See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281352208

Cu–Ni processing improvements at First Quantum’s Kevitsa mine

Article  in  Minerals Engineering · August 2015

DOI: 10.1016/j.mineng.2015.08.005

CITATIONS READS
8 623

3 authors, including:

Benjamin Musuku Ishamael Muzinda

Aalto University Aalto University
4 PUBLICATIONS   13 CITATIONS    1 PUBLICATION   8 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Benjamin Musuku on 09 March 2020.

The user has requested enhancement of the downloaded file.

 Cu-Ni processing improvements at First Quantum’s Kevitsa mine.

Benjamin Musukua, Ishmael Muzindaa and Barry Lumsdenb*.

a
FQM Kevitsa Mining Oy, Kevitsantie 730, 99670 Petkula, Finland.
b
Ausmetec Pty Ltd, 47/176 South Ck Rd, Cromer, NSW 2099, Australia
*
Corresponding Author. Tel.:61 2 99715716:

Email address: barrylumsden@ausmetec.com.au

ABSTRACT

First Quantum’s Kevitsa Mine is located in northern Finland and mines a large, low
grade, polymetallic ore body. The processing rate is currently 6.7 million tonnes per annum
(mtpa) up from the design rate of 5.5 mtpa. Payable metals include nickel, copper, gold and
PGM group metals. Profitability, at high volume, low grade operations like Kevitsa, is highly
geared to metal prices and process improvements. Unusually, for a high tonnage low grade
operation, metal losses at Kevitsa are primarily in the finer fractions.
Since the operation commenced there has been a focus on maximising the separation
process so as to improve the Cu-Ni process profitability. This has included:
1. Optimising the flowsheet
2. Installing column cells
3. Improving fines recovery
This paper will outline these process improvements, their introduction and efficacy.

1. Introduction

Kevitsa Mining Oy is a Ni-Cu-PGE project in Finland and is owned by First Quantum

Minerals. The deposit was acquired from Scandinavian Minerals Limited in 2008 and a
decision to develop was made in November 2009. Construction of the mine facilities
including the processing plant began in 2010 and was completed in May 2012.
Commissioning of the processing plant began in earnest in June of the same year leading to
production of the first concentrate. Commercial production status was achieved well ahead of
schedule and by August the project had moved from commissioning to commercial
production status. Currently the operation is processing 6.7 mtpa up from the design
throughput of 5.5 mtpa and a permit to operate at 10 mtpa has since been awarded giving
room for further expansion.

1.1.Mineralogy

The Kevitsa Ni-Cu-PGE deposit is disseminated in style while having some minor
sulphide veins. Pyrrhotite is the main sulphide mineral followed by chalcopyrite and
pentlandite. The sulphide grain size is fine to medium. In near surface parts of the
mineralisation, pyrrhotite is partly replaced by pyrite and pentlandite by millerite and
heazlewoodite (Kojonen et al., 2008). Major and minor sulphide minerals in the ore are listed
with some of their characteristics in Tables 1 and 2.

Table 1
Major sulphide minerals for the Kevitsa ore
Mineral name Emipirical formula Chemical formula Crystal system Hardness
Pentlandite Fe2+4.5Ni4.5S8 (Fe,Ni)9S8 isometric 3.5-4
Pyrrhotite Fe2+0.95S Fe1-xS hexagonal 3.5-4
2+
Chalcopyrite CuFe S2 CuFeS2 tetragonal 3.5
2+
Troilite Fe S FeS hexagonal 3.5-4
Talnakhite Cu9Fe2+6Ni2S16 Cu9(Fe,Ni)8S16 isometric 4-4.5

Table 2
Minor sulphide minerals for the Kevitsa
Mineral name Emipirical formula Chemical formula Crystal system Hardness
Heazlewoodite Ni3S2 Ni3S2 trigonal 4
Millerite NiS NiS trigonal 3-3.5
Nickeline NiAs NiAs hexagonal 5.5
Maucherite Ni11As8 Ni11As8 tetragonal 5
Gersdorffite NiAsS NiAsS isometric 5.5
Cubanite CuFe2+2S3 CuFe2S3 orthorhombic 3.5
Bornite Cu5Fe2+S4 Cu5FeS4 orthorhombic 3
Mackinawite Fe2+0.75Ni0.25S0.9 (Fe,Ni)S0.9 tetragonal 2.5

Being a low grade deposit, gangue minerals dominate the ore treated, with the pyroxene
group (including altered forms) being the major non sulphide gangue (NSG). Sulphides
contribute less than 5% of the total mineralogy with chalcopyrite and cubanite being the main
Cu minerals and pentlandite and to a lesser extent millerite the main Ni sulphide minerals.
Sulphide gangue minerals in the ore are pyrrhotite (which exists in both hexagonal and
monoclinic forms, with varying proportions). As much as 20% of the nickel in the plant feed
is contained in non-sulphide minerals. Similarly, up to 10% of the copper in feed can be in
non-sulphide copper minerals
Variability in feed grades, gangue and sulphide minerals presents challenges to the
concentrator operations. Low metal feed grades result in a commercially unforgiving ore,
with little room to maneuver to manage the substantial variability and still maintain
profitability. This is unlike more commercially forgiving copper-nickel ores from other sites
where the nickel and copper content in the ore can be 5-10 times the grades at Kevitsa
(Lawson et al., 2014).
A recently published review of different nickel processing operations worldwide
highlighted ore variability characteristics (Lawson et al., 2014). They categorise nickel –
chalcopyrite ores into 4 main types:
 1. Nickel and copper in a pyrrhotite bearing sulphide deposit
2. Nickel and copper in a pyrrhotite bearing sulphide deposit with moderate to
high pyrrhotite to pentlandite ratios
3. Low grade nickel and copper in a high grade pyrrhotite deposit that are
uneconomic
4. Low grade nickel and copper in a low grade sulphide that can be economically
processed.
The Kevitsa ore would be categorised as type‘4’ with average sulphur in feed of about
1%-1.5%, and pyrrhotite in plant feed of around 2%.

1.2.Kevitsa Processing

1.2.1. Crushing and Grinding

The comminution circuit includes a crushing and autogenous grinding (AG) circuit as
shown in Fig. 1.

Fig. 1. Kevitsa comminution circuit flowsheet.

Crushing consists of 2 stages; primary crushing using a Sandvik CG820 gyratory crusher.
The secondary and tertiary crushing is done in two MP800 Metso cone crushers fitted with
different bowl and mantle liners in line with their duties. Primary crusher product is screened
on a double deck vibrating screen having a 100 mm top deck and a 25 mm bottom deck. Feed
to the secondary and tertiary crushers is the - 100 mm + 25 mm fraction which is presplit into
- 60 mm and + 60 mm fraction by a single deck scalping screen. The - 60 mm fraction feeds
the tertiary crusher whilst the + 60 mm fraction is feed to the secondary crusher. Some of this
fraction (+ 60 mm) is drawn to provide the secondary mill with grinding media as pebbles.
The + 100 mm, - 25 mm and products from both the secondary and tertiary crushers are
deposited on a stockpile from which four feeders draw to feed the 2 AG mills in parallel.
Primary milling is 100% autogenous and is carried out in 2 primary 7 MW AG mills of
8.5 m x 8.5 m dimensions. Discharge from the 2 AG mills is screened on a vibrating screen
fitted with 10 mm x 40 mm aperture panels with oversize rejects being conveyed back to the
secondary and tertiary crushers for crushing. The screen under size is split into 2, one stream
feeding a cyclone cluster in closed circuit with the mill and the other stream feeding the
secondary mill. The secondary mill is 14 MW and is of similar dimensions to the primary
mills with a grate discharge. Pebbles are fed to it as grinding media and the mill is in closed
circuit with a cyclone cluster. The overflows from the 3 mill cyclones are combined in a
surge tank that feeds flotation.

1.3.2. Flotation
The Kevitsa concentrator employs a copper-nickel sequential flotation circuit see Fig. 2.

Fig. 2. Kevitsa flotation circuit flowsheet.

Cu is floated first with a Cu selective collector and a high pH (>12) is employed to reduce
Ni flotation to copper concentrate. The circuit has a roughing and scavenging stage with the
concentrates from these banks reporting to the cleaning circuit. There are 4 stages of cleaning
in total with the column cell as the final cleaning stage. The depressant, carboxy methyl
cellulose (CMC) is added to both the roughing and cleaning circuits to reduce flotation of
NSG. The cleaner tails are open to final copper tails and combine with copper rougher
scavenger tails to feed the Ni circuit. Plant copper recovery to copper concentrate and copper
concentrate grade are as expected for an ore with low feed grades, but copper recovery is
constrained by the necessity to maximise copper-nickel selectivity in the copper circuit.
Selectivity against nickel is good with on average less than 4% of nickel sulphide reporting to
the copper concentrate
The feed to Ni flotation is at a high pH (10.8) and requires adjustment to pH 10 by
sulphuric acid to improve Ni flotation kinetics. A selective xanthate is used as collector to
minimise pyrrhotite recovery, with CMC as the NSG depressant. The circuit has 4 cleaning
stages treating mainly rougher scavenger concentrates and Ni column tail. The Ni column is
fed with rougher concentrate producing a final concentrate grade in one cleaning stage. Like
the copper circuit, the Ni cleaner circuit tail is open to the final tail. In the nickel circuit the
main mineral separation challenge is the pentlandite-pyrrhotite separation, and to a lesser
extent the pentlandite-gangue separation. The average sulphide nickel in feed is around
0.22% at Kevitsa, which is less than the tail grade at most nickel flotation operations.
Because of the complexity of the cleaner circuits for both copper and nickel, but
especially for nickel, there are large recirculating loads and long residence times in the
cleaner circuit. Moreover, there is active depression of gangue and pyrrhotite with flotation at
high pH and multiple dosing of chemical depressants.
Ni tails are subjected to an unselective bulk float for sulphides to generate a high sulphur
tail which is impounded in a lined pond for environmental reasons. The final tails, low in
sulphur are deposited in the bigger, unlined pond from where process water is reclaimed for
reuse in the process. The circuit employs a stronger xanthate and frother.
The PGMs are strongly associated with the Ni minerals with gold reporting
predominantly to the Cu concentrate The PGMs and Au are payable in both the copper and
nickel concentrate.

1.3.Kevitsa Performance

The performance of the Kevitsa plant has steadily improved since commissioning with
the plant results summarized in Table 3.

Table 3
Summary of the performance of the Kevitsa concentrator to date2012-2014.
Kevitsa nickel-copper-PGE operation 2012 2013 2014
Ore tonnes milled (000’s) 3,138 6,314 6,711
Nickel ore grade processed (%) 0.2 0.2 0.2
Nickel concentrate grade (%) 10.0 9.2 9.5
Nickel recovery (%) 56 63 61
Nickel production (tonnes) 3,874 8,963 9,433
Copper ore grade processed (%) 0.3 0.3 0.3
Copper recovery (%) 83 83 87
Copper concentrate grade (%) 22.6 23.0 24.1
Copper production (tonnes) 8,093 14,775 17,535
Gold production (ounces) 6,309 11,723 12,844
Platinum production (ounces) 13,753 30,403 34,090
Palladium production (ounces) 12,131 24,639 25,990
 1.4.Optimising the Flowsheet

The Kevitsa concentrator flotation flow sheet as designed is illustrated in Fig. 3 and has
gone through several changes to arrive at the current flowsheet shown in Fig. 2.

Fig. 3. Kevitsa design flotation circuit flowsheet.

The original flotation flowsheet is a typical, conventional flowsheet but was designed
without detailed information about mineralogical characteristics of the ore body and the
extent of its variability (Lotter et al., 2014). One could observe that the design of the
flowsheet was more suited for a Cu concentrator than for Cu-Ni sequential flotation circuit
with all the recycle streams built into it.
Failure by the 2 (Cu and Ni) circuits to meet the targeted recovery and concentrate grades
then necessitated the need to evaluate the original flowsheet to identify process improvement
opportunities. The initial step was to establish the base case performance of the concentrator
as a benchmark through a survey. For this survey the established, Statistical Benchmark
Survey, as described by Lotter (2005) was employed with the assistance of the Xstrata
Process Support (XPS) team. This survey was carried out in July and August of 2013 with the
concentrate grade and recovery performance shown in Table 4.

Table 4
Statistical Benchmark Survey Results.
Overall Cu Cu Ni Ni Final concentrate grade
recovery Recovery Recovery Recovery Recovery Cu % Ni%
(%) to Cu (%) to Ni (%) to Cu (%) to Ni (%)
Copper 78.2 65.4 12.8 - - 22.5 0.97
Nickel 50.7 - - 3.1 47.6 2.7 9.1
 Further to this the survey samples were submitted for mineralogical analysis with
QEMSCAN and sample streams mass balances by BILMAT. The findings of this survey are
summarised below:
1. Cu sulphide minerals recovery size range is wider than for Ni sulphide
minerals whose recovery falls off on both ends of the fines and coarse (<15µm
and >75µm).
2. Liberated Cu and Ni sulphide minerals in the fine size range are lost to
tailings
3. Some mineral losses were as binary composites between the sulphide
minerals and gangue in the finer size fractions (<15µm and >13µm) and the
coarser fraction (>106µm).
4. Ni sulphides reporting to the Cu concentrate were liberated (true
flotation or entrainment) or as binary composites with Cu sulphides or
floatable NSG.
5. High levels of floatable NSG in both concentrate more so in Ni final
concentrate
6. High levels of pyrrhotite in Ni final concentrate (41%)

Results of the mass balancing by BILMAT with all the intermediate streams gave an
insight into the mass flows and contained metal recycling with the numerous recycle streams
as show in Table 5.

Table 5
Bilmat balance assay grades and recoveries in Kevitsa circuits
%Mass % Ni % Cu % Fe %S % Mg % Si
Feed 100.00 100.0 100.0 100.0 100.0 100.0 100.0
Cu Concentrate 0.83 3.2 66.0 3.0 17.7 0.2 0.2
Cu Ro Conc 2.12 11.9 67.2 5.2 22.7 1.4 1.4
Cu Scav Conc 3.24 11.2 22.7 4.6 12.5 2.9 2.9
Cu Clnr 1 Tls 4.53 19.9 23.8 6.7 17.6 4.1 4.1
Cu Scav Tailing 99.17 96.8 34.0 97.0 82.3 99.8 99.8
Ni Concentrate 1.37 49.1 12.5 6.8 31.8 0.3 0.3
Ni Ro Feed 143.39 342.8 126.6 222.2 482.3 131.4 131.4
Ni Ro Conc 10.49 114.1 32.0 36.8 135.4 6.0 6.0
Ni Scav Conc 35.09 181.0 73.1 95.3 296.4 25.8 25.8
Ni Clnr 1 Tls 9.12 65.0 19.5 30.0 103.6 5.7 5.7
Ni Clnr 2 Conc 2.11 66.8 14.4 10.4 46.5 0.5 0.5
Ni Clnr 3 Tls 0.75 17.7 1.9 3.5 14.7 0.2 0.2
Thick & Regrind Mill 2.11 66.8 14.4 10.4 46.5 0.5 0.5
Ni Scav Tailing 97.80 47.7 21.5 90.1 50.5 99.6 99.6
Final Tail 96.23 40.1 18.2 85.8 37.5 98.4 98.4
S Float Conc 1.58 7.6 3.2 4.3 13.1 1.2 1.2
 The mineral assays and distributions (Cp is chalcopyrite, Po is pyrrhotite, Pn is
pentlandite) are also shown in Table 6.

Table 6
Bilmat balance mineral assays and recoveries in Kevitsa circuits
%Mass %Cp %Pn %Po %NSG
Feed 100.00 100.0 100.0 100.0 100.0
Cu Concentrate 0.83 66.0 3.7 4.7 0.2
Cu Ro Conc 2.12 67.2 13.8 9.7 1.3
Cu Scav Conc 3.24 22.7 12.6 9.0 2.9
Cu Clnr 1 Tls 4.53 23.8 22.8 14.1 4.0
Cu Scav Tailing 99.17 34.0 96.3 95.3 99.8
Ni Concentrate 1.37 12.5 57.7 31.8 0.2
Ni Ro Feed 143.39 126.6 380.3 630.3 130.8
Ni Ro Conc 10.49 32.0 133.2 171.4 5.8
Ni Scav Conc 35.09 73.1 208.5 395.3 25.4
Ni Clnr 1 Tls 9.12 19.5 75.4 139.6 5.6
Ni Clnr 2 Conc 2.11 14.4 78.5 49.4 0.4
Ni Clnr 3 Tls 0.75 1.9 20.7 17.6 0.2
Thick & Regrind Mill 2.11 14.4 78.5 49.4 0.4
Ni Scav Tailing 97.80 21.5 38.6 63.5 99.6
Final Tail 96.23 18.2 29.8 46.0 98.5
S Float Conc 1.58 3.2 8.7 17.5 1.2

Armed with this and other data from prior QEMSCAN and MLA reports a circuit
reconfiguration was embarked on at Kevitsa with the following changes being made.
For the copper circuit:
1. Cu rougher concentrate redirected to feed Cu cleaner 2
2. Cu Cleaner 1 tail opened to Cu tails
For the nickel circuit:
1. 7 of the 14 Ni rougher/scavenger cells were bypassed (1 line bypassed)
2. Ni rougher concentrates redirected to feed Ni cleaner 3
3. Ni scavenger concentrate redirected to feed Ni cleaner 1
4. Ni cleaner 3 tails redirected to Ni cleaner 2 feed
5. Ni cleaner 1 tails opened to Ni final tails
6. Ni regrind thickener bypassed and dewatering cyclones installed in its place
7. Ni cleaning circuit shortened to 4 cleaning stages

The changes targeted reduction of the cleaner recycle that recycled both valuable minerals
and gangue leading to an increased chance of losing the valuable minerals to tail and the
entrainment/entrapment of gangue into the concentrate streams. Another target was to direct
the concentrate streams to feed nodes in the circuit of similar metal grades, to prevent
pollution of high grade streams and keeping such streams from nodes close to the exit point
of the circuit.
 1.5.Column Cells

The Kevitsa mills are operated to achieve a product p80 of between 75 μm-80 μm.
Flotation feed has a high proportion of fines, 50% below 25 μm. Previous works by XPS
showed 60% of metal losses in tails are fully liberated and below 25 μm. The presence of
varying amounts of gangue minerals such as talc, serpentine and amphibole poses a challenge
to maintaining targeted concentrate grades. Flotation column cells were installed on the
copper and nickel circuits (Fig. 2.), after a successful plant trial with a pilot flotation column
that demonstrated a cleaner column concentrate.
A column flotation cell has a reduced surface area to cell volume ratio to promote froth
stability, a froth washing system to minimize the entrainment of impurities, quiescent
flotation conditions to promote selectivity and an air sparging system to generate a supply of
uniform bubbles. Column flotation has become widely accepted throughout the mineral
industry (Sandvik et al., 1991). Under some conditions, their special design features ensure
outperformance of mechanical cells. Column cells have a much smaller surface area per unit
volume of capacity than mechanical cells which promotes froth crowding. This results in
higher froth densities, and when coupled with stabilizing wash water addition, permits froth
depth of 1 m–2 m to be maintained (Eric, 2014). A deep froth, coupled with wash water,
allows for better rejection of entrained hydrophilic minerals to column concentrate. The
slurry in the column is pneumatically agitated through the introduction of controlled size of
bubbles in the lower section of the column. Some columns are equipped with cavitation
system which improves thorough mixing of slurry and bubbles for better flotation
performance (Eric, 2014).With the low intensity mixing only caused by a rising bubble
swarm, the velocities of both bubbles and particles during attachment are slow in quiescent
conditions thereby providing sufficient contact time.
These characteristics of the column cell lent themselves to the cleaner challenges at
Kevitsa; improving the Cu-Ni and Cu-NSG separation in the copper circuit and the Ni-Po and
Ni-NSG separation in the nickel circuit.

1.6.Fines Recovery

The relatively poor recovery of < 20 µm sulphide minerals by flotation has been
recognised and well documented (Trahar, 1981; Jameson et al, 2007). While the poor
recovery of < 10 µm chalcopyrite is well known, Lawson et al. (2014) drew attention to the
even poorer recovery of < 10 µm pentlandite. Solutions to the problem have been more
limited, particularly solutions that do not reduce the selectivity of the flotation separation. So
various solutions such as: bubble size reduction (Finch and Dobby, 1990), or increasing
particle velocity (Ahmed and Jameson, 1985) have been shown to increase recovery but at
the cost of selectivity. Other flotation strategies to increase fines recovery, such as selective
aggregation or split flotation have shown some improvement, but at a relatively high cost
(Song et al., 2000; Holder, 1994; Torrisi, 2003).
Magnetic conditioning has been shown to selectively increase the flotation recovery of <
20 µm chalcopyrite, pentlandite and other sulphides (Zoetbrood et al., 2010; Engelhardt et al.,
2005; Holloway et al., 2008). Moreover, Wilding and Lumsden (2011) showed that it can
increase the selectivity between two paramagnetic minerals that are sequentially floated.
 The Kevitsa flotation feed size is not particularly fine for a complex sulphide separation
process, but the nickel sulphide losses are overwhelmingly fine, with more than 80% of
nickel in tail <25 µm. The <25 µm copper in tail is only about half as much as the <25 µm
nickel in tail. Therefore, at Kevitsa improving nickel recovery must focus on improving the
fine pentlandite recovery. While magnetic conditioning has been shown by Engelhardt et al.
(2005) to improve chalcopyrite recovery and by Zoetbrood et al. (2010) to increase
pentlandite recovery, at Kevitsa the nickel in feed is about twice the value of copper in feed,
therefore, nickel recovery is the metallurgical and financial focus.
The nickel sulphide in plant feed reports to either the copper concentrate, nickel
concentrate or nickel tail. Generally, the proportion reporting to these streams was <5% to
copper concentrate, around 60% to nickel concentrate and the remaining 35% to tail. The
strategy in installing the magnetic conditioning units was to reduce the misplacing of
pentlandite to the copper concentrate and to the final tails.
Some of the pentlandite recovered to the copper concentrate was there by entrainment,
therefore, it was decided to magnetically condition the copper cleaner feed to homogeneously
aggregate the chalcopyrite to improve its flotation, and to homogeneously aggregate the
pentlandite to reduce its recovery by entrainment to the copper concentrate.
Magnetic conditioning was also installed in the nickel rougher and first nickel cleaner to
aggregate and reduce nickel sulphide losses to the nickel tail.

2. Results and discussion

2.1.Flowsheet Reconfiguration

Changes to the flowsheet have been ongoing for the duration the concentrator has been in
operation. Therefore, a second statistical benchmarking survey to measure the impact of the
reconfiguration changes has been deferred. To this effect operational performance data was
analysed statistically, comparing plant results before and after the flowsheet changes to
evaluate the impact of the changes on recovery and concentrate grades for Cu and Ni.
The flowsheet changes outlined in the introduction were implemented and the data for
both the copper and nickel circuit were collected so that the relative performance of the plant
before and after these changes were made could be analysed.
Table 7 compares the results in the copper circuit before and after circuit reconfiguration.

Table 7
Copper circuit results before and after circuit reconfiguration
% Cu Upgrade
Rec Ratio
Before circuit reconfiguration 71.7 80.7
After circuit reconfiguration 71.3 91.8
Difference 0.3 11.1
Level of confidence low >99%
 The upgrade ratio (concentrate grade/feed grade) was used so as to normalise the data for
variations in feed grade. There was a substantial increase in the upgrade ratio with no change
in recovery for the period after the circuit was reconfigured. Subjecting the data to a tt- test
(Napier-Munn, 2010),, this observation is validated showing no difference in copper recovery
whilst the upgrade ratio improved by 11 units with greater than 99% % level of confidence.
For the Ni circuit Pn:Po selectivity is the major challenge. The
The impact of shortening
residence time to limit pyrrhotite recovery is illustrated in Fig. 4 where the Po /Pn selectivity
graphs are shown for Ni roughing tests on Kevitsa ore,, under different conditions. As
residence time increases Pn and Po recovery increases, with eeachach successive point on the
curve showing a longer residence time. As the recovery of Pn increases with time, the
recovery of Po increases proportionally,
proportionally, until around 70% Pn recovery, where the relative
rate of Po recovery increases while the relative Pn recovery rate reduces and the Pn:Po
selectivity deteriorates markedly.
To minimise the recovery of Po relative to Pn, the flotation time had to be shortened to
the time where about 70% Pn recovery occurs. This avoidsavo the longer flotation times where
high recovery of Po occurs for marginal gain in the Pn recovery. For this reason and because
re-routing
routing of the recycling streams reduced Ni rougher feed flow, the plant rougher
rougher-scavenger
residence time was cut by half (one line taken off line).
line). Design residence time was 90
minutes at 5 mtpa and currently the residence time is 33 minutes at the current 6.77 mtpa
which is 37% of the design residence
residenc time.

Fig. 4. Pn-Po selectivity with flotation time.

Data was collected for the nickel circuit before and after the circuit reconfiguration and
Table 8 details the data.
Table 8
Nickel circuit results before and after the circuit reconfiguration
% Ni
Rec Upgrade Ratio
Before circuit reconfiguration 60.1 43.3
After circuit reconfiguration 59.5 45.1
Difference 0.6 1.8
Level of confidence low 88%

A similar analysis of the Ni circuit data was not so distinct, with a marginal drop in
recovery and an increase in upgrade ratio of -0.6% and 1.8 respectively. For each of these
changes the level of confidence was low and thus the circuit reconfiguration produced no
change in nickel circuit performance. However, it was noted that a similar performance was
achieved before and after the change, but with 50% less residence time (lower operating
cost).
With improving Po/Pn selectivity on the rougher and scavengers being one of the
objectives of the changes to the flow sheet, through the reduction of the flotation time, data
on the Po/Pn ratio in the feed to flotation versus Po content in the Ni final concentrate was
compared before and after circuit reconfiguration. In the absence of XRD analysis to quantify
the Po content in the final concentrate the contained minerals in the samples were calculated
using metal assays obtained by making the following assumptions (Kelebek and Tukel,
1999).
1. The sulphide minerals contained in the ore are chalcopyrite,
pentlandite and pyrrhotite and all other sulphide minerals occur in such low
quantities as to be deemed negligible and thus are ignored.
2. All the copper is contained in chalcopyrite whilst the nickel mineral is
solely pentlandite.
3. The total sulphur is distributed between the 3 major sulphide minerals.
Using minerals compositions for the minerals the minerals content including pyrrhotite
are calculated.
Fig. 5 shows the significant improvement in Pn/Po selectivity achieved with the circuit
change. It can be seen that there was a 7.3% drop in Po in final concentrate following the
reconfiguration.
 60

50

40
Ni Final Conc Po %

30

20

10

0
1 2 3 4 5 6 7
Feed Po/Pn Ratio

Before After Linear (Before) Linear (After)

Fig. 5. Po/Pn Selectivity change after circuit reconfiguration in the Ni circuit.

Subjecting the data to the t-test, the reduction in Po content is validated to a greater than
99% level of confidence and supports the laboratory test findings that showed limiting
flotation time gives an increase in the selectivity of Pn against Po.
With reduced recycle of material from the cleaning circuit to the rougher scavenger
circuit the volume of gangue being carried into the cleaning circuit was reduced, allowing for
the required cleaning capacity to be reduced, and thus the 5th nickel cleaning stage was
bypassed.
Bypassing of the Ni regrind feed thickener and replacing it with dewatering cyclones was
done to minimise concentrate hold up in the thickener which exposed the pentlandite mineral
to possible oxidation which would in turn reduce its floatability. With the dewatering
cyclones the targeted mill feed densities were met with no concentrate hold up.

2.2.Column Cells

2.2.1 Copper Circuit

Four months of plant data was collected, 2 months before and 2 months after the
installation of the flotation column. Graphs of upgrade ratio versus recovery were plotted
separately for Cu circuit. The impact of varying feed grades was normalised by using the
upgrade ratio. The feed to the column was established following a circuit survey.
 90

85

80

Recovery (%) 75

70

65

60

55

50
40 80 120 160 200
Upgrade ratio

Before After

Fig. 6. Cu Upgrade – recovery results before and after column installation

Fig. 6 shows a plot of upgrade ratio against Cu recovery in the copper circuit. The
statistical analysis of this data is in Table 9. Copper recovery improved by 1.52% Cu at 99%
confidence level. The column cell demonstrated a remarkable impact on gangue rejection to
copper concentrate. The recovery of fines to the column concentrate also improved compared
to the mechanical cells. The challenge now however, is to optimise the operation of
mechanical cleaning cells which were producing a low final concentrate. Further circuit
reconfigurations have continued.

Table 9
Recoveries and upgrade ratio of Cu in Cu circuit before and after column installation
%Cu recovery Upgrade ratio
After column installed 74.03 89.25
Before column installed 72.51 82.51
Difference 1.52 6.73
Paired level of confidence 99% 94%

2.2.2 Nickel Circuit

The results for the nickel circuit are shown in Fig. 7 and detailed in Table 10.
 75

70

Recovery (%) 65

60

55

50

45

40
20 30 40 50 60 70
Upgrade ratio
Before After

Fig. 7. Ni Upgrade – recovery results before and after column installation

Fig. 7 graphs Ni upgrade ratio against Ni recovery in the Ni circuit before and after the
column cell installation. Ni recovery improved by 1.83% to 99% confidence level. The major
diluents in Ni final concentrate are NSG and pyrhottite. The flotation column had a positive
impact on rejecting NSG minerals to the concentrate, however, the space was filled by
pyrhottite. The overal impact of column on Ni final concentrate grade was hardly seen as
shown by only a 2 unit increase in the upgrade ratio. The results are summarised in Table 10.
This is attributed to the substitution of NSG with pyrhottite and low quality concentrate
coming from the mechanical cells. Plant trials with different types of pyrhottite depressants
have continued.

Table 10
Recoveries and upgrade ratio of Ni in Ni circuit only befor and after column installation
%Ni recovery Upgrade ratio
After column installed 59.26 40.13
Before column installed 57.43 38.09
Difference 1.83 2.04
Paired level of confidence 99% 98%

2.3.Fines Recovery

A statistical ON/OFF paired‘t’ test (Napier - Munn, 2010) was undertaken with the
magnetic conditioning in each location (copper cleaner, nickel rougher and nickel cleaner)
ON or OFF concurrently. The magnetic conditioning operated ON for two days and then OFF
for two days with consecutive ON and OFF days paired (Day 1 OFF compared with prior
Day 2 ON and Day 2 OFF compared with latter Day 1 ON). The aim of the magnetic
conditioning testwork was to increase the nickel recovery to the nickel concentrate by
reducing nickel losses to copper concentrate and final tail.

2.3.1 Copper Circuit Results

However, early in the testing it became apparent that not only was there a reduction in
nickel lost to copper concentrate but also there was an increase in copper recovery to copper
concentrate. The copper results for the 44 pairs of ON/OFF results for the first 3 months in
the copper circuit are given in Table 11.

Table 11
Paired results for copper circuit; copper concentrate results.
Grade Recovery Grade Recovery
%Cu %Cu %NiS %NiS
Magnetic conditioning 25.7 69.6 1.06 3.6
ON
Magnetic conditioning 25.3 68.5 1.32 4.4
OFF
Difference 0.4 1.1 0.26 0.8
Level of Confidence low 94% 99% 98%

This is showing an improved copper grade recovery response-higher copper recovery at a

similar copper concentrate grade. Moreover, there is improved copper-nickel sulphide
selectivity, with higher copper recovery at a lower nickel sulphide recovery to the copper
concentrate. This is consistent with homogeneous aggregation of the paramagnetic minerals,
also found by Wilding and Lumsden (2011), with a reduction in fine nickel recovery due to
less entrainment of homogeneously aggregated nickel sulphide and an increase in copper
recovery due to selective homogeneous aggregation of the copper sulphide. The reduction in
nickel losses to the copper concentrate while 0.8% in absolute recovery is about a 20%
reduction in nickel sulphide recovered to the copper concentrate, comparable to the
approximately 20% reduction in nickel sulphide grade in the copper concentrate.
Using a financial model for the Kevitsa plant and the method of Long et al. (2012) the
total financial benefit can be calculated to a level of confidence using the paired‘t’ test. This
combines the extra copper recovery and the extra nickel recovery to nickel concentrate. It is
conservatively assumed, that only 60% of the nickel sulphide rejected from the copper
concentrate, is recovered in the nickel concentrate.
The financial benefit with magnetic conditioning in the copper cleaner is USD 0.73 per
tonne of plant feed with a level of confidence of 99%.

2.3.2 Nickel Circuit Results

After several months of testwork in the nickel circuit it was observed that the change in
nickel circuit performance took 24 hours to be clearly apparent. There was an equilibrium
period where the mixing of ON and OFF data occurred. A review of the literature on
pentlandite flotation kinetics and the Kevitsa circuit/mineralogy, supported the observation
that an equilibrium period is required. Natarajan and Nirdoash (2006) have shown that the
flotation rate constant for pentlandite is about 3 times lower than the flotation rate constant
for chalcopyrite. Lawson et al. (2014) also found that pentlandite flotation is much slower
than chalcopyrite flotation. Moreover, the pentlandite in plant feed at Kevitsa is finer than the
chalcopyrite and this would further lower the flotation kinetics of pentlandite relative to the
chalcopyrite. Given the four stages of cleaning in the nickel circuit, with all tails streams
(except Ni cleaner1 tail) recycled back to the previous cleaner stage, a regrind stage, very
large recirculating loads and the lower flotation rate constant for pentlandite, it is not
surprising that is takes a period of time for a change in the rougher and first cleaner circuit to
progress through to the final concentrate. Hence, the necessity for an equilibrium period after
a plant change before measuring the effect of that plant change.
From a statistical perspective, the disadvantage of allowing an equilibrium period is that it
halves the number of pairs of data. The number of pairs is important when carrying out paired
t tests, and halving the number of pairs in effect doubles the length of the test. Therefore,
after the first three months of testing, the magnetic conditioning in the copper circuit
remained ON, while the testing in the nickel circuit continued into 2015. In all 40 pairs of
ON-OFF results were collected that allowed an equilibrium period. The results are in Table
12.

Table 12
Paired results for nickel circuit; nickel concentrate results.
Equilibrium Recovery Recovery Combined recovery
%Cu %NiS %Cu+%Ni
Magnetic 13.8 58.9 72.7
conditioning ON
Magnetic 12.8 57.6 70.4
conditioning OFF
Difference 1.0 1.3 2.3
Level of Confidence 91% 91% 98%

The improvement in nickel and copper is to a high 91% level of confidence. However,
combining the total circuit improvement in the combined recovery of the two payable metals
copper and nickel, which is the true total difference in circuit performance due to magnetic
conditioning results in a 2.3% increase in total metal recovery to a 98% level of confidence.
This higher confidence is due to the lower relative standard deviation for the combined metal
recovery paired difference, compared to the relative standard deviation of the individual
metal recoveries and also the combined recovery difference is larger.
However, combining the two recoveries may give a better metallurgical measure of the
benefit to the circuit due to magnetic conditioning, but does not take into account the true
economic benefit to the circuit. The value of nickel in the nickel feed is substantially higher
than the copper because it is higher grade and higher value. Using the methodology of Long
et al. (2012) and a financial model for Kevitsa’s nickel circuit and combining the
metallurgical benefits of copper recovery and nickel recovery to the nickel concentrate shows
that the increases in metal value due to the magnetic conditioning in the nickel circuit is:
USD0.76 per tonne of plant ore feed to 97% level of confidence. If the increase in PGM
recovery is included with the increase in nickel and copper recovery then the increase in
metal value in the nickel concentrate due to the magnetic conditioning is: USD0.98 per tonne
of plant ore feed to 97% level of confidence. This is a substantial result given the variability
in the mineralogy. These results are tabulated in Table 13.

Table 13
Paired results for nickel circuit; modeled increase in USD value of metal recovered.
USD Metal value Cu and USD Metal value Cu, Ni and
Ni PGM
Magnetic conditioning ON 26.38 30.80
Magnetic conditioning 25.62 29.82
OFF
Difference 0.76 0.98
Level of Confidence 97% 97%

3. Conclusion

Kevitsa concentrator flowsheet was a generic design and on commissioning, it failed to

address the ore mineralogical factors and substantial ore variability. This created
opportunities for process performance evaluation and investigation into flowsheet suitability
through sampling surveys and laboratory flotation tests. The subsequent analysis of samples
by a combination of QEMSCAN and MLA’s identified process improvement opportunities.
The result of all this work was the reconfiguration of the circuit to eliminate recycling
streams, inappropriate equipment and adjusting Ni rougher flotation residence time for better
selectivity between the main sulphide gangue and the Ni sulphide mineral. The overall result
was an improved upgrade ratio in the copper circuit and less metal sulphide gangue recovery
to the nickel concentrate. Additionally, extra roughing/scavenger and cleaner capacity are
now available for the expansion that Kevitsa has planned (10 mtpa). With fewer flotation
cells in operation, power and maintenance savings were also realized, although not
quantified.
Installation of a column cell in the Cu circuit has improved copper recovery to copper
concentrate by 1.5% at 99% level of confidence. The concentrate grade has increased even
with varying feed grades. Ni recovery in Ni circuit has improved by 1.83% at 99%
confidence level, however, the concentrate grade has not benefited as much due to improved
flotation of pyrrhotite with the introduction of the column.
The magnetic aggregation has increased copper recovery to copper concentrate and
copper and nickel recovery to nickel concentrate to high levels of confidence. This
improvement has had a good effect on the financial returns to the operation. Testwork is now
completed.
Overall, since commissioning step changes have been made in the Cu-Ni processing at
Kevitsa. These have substantially increased the financial returns of the project. This process
improvement is ongoing and is driven by the need to maintain profitability in a low grade
deposit at a time of reduced metal prices.
Acknowledgements

The authors would very much like to thank Kevitsa Mining Oy and First Quantum Minerals
for their support to research and present this paper. Also thank you to all the Kevitsa
employees who have contributed to this work.

References

Ahmed, N., Jameson, G., 1985. The effect of bubble size on the rate of flotation of fine
particles. Inter. J. Miner. Process., 14,196-215.

Engelhardt, D., Ellis, K, Lumsden, B., 2005. Improving fine sulphide mineral recovery –
plant evaluation of a new technology. In: Proceedings Centenary of Flotation Symposium,
Brisbane, June 2005, pp. 829-834.

Eric, B. W., 2014. Flotation technology for coarse and fine particle recovery. In: Congreso
Internacional de flotation de Minerals, Peru August 2014.

Finch, J., Dobby, G., 1990. Column Flotation, Pergamom Press, Oxford.
Holder, R., 1994. Improvement in copper and silver flotation at Hellyer using high energy
conditioning. In: Proceedings of the Fifth Mill Operators’ Conference, 1994. Roxby Downs,
October 1994, pp. 153-159.

Holloway, B., Clarke, G., Lumsden, B., 2008. Improving fine lead and silver flotation
recovery at BHP-Billiton’s Cannington mine. In: 40th Annual Meeting of the Canadian
Mineral Processors 2008, Ottawa, January 2008, pp. 347-362.

Jameson, G., Nguyen, A., Ata, S., 2007. The flotation of fine and coarse particles, in:
Fuerstenau, M., Jameson G., Yoon R., (Eds.), Froth Flotation a Century of Innovation.
Society for Mining, Metallurgy and Exploration Inc., Littleton, pp. 339-372.

Kelebek,S., Tukel,C., 1999. The effect of sodium metabisulphite and triethylenetetramine

system on pentlandite-pyrrhotite separation. Int. J. Min. Proc., 57, 135-152.

Kojonen, K., Laukkanen, J., Gervilla, F., 2008. Applied mineralogy of the Kevitsa nickel-
copper-PGE deposit, Sodankyla, northern Finland. In: Proceedings of the Ninth International
Conference of Applied Mineralogy, 2008. Brisbane, September 2008, pp. 605-614.

Lawson, V., Hill, G., Kormos, L., Marrs, G., 2014. The separation of pentlandite,
chalcopyrite, pyrrhotite and gangue in nickel projects throughout the world. In: Proceedings
of the Twelfth Mill Operators’ Conference, 2014. Townsville, October 2014, pp. 153-162.

Long, G., Taylor, S., Lumsden, B., 2012. Statistical plant trials – efficiently combining
concentrate grade and recovery results. In: Proceedings of the Eleventh Mill Operators’
Conference, 2012. Hobart, October 2013, pp. 355-362.
 Lotter, N.O., Whiteman, E., Bradshaw, D.J., 2014. Modern plant practice of laboratory
flotation testing for flowsheet development-A review. Minerals Engineering., 66-68 , (2014),
2-12.

Lotter, N., 2005. Statistical Benchmark Surveying of Production Concentrators, Ph.D. Thesis,
McGill University, Montreal, August 2005.

Napier- Munn, T., 2010. Designing and analysing plant trials, in: Greet, C., (Ed.), Flotation
Plant Optimisation. The Australian Institute of Mining and Metallurgy, Melbourne, pp. 175-
190.

Natarajan, I., Nirdosh, R., 2006. A comparative study of kinetics of flotationof a copper-
nickel ore by N-hydrocinnamoyle-N-phenylhydroxylamine (HCNPHA) vis-à-vis potassium
amyl xanthate (PAX). In: Proceedings of the International Seminar on Mineral Processing
Technology 2006, Chennai, 2006. pp.139-148.

Sandvik, K.L., Nybo,A.S., Rushfeldt, O., 1991. Reverse flotation to low impurity levels by
column flotation. In: Column 91, International Conference on Column Flotation, Sudbury,
June 2-6 1991.

Song, S., Lopez-Valdivieso, A., Reyes-Bahena, J., Bermjo-Perez., H Trass O., 2000.
Hydrophobic flocculation of galena fines in aqueous suspensions, Journal of Colloid and
Interface Science, 227, 272-281

Torrisi, C., Smith, R., 2003. Bringing split float to reality – The Cannington experience. In:
Proceedings of the Eighth Mill Operators’ Conference, 2003. Townsville, July 2003, pp.73-
80.

Trahar, W., 1981. A rational interpretation of the role of particle size in flotation. Int. J. Min.
Proc., 8, 289-327.

Wilding, J., Lumsden, B., 2011. Implementation of magnetic conditioning in two stage
sequential Cu-Zn flotation separation. In: Proceedings of the Conference of Metallurgists,
2011. Montreal, October 2011, pp.139-148.

Zoetbrood, D., Vass, P., Lumsden, B., 2010. Magnetic conditioning of pentlandite flotation –
plant evaluation. In: Processing of Nickel Ores and Concentrates 2010, Falmouth, June 2010.

View publication stats