24
Mineralogical Modelling for Flowsheet Design Elizabeth Whiteman Presented by Lori Kormos November 18, 2015
Mineralogical Modelling for Flowsheet Design
Elizabeth Whiteman Presented by Lori Kormos
November 18, 2015
Elizabeth Whiteman
• Graduate of University Queensland, Materials Engineering, 2002
• Work History
– CSIRO – Minerals Division
– Intellection
• Part of team that developed QEMSCAN software
• Manager of Customer Support
– XPS since 2009
• Process Mineralogy team
• Mineralogical and Flotation Programs
• Paper Presented at MEI Flotation ’15, Cape Town - “A Practical Process Mineralogy
Approach to Advancing the Flowsheet for the Kamoa Project”
– Conference Proceedings
– Minerals Engineering
Case Study – Kamoa Project
– Owned by Ivanhoe Mines Ltd.
– Located in the Katanga Province of the DRC, Africa
– Large, high grade copper sulphide deposit and is an extension of the Central African
Copperbelt
Introduction – Objective of Paper
– Describe the use a Process Mineralogy centric methodology as a valid predictive tool in
flowsheet design
• Complex ore bodies can undergo many stages of empirical flotation testing
• Mineralogical data is used in this methodology to eliminate multiple stages of empirical
testing and focus on requirements of the ore
• By focusing on ore requirements, we have removed inherent limitations in flotation
equipment i.e. building flowsheet around Denver cell for example
• Flowsheet design and simulation is based entirely on mineralogical measurement data
collected from simple kinetic flotation test
Process Mineralogy and Kamoa
• Kamoa is a complex ore body …
• Cu mineralogy includes chalcopyrite, bornite, covellite, chalcocite (as well as oxides,
carbonates and native Cu)
– Ratios of sulphide mineralogy change between supergene and hypogene horizons
– Variability is high within the horizons geospatially
– Ratios affect feed grade and flowsheet response models and predictions
• Cu mineralogy grain size
– This is consistent between all ores previously tested
– Some Cu sulphide mineralogy ~50µm with remainder always around an 8-10µm grain size
• Flowsheet development has targeted fine grinding and mixed collector suite to handle
variation in sulphide mineralogy
What we already know…
Previous Kamoa WorkExisting MF2 Flowsheet (Lotter et. al. 2013)
Cu Grade % Cu Recovery % SiO2 Grade %Feed 3.3 - 3.9
Final Concentrate 32 - 45 83 - 85 19 - 26Scavenger Tails 0.5 - 0.7 11 - 14
• MF2 arrangement complex and $$$ intensive (capex and opex)• Current Results
• Required a new approach to answer the following questions outlining new objectives 1. Can we get beyond 83% recovery with high feed grades? 2. Are there sufficient liberated Cu sulphides in early stages of roughing to warrant a
separate cleaner circuit that does not require regrinding? 3. Can a single stage grind effectively replace the more complex and expensive MF2
arrangement? What would be a suitable primary grind? 4. Can the SiO2 dilution be reduced to near 14% and Cu grade maintained above 28%
regardless of Cu sulphide mineralogy?5. Can tailings grade of 0.4% Cu be achieved? Scavenger Tails are mostly coarse locked
Cu sulphides – How do we minimize this?
Previous Kamoa WorkExisting MF2 Flowsheet Typical Results (Lotter et. al. 2013)
Cu Grade % Cu Recovery % SiO2 Grade %
Final Concentrate 32 - 45 83 - 85 19 - 26Feed 3.3 - 3.9
Final Concentrate 32 - 45 83 - 85 19 - 26Final Concentrate 32 - 45 83 - 85 19 - 26Final Concentrate 32 - 45 83 - 85 19 - 26Scavenger Tails 0.5 - 0.7 11 - 14Scavenger Tails 0.5 - 0.7 11 - 14
ApproachHow to Achieve New Targets?
• New methodology of combining kinetic flotation test with comprehensive mineralogy to effectively design the new flowsheet with no empirical testing
• Kinetic floats at 150µm, 106µm, 75µm, 53µm and 38µm were performed on new composite material
• Mineralogy was completed on three best performers.
Kinetic FloatResults
Improved Recovery with Finer Grind
0
5
10
15
20
25
30
35
40
0 20 40 60 80 100
Gra
de %
Cu
Cu Recovery %
150µm
106µm
75µm
53µm
38µm
Impr
oved
Gra
des
with
Fin
er G
rind
106µm, 53µm and 38µm duplicated for mineralogical measurement
28% Cu target
MineralogyProcedure
• Each Concentrate and the Rougher Tailing of the 38µm, 53µm and 106µm grind was sized and prepared for mineralogy by QEMSCAN
• Kinetic mass and value balance data was used to calculate size-by-size mineral recoveries
• Given close textural association of the individual Cu sulphide minerals – these were combined into one grouping of “Cu sulphides” which is a true depiction of bulk sulphide
liberation required for this ore – This also simplifies liberation for subsequent modelling and simulation
BackgroundChalcopyriteBorniteChalcociteCovelliteAzurite/MalachiteChrysocollaNative Cu and Cu OxideCarrolitePyriteGangue
}} BackgroundCu SulphidesCu Silicates/OxidesPyriteNSG
Recovery by Liberation
Are there sufficient liberated Cu sulphides in early stages of roughing to warrant a separate
cleaner circuit that does not require regrinding?
0
10
20
30
40
50
60
70
80
90
100
1 10 100
reco
ver
y %
Cu
su
lph
ide
wrt
to
Fee
d
Particle Size (µm)
38µm Grind 10 Minute Float Time
0
10
20
30
40
50
60
70
80
90
100
1 10 100
reco
ver
y %
Cu
su
lph
ide
wrt
to
Fee
d
Particle Size (µm)
53µm Grind 10 minute Float Time
0
10
20
30
40
50
60
70
80
90
100
1 10 100
reco
ver
y %
Cu
su
lph
ide
wrt
to
Fee
d
Particle Size (µm)
106µm Grind 10 Minute Float Time
Locked (<30%)
Low Grade Middling (30-80%)
High Grade Middling (80-95%)
Liberated (>95%)
Free (100%)
Kinetics are similar
between grinds for
liberated and free Cu
sulphides
>90% of liberated Cu
sulphides are
recovered by 10
minutes of flotation
Recovery by LiberationCan the SiO2 dilution be reduced and Cu grade maintained above 28% regardless of Cu sulphide mineralogy?
Recovery in particle sizes <10µm
Can tailings grade of 0.4% Cu be achieved? How do we minimize coarse locked Cu sulphide losses?sulphide losses?
0
10
20
30
40
50
60
70
80
90
100
1 10 100
reco
very
% C
u su
lphi
de w
rt to
Fee
d
Particle Size (µm)
38µm Grind 40 Minutes Flotation Time
0
10
20
30
40
50
60
70
80
90
100
1 10 100reco
very
% C
u su
lphi
de w
rt to
Fee
d
Particle Size (µm)
53µm Grind 40 Minute Float Time
0
10
20
30
40
50
60
70
80
90
100
1 10 100
reco
very
% C
u su
lphi
de w
rt to
Fee
d
Particle Size (µm)
106µm Grind 40 minute Float Time
Locked (<30%)
Low Grade Middling (30-80%)
High Grade Middling (80-95%)
Liberated (>95%)
Free (100%)
Recovery by Liberation
20
30
40
50
60
70
20
30
40
50
60
70
80
Middling flotation improved to >80% by 40 minutes
Locked Cu sulphide recovery plateaus around 50% regardless of primary grind
Regrind stage still necessary to improve recovery >83% and minimize SiO2 recovery to concentrate
38µm 53µm 106µm% Cu Loss 8.33 10.75 12.56
% Cu Loss >25µm 3.23 6.83 8.58CuS Grain Size µm 8 8 8
Flowsheet Simulation
• Simulations completed on mineralogical dataset
• Liberation data suggests a bypass without the need for regrinding is possible
– Simulation looks at 3 minute and 10 minute bypass concentrate
• Can tailings grade of 0.4% Cu be achieved? How do we minimize coarse locked Cu
sulphide losses?
– Finer primary grinding than 38µm not economical
– Simulation assesses scalping of coarse particles for reprocessing and subsequent tailings grade
• Can the SiO2 dilution be reduced to near 14%?
– Simulation assesses SiO2 recovery by liberation and particle size and models bypass concentrate
cleaning potential
• Can a single stage grind effectively replace the more complex and expensive MF2
arrangement? What would be a suitable primary grind
– Simulation assesses all of the above criteria at the 3 grinds of 106µm, 53µm and 38µm
Flowsheet SimulationSimulation 1
Rougher Tailing Discard Final Tailing -25µm (-38µm)
Rougher Concentrate
To Final Concentrate
Bypass Cleaner Tailing
Rougher Tailing +25µm (+38µm)
Rougher Float10 minutes
Particle Scalp 25µm (38µm)
Bypass Cleaner Flotation
Regrind and Scavenger Float
Feed
Flowsheet SimulationSimulation 2 and 3
• Target for final concentrate increment: – >28% Cu – <14% SiO2
• Remember SiO2 recovery is all liberated and <10µm indicating entrainment
3 minute modelled concentrate 10 minute modelled concentrate% Cu % SiO2 % Cu Recovery % Cu % SiO2 % Cu Recovery
38µm 40 9 66 38µm 37 13 7753µm 42 7 59 53µm 40 10 70
106µm 39 10 54 106µm 38 12 63
3 minute concentrate 10 minute concentrate% Cu % SiO2 % Cu Recovery % Cu % SiO2 % Cu Recovery
38µm 25 31 71 38µm 19 39 8453µm 29 26 65 53µm 22 35 81
106µm 20 38 77 106µm 20 38 77
Flowsheet SimulationBy Pass Concentrate
% Cu % Cu Recovery% SiO2
312638
3 minute modelled concentrate% Cu
404239 10
% SiO2
393538
3 minute modelled concentrate% Cu Recovery
39 10
3 minute modelled concentrate
39 10
% SiO2
97
39 10
10 minute modelled concentrate
37 1340 1038 12
% SiO2
37 1340 1038 12
10 minute modelled concentrate
37 1340 1038 12
10 minute modelled concentrate
37 1340 1038 12
% Cu37 1340 1038 12
• Target for final scavenger/rougher discard tails: – <0.4% Cu – <11% Cu Loss
• Remember Cu losses are mostly locked in coarse particle sizes
Rougher Tail after 40 minutes% Cu % Cu Loss
38µm 0.5 953µm 0.7 11
106µm 0.7 12
Rougher Tail After 10 minutes% Cu % Cu Loss
38µm 0.7 1653µm 1.0 20
106µm 1.1 23
Flowsheet SimulationRougher Middling and Tail
% Cu % Cu Loss0.5 9
% Cu % Cu Loss0.5 90.70.7
% Cu % Cu Loss% Cu % Cu Loss0.71.01.1
10 minute Rougher Tail -25µm% Cu % Cu Loss
38µm 0.5 1053µm 0.7 8
106µm 0.8 9
40 minute Rougher Tail -25µm% Cu % Cu Loss
38µm 0.4 653µm 0.4 4
106µm 0.4 4
% Cu % Cu Loss0.4 60.4 40.4 4
% Cu % Cu Loss0.4 60.4 40.4 4
% Cu % Cu Loss
0.7 80.8 9
% Cu % Cu Loss0.50.7 80.8 9
Flowsheet SimulationRecommendations from mineralogy
Rougher TailingDiscard Final Tailing -25µm (-38µm)
Rougher Concentrate
To Final Concentrate
Bypass Cleaner Tailing
Rougher Tailing +25µm (+38µm)
Rougher Float 3 minutes
Rougher TailingRougher Mids3-40 minutes
Bypass Cleaner Bypass Cleaner
Flotation
Bypass Cleaner Regrind and Scavenger Float
Feed
Rougher Tailing
Particle Scalp 25µm (38µm)
Cu: 41.8% Rec %: 58.6% SiO2: 7.3% MP: 6.1%
Cu: 0.40% Loss: 4.0%
Cu: 3.67% R: 37.5% MP: 44.5%
Regrind to 10µm
53µm Primary Grind
By pass cleaning needs to be an entrainment controlled process
Cu: 3.67%
R: 37.5%
MP: 44.5%
Regrind to 10µm
Flowsheet Simulation
Did we achieve our project goals with this mineralogical design method?
Can we get beyond 83% recovery with high feed grades?
Is there sufficient liberated Cu sulphides in early stages of
roughing to warrant a separate cleaner circuit that does not
require regrinding?
Can a single stage grind effectively replace the more
complex and expensive MF2 arrangement? What would
be a suitable primary grind
Can tailings grade of 0.4% Cu be achieved? How do we
minimize coarse locked Cu sulphide losses?
Can the SiO2 dilution be reduced?
Actual Flowsheet
Single-Stage Grind to optimum d80 size 53 um
Rougher Flotationt=5’ 30% solids
Rougher Tailings <53 µm
RougherTailings >53 µm
70% Collector Dose in Mill for 5 mins
t=35’
ScavReclnrTails
Scav ClnrTails
Scavenger Cleaner
Scav ClnrScavenger Recleaner
ScavReclnr
Saleable Concentrate
Bypass Cleaner
Bypass Recleaner
10 um
12% Solids
12% Solids
12% Solids
12% Solids
Scavenger ReclnrConcentrate
t=13’
t=10’
ScalpingCyclone
t=5’
t=3’ t=2’
Bypass ReclnrConcentrate
Concentrate
Concentrate
t=5’
t=2’
t=3’
Regrind Mill
53 um
Actual vs. Simulation
How do they compare?
• Actual Flowsheet Results
– 5 minute by-pass concentrate followed by low density cleaning
– Middling flotation to 40 minutes
– 53µm scalp with oversize to regrind feed
• Simulated Flowsheet Results from Mineralogical Data
– 3 or 10 minute by-pass concentrate
– Middling flotation to 40 minutes
– 25µm scalp with oversize to regrind feed
Cu Grade % Cu Recovery % SiO2 Grade % MP %
5 min By-pass Concentrate 41.34 65.3 10.9 8.57
Final Concentrate 38.99 88.3 14.56
Rougher Tails 0.35 4.9
Cu Grade % Cu Recovery % SiO2 Grade % MP %
3 min By-pass Concentrate 41.8 58.6 7.3 6.1
10 min By-pass Concentrate 39.5 70.3 10.3 7.8
Final Concentrate - - -
Rougher Tails 0.4 4.0
Conclusions
• A simple combination of kinetic floats with size-by-size mineralogy was used
• Designed a flowsheet based on the mineralogical data and the kinetics of key minerals by
particle size and liberation
• Enhanced laboratory testwork by:
– Effectively replacing empirical flotation testing and accurately predicting the physical
response of the ore at set target grind
– Removed inherent laboratory equipment limitations and identified the process required
– Guiding physical testing to begin at demonstration and optimization rather than discovery
• Process mineralogy can be used as a valid predictive tool in process design
Acknowledgements
• Management of XPS Consulting & Testwork Services
• Management of Ivanhoe Mines Ltd
