You are on page 1of 8

HOME

Building a Geometallurgical Model


in Iron Ores using a Mineralogical
Approach with Liberation Data
P Lamberg1, J Rosenkranz2, C Wanhainen3, C Lund4, F Minz5,
A Mwanga6 and M Amiri Parian7

ABSTRACT
A geometallurgical model is currently built in two different ways. The first and the most
common way relies on geometallurgical testing, where a large number of samples are analysed
for metallurgical response using small-scale laboratory tests, eg Davis tube testing. The second,
mineralogical approach focuses on collecting mineralogical information over the orebody and
building the metallurgical model based on mineralogy. At Luleå University of Technology,
Sweden, the latter method has been adopted and taken further in four ongoing PhD studies.
The geological model gives modal composition by the help of element-to-mineral conversion
and Rietveld X-ray diffraction. Texturally, the orebody is divided into different archetypes, and
liberation measurements for each of them are carried out in processing fineness using IncaMineral,
a SEM-based technique. The grindability and liberation spectrum of any given geological unit
(sample, ore block, domain) are extrapolated from the archetypes. The process model is taken
into a liberation level by mass balancing selected metallurgical tests using the particle tracking
technique. The approach is general and can be applied to any type of ores. Examples of ongoing
studies on iron and massive sulfide ores are given.

INTRODUCTION
A geometallurgical model combines geological and In geometallurgical programs, the weakest points are
metallurgical information into a spatial and predictive tool normally in inadequate information collected from drill cores
to be used in production planning and management in the and the small number of samples collected for variability
mining industry. A geometallurgical program is an organised testing. In laboratory tests, quite a number of small samples
way to establish such a model and usually it involves the should represent large tonnages of the ore. Commonly some
following steps: tens of carefully selected and prepared samples are tested
•• collection of geological data (Sammelin-Kontturi, Wanhainen and Martinsson, 2011;
Schouwstra et al, 2010; Johansson and Wanhainen, 2010;
•• collection of samples for metallurgical testing
Morrell, 2009; Philander and Rozendaal, 2008, 2010), but
•• laboratory testing, ie ore variability testing
there are examples where the whole program is based on
•• developing geometallurgical domains
less than ten samples (David, 2007; Suazo, Kracht and Alruiz,
•• developing a model for metallurgical parameters across 2010). This sets high requirements for the sample selection,
the geological database sampling and sample preparation to avoid the sampling error
•• developing a process model, ie metallurgical model, rising so high that it limits the usefulness of collected data
which uses the metallurgical parameters of the geo model (Gy, 1982; Pitard, 1989a and 1990b). There lies also a dilemma
•• plant simulation using the geo and process models in selecting and preparing metallurgical samples based on
•• calibration of the models via benchmarking for existing geological information: tested samples should represent the
operations (Bulled and McInnes, 2005; David, 2007; Dobby full variability of the ore in terms of metallurgical response
et al 2004). and this can be known only after the tests have been done.

1. Professor in Geometallurgy, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: pertti.lamberg@ltu.se
2. Professor in Mineral Processing, Luleå University of Technology, SE-971 87 Luleå, Sweden. jan.rosenkranz@ltu.se
3. Senior Lecturer in Ore Geology, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: christina.wanhainen@ltu.se
4. PhD student, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: cecilia.lund@ltu.se
5. PhD student, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: friederike.minz@ltu.se
6. PhD student, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: abdul.mwanga@ltu.se
7. PhD student, Luleå University of Technology, SE-971 87 Luleå, Sweden. Email: mehdi.parian@ltu.se

THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013 317
P LAMBERG et al

To ensure that samples are representative, new rock processing of the ore. The production model returns figures
measurement and analysis techniques are needed. The like the amount of final products in a given time, production
challenge is that measurements and analyses have to be value and production costs.
done for a very big number of samples, from thousands to The process model will be described first because it sets
hundreds of thousands. Techniques must be fast, inexpensive the requirement for the geological model. In the mineralogy-
and preferentially fully automated. This is an area of rapid based process model, three different levels are possible. The
technological development and there are logging systems first level, called 1D, uses mineralogical information on an
available for automated lithology identification (eg optical unsized basis, ie no information on particle sizes. The model
borehole imaging, OBI), chemical analysis directly from the describes quantitatively, with mathematical equations how
drill cores ore drilling sludge, measuring rock mechanical each mineral behaves in the process. An example of such a
properties when drilling or from drill cores (AWD analysis model is the flotation model developed for the Collahuasi
while drilling; Vantandoost, Fullagar and Roach, 2008) and copper deposit in northern Chile (Suazo et al 2009). As this
semi-quantitive analysis of mineralogy with reflectance level doesn’t handle particle sizes, the grinding fineness of the
spectroscopy (CSIRO, 2011; Haavisto and Kaartinen, 2009; Da process is fixed.
Costa et al, 2009; Pirard et al, 2008).
Particle size is of paramount significance in mineral
Since none of the above-mentioned methods can directly processing. This is taken into account in the second modelling
measure metallurgical response small-scale laboratory tests level (2D), where the metallurgical functions are given on
called geometallurgical tests have been developed. They are minerals by size basis. For example, in a flotation circuit the
fast, inexpensive and can be done with small sample amounts. recovery function parameters are given for each mineral for
Examples of such for comminution characterisation are GeM each particle size class or the function includes the particle
Comminution Index (GeMCi), GeM RBT Lite (GeM Rotary size as a parameter. This model level enables playing with the
Breakage Tester, Lite) and EQUOtip (portable hardness tester; grinding fineness to find the best economic solution between
JKTech, 2010). For flotation, a test called JKMSI (JK Mineral the grinding fineness, throughput and recovery. This level is
Separability Indicator) has been developed (Bradshaw, 2010) common in the designing, studying and optimising of mineral
and for magnetic separation a Davis tube test has been known processing circuits but examples in geometallurgy are very
for a long time (Niiranen and Böhm, 2012). For leaching, few.
different basic and diagnostic tests are available (Lamberg et Minerals don’t behave independently in processes because
al, 1997). The basic idea of geometallurgical tests is that it is they occur as particles with varying composition and size.
possible to run them for a large number of samples and the The purpose of the comminution circuit is not to reduce the
database would be a basis for a geometallurgical model, ie particle size but to liberate the minerals adequately for the
they would partly replace the variability testing. production of saleable concentrate with high recoveries. The
Side-by-side with the development of new scanning, third level (3D) takes into account the mineral liberation and
analysis and testing technology there has been progress in is called mineral by size by liberation level. This level enables
quantitative mineralogical tools. Mineralogical studies have optimising the comminution in terms of liberation, not only
always formed a basis for flow sheet development and have by size, as in the previous level. The metallurgical functions
been included in geometallurgical programs. The problem must be based on liberation or particle composition. This
in the efforts of giving mineralogy a bigger role to play in can be regarded as a problem since unit process models of
geometallurgy is that mineralogical analyses are tedious and liberation level are rare.
costly, which makes them unsuitable for mass analysis for tens Based on the level of the process model, the information
of thousands of samples. Also lack of methods to incorporate from the geological model varies. In the unsized mineral level
several mineralogical parameters like mineral textures in the the geological model must give accurately mineral grades,
geometallurgical models. ie modal mineralogy. In the second level the geological
In 2010 a geometallurgical research group was established model must, in addition to modal mineralogy, describe also
at Luleå University of Technology. This paper summarises the behaviour in grinding, ie grindability. In the third level
the result of the groups in regard to applying a mineralogical the geological model must include information to forecast
work approach in geometallurgy. Firstly it describes the mineral liberation. These challenges are discussed in the
mineralogical approach in general and then it gives examples following section with examples.
from ongoing research projects from different ore types.
GEOLOGICAL MODEL
MINERALOGICAL APPROACH
The mineralogical approach in building a geometallurgy Modal mineralogy
model means that the full model is based on mineralogy Accurate information on modal mineralogy is a minimum
(Lamberg, 2011). The model is here divided into three parts. requirement for the geological model in the mineralogical
The first part is the geological model, the second is the process approach. This information includes mass proportion of
model and the third is the production model. The geological each (important) mineral in the ore but also their chemical
model gives quantitative information of minerals in such a composition. For example, in massive sulfide ores it is
way that other information, like elemental grades or lithology, common that sphalerite shows significant variation in its
is not needed. The process model must be capable of taking chemical composition within the orebody.
the information from the geological model and forecasting For the modal analysis, state-of-the-art techniques include
the metallurgical response for any given geological unit automated mineralogy, quantitative X-ray diffraction and
(sample, ore block, geometallurgical domain). The process element to mineral conversion (Lamberg et al, 2013; Lund,
model must deliver at least throughput and recoveries for Lamberg and Lindberg, 2013). Each of these methods
the main minerals and commodities. These two models are has advantages and disadvantages (Table 1). Automated
combined in a production model which is capable of handling mineralogy is the most reliable but it is time-consuming and
production time frame and different scenarios for mining and expensive. Quantitative X-ray diffraction can be very fast and

318 THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013
BUILDING A GEOMETALLURGICAL MODEL IN IRON ORES USING A MINERALOGICAL APPROACH WITH LIBERATION DATA

TABLE 1
Methods for determining modal composition in geological and process samples.

Technique Advantage Disadvantage Hardware/software References


Automated mineralogy Reliable, works with Tedious, requires sizing of samples MLA, QEMSCAN, IncaMineral Fandrich et al, 2007
all types of samples
Quantitative X-ray diffraction Fast High detection limit, does not work for Panalytical Empyrean and High Score Castendyk et al,. (2005);
non-crystalline phases, requires expertise Plus; Bruker D8 Advance and Topas Hastenes (2012); Paine (2005)
Element to mineral conversion Fast, inexpensive, low Requires additional selective methods HSC Chemistry, SimFloat, Bilco Whiten (2008), Lamberg et al
detection limit when complex mineralogy (1997)

inexpensive but it has problems in high detection limits and in The combination of element to mineral conversion with
reliability if the number of phases is high. Element to mineral quantitative X-ray diffraction by Rietveld refinement is a
conversion uses chemical assays and information on the possible solution for reliable modal analysis in Malmberget
chemical composition of minerals and mathematically solves as well as in other ores with relatively simple mineralogy.
the mineral grades. The method is fast and inexpensive, and However, in ores where minerals of interest occur as trace
detection limits can be as low as for chemical assays, ie parts quantities and distribution of elements of interest is complex,
per million level, but the method fails in complex mineralogy. more mineralogical information is required from the sample.
However, in geometallurgy it is often possible to simplify An example of such a case is Rockliden, the study target of
mineralogy for example by combining gangue silicates. Minz, et al 2013. Rockliden is a massive sulfide ore located
The Malmberget iron ore is one of the case study targets south of the Skellefte volcanic-hosted massive sulfide (VMS)
in the PhD projects of Lund (Lund, Lamberg and Lindberg,
2013), Mwanga and Parian (Lamberg et al, 2013). The
Kiirunavaara and Malmberget iron ores operated by LKAB
contribute about 90 per cent of the iron ore production in
Europe. The annual production in Malmberget is about 14 Mt
of ore at 43.1 per cent Fe (LKAB, 2011). In the Malmberget
ore field more than 20 different orebodies of both magnetite
and hematite are found over an area of 2.5 × 5 km2. The ore
is hosted by volcanic rocks of Svecofennian age (c 1.9 Ga,
Bergman, Kübler and Martinsson, 2001).
Lund, Lamberg and Lindberg (2013) developed a practical
way to calculate the mass proportion of nine minerals from
routine X-ray fluorescence analysis of Malmberget iron ore
samples. The calculation assumes that the sample consists
of these nine minerals and that the chemical composition
of the minerals doesn’t differ from the average of electron
microprobe analyses (Lund, Lamberg and Lindberg, 2013).
Instead of calculating all minerals in one go it was found that
a better result was reached when calculation was divided
into four rounds in HSC Chemistry. In comparison with
modal analysis by QEMSCAN it was found that the relative
standard deviation of element to mineral conversion is about
12 per cent (Figure 1). This is reasonably high and the quality
is not good enough for using mineral grades as a basis for
resource estimate. Lund, Lamberg and Lindberg (2013)
regarded that quality as good enough to be used in defining
geometallurgical domains.
To improve the quality of the element to mineral conversion
and to overcome the trouble that if there are new minerals
present in the sample the technique fails, Lamberg et al (2013)
studied the applicability of quantitative X-ray diffraction using
Rietveld refinement. It was found that the modal analysis by
X-ray diffraction with Rietveld refinement is poorer than
by element to mineral conversion (Figure 1; Lamberg et al,
2013). However, by combining these techniques and using
a weighted non-negative least squares technique in HSC
Chemistry the result improved significantly, thus approaching FIG 1 - Comparison of the result for analysing magnetite grade in the samples by
the level required for a geometallurgical model. These results QEMSCAN and element to mineral conversion (above; Lund et al 2013) and with
are preliminary and the study continues with comparison of X-ray diffraction + Rietveld (y-axis) and element to mineral conversion (x-axis;
different X-ray analysis setups Rietveld refinement software below; Lamberg et al 2013). Plot marks in the lower graph indicate the quality of
and comparison to automated mineralogy. the Rietveld refinement by R-value; the lower the value, the better the analysis.

THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013 319
P LAMBERG et al

district. It has a relatively high Sb content and in the flotation Abdul Mwanga is developing a small-scale method for
tests Sb distributes uncontrollably (ie without a clear pattern) measuring comminution characteristics for geometallurgy. So
between Cu-Pb concentrate, Zn concentrate and flotation far, the grindability characterisation has focused on finding
tailings. Minz et al (2013) found that the Sb mineralogy links between mineralogy, grinding energy and particle
is very complicated, including at least the following five size distribution. Tests with Malmberget ore have shown
main Sb phases: tetrahedrite, bournonite, meneghinite, that in the grinding product the mineral grades vary by
boulangerite and gudmundite. In the flotation tests, however, size, indicating a non-random breakage. Magnetite grade is
these minerals behave in a logical manner depending on the highest in the middle size fractions, whereas all gangue
their chemical composition and associating minerals. The minerals are enriched in the fine size fractions and tremolite
estimation of the mass proportion of the Sb phases and other also in the coarsest fraction (Figure 2). In the grinding model
main sulfides (pyrite, pyrrhotite, chalcopyrite, sphalerite and the forecasting of overall particle size distribution (PSD) and
galena) can’t be done with simple bulk chemical assays and mineral grades by size (2D) has been decoupled. Using the
XRD, as for Malmberget. One possible solution is to find a relatively simple assumption that:
link between the main sulfide minerals and Sb minerals and •• relative mass proportions of minerals are retained in
add paragenetic constraints in the mass balance equations. narrow size fractions
For example, in chalcopyrite-dominating ores the main Sb •• overall mineralogical mass balances are conserved
mineral is tetrahedrite. •• the particle size distribution is received from the PSD
model, the mineral composition of the mill discharge by
Grindability size fraction can be forecasted accurately (Figure 3).
A geometallurgical model should be capable of forecasting
the behaviour of the ore in comminution. Industrial Mineral textures and liberation
comminution circuits are commonly operated to produce a To reach the 3D level in geometallurgy, the geological model
given fixed particle size distribution, measured eg with 80 per must somehow describe the liberation characteristics of the
cent passing, and this is regulated mainly by changing the ore when it is crushed and ground. The link between the
throughput. Therefore the comminution characteristic of the geological information and liberation distribution is mineral
ore, ie grindability, controls the production volume. textures. Lund, Lamberg and Lindberg (2013) and Lamberg

FIG 2 - Mineral composition by size in the mill feed (above) and mill discharge (below). The P80 per cent value for the mill feed is 975 μm and for the mill
discharge 165 μm. Mineral abbreviations: Mgt – magnetite, Ap – apatite, Ab+Or – albite and orthoclase, Bt – biotite, Tr+Di – tremolite and diopside.

320 THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013
BUILDING A GEOMETALLURGICAL MODEL IN IRON ORES USING A MINERALOGICAL APPROACH WITH LIBERATION DATA

‘association index’ is introduced. It is calculated for each


target mineral separately and, for example, the ‘association
index of magnetite for albite’ describes how commonly
magnetite is associated with albite when the degree of
liberation of magnetite and the mineral grade of albite are
considered. It is calculated for the particle population where
the fully liberated particles of the target mineral are first
removed. Then the mass proportion of magnetite associations
are calculated, ie how much of magnetite is occurring with
the other minerals. These values are divided by the mineral
grades of the other minerals, which are normalized to 100
per cent. For example, if 50 per cent of the magnetite is
associated with albite and the albite grade is 50 per cent
then the value is one and there is no preference for albite.
FIG 3 - Comparison mineral grades in size fractions of the mill discharge However, if 75 per cent of magnetite is associated with albite
showed in Figure 2; simulated (y-axis) versus measured (x-axis). and the albite grade is 50 per cent then the value is 1.5, and
The simple model developed can forecast the mineral grades by size accurately it indicates that magnetite is preferentially associated with
in the mill discharge. Mineral abbreviations as in Figure 2. albite. Finally, if only ten per cent of magnetite is associated
with albite and the albite grade is still 50 per cent then the
and Lund (2012) found that in the Malmberget iron ore the association index is 0.2 and magnetite is rarely with albite.
liberation size (the size where 95 per cent of the minerals are Reflecting on primary mineral textures, the association index
liberated) of magnetite and hematite is controlled mainly by
is expected to be very high if a mineral occurs as inclusions in
the grain size of the ore minerals, and the mineral association (ie
another mineral, whereas if two minerals are not in contact
the minerals commonly occurring in magnetite-rich particles)
with each other the index should be zero.
is controlled not only by modal mineralogy but also by the
ore textures. Lund et al (2013) and Lamberg and Lund (2012) Figure 5 shows the association index of five different
divided the breccia ore of Malmberget into different textural samples representing three different textures. The association
types by macroscopical and microscopical observations. Five indexes are almost similar in the Fsp–textures (feldspar;
samples representing three different textures were crushed Pz_Fsp and Fa_Fsp) and Ap-textures (apatite; Fa_Ap and
to a given fineness, samples were sized and resin mounts Pz_Ap) are quite identical, confirming that the assumption
were prepared. Size fractions were analysed by QEMSCAN that they represent similar textures holds. The Ap-texture is
for mineral liberation. Liberation data was processed using characterised by magnetite being preferentially associated
binning algorithm of the particle tracking. In this way each with apatite, whereas in the Fsp-types the association of
sample had identical particle classes but their mass proportion magnetite with apatite is rare. The Amp-type (amphibole)
were different. For example, each sample had a particle class is closer to the Ap-type and the question to be answered is:
of ‘fully liberated magnetite of size fraction 38 - 75 µm, but its are these textures so different that it justifies the existence of
mass proportion varied between the samples. separate Ap- and Amp-types?
Liberation is dependent on modal mineralogy. Generally Lamberg and Lund (2012) developed preliminary
the degree of liberation increases as the mineral grade systematics to convert textural information into particles.
increases (Figure 4). The relative mass proportion of This is done through model samples, archetypes, which are
association with two minerals is also directly related to their collected into a library as follows:
grades. For example, in the ternary system magnetite-apatite-
•• For each textural type a representative sample or set of
tremolite the mass proportion of magnetite associated with
tremolite increases as the tremolite grade increases. In order samples is selected.
to decouple the association and modal abundance a new

FIG 5 - Association index for magnetite in size fraction 150 - 300 μm in five
samples representing three different textures (Fsp, Amp and Ap). Values
above one indicate that target mineral is preferentially associating with
this mineral and values lower than one that the association of the target
mineral with other mineral is rarer than mineral grades would suggest. From
FIG 4 - Liberation degree of magnetite in a size fraction 150 - 300 μm against the diagram one can read that magnetite is preferentially associating with
the magnetite grade in the same fraction in four samples representing three apatite in Fa_Amp, Fa_Ap and Pz_Ap samples, whereas this association is
different textures (Fsp, Amp and Ap). rare in Fa_Fsp and Pz_Fsp samples.

THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013 321
P LAMBERG et al

FIG 6 - Mineral grade and recoveries in Mörtsell dry magnetic separation (cobbing) test done for the low-grade magnetite ore of Fsp type (see text).

•• The samples are ground close to the processing fineness PROCESS MODEL
and sized.
The process models using the 1D or 2D levels are common
•• Liberation analysis is done by size.
in minerals processing. The developing of the 3D model can
•• As the liberation measurement is not possible or practical be done for separation processes by measuring liberation in
in very fine or coarse particle sizes, the liberation all the streams and mass balancing the liberation data using
distribution of these size fractions is estimated by an
the particle tracking technique (Lamberg and Vianna 2007).
extrapolation technique described by Lamberg and
Even though this is tedious to do, it will make the process
Vianna (2007).
model more generic. The hypothesis behind it is that: similar
•• For each sample the particles (of the liberation analysis) particles (in size, shape and composition) will behave in the
are classified in a systematic way using a technique
process in the same manner regardless of which part of the
developed by Lamberg and Vianna (2007).
orebody they come from. Therefore, once the model has been
•• The classification scheme should be established in such a created it can be used for the whole deposit.
way that all modal and textural types are considered and
correspondingly all possible particle types in all of the Lamberg and Lund (2012) and Lund et al (in prep)
archetypes are presented. developed a 3D model for a dry magnetic separation. Figure 6
•• This results in each archetype having similar particles; shows the distribution of minerals on an unsized basis
only their relative abundance will vary from one sample between the concentrate and tailings in one representative
to another. It is possible that the mass proportion of some test. In a particle size of P80 = 1 mm about 30 per cent of the
particles is zero in some archetypes. material is rejected into the tail (MagTail) with about six per
The archetypes are then used in the geometallurgical system cent magnetite losses, ie the recovery of magnetite into the
in the following way (Lamberg and Lund 2013): magnetic concentrate is 94 per cent (Figure 6).
•• The geological model provides the modal mineralogical It is obvious that magnetite-rich particles enter into the
composition and the textural type for each sample or ore magnetic concentrate, whereas particles rich in gangue
block minerals are found in the tailings, and the particle tracking
•• Based on textural type a corresponding archetype is technique gives quantitative information on the behaviour
selected. The archetype lists all the particles possibly of particles. Figure 7 shows the recovery of magnetite-albite
present and gives their mass proportions by size. binaries into the magnetic concentrate by size and as a
•• The particle size distribution comes from a conventional function of magnetite grade. Interestingly, the recovery curve
grinding model. is upwardly convex in fine particle sizes and straightens
•• The mass proportion of particles in the sample is towards the coarse particles.
calculated by refining the mass proportion of particles Table 2 shows the comparison of dry magnetic separation
in the archetype so that modal mineralogy can be back results from an actual test and from a simulation. For the
calculated from the particles and given by the geological simulation the liberation data of the Pz_Fsp sample was used
model. and the process model parameters were taken from the test
A simple algorithm called breakage model, developed by done for the Fa_Fsp sample. Good matching between the
Lamberg and Lund (2012) is used. measured and simulated values validates the model and

322 THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013
BUILDING A GEOMETALLURGICAL MODEL IN IRON ORES USING A MINERALOGICAL APPROACH WITH LIBERATION DATA

Bradshaw, D J, 2010. Development of a new tool for process


mineralogy, in Proceedings Process Mineralogy ’10.
Bulled, D and McInnes, C, 2005. Flotation plant design and production
planning through geometallurgical modelling, in Proceedings
Centenary of Flotation Symposium, pp 809-814 (The Australasian
Institute of Mining and Metallurgy: Melbourne).
Castendyk, D N, Mauk, J L and Webster, J G, 2005. A mineral
quantification method for wall rocks at open pit mines, and
application to the Martha Au–Ag mine, Waihi, New Zealand.
Applied Geochemistry, 20:135–156.
CSIRO HyLogging Systems [online], 2011. Hyperspectral
mineralogical logging and imaging of drill core and chips.
Available from: <http://www.csiro.au/files/files/py3w.pdf>.
Da Costa, G M, Barron, V, Ferreira, C M and Torrent, J, 2009. The use
of diffuse reflectance spectroscopy for the characterisation of iron
ores, Minerals Engineering, 22:1245-1250.
FIG 7 - The recovery of magnetite-albite binary particles into the magnetic
David, D, 2007. The importance of geometallurgical analysis in plant
concentrate by size fraction as a function of magnetite grade in the particle. study, design and operational phases, in Proceedings Ninth Mill
Operators’ Conference, pp 241-247 (The Australasian Institute of
TABLE 2 Mining and Metallurgy: Melbourne).
Comparison of the dry magnetic separation result (measured) with the
Dobby, G, Bennett, C, Bulled, D and Kosick, G, 2004. Geometallurgical
simulated one. For the simulation, the liberation data of the Pz_Fsp sample was
modelling: the new approach to plant design and production
used and the process model parameters were derived from Fa_Fsp samples forecasting/planning and mine/mill optimisation, in Proceedings
of 36th Annual Meeting of the Canadian Mineral Processors.
Simulated Measured
Fandrich, R, Gu, Y, Burrows, D and Moeller, K, 2007. Modern SEM-
Fe % 63.7 63.9 based mineral liberation analysis, International Journal of Mineral
Fe recovery % 89.6 90.2 Processing, 84:310–320.

Si % 3.10 3.06 Gy, P M, 1982. Sampling of particulate materials: theory and practice,
Developments in Geomathematics 4 (Elsevier).
P% 0.13 0.09
Haavisto, O and Kaartinen, J, 2009. Multichannel reflectance spectral
assaying of zinc and copper flotation slurries, Int J Miner Process,
verifies the hypothesis that similar particles behave in the 93:187–193.
same manner in the process regardless of their origin.
Haestnes, K H and Sørensen, B E, 2012. Evaluation of quantitative
A big challenge in the particle-based models and simulation X-ray diffraction for possible use in the quality control of
is the grinding model. In an ongoing work the model is granitic pegmatite in mineral production, Minerals Engineering,
divided into two parts. The first one predicts the throughput volume 39, pp 239-247.
and particle size distribution in a traditional way. The other JKTech, 2010. JKTech’s monthly e-Newsletter, December.
part forecasts the liberation by using a hypothesis that the Johansson, B and Wanhainen, C, 2010. Flotation and leach tests
liberation distribution within narrow size fractions is retained. performed within a geo-metallurgical project on gold in the Aitik
Cu-Au-Ag-(Mo) deposit, in Proceedings Conference in Minerals
CONCLUSIONS Engineering, pp 61-72.

At Luleå University of Technology a mineralogical Lamberg, P, 2011. Particles: the bridge between geology and
approach has been adopted and further developed in the metallurgy, in Conference in Minerals Engineering, Preprints 1-16.
geometallurgical research projects. Preliminary results Lamberg, P, Hautala, P, Sotka, S and Saavalainen, S, 1997.
from the ongoing studies show that it is possible to base the Mineralogical balances by dissolution methodology, in Proceedings
Short Course on ‘Crystal Growth in Earth Sciences’, pp 1-29.
geometallurgical model on mineralogy. However, more work
is needed to verify the results and to evaluate if the accuracy Lamberg, P and Lund, C, 2012. Taking liberation information into
is good enough. a geometallurgical model: case study Malmberget, Northern
Sweden, in Proceedings Process Mineralogy ’12 Conference.

ACKNOWLEDGEMENTS Lamberg, P and Vianna, S M S, 2007. A technique for tracking


multiphase mineral particles in flotation circuits, in Proceedings
The research projects described here have received VII Meeting of the Southern Hemisphere on Mineral Technology (eds:
financial support from the Centre of Advanced Mining and R M F Lima, A C Q Ladeira, C A Da Silva), pp 195-202.
Metallurgy (CAMM), Hjalmar Ljundbohm Research Centre Lamberg, P, Parian, M, Mwanga, A and Rosenkranz, J, 2013.
(HLRC), LKAB and Boliden Mineral AB. Cooperation with Mineralogical mass balancing of industrial circuits by combining
researchers in the ProMinNET network has greatly helped in XRF and XRD analyses, Preprints, Conference in Minerals
the research projects. Engineering 2012.
LKAB (Luossavaara-Kiirunavaara AB), 2011. Annual Report and
REFERENCES Sustainability Report 2011, 122 S (in Swedish).
Bergman, S, Kübler, L and Martinsson, O, 2001. Description of Lund, C, Lamberg, P and Lindberg, T, 2013. Practical way to quantify
regional geological and geophysical maps of northern Norrbotten minerals from chemical assays at Malmberget iron ore operations:
County (east of the Caledonian orogen), Geol Surv Sweden Ba, an important tool for the geometallurgical program, Minerals
56:110. Engineering, in press.

THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013 323
P LAMBERG et al

Minz, F, Bolin, N-J, Lamberg, P and Wanhainen, C, 2013. Detailed Pirard, E, Bernhardt, H-J, Catalina, J C, Brea, C, Segundo, F and
characterisation of antimony mineralogy in a geometallurgical Castroviejo, R, 2008. From spectrophotometry to multispectral
context at the Rockliden Ore Deposit, North-Central Sweden, imaging of ore minerals in visible and near infrared (VNIR)
Minerals Engineering, submitted. microscopy, Proceedings ICAM 2008 – Ninth International Congress
for Applied Mineralogy, pp 57-62 (The Australasian Institute of
Morrell, S, 2009. Getting optimum value from ore characterisation
Mining and Metallurgy: Melbourne).
programs in design and geometallurgical projects associated
with comminution circuits, in Proceedings Tenth Mill Operators’ Pitard, F F, 1989a. Pierre Gy’s sampling theory and sampling practice,
Conference, pp 167-170 (The Australasian Institute of Mining and volume I, Heterogeneity and Sampling (CRC Press: Boca Raton).
Metallurgy: Melbourne). Pitard, F F, 1989b. Pierre Gy’s Sampling Theory and Sampling Practice,
Niiranen, K, and Böehm, A, 2012. A systematic characterisation of the Sampling Correctness and Sampling Practice, volume II (CRC Press:
orebody for mineral processing at Kiirunanvaara iron ore mine Boca Raton).
operated by LKAB in Kiruna, Northern Sweden, in Proceedings Sammelin-Kontturi, M, Wanhainen, C and Martinsson, O, 2011.
XXVI International Mineral Processing Congress (IMPC), paper Gold mineralogy at the Aitik Cu-Au-Ag deposit, Gällivare area,
1039, pp 03855-03864. northern Sweden, GFF, 133: 1-12
Paine, M, König, U and Staples, E, 2011. Application of rapid Schouwstra, R, de Vaux, D, Hey, P, Malysiak, V, Shackleton, N, and
X-ray diffraction (XRD) and cluster analysis to grade control of Bramdeo, S, 2010. Understanding Gamsberg: a geometallurgical
iron ores, in Proceedings, 10th International Congress for Applied study of a large stratiform zinc deposit, Minerals Engineering,
Mineralogy (ICAM), pp 495-501. 23:960–967.
Philander, C and Rozendaal, A, 2008. Geometallurgical challenges Suazo, C J, Kracht, W and Alruiz, O M, 2010. Geometallurgical
of Namakwa Sands - A South African titanium-zirconium heavy modelling of the Collahuasi flotation circuit, Minerals Engineering,
minerals mine, in Proceedings ICAM 2008: Ninth International 23:137–142.
Congress for Applied Mineralogy, pp 459-464 (The Australasian
Institute of Mining and Metallurgy: Melbourne). Whiten, B, 2008. Calculation of mineral composition from chemical
assays, Mineral Processing and Extractive Metallurgy Review,
Philander, C and Rozendaal, A, 2010. The hards liberation project of volume 29, pp 83-97.
Namakwa Sands, chronicles of a geometallurgical success, Process
Mineralogy ’10.

324 THE SECOND AUSIMM INTERNATIONAL GEOMETALLURGY CONFERENCE / BRISBANE, QLD, 30 SEPTEMBER - 2 OCTOBER 2013

You might also like