Modern Process Mineralogy: Two case studies

(1)

Modern Process Mineralogy: Two case studies

^q

N.O. Lotter

^⇑

, L.J. Kormos, J. Oliveira, D. Fragomeni, E. Whiteman

Xstrata Process Support, 6 Edison Road, Falconbridge, Ontario, Canada P0M 1S0

a r t i c l e i n f o

Article history:

Available online 10 March 2011 Keywords:

Process Mineralogy Ore characterisation Liberation Flotation

a b s t r a c t

Process diagnosis, ﬂowsheet design and optimisation are most effectively and efﬁciently achieved through the use of metallurgical testwork combined with modern quantitative mineralogical techniques.

The integration of these two areas of study form the discipline known as process mineralogy. A brief history of the discipline is described along with the program now in place at Xstrata Process Support (XPS).

Representative sampling protocols for orebodies, plant or test products, the use of geometallurgical unit classification, stratified sampling, high confidence metallurgical test programmes, concentrator sampling audits (Benchmark Surveys) and the use of quantitative mineralogy (QEMSCAN and EPMA) are key com- ponents of the strategy. Two case studies from Xstrata Nickel’s Nickel Rim South Mine in Sudbury and its Raglan Concentrator in Quebec are described to show how mineralogical data can be integrated into metallurgical programs to assist mineral processing engineers to design and optimise flowsheets and how the use of quantitative mineralogy can be used to benchmark plant performance and enable predictions of performance ahead of plant changes.

1. Introduction

1.1. History of process mineralogy

Historically, the design of concentrators treating sulphide ores was treated with the tools available at the time – for example, pio- neer work byBond (1952), for work index and mill sizing, and the various empirical sampling and testing of drill core and other geological samples or specimens, to arrive at an expectation of concentrator performance after commissioning. Others such asRestarick (1976), pioneered a series of plant survey or sampling structures that would produce a diagnosis of the ﬂowsheet limitations. As these tools were advanced, and newer inventions and methods developed, this endeavour reached a breakthrough point in the early eighties when two developments occurred. One was the real- isation that the integrated approach using mineralogy and mineral processing would produce a synergy; the other, the development of a quantitative automated mineralogical measurement platform with Quantitative Evaluation of Minerals by Scanning Electron Microscope (QEMSEM), and later, the Mineral Liberation Analyser (MLA).

For the operating concentrator, the question had always arisen as to what further gains could be made by way of recovery, selec- tivity or concentrate grade whilst sustaining or increasing throughput. At the time, such a concentrator had at best a platform of monthly composites of feed, concentrate and tailings which would be sized, weighed and assayed, producing size-by-size recovery models, sometimes supported by optical microscopy, informing the grinding and classiﬁcation circuit of possible improvements.

Some investigators saw the link between the type of size distribution produced by the grinding circuit, and the limitations or opportunities that this presented to the flotation circuit (McIvor et al., 1990; McIvor and Finch, 1991). Often, a programme of flotation tests would support these activities. Others, such as Trahar (1981), showed the value of describing flotation performance on a size class basis. It was common to see a series of publications of these endeavours across the life-of-mine, often concluding their best result in the last 5 years of the operation before shut-down after thirty or so years of production (Shannon et al., 1993, for example). The impact of these practices was reviewed byMackey and Nesset (2003), showing that using the McNulty models for start-up, projects could be ranked by order of the maturity of the technology to be used; by the actual delivery of saleable metal against design in the first few years of operation; and how this im- pacted the nett present value delivery (NPV) of the project.

The connection between mineralogy and metallurgical performance in a plant was recognised long ago (Petruk, 1976; Petruk and Hughson, 1977; Cabri, 1981; Petruk and Schnarr, 1981; Peyerl, 1983; Petruk, 1988, for example) as was the need to provide diagnostic sampling techniques of a plant (Restarick, 1976) and 0892-6875/$ - see front matterÓ2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.mineng.2011.02.017

qThis paper was originally presented to MEI conference ‘‘Nickel ‘10’’, Falmouth, June 2010. The Nickel Rim Case Study was ﬁrst presented to the Canadian Mineral Processors Conference, Ottawa, January 2010. Now presented to the journal with changes and permission of the CMP.

⇑Corresponding author. Tel.: +1 705 699 3400.

E-mail address:[email protected](N.O. Lotter).

Contents lists available atScienceDirect

Minerals Engineering

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / m i n e n g

(2)

to improve the statistical reliability of mineralogical and process measurements (Henley, 1983; Lotter, 1995, 2005). The development of QEMSEM (and the second generation QEMSCAN), and the later development of the MLA formed the breakthrough plat- forms into what is now known as modern Process Mineralogy. At Falconbridge Limited, for example, this vision was taken into a project to develop the opportunity and deliver value into operations using this new integrated approach, in which an internal rate of return of 92% p.a. was shown for the investment in the laboratory equipment, sampling, and cost of plant modiﬁcations (Lotter et al., 2002). In this case, the Process Mineralogy platform was designed using geology, sampling, mineralogy and mineral processing.

Numerous examples of equivalent development and applications have since been published, for example Lotter et al. (2003), Fragomeni et al. (2005), Charland et al. (2006), Dai et al. (2008), McKay et al. (2007), andTriffett et al. (2008).

1.2. Modern Process Mineralogy

As it stands at the time of writing, modern Process Mineralogy has been advanced to a more integrated practice as a hybrid discipline. The salient features of its structure follow.

1.2.1. Representative sampling

Representative sampling is a key component to any testwork programme. Although historically bulk samples from one location within a deposit were deemed appropriate for process diagnosis and flowsheet development, it is now recognised that a set of unoxidised drill core samples, representative of the resource in terms of space, grade, grade distribution and lithology will better quantify process performance. In cases where random sampling methodologies are required, Gy’s sampling models (Gy, 1979) are applied in conjunction with an understanding of preliminary geometallurgical unit definitions to help streamline minimum sample mass requirements. This approach stratifies the compound distribution (Cochran, 1946) and reduces the minimum sample mass required for a given fundamental variance (Lotter, 2010).

Samples collected for the metallurgical testwork programme are crushed and screened in two stages to minimise the generation of fines and blended using an odds and evens blending (Middleditch and Lotter, 2008) or equivalent (Middleditch et al., 2009) protocol to produce individual charges for flotation or other mineral separation testing. A sample of blended charge is subsampled by spinning riffler to produce replicate subsamples with minimal variance for mineralogical assessment and external reference distribution which defines a robust estimate of the sample mean grade (Lotter and Fragomeni, 2010). This respects Gy’s safety line of 1979, whereby subsamples may be taken from the lot in a relationship between primary sample and subsample mass and particle topsize.

1.2.2. Geometallurgical units

Geometallurgical units (Lotter et al., 2003; Fragomeni et al., 2005) can be defined as an ore type or group of ore types that pos- sess a unique set of textural and compositional properties from which it can be predicted they will have similar metallurgical performance. Sampling of an orebody based on geometallurgical units will define metallurgical variability and allow process engineers to design more robust flowsheet options. This variability can be muted when samples from different geometallurgical units are blended and tested as one sample. Composites are created by ensuring grade and grade distributions from a specific area defining the geometallurgical unit within a resource are maintained.

The method used to divide an orebody into geometallurgical units is based on a review of geological data including host rock, alteration, grain sizes, texture, structural geology, grade, sulphide mineralogy and metal ratios with focus on characteristics which

are known to affect metallurgical performance. The foregoing list is, however, not complete and also uses hardness testing and the grade/recovery curve as characterising parameters (Fragomeni et al., 2005for example). Statistical analysis is often used to help define preliminary units. In addition, it is recommended that a variability program based on smaller samples from throughout a geometallurgical unit is completed prior to finalising the divisions between geometallurgical units. This approach will quantify the range in performance that can be expected from within a unit, and provides a cross check that the geometallurgical unit definition is robust.

1.2.3. Strategy for mineralogical measurement

Scoping or pre-scoping level studies are often labelled Ore characterisation programs. Corresponding mineralogical work is usu- ally done on coarse samples so that in situ textures and mineral grain sizes can be defined prior to grinding. Once flotation testing has been completed on an ore, mineralogical characterisation of milled feeds, concentrate and tailings provides additional detail which can further define opportunities for improved metallurgical performance. These measurements can take place at either the scoping or pre-feasibility stage of the program and when completed will provide additional information to help optimise the flowsheet.

Similar strategies are followed when diagnosing plant prob- lems. Samples representing speciﬁc areas of a mine can be sampled to better understand the reasons for inferior metallurgical performance, or plant streams can be sampled to measure the mineralogical characteristics relating to a speciﬁc metallurgical result.

1.2.4. Quantitative mineralogy

Mineralogical data are collected using QEMSCAN and EPMA.

QEMSCAN is an automated system that produces mineral maps (colour coded by mineral), through collection of rapidly acquired energy dispersive X-rays and/or by identification based on back- scattered electron images. The mineral maps describe the texture and mineral associations in each of the samples. In addition to the coloured map, the output of the QEMSCAN measurement in- cludes a quantitative measure of modal mineralogy, mineral grain size, mineral liberation and element deportment by mineral. A key component of the mineralogical result is that in addition to all economic minerals of interest, each gangue mineral is also identified and quantified.

In any automated mineralogical system, compositional information for every mineral is a required input to allow elemental deportment calculations. Although textbook compositions will provide a basis for these calculations, elements that occur as low level solid solution values are not captured when using textbook compositions. In addition, variations in composition can occur for the same mineral type from different deposits. Accurate deportment assessments require quantitative compositional analyses.

Quantitative compositional analysis at XPS is completed using a Cameca SX-100 Electron Microprobe. Compared to energy dispersive spectrometry (EDS), Electron Probe Microanalysis produces higher electron beam currents and increased beam stability, cou- pled with higher resolution wavelength dispersive spectrometry (WDS). These features allow for improved detection limits and accuracy of the resulting analysis. Detection limits can be as low as 100 ppm which provides detailed trace element compositions within the various mineral species. Individual grains and textures as small as 2–5

l

m are targeted for analysis and care is taken whenever possible to ensure proportional analysis relative to species availability and grain size. Resulting detailed compositional data is then input back into the QEMSCAN software, in order to re- ﬁne the ﬁnal elemental deportment calculations. Compositional data from microprobe analysis can also be used to update the

(3)

Species Identiﬁcation Protocol (SIP) within QEMSCAN and mineralogical measurements can be reclassiﬁed in order to update overall modal and deportment data. The composition of minerals varies from deposit to deposit, and often from geomet unit to geomet unit within a deposit. In order to produce accurate deportment information, EPMA analysis on each geomet unit within a deposit is necessary.

1.2.5. Metallurgical testing

In addition to the tools described to measure mineralogical characteristics, XPS utilises several mineral processing procedures in an attempt to correlate mineralogy with the metallurgical performance. These procedures use replicate flotation as described in high confidence flotation testing (Lotter and Fragomeni, 2010) and factorial design of experiments to quantify and minimise fundamental error and allow for distinguishing real differences in metallurgical performance between the geometallurgical units described.

1.2.5.1. High confidence flotation testing. High confidence flotation is based on two principles: one, to ensure that the ore sample is representative and has been well-blended and representatively subsampled; two, to perform a sufficient number of flotation tests with appropriate quality controls, so as to improve the reproducibility of the test data and reduce the error level. The use of representative samples, as previously discussed, is a necessary prerequisite for these high confidence tests.

The crushed mill feed sample is blended using a spinning rifﬂer with an odds and evens procedure as referenced in the earlier dis- cussion. After blending, a total of 10–12 subsamples are taken from the blended 2 kg charges. The ten to twelve pulverised samples are pooled to form an external reference distribution, which estimates the sample mean grade. It is a requirement that the relative standard deviation of these measurements does not exceed a set criterion.

A minimum number of replicate flotation tests is required for each high confidence flotation test observation. This is so as to en- gage the powerful averaging effects of the Central Limit Theorem, which in this context reduces the random errors and leaves resid- ual error normally-distributed (Grant and Leavenworth, 1988). It is an additional requirement that the first rougher concentrate masses agree within a prescribed relative standard deviation. As the procedure became applied to a wider range of base metal ores, this rule was rewritten so as to determine this point of measurement was made according to the cumulative recovery, and not the cumulative flotation time (Fragomeni and Lotter, 2002). The sets of reconciled head grades resulting from the mass and value balances of the flotation tests are pooled to form the internal reference distribution. The sample mean and standard deviation are calculated. The external reference distribution and the internal reference distribution means are compared as a percentage ratio called the call factor. This call factor must lie within a range of 96.73–103.27%. If not, the metal balance lies outside the 95% confidence limits and the flotation test is rejected and repeated.

1.2.5.2. Factorial design of experiments. Factorial design or design of experiments (DOE) is documented elsewhere in detail (Box et al., 1978). The use of DOE is signiﬁcant when used in conjunction with quantitative mineralogy as mineralogical features can be manipu- lated using a combination of mineral processing treatments such as grind size or reagent type and quantity. Frequently, XPS uses DOE laboratory scale testing in a three variable, two level DOE as a suitable format. The factorial cube involves eight test points, plus midpoint replication.

The sequence of these tests is randomised so as to avoid any operator preference. Data are interpreted in terms of main effects

and interactions on key performance measurements such as tailings grade, concentrate grade or recovery. This analysis is done using Mini-Tab or Stat-Ease software, however the same results can be calculated by hand.

The key advantage of this testwork layout is that each main effect is calculated from eight observations. Therefore, the standard error of that main effect is much lower than with conventional sin- gleton or duplicate tests. It is thus possible to test for small differences in recovery and establish, for example, the best combination or grind size and reagent addition as suggested by the mineralogy.

Typically, optimised conditions are tested using the high conﬁ- dence testing protocols to further increase conﬁdence in results.

Modelling of the response surface within the cube forms the ﬁnal step of the data interpretation. This often ﬁnds an internal (un- tested) point in the cube which offers superior performance (Deng et al., 2010). In certain cases, a partial factorial in the form of a La- tin Square may be performed where prior information has already been developed.

1.2.5.3. Flotation mini-pilot plant. Typically, high confidence flotation testing and DOE procedures are performed on a batch basis in either rougher or open circuit cleaner formats. Following batch testing and correlation between quantitative mineralogy and metallurgical testing, the geometallurgical units and flotation conditions are confirmed. Once confirmed, these geometallurgical units can be tested in a locked cycle test or in continuous mode using a Mini Flotation Pilot Plant (MPP) (Fragomeni et al., 2006).

The inherent advantages of the MPP over conventional piloting include smaller sample size, reduced time to steady state and abil- ity to test samples on a geometallurgical unit basis. This smaller sample size is made possible by stratified sampling. The MPP requires significantly lower sample mass and can operate a continuous flowsheet test including regrinding and column cleaning at as low as 8 kg/h and 600 kg of ore sample. Typically, this sample mass is available from pre-feasibility level exploration drilling programs, enabling the creation of a representative composite rather than using a bulk sample from one location within a resource. Testing of individual geometallurgical units or specific production periods are possible. MPP testing has been used for evaluation of reagent changes, (DiFeo, 2006) testing of various flowsheet concepts on a continuous scale, (Yu and Fragomeni, 2006) or to define design basis parameters for scale up and design (Ouellet and Fragomeni, 2007),

1.2.5.4. Statistical Benchmark Surveying – a sampling strategy.At plant operations scale, a procedure was developed and validated whereby a representative suite of ﬂowsheet samples could be taken from an operating concentrator for mass and value balancing, followed by QEMSCAN measurement across a series of closed size fractions. This became known as Statistical Benchmark Surveying, and made the connection between the operating plant and the microscopic-scale measurements of the minerals. The key in this system is to tie the tests of representativity to measurements of ore grade, so that the accepted set of ﬂowsheet samples correspond to treatment of ore of similar grade to that which is typically milled at that operation in the 3-month period surrounding the survey.

Ultimately two forms of this survey were developed. One, the standard benchmark form, or ‘‘Statistical Benchmark Survey’’, described the flowsheet behaviours under typical mining and milling conditions (Lotter, 2005; Lotter and Laplante, 2007a). This is used to update a concentrator’s performance across several years. The other, called the ‘‘Campaign Survey’’, was specifically designed to capture a suite of flowsheet samples for a known ore type that had been specifically mined and stockpiled for a demonstration run that would last between a week and 10 days (Lotter and Laplante, 2007b). Both systems operate at the 95% confidence level.

(4)

2. Case studies

2.1. Case study I: Nickel Rim South

The Process Mineralogy group of Xstrata Process Support (XPS), at the time Falconbridge, connected at an early stage with the exploration team who discovered this deposit. The majority of the deposit is made up of geometallurgical units which, prior to this discovery, accounted for a small percentage of the Strathcona concentrator feed. This triggered a number of characterisation and ﬂowsheeting studies to select the best processing option for this ore, including a redesign of the Strathcona Mill.

2.2. Case study II: Raglan

XPS, at the time, the Metallurgical Technology Group of Falconbridge, approached the Raglan management at the time of commissioning their operation in early 1998, with a view toward supporting and optimising their operation from a Process Mineral- ogy platform. Since this concentrator had been designed using conventional methods, an opportunity was perceived by the use of Process Mineralogy. The case study reviews the track record of these activities.

2.2.1. Nickel Rim South

Nickel Rim South (NRS), a high grade copper–nickel–PGE deposit with total proven and probable reserves of 9.6 Mt at 1.57% Ni, 2.85% Cu, 1.2 g/t Pt, 1.35 g/t Pd, 10.2 g/t Ag and 0.46 g/t Au (Xstrata Nickel, 2008), is located in the Sudbury region, approximately 400 km north of Toronto, Ontario, Canada. The deposit was discovered in 2001, after which several phases of mineralogical and metallurgical work were undertaken to support flowsheet optimisation prior to the start of production in 2009. During the exploration drilling phase, the project team identified several options for processing of the ore. These included: construction of a new concentrator, processing of the ores through the existing Strathcona concentrator or retrofitting the Strathcona Flowsheet to provide improved performance. The metallurgical program was designed to test all of these options in the context of on-going exploration drilling, resource definition and economic analysis.

2.3. Geology, mineralogy and geometallurgical unit deﬁnition Mineralisation in the NRS deposit occurs within the breccias along the main contact of the Sudbury Igneous Complex and in the underlying footwall rocks. Metal zonation and different host rocks result in several unique ore types or geometallurgical units, which as a result of their nature, were expected to show different metallurgical responses. A cross-section of the deposit is shown in Fig. 1.

A total of six distinct geometallurgical populations were deﬁned within the deposit based on their mineralogical characteristics.

These include four ores within the primarily Cu bearing, PGM Footwall:

Main (core).

Upper.

Fringe.

Low Sulphur PGM.

Along with two ores from the Ni bearing, higher pyrrhotite contact zone:

Sublayer Breccia.

Footwall Breccia.

Fig. 2shows typical textures from each of the geometallurgical units, as imaged by QEMSCAN. The scale of each image is approximately 3.5 mm3.5 mm.

Images show that the Main Footwall and Upper Footwall textures are coarsest. Sulphide textures in all other geometallurgical units are ﬁner grained and more commonly associated with silicate gangue.

A summary of the key mineralogical features within each ore type is presented inTable 1below.

2.4. Metallurgical testing of geometallurgical units

Representative composites for each of the geometallurgical units were created from drill core located throughout the resource.

As a benchmark, high conﬁdence replicate ﬂotation tests based on the Strathcona Flowsheet at the time of testing (Fig. 3), were

Fig. 1.NRS schematic cross-section looking Northwest.

(5)

completed for each unit and the results are presented inFig. 4. The x-axis in this case uses the total of nickel and copper grades rather than just nickel because of the dominance of copper in this ore.

Trial of the conventional grade/recovery curve i.e. nickel grade versus nickel recovery, produced a parabolic grade/recovery curve showing that the copper ﬂotation dominates the early part of the ﬂotation process at the expense of nickel. In order to obtain sensi- ble decisions on the ranking of the grade/recovery curves, the subtotal grades of nickel and copper had to be used in this instance.

The results showed each geometallurgical unit has a unique grade/recovery response. The Upper Footwall geometallurgical unit is characterised by a much higher pyrrhotite grade; the poor grade/recovery curve reﬂects the higher concentration of pyrrho-

tite in the concentrates. The Main Footwall constitutes the bulk of the deposit and is characterised by the best grade/recovery performance due to its high grade and coarse grained nature. The Fringe Footwall geometallurgical unit grade/recovery performance is good, but exhibits low base metal recovery to concentrates as a result of lower feed grade.

Lower PGE recoveries were identiﬁed for the Fringe ore type which is attributed to the mineralogy of the unit, i.e. loss of PGMs is related to ﬁne disseminated textures within silicate gangue.

These lower recoveries were also observed when a blend of footwall geometallurgical units was processed. A typical precious metal distribution by size fraction in the rougher–scavenger tails for a footwall blend is shown inFig. 5. PGE, Au and Ag losses are

Main Footwall Upper Footwall Fringe Footwall

Low Sulphur PGM Sublayer Breccia Footwall Breccia

Pentlandite Chalcopyrite Bornite

Pyrrhotite Feldspar Quartz

Chlorite Pyroxene Magnetite

Fig. 2.Geometallurgical unit textures as measured by QEMSCAN.

Table 1

Salient features of the geomet units in the NRS deposit.

Ore type Stratigraphic position/host rock Minerals and texture

Presence of pyrrhotite/Ni in pyrrhotite

Presence of PGM

Footwall Main (core)

Granitic Footwall Coarse MS veins of Cpy and Pn

Low levels (3–6 wt.%), Ni in solution is low (0.05 wt.% Ni)

Elevated levels dominated by michenerite and maslovite/

moncheite Upper

zone

Footwall Breccia and Granitic Footwall

Coarse MS veins of Po, Cpy and Pn

20–25% Po, Ni in solution averages 0.4% Ni

Low PGM content Fringe Along margins of main zone, in

Granitic Footwall

Disseminated or stringers of Cpy, Bn, Mill, Pn

No pyrrhotite Elevated and more varied species including michenerite, maslovite/monchiete, froodite, sperrylite, niggliite; ﬁner than in main geomet unit

Low Sulphur PGM

Lower extremites of Fringe in Granitic Footwall

Fine Disseminated Sulphides (2%)

No pyrrhotite Elevated PGM content

Contact Sublayer Breccia

Hosted by Maﬁc Norite Breccia (elevated MgO) at top of stratigraphic sequence

Blebby to SMS; Po, Pn and minor Cp

20–25% pyrrhotite. High levels of Ni in solution (average 1.07%)

Very low PGM content

Footwall Breccia

Hosted in Felsic to intermediate Footwall Breccia directly below Sublayer

Blebby, SMS and MS

20–25% pyrrhotite. High levels of Ni in solution (average 1.07%)

Very low PGM content

(6)

found predominately in the coarse size fraction (53–106

l

m) and are locked in gangue.

Contact ores have been subdivided into two geometallurgical units: Sublayer Breccia and Footwall Breccia. When the Mg-bearing Sublayer ores are processed in isolation from the Footwall Breccia ores, inferior metallurgical performance is observed. Froth collapse and low mass recoveries to primary concentrate result in delayed ﬂotation of liberated pentlandite and chalcopyrite and lead to higher losses to tailing. Mineralogical assessment of the ores has shown that orthopyroxene, and associated alteration minerals of anthophyllite and talc is unique to this ore, and has contributed to the poor performance. Gangue depressant testing was later initiated to improve the metallurgical performance of Sublayer ores.

The concept was lab tested and piloted using the MPP. Results showed that a signiﬁcant increase in kinetics and an improvement in both grade and recovery were achieved when these ores were processed using CMC. An example of the increase in kinetics is shown inFig. 6.

2.5. Assessment of new concentrator option

In the assessment of a new concentrator option, XPS was commissioned with the enviable task of studying a greenfields flowsheet for the Footwall ores. A ‘‘Virtual Flowsheet’’ based specifically on the mineralogical characteristics of the NRS geometallurgical units was designed and tested. Grain size averages Cu and Ni deportments, liberation data and mineralogical associations determined by QEMSCAN measurements, were used as an input into the grinding and flotation strategy. An example of average grain sizes for chalcopyrite and bornite are shown inFig. 7. The dashed line indicates the typical p80 of the rougher flotation feed.

It is recognised that all four ore types will be present in mill feed in various and estimated quantities. Therefore, the proposed ﬂow- sheet included a coarse primary grinding step followed by ﬂash or

‘‘scalp’’ flotation to facilitate fast floating chalcopyrite recovery, minimise over-grinding of sulphides and coarser brittle PGMs followed by extra grinding stages designed to better liberate fine PGM–gangue textures common to the Fringe Zone. The basic concept flowsheet is summarised in Fig. 8. Although not presented here, subsequent cleaning stages on each increment of concentrate were tailored to accommodate varying grades and cleaner circuit requirements.

Results of the testwork program indicated that a coarse primary and secondary grind improved concentrate grade at the same recovery. Ni and Cu metallurgy were very similar to the 2

3

4

C1

C2

C3

Tailings 1^stRo

2^ndRo

Scav 1

Fig. 3.Standard Strathcona rougher/scavenger ﬂowsheet. (1) Lime, PIBX. (2) Lime, PIBX, Dowfroth 250. (3) PIBX. (4) Sulphuric acid, copper sulphate, PIBX.

Ni+Cu Rougher Grade vs Ni Recovery

5 10 15 20 25 30 35

0 20 40 60 80 100

% Ni Recovery

% Ni + Cu Grade

Upper Footwall Main Footwall Fringe Footwall Contact Sublayer Contact Footwall Breccia

Fig. 4.Rougher grade recovery performance by geometallurgical unit.

PM Distribution vs. Size Fraction

0 5 10 15 20 25 30 35 40 45 50

-8 -25/+8 -53/+25 -106/+53 +106

Distribution (%)

Ag Au Pt Pd

Fig. 5.PGE, Au, Ag distribution by size fraction in the rougher-scav tails.

0 10 20 30 40 50 60 70 80 90 100

0 5 10 15 20

Flotation Time, min

Ni Recovery %

Footwall Breccia Sublayer Breccia Sublayer Breccia w CMC Fig. 6.Impact of CMC on Sublayer Breccia ore.

0 50 100 150 200 250 300 350 400

Cp- Contac t Cp -Upper FW Cp –Main FW Cp - Fringe FW Bo- Fringe FW

Geometallurgical Unit

Grain Size (micrometers)

Fig. 7.Chalcopyrite/bornite grain size averages by geometallurgical unit.

(7)

performance obtained using the Strathcona Flowsheet and therefore no benefit of flash or scalp flotation in improved fines recovery could be identified. The PGM, Ni and Cu recoveries associated with Fringe Footwall ores and the Low Sulphur PGM ores were improved. Results of testing a composite sample containing Fringe Zone were less clear and muted the effect, however when the geometallurgical units are processed separately the results are obvious as summarised inTable 2.

2.6. Blending tests and the Cu pre-ﬂoat concept

Xstrata Nickel’s Strathcona Mill processes ore from up to ﬁve mines with combinations of Contact and Footwall ore types at varying tonnages including several custom feeds. During the development of the NRS ﬂowsheet, several production scenarios were being investigated as a function of metal prices and mining costs.

This, in combination with a more clear definition of the NRS deposit tonnage value, moved the flowsheet development effort to focus on a Strathcona Mill retrofit option rather than a new greenfield concentrator for Footwall ores. The geometallurgical unit definition was key in defining a flowsheet that would afford flexibility without sacrificing metallurgical performance.

The subsequent phase of testing along with plant experience showed that blending of contact and footwall ores resulting in higher than a 1:1 ratio of Ni:Cu in the feed resulted in higher losses of nickel and copper to the pyrrhotite tailing. A Cu pre-ﬂoat option was introduced to allow for varied Contact and Footwall tonnage rates and limit the metal losses to the pyrrhotite tailings. This option required separate grinding circuits for Contact type and Footwall type ores as shown inFig. 9.

The results of the laboratory scale testwork are shown inTable 3, which deﬁnes the recovery gains with and without Cu pre-ﬂoat

This ﬂowsheet also allows for a tailored grinding strategy for Foot- wall Cu ore to take of advantage of the coarse grain sizes.

2.7. Mini-pilot plant testing

A series of MPP campaigns were completed to demonstrate the performance of both a footwall blend and a contact/footwall blend with the Strathcona Flowsheet, Staged Grind Flowsheet and the Strathcona Flowsheet with the addition of a Cu pre-ﬂoat (Yu and Fragomeni, 2005, 2006). A total of seven different ores and ﬂow- sheet concepts were investigated over two separate 1 week, 24 h/

day campaigns. The samples were primarily drill core from surface drilling and totalled only 5 tonnes of sample material. The footwall blend which included samples from Upper (6%), Main (46%), Fringe (22%) and 26% dilution was prepared according to the resource model available at the time of sampling. The footwall/contact sample was prepared with at 50/50 blend of footwall and contact ores.

The Low Sulphur PGM geometallurgical unit was not included within the resource model at this time, and therefore was not included in the composites prepared for piloting.

Given the low contribution of Fringe PGE units, the Staged Grind PGM flotation recovery gains were not significant and could not justify the capital expense of additional grinding capacity. How- ever, the piloting did show that the addition of a copper pre-float into the Strathcona Flowsheet would not only increase Ni and Cu recoveries but that Pt and Pd recoveries would improve as well.

This was attributed to the fact that the pre-float promoted chalcopyrite and pentlandite flotation in the front end of the circuit, and allowed PGM grains locked in silicate to be more effectively floated with the longer rougher scavenger-retention time. This flowsheet then allows for regrinding of locked PGM and cleaner flotation in the current cleaner circuit.

The final flowsheet concept selected included a retrofit to the existing Strathcona concentrator to process NRS ores based on pilot plant metallurgical results. The Strathcona Flowsheet retrofitted with a Cu pre-float and increased Cu/Ni separation capacity provides flexibility in grind size and blend ratio requirements while allowing the plant to increase throughput without a corresponding increase of metal lost to tailing. The retrofit option also resulted in an increase to Pt and Pd recoveries as compared to the Strathcona Flowsheet without the Cu pre-float retrofit.

2.8. Plant implementation

Following 3 years of shaft sinking and mine development, Ni Rim South began delivering development ore to Strathcona Mill Flash

Rougher 155µm

to

283µm 53µm

to

155µm 38µm

Regrind

Fig. 8.New concentrator concept: staged grind approach.

Table 2

Recovery comparison between Strathcona and Staged Grind Flowsheet.

Flowsheet Recovery

Ni% Cu% Pt% Pd%

Fringe

Strathcona 83.4 92.5 78.7 85.4

Staged grind 88.9 97.6 94.2 95.9

Difference 5.5 5.1 15.5 10.5

Low Sulphur PGM

Strathcona 61.8 74.6 65.8 68.4

Staged grind 83.8 91.5 89.5 92

Difference 22.0 16.9 23.7 23.6

(8)

in May 2009. The metallurgical program, supported by quantitative mineralogical data, recommended three retroﬁts to the concentrator, all of which have been implemented in preparation for full commercial production in 2010.

The first was the inclusion of the Cu pre-float, designed to limit Ni and Cu losses to the pyrrhotite tailing when processing blends of both contact and footwall feeds, and which was shown to improve Pt and Pd recoveries through selective flotation and subsequent grinding and flotation of fine PGM textures.

The second was the introduction of a CMC system, which will effectively deal with the Mg-bearing Sublayer Breccia ores. In an effort to further address this metallurgical issue, the mine has been encouraged to blend Sublayer Breccia with more abundant Foot- wall Breccia ores. This blending has been shown to improve performance of the Sublayer ores by virtue of froth stability in the primary rougher.

Finally, additional Cu/Ni separation capacity has been installed to allow for over 80% Cu recovery to Cu concentrate and allow for a signiﬁcant increase in Footwall ore tonnage rates.

The geometallurgical unit definition, lab scale flotation testwork, piloting and design basis were performed using exploration drill core and was essentially complete nearly 2 years before the mine commercial production. This has allowed for engineering and implementing of retrofits at the Strathcona Mill and the resulting flexibility and recovery gains forecast.

Currently, XPS is investigating the Stage Grind concept once again as the Low S High PGM GeoMet resource is being deﬁned.

Based on past testwork, various options are being considered including rougher–scavenger tailings regrinding and ﬂotation and the use of mixed collectors to promote middling recovery in the rougher scavenger circuit (Lotter and Bradshaw, 2009). Quantita- tive mineralogy will be used in conjunction with these testwork programs to deﬁne targets and quantify improvements in the metallurgical performance.

2.8.1. Raglan

The Raglan operation is located in the Ungava peninsula of north- ern Québec, and was commissioned into production in January 1998. The initial measured treatment capacity was approximately 108 tonnes per operating hour or 850,000 tpa, treating ore at a grade of 2.98% Ni. The commissioned grind at rougher ﬂoat feed level was

equivalent to ad80size of 68

l

m from a SAG/ball mill circuit with in- circuit crushing. Potassium Amyl Xanthate (PAX) was the standard xanthate used since operations were commissioned. The standard PAX dosage rate in the ﬂotation circuit was 300 g/t of ore milled.

A bulk concentrate for shipment to Sudbury was produced at a grade of 16% Ni and at a recovery of approximately 86.8%. These results closely matched their design equivalents, which were 100 tph milled with 87% nickel recovery at a 16% nickel grade in concentrate.

From commissioning, several projects were identiﬁed and successfully implemented so as to increase capacity to 1 million tpa, and to improve metallurgical performance (Langlois and Holmes, 2001;

Lotter et al., 2002; Fragomeni et al., 2005). A practice of surveying this circuit to benchmark the progress in the operation was implemented.

2.9. Geology, mineralogy and geometallurgical unit definition The orebody is hosted by an alternating succession of thick komatiic peridotite flows of Archean age, emplaced in a complex net- work of overlapping lava channels (Lesher, 1999). The host peridotites are pervasively serpentinised. Sulphide mineralisation consists of pyrrhotite, pentlandite and chalcopyrite which occur as discrete zones across a 60 km strike length. Although the zones are variable in terms of size, shape, and grade, three main geometallurgical units are consistently identified in most of the zones.

Each of these geometallurgical units have unique mineralogy and mineral processing characteristics (Fragomeni et al., 2005). These are:

Massive Sulphides.

Net-Textured Sulphides.

Disseminated Sulphides.

Most zones are dominated by Net-Textured Sulphides, with lesser amounts of both Massive Sulphides and Disseminated Sulphides.Fig. 10shows typical textures from each of the geometallurgical units, as imaged by QEMSCAN. The scale of each image is approximately 3.5 mm3.5 mm.

Mineralogical characterisation of numerous samples from the major zones within the Raglan deposit has been completed over Footwall Ore Contact Ore

Cu Pre-float Conc

Pri Ro Conc

Sec Ro Conc

Scav Conc

Scav Tails P56 75µm P56 75µm

Fig. 9.Schematic ﬂowsheet for footwall/contact blend with Cu pre-ﬂoat.

Table 3

Comparison of recoveries with and without Cu pre-ﬂoat.

Ore blend (contact/footwall) With Cu pre-ﬂoat Without Cu pre-ﬂoat Difference

Ni recovery Cu recovery Ni recovery Cu recovery Ni recovery Cu recovery

50/50 85.0 97.4 79.1 95.4 5.9 2.0

25/75 89.8 98.4 88.0 97.1 1.7 1.3

(9)

a 10 year period starting in 1998. Grain size data for each of the geometallurgical units has been combined and is presented in Table 4.

The knowledge gained in the ore characterisation studies has provided a fundamental understanding of how the ores react to milling and ﬂotation. A concurrent program of benchmarking the concentrator four times over the same 10 year period has enabled the company to identify opportunities for improved plant performance and to be in a position to predict metallurgical behaviour of new orebodies or the effect of a process change before implementation.

2.10. Statistical Benchmark Surveys

Four Statistical Benchmark Surveys at the Raglan Concentrator supported by size by size mineralogical measurements have provided a detailed characterisation of metallurgical performance over time. The surveys took place in 1998, 2000, 2003 and 2008. A summary of the survey results for these years, together with the salient features of operational changes, is shown inTable 5.

The 1998 survey was completed 6 months after plant commissioning and was focused on describing the actual flowsheet behaviour in detail so as to identify those sections of the flowsheet that may be improved. The as-commissioned flowsheet is shown in Fig. 11.

The sampling campaign used the early prototype survey design, drawing ﬁve independent 2-h surveys taken across a week of operations. A series of composite samples were prepared from four

of these. One of the five surveys was disqualified because the grade of ore milled fell outside the acceptance limits of the external reference distribution. The closed mass and value balance produced by Bilmat released subsamples of the flowsheet streams to sizing and measurement by QEMSCAN. More details of this survey may be seen inLotter et al. (2002). The key findings were:

1. The first rougher concentrate was sufficiently high grade to be bypassed directly to the final concentrate. This change would increase the retention time in the cleaner circuit and remove some of the fast-floating species from the cleaner.

2. The recleaner tailings and cleaner tailings had different sulphide abundances and liberation patterns, but were recycled to the same matchpoint, namely the scavenger circuit feed.

The recleaner tailings were thus rerouted to a new matchpoint at the Cleaner Feed due to its relatively high abundance of liberated sulphides.

3. A gangue depressant strategy for the rougher ﬂoat was necessary to control talc ﬂotation.

Massive Sulphide Net-Textured Sulphide Disseminated Sulphide

po pn cpy serpentine

Fig. 10.Geometallurgical unit textures as measured by QEMSCAN.

Table 4

Summary of grain sizes: major sulphides in Raglan geometallurgical units (inlm).

Geomet unit Pentlandite Chalcopyrite Pyrrhotite

Massive Sulphides 291 106 386

Net-Textured Sulphides 78 43 68

Disseminated Sulphides 68 36 60

Table 5

Summary of survey results.

dtph Grade Recovery Comments

Ni Cu Ni Cu

1998 Survey

Feed 105.6 3.1 1.0 100.0 100.0 Normal ore

Final conc 19.6 14.7 4.3 87.6 84.4 Regrind not implemented

Tailing 86.0 0.5 0.2 12.4 15.6

2000 Survey

Feed 116.0 3.6 0.9 100.0 100.0 High talc (problematic) ore

Final conc 21.5 17.1 3.8 87.4 78.8 Regrind installed

Tailing 94.5 0.6 0.2 12.6 21.2

2003 Survey

Feed 114.2 2.7 0.8 100.0 100.0 Future ore blend

Final conc 14.9 17.9 5.0 86.3 81.1 (Lower grade than previous surveys)

Tailing 99.3 0.4 0.2 13.7 18.9 Regrind in use

2008 Survey

Feed 166.2 2.3 0.7 100.0 100.0 Increased throughput

Final conc 19.9 16.2 4.5 84.3 77.6 Combined with lower grade ores

Tailing 146.4 0.4 0.2 15.7 22.4 Regrind in use

(10)

4. The cleaner tailings stream was dominated by ﬁne-grained locked and middling particles, and should be reground before being presented to the scavenger ﬂoat. This had caused a high circulating load in the cleaner scavenger circuit.

These changes led to the modiﬁed ﬂowsheet as shown inFig. 12.

The cumulative gains from these changes amounted to an increase in concentrate grade from 16% to 18% Ni, and recovery gains of 2.1% Ni, 1.5% Cu, 1.9% Pd, and 4.1% Pt.

The 2000 Benchmark Survey was designed as a campaigned run in which the metallurgical performance of a high talc ore was characterised. Despite very high head grades as a result of an increase in the proportion of massive sulphide material, the mineralogical characterisation showed that the high talc ores were texturally more complex that low talc ores. As a result, liberation in the feed dropped when processing these ores. Stronger mass recoveries which resulted in the ﬂotation of particles of complex sulphide–

silicate textures allowed the concentrator to maintain good overall recoveries. It was also concluded that the higher talc grades in the ore had little impact on the metallurgical performance, indicating the addition of a gangue depressant had a positive impact on the process.

The 2003 survey was also based on a campaigned run where the ore selected for mining and milling during the survey was deliber- ately conﬁgured so as to mimic the future ore mix of geometallurgical units (Lotter and Laplante, 2007b). The change from the ore mix in the early years of the operation to that of the future is basi- cally a reduction in the amount of Massive Sulphides and an increase in the amount of Disseminated Sulphides (Fragomeni et al., 2005).

Finally, the 2008 survey was designed to benchmark the concentrator after the plant had undergone an expansion to increase the throughput from 114 t/h in 2003 to 166 t/h in 2008 or 1.3 Mtpa. The corresponding primary grind moved from ad80size Rougher Flotation

Final Concentrate

Final Tailings Scavenger Flotation

Autogenous Mill

Secondary Mill

Fig. 11.As-commissioned Raglan ﬂowsheet, January 1998.

Rougher Flotation

To Final Concentrate

Final Tailings

Scavenger Flotation Autogenous Mill

Secondary Mill

To Final Concentrate

Regrind Mill To Final

Fig. 12.Modiﬁed Raglan ﬂowsheet, September 1998–September 2000.

(11)

of 68

l

m to 90

l

m. At the same time, head grades were dropping.

These two factors were expected to have a negative impact on metallurgical performance. This Benchmark Survey was designed to understand these impacts and identify opportunities where performance could be improved.

Fig. 13shows the size by size liberation of Ni Fe sulphide in the flotation feed for each of the four Benchmark Surveys. The definition of a liberated particle used in this study is any particle containing between 90% and 100% by area of the mineral of interest, in this case Ni Fe sulphide. The percentage of liberated particles is marked on the charts in red. Middling particles are defined as those containing more than 30% but less than 90% by area of the mineral, whilst locked particles are defined as those containing less than 30% by area of the mineral. The total liberated value is shown in text on the right side of each of the four charts.

Typical Ni Fe sulphide liberation in feed achieved prior to Rag- lan expansion efforts, is between 75% and 80% as characterised by the 1998 and 2003 surveys. The 2000 Benchmark Survey was completed on problematic, high Mg-bearing ores. These ores were texturally more complex than typical ores and resulted in a corresponding drop in feed liberation. The drop in liberation for the 2008 Ni Fe sulphide rougher ﬂotation feed, to 69.6% is a result of increased throughput in the plant, and lower head grades as compared to the 1998 and 2003 surveys.

The impact of the mill expansion is most clearly seen in the corresponding liberation analysis of the rougher tailing.Fig. 14shows the size by size Ni Fe sulphide liberation in each of the four surveys.

The comparison shows that there is a significant increase in liberated losses in the rougher tailing stream in the 2008 survey compared to data collected in previous surveys. Most obvious, is the increase in the slow floating ultrafine liberated particles. There is also an increase in the amount of liberated Ni Fe sulphide particles in the fractions that normally float very quickly. All previous surveys show very limited recoveries of liberated particles in the mid size fraction range to the rougher tailings.

It was concluded that additional rougher retention time was needed to recover liberated particles in mid and ﬁne size fractions as well as locked/middling particles currently lost to the rougher tailing.

In addition to assessing ways to increase flotation plant retention time, a test program designed to find a new collector suite capable of recovering more of the slower floating liberated pentlandite along with the complex sulphide–silicate middling particles was undertaken at XPS. After a series of laboratory scale flotation tests to identify better collector suites that would recover the fine liberated pentlandite more effectively, Sodium Isobutyl Xanthate (SIBX) was chosen to replace Potassium Amyl Xanthate (PAX). This was taken to a plant scale trial in 2007 using a version

+106+53

+25CS1-3

CS4+5

CS6CS7

Liberated

Middling

Locked 0

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Feed -1998

Liberated Middling Locked

+106+53

+25CS1-3

CS4+5

CS6CS7

Liberated

Middling

Locked 0

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Feed - 2000

+106+53

CS1-2

CS3CS4-5

CS6CS7

Liberated

Middling

Locked 0

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

-212/+106

-106/+53

CS1-2

CS3CS4-5

CS6CS7

Locked Middling

Liberated

0 5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Locked Middling Liberated

78.9%

69.6%

65.3%

76.1%

Fig. 13.Ni Fe sulphide rougher feed liberation comparison: 1998, 2000, 2003 and 2008 surveys.

(12)

of replicated blocking proposed by Napier-Munn (Napier-Munn, 1995; Lotter et al., 2009). The trial was successful and demon- strated a range of paymetal recovery improvements: Ni: 1.00%;

Cu: 0.96%; Pt: 5.07%, and Pd: 1.90%. The key to this result was in recognising that the first rougher concentrate bypasses the cleaner circuit and proceeds directly to final concentrate. Thus, any sign in the laboratory flotation work that showed a higher yield of pay- metals to this concentrate was positive. The SIBX increased the nickel recovery to this stream from 23.5% (using the PAX) to 44.2% (using the SIBX).

Further endeavour in 2009 addressed the topic of mixed collectors for synergistic performance gains in ﬂotation (Lotter and Brad- shaw, 2009). A mixed collector programme was initiated at XPS to provide a fundamental platform upon which better (mixed) collector systems could be formulated and taken to plant trials. One of these laboratory bench scale projects, reported in 2010 by Deng et al., improved this result for Raglan by the addition of Sodium Ethyl Xanthate (SEX) to the SIBX. This work, which will be taken to plant trial in 2010, further advanced the yield of nickel to ﬁrst rougher concentrate by 3.3%, equivalent to a 1% gain in overall recovery.

It has thus been shown that modern Process Mineralogy can now advise on recovery opportunities for ﬂoatable liberated sulp-

hides or complex middling particles, leading to reagent suite advancement and delivery of more effective collector formulation.

3. Conclusions

The use of process mineralogy has been successfully implemented at XPS. The integration of mineralogical and metallurgical information into testwork programmes and plant surveys has led to accurate process diagnosis, optimisation of concentrator performance and in ﬂowsheet design. The work is predicated upon integration of geological knowledge into the design and execution of the program, representative sampling, replicate ﬂotation testing, piloting, plant surveying along with high quality mineralogical data. The result is a reliable approach that enhances the diagnostic power of the data and strengthens predictive capabilities for each process.

Acknowledgement

The authors would like to thank the management of Xstrata Nickel for permission to publish this paper.

+106

+53

+25

CS1-3

CS4+5

CS6

CS7

Liberated Middling Locked 0

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Tailing -1998

+106

+53

+25

CS1-3

CS4+5

CS6

CS7

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Tailing - 2000

+106

+53

CS1-2

CS3

CS4-5

CS6

CS7

5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Tailing - 2003

-212/+106

-106/+53

CS1-2

CS3

CS4-5

CS6

CS7 ^Locked

Middling Liberated

0 5 10 15 20 25 30

% Ni Fe Sulphide

Size Fraction

Ni Fe Sulphide Liberation in Rougher Tailings - 2008 Locked Middling Liberated

39.2%

8.0% 7.4%

5.9%

Fig. 14.Ni Fe sulphide rougher tailing liberation comparison: 1998, 2000, 2003 and 2008 surveys.

(13)

References

Bond, F.C., 1952. The third theory of comminution. Trans. AIME Miner. Eng. 193, 484–494.

Box, G.E.P., Hunter, W.G., Hunter, J.S., 1978. Statistics for Experimenters. Wiley, United States of America.

Cabri, L.J., 1981. Relationship of mineralogy for the recovery of PGE from ores. In:

Cabri, L.J. (Ed.), Platinum-Group Elements: Mineralogy, Geology, Recovery, Special vol. 23. Canadian Institute of Mining, Metallurgy and Petroleum, pp.

233–250.

Charland, A., Kormos, L.J., Whittaker, P.J., Arrué-Canales, C.A., Fragomeni, D., Lotter, N.O., Mackey, P, Anes, J., 2006. A case study for the integrated use of automated mineralogy in plant optimisation: the Montcalm concentrator. In: Proc.

Automated Mineralogy, MEI Conference, Brisbane, July 2006.

Cochran, W.G., 1946. Relative accuracy of systematic and stratiﬁed random samples for a certain class of populations. J. Ann. Math. Stat. 17, 164–177.

Dai, Z., Bos, J.-A., Lee, A., Wells, P., 2008. Mass balance and mineralogical analysis of ﬂotation plant survey samples to improve plant metallurgy. In: Minerals Engineering, ‘‘Flotation 2007’’, special ed., vol. 21, pp. 826–831.

Deng, T., Yu, S., Lotter, N.O., Di Feo, A., 2010. Laboratory testwork of mixed xanthates for the Raglan ore. In: Proc. Canadian Mineral Processors, Ottawa, January 2010, Paper No. 16, pp. 253–268.

DiFeo, A., 2006. Mini Pilot Plant Craig Ore – DNBX. Falconbridge Technology Centre Internal Memo.

Fragomeni, D., Lotter, N.O., 2002. Personal Communication. Falconbridge.

Fragomeni, D., Boyd, L.J., Charland, A., Kormos, L.J., Lotter, N.O., Potts, G., 2005. The Use of End-Members for Grind-Recovery Modelling, Tonnage Prediction and Flowsheet Development at Raglan, Paper No. 6. Canadian Mineral Processors, Ottawa. pp. 75–98.

Fragomeni, D., Hoffman, M., Kelly, A., Yu, S., Lotter, N.O., 2006. Flotation mini pilot plant experience at Falconbridge limited. In: Proceedings Mineral Process Modelling, Simulation and Control, Laurentian University, Sudbury, Canada, pp.

329–355.

Grant, E.L., Leavenworth, R.S., 1988. Statistical Quality Control. McGraw-Hill. pp.

213–215.

Gy, P.M., 1979. Sampling of Particulate Materials, Theory and Practice. Elsevier, Amsterdam.

Henley, K.J., 1983. Ore-dressing Mineralogy: A Review of Techniques, Applications and Recent Developments, Special Publication, vol. 7. Geological Society of South Africa. pp. 175–200.

Langlois, P., Holmes, J., 2001. Process Development at Raglan Concentrator, Paper 30. Canadian Mineral Processors. pp. 427–451.

Lesher, C.M., 1999. Komatiic Peridotite-hosted Ni–Cu(PGE) Deposits of the Raglan Area, Cape Smith Belt, New Québec. Laurentian University Mineral Exploration Research Centre. pp. 205.

Lotter, N.O., 1995. A Quality Control Model for the Development of High-Conﬁdence Flotation Test Data. M.Sc. Chem. Eng. Thesis, University of Cape Town.

Lotter, N.O., 2005. Statistical Benchmark Surveying of Production Concentrators.

Ph.D. Met. Eng. Thesis, McGill University, Montréal.

Lotter, N.O., 2010. Stratiﬁed Sampling of Drill Core, Paper No. 11. Canadian Mineral Processors, Ottawa. pp. 163–179.

Lotter, N.O., Bradshaw, D.J., 2009. The formulation and use of mixed collectors in sulﬁde ﬂotation. In: Proc. Flotation ’09, MEI, Cape Town, November, 2009.

Lotter, N.O., Fragomeni, D., 2010. High-conﬁdence ﬂotation testing at Xstrata process support. J. Miner. Metall. Process. 27 (1), 47–54.

Lotter, N.O., Laplante, A.R., 2007a. Statistical benchmark surveying of production concentrators. Miner. Eng. 20, 793–801.

Lotter, N.O., Laplante, A.R., 2007b. The campaign survey model – a case study at the Raglan Mine, Québec. Miner. Eng. 20, 480–486.

Lotter, N.O., Whittaker, P.J., Kormos, L.J., Stickling, J.S., Wilkie, G.J., 2002. The development of process mineralogy at Falconbridge limited, and application to the Raglan mill. CIM Bull. 95 (1066), 85–92.

Lotter, N.O., Kowal, D.L., Tuzun, M.A., Whittaker, P.J., Kormos, L.J., 2003. Sampling and ﬂotation testing of Sudbury Basin drill core for Process Mineralogy modelling. Miner. Eng. 16, 857–864.

Lotter, N.O., Comeau, G., Kormos, L.J., Fragomeni, D., Di Feo, A., 2009. Plant Scale Trial of Isobutyl Xanthate at Raglan Concentrator Using Reference Distributions, Paper 6. Canadian Mineral Processors, Ottawa. pp. 79–105.

Mackey, P.J., Nesset, J.E., 2003. The Impact of Commissioning and Start-up Performance on a Mining/Metallurgical Project, Paper No. 21. Canadian Mineral Processors, Ottawa. pp. 331–347.

McIvor, R.E., Finch, J.A., 1991. A guide to interfacing of plant grinding and ﬂotation operations. Miner. Eng. 4 (1), 9–23.

McIvor, R., Lavallee, M., Wood, K., Blythe, P., 1990. Functional performance characteristics of ball milling. In: 22nd Annual Meeting of the Canadian Mineral Processors, Ottawa, ON, Canada, January 16, 1990, pp. 100–126.

McKay, N., Wilson, S., Lacouture, B., 2007. Ore characterisation of the Aqqaluk deposit at Red Dog. In: 39th Annual Meeting of the Canadian Mineral Processors, Ottawa, Paper No.5, January 23–25, 2007, pp. 55–74.

Middleditch, D., Lotter, N., 2008. Laboratory Quality Standards Manual, Xstrata Process Support Internal Document, May 13, 2008, pp. 14–16.

Middleditch, D., Carriere, P., Danyliw, J., 2009. Crushing Plant Manual, Xstrata Process Support Internal Document, October, 2009, pp. 22–26.

Napier-Munn, T.J., 1995. Detecting performance improvements in trials with time- varying mineral processes – three case studies. Miner. Eng. 8 (8), 843–858.

Ouellet, N., Fragomeni, D., 2007. Design Basis Summary for the Kabanga Plant, XPS Final Report, November 17, 2007.

Petruk, W., 1976. The application of quantitative mineralogical analysis of ores to ore dressing. CIM Bull. 767, 146–153.

Petruk, W., 1988. The capabilities of the microprobe Kontron image analysis systems: application of mineral beneﬁciation. Scan. Microsc. 2, 1247–1255.

Petruk, W., Hughson, M.R., 1977. Image analysis evaluation of the effect of grinding media on selective ﬂotation of two zinc–lead–copper ores. CIM Bull. 787, 128–

135.

Petruk, W., Schnarr, J.R., 1981. An evaluation of the recovery of free and unliberated mineral grains, metals and trace elements in the concentrator of Brunswick Mining and Smelting Corporation Limited. CIM Bull. 833, 132–159.

Peyerl, W., 1983. The Metallurgical Implications of the Mode of Occurrence of Platinum-group Metals in the Merensky Reef and UG-2 Chromitite of the Bushveld Complex, Special Publication, vol. 7. Geological Society of South Africa.

pp. 295–300.

Restarick, C.J., 1976. Pulp sampling techniques for steady state assessment of mineral concentrators. In: Proceedings, Sampling Symposium, Australasian Institute of Mining and Metallurgy, pp. 161–168.

Shannon, E.R., Grant, R.J., Scott, D.W., 1993. Back to basics – the road to recovery milling practice at Brunswick. In: Proc. 25th Canadian Mineral Processors, Paper No. 2, Ottawa, January, 1993.

Trahar, W.J., 1981. A rational interpretation of the role of particle size in ﬂotation.

Int. J. Miner. Process. 8, 289–327.

Triffett, B., Veloo, C., Adair, B.J.I., Bradshaw, D.J., 2008. An investigation into the recovery of molybdenite in the Kennecott Utah copper bulk ﬂotation circuit. In:

Minerals Engineering, ‘‘Flotation 2007’’, special ed., vol. 21, pp. 832–840.

Xstrata Nickel, 2008. Ore Reserves and Mineral Resources. Xstrata Nickel Publication.

Yu, S., Fragomeni, D., 2005. Nickel Rim South Mini Pilot Plant (MPP) Flotation Campaign Falconbridge Technology Center Internal Report, June 2005.

Yu, S., Fragomeni, D., 2006. Nickel Rim South Mini Pilot Flotation Plant Campaign Report, Falconbridge Technology Center Internal Report, September 2006.