The geomet program has been successful in context of original objectives

• The original Orientation Study samples really were samples where the process
extremes were observed

• The selected grind size was appropriate

• A viable process flow sheet was developed for each end member rock texture

• An economic signature was developed for each orientation sample

• The mineralogy that controls the favoured process flow sheet was diagnosed,
then procedures to map them back in the deposit in a spatial context was
developed

• Process defined ore domains were developed across all major geological
structures in the deposit

• Sampling is consistent enough to make a production procedure

• A viable link between the geometallurgy campaign and the conventional

process modelling characterization was established
 Define geometallurgical
objective(s)

Conduct a comprehensive multivariate

statistical study on all available existing data.

Select end member rock

textures. Orientation Samples

Define Useful process targets

based on characterization

Experimental
Design

Conduct experimental Data

test work QA/QC

Conduct process Do a mineralogical

Determine what
engineering simulations for reconciliation for each
mineralogy controls
each process path for each process separation test
each process path
Orientation Sample and each process path.

Compare process path

For all Orientation Samples effectiveness for each
together in one data set, individual Orientation Sample
select the 2 or 3 most For each individual
effective process paths Orientation Samples,
Compare process path effectiveness
select the most
for all Orientation Samples.
effective process path
GEOMET TO OPERATION
Geometallurgy Laboratory scale Pilot Scale/Production scale

Behavior 1 Behavior A

Behavior 2 Behavior B

Behavior 3 Behavior C

Behavior 4 Behavior D

Behavior 5 Behavior E

Mineral Instrumentation
signatures Measurement
 Comparative Comminution
Hardness Index Tester HIT
(Toni Kojovic– SimSAGe Pty Ltd )

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.
Comparative Comminution Hardness Test Geopöyrä
(Marcos Bueno – University of Oulu and R. Chandramohan)

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)

Broken product real time t10 (%)

PSD sizing measurement Calculation of A*b

A = 49.1 50

(optical system)
40

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)

Torvela, J., Bueno, M., Liedes, T. and Luukkanen, S. (2020): The Geopyörä
Breakage Test, University of Oulu, Minerals Engineering journal “in press”
• GTK-Mintec
• 5 tph pilot plant
• Comminution/gravity sep/magnetic sep/flotation
 90
Machine • Outotec HSC Digital Twin System Particle Size Distribution
Learning AI (dry crushed rock)
• Outotec SMART flotation cells
Mineral Characterization
Data Lake Particle • Geopörä system (XRF)
Library Tracking System
• MetheOre dry PSD SAG feed Particle Size Distribution
• Malvern wet PSD slurry (wet slurry)
PLC Server PLC Server PLC Server PLC Server Mineral Characterization
Bank Bank Bank Bank (Raman Spectroscopy)
Cyclone Overflow
CycloneTrac
Product
Cyclone Froth Camera
Rod Mill Feed (Frothvision)
Feed
Cyclone
Rod Mill Nest
Rod Mill
Product
Cyclone
Underflow

Ball Mill
Ball Mill Flotation
Feed Concentrate
Ball Mill Flotation
Product Cell
Flotation
Tails
INTEGRATE FOUR PARADIGMS 91

GTK instrumented pilot

plant, running steady
state and dynamic state

Next
generation
process
models

Technical outcome exchange

Outotec HSC Metso/JKMRC across all 4 paradigms
theoretical modelling empirical modelling

30.5.2020
Direct link
http://tupa.gtk.fi/raportti/arkisto/72_2019.pdf
QUESTIONS
93

?
 KIITOS
[email protected]
www.gtk.fi
WHERE MOST CURRENT GEOMET PROGRAMS FALL
95

OVER
Data Collection
• Samples collected without spatial coordinates in the ore body
• Tests done on parcel of rock in non-representative way
• Not enough samples collected
• Test work based on composites that mask variability
• Different tests done on wildly separate parcels of rock with very few or no rock samples with more
than one test type (for example A*b and BMWi)
• The wrong hypothesis used to collect data
• No assay data collected with metallurgical testing
• Tests done years apart by different people and laboratories (different methods of reporting)
WHERE MOST CURRENT GEOMET PROGRAMS FALL
96

OVER
Analysis
• Test data not related to phenomenon being modelled
• Too many things being modelled at once, confusing the outcome
• Analysis done in isolation to the rest of mining process due to mining culture (silo effect)

30.5.2020
CONCLUSIONS
97

• Strong experimental design is a must

• Clear objectives, with clear milestones
• Clear agreement on what success looks like
• Clear agreement on what the scope and budget allocated is
• Project sovereignty has to be defended
• Mission creep can be lethal
• Agreeing to change things half way through can lead to running out of budget at the end
• Agreeing to help others who are not invested in the final outcome is the greatest risk for this

• Regular communication is the key success

• Regular and comprehensive documentation
• ‘Upward’ management and all stake holders is required to keep the integrity of the project
CASE STUDY P 98
CASE STUDY P
99
CASE STUDY P 100

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

 CASE STUDY P 101

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

CASE STUDY P 102
CASE STUDY P 103
CASE STUDY P 104
 PRINCIPLE COMPONENT ANALYSIS (PCA)
105

• The PCA approach is to spatially model classes created during

principal component analysis:
• Each class has a given distribution.
• Assign processing response and variability to each block by association.

• Problem/Issues:
• Class based characterisation has been set up to group based on fundamental controls, 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.

30.5.2020
 PRINCIPLE COMPONENT ANALYSIS (PCA) 106

• Principal Component Analysis (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.

• First principal component accounts for as much data variance as possible and each subsequent
principal component accounts for remaining data variance.

30.5.2020
PRINCIPLE COMPONENT ANALYSIS (PCA) – EXAMPLE A 107

Barite

Silica

Sphalerite

Galena

Fe sulphides

Organic carbon

N=20,250

Data Desk
software

Mineralogy calculated from assays for a massive sulphide system