Date post: | 23-Nov-2023 |
Category: |
Documents |
Upload: | khangminh22 |
View: | 2 times |
Download: | 0 times |
A process mineralogy study of grinding
characteristics for the polymetallic orebody,
Lappberget Garpenberg
Gustav Lood Stark
Natural Resources Engineering, master's
2021
Luleå University of Technology
Department of Civil, Environmental and Natural Resources Engineering
Abstract
Most of the high-grade ores have been depleted globally, thus the effective processing of the low-grade and
complex ores require a comprehensive mineral characterization through the process mineralogy/
geometallurgical approaches. 30-70 % of the total energy consumption in mining comes from the
comminution step in mineral processing. This study, is aimed to investigate how different mineral domains
in Lappberget, Garpenberg affect the grinding energy and throughput of an autogenous grinding mill (AG)
and how blending different mineralogical domains will have an effect on throughput. The results were
obtained through automated mineralogy using a Zeiss Sigma 300 VP at the QANTMIN scanning electron
microscope (SEM) laboratory (Luleå University of Technology) and an in-house grindability test developed
by Boliden Mineral AB. There is approximately a multiple of three times differences in the amount of
energy consumption and throughput between the hardest and softest mineralogical domains. This difference
is attributed to mineral composition of the individual domains and mineral characteristics. Blending
different samples indicate that a higher throughput can be achieved and one possible hypothesis is that the
harder minerals act as grinding media.
Behovet av en bättre förståelse för hur olika mineralogiska attribut och olika mineral påverkar de olika
delarna av mineralseparationsprocesserna, drivs av ett högre behov av metaller och mer effektiv
gruvproduktion. 30-70 % av den totala energikostnaden vid gruvbrytning uppstår vid krossning- och
malning-steget i mineralprocessen. Den här studien syftar till att undersöka hur olika mineralogiska
domäner inom Lappberget, Garpenberg påverkar malningsenergin samt genomsättningen för en autogen
kvarn, och hur blandningar av olika mineralogiska domäner påverkar genomsättning. Resultaten togs fram
med hjälp av automatisk mineralogi med ett Zeiss Sigma 300 VP vid QANTMIN SEM laboratoriet på
Luleå tekniska universitet, och en malbarhetstest framtagen av Boliden Mineral AB. Det är ungefär tre
gånger så stor skillnad mellan de olika mineralogiska domänerna vad gäller energikonsumtion samt
genomsättning, och skillnaden härrör från mineralsammansättningen för de individuella domänerna och
mineralens karaktär. En blandning av olika prover leder till att en högre genomsättning kan uppnås och en
möjlig hypotes är att hårdare mineral agerar som malmedia.
Keywords: Automated mineralogy, autogenous milling, grinding characteristics, mineral
characterization, process mineralogy/geometallurgy, blending strategies, energy consumption, throughput,
scanning electron microscope (SEM)
Table of Contents
1 Introduction ................................................................................................................................................ 1
2 Literature Review ....................................................................................................................................... 2
2.1 Comminution .......................................................................................................................................... 3
2.1.1 Crushing ........................................................................................................................................... 5
2.1.2 Grinding ........................................................................................................................................... 6
2.1.2.1 Tumbling mill ........................................................................................................................... 8
2.1.2.2 Stirred Mills .............................................................................................................................. 9
2.1.2.3 Mill Liners .............................................................................................................................. 11
2.1.2.4 Grinding Circuits..................................................................................................................... 12
2.1.3 Energy consumption ...................................................................................................................... 14
2.1.4 Grindability .................................................................................................................................... 16
2.2 Mineral Characterization ...................................................................................................................... 17
3 Experimental Work .................................................................................................................................. 18
3.1 Sample Selection ............................................................................................................................... 18
3.2 Sample Preparation ........................................................................................................................... 18
3.2.1 Grindability test ......................................................................................................................... 18
3.2.2 Analytical methods .................................................................................................................... 19
3.3 Equipment ......................................................................................................................................... 20
3.4.1 Experimental procedure for a grindability study ........................................................................... 21
3.4.2 Blending test procedure ................................................................................................................. 23
3.5 Sample characterization .................................................................................................................... 23
3.5.1 X-Ray Fluorescence analysis ..................................................................................................... 24
3.5.2 Optical microscopy .................................................................................................................... 25
3.5.3 Automated SEM ......................................................................................................................... 25
3.6 Equations........................................................................................................................................... 27
4 Results and Discussion ............................................................................................................................ 29
4.1 Grindability results ............................................................................................................................ 29
4.1.1 Blending ore types and the effect on energy consumption and pebble wear rate ...................... 31
4.2 Mineral characterization ................................................................................................................... 33
4.2.1 Blending of two different mineralogical domains. .................................................................... 43
4.3 Relationship between mineralogy, ks-value, and energy .................................................................. 44
4.4 Back-calculations vs XRF data ......................................................................................................... 49
4.5 Source of errors ................................................................................................................................. 49
4.5.1 Sample preparation errors .......................................................................................................... 49
4.5.2 Analytical methods errors .......................................................................................................... 50
5 Conclusions and recommendations .......................................................................................................... 53
References ................................................................................................................................................... 55
Appendices ................................................................................................................................................. 58
Appendix A: Assay data from XRF and automated mineralogy ...................................................... 58
Appendix B: Mineral grade vs. ks-value ............................................................................................. 61
Table of Figures
Figure 1 Overview of the Garpenberg area, from Tiu et al.,( 2019). ............................................................ 2 Figure 2 Simplified mineral processing flowsheet, adapted from Wills & Finch (2015). ............................ 4 Figure 3 Different breakage mechanism (a) impact or compression, (b) chipping, (c) abrasion. Adapted
from (Wills and Napier-Munn, 2006). .......................................................................................................... 5 Figure 4 Three stage crushing closed circuit, adapted from Metso, 2015. ................................................... 6 Figure 5 Schematic over HPGR, reduction ratio of 25, adapted from Metso 2015. ..................................... 7 Figure 6 Different types of grinding mills, from Metso 2015. ...................................................................... 8 Figure 7 Tumbling mill movement with descriptive text and example of a tumbling mill. Adapted from Will
& Finch 2015 and Metso, 2015. .................................................................................................................... 9 Figure 8) Stirred mills. Adapted from Metso 2015. .................................................................................... 11 Figure 9 Different types of mill liners, adapted from Wills & Finch 2015................................................. 12 Figure 10 Increasing grinding capacity through applying a closed circuit. Green circle, circles the more
energy efficient solution. Adapted from Wills & Finch 2015. ................................................................... 13 Figure 11 Two different AG/SAG mill circuit, adapted from Gupta & Yan, 2006. ................................... 14 Figure 12 An imaginary example of the different reduction characteristics plotted on a logarithmic paper,
adapted from Hukki, 1961. ......................................................................................................................... 15 Figure 13 Total amount of energy and amount Cu produced over time. The orange arrow indicates that the
total amount of energy have diverged from the total tonnes Cu produces, adapted from Calvo et al., 2016.
.................................................................................................................................................................... 16 Figure 14 (A) Inside of Sala mill. (B) Sala mill at Boliden pilot plant. (C) To the left, mechanical sieve and
to the right laboratory jaw crusher. ............................................................................................................. 20 Figure 15 (A) Inside of laboratory AG mill drum. (B) Laboratory AG mill used in grindability test. ....... 21 Figure 16 AccuPyc II 1340 used for density measurement. ....................................................................... 21 Figure 17 Grindability test developed by Boliden Mineral AB, repeated for three different times (60, 90
and 150 min). .............................................................................................................................................. 22 Figure 18 Schematic overview of the ore characterisation process. ........................................................... 23 Figure 19 XRF at Boliden pilot plant with sample holders. ....................................................................... 24 Figure 20 (A) Polished epoxy mounts, (B) Optical microscope, (C) Tarnished chalcopyrite with an Ag2S
film (50 times the microscopic magnification). .......................................................................................... 25 Figure 21 (A) SEM for automated mineralogy at QANTMIN LTU, (B) Randomized pattern of analyzed
fields, (C and D) Galena (bright/blue phase) enclosed in enstatite (dark grey/green phase) in domain MN-
MSPG, size fraction >90 µm. ..................................................................................................................... 27 Figure 22 Pebble wear rate, four different ore domains.............................................................................. 29 Figure 23 Grinding product for different ore domains................................................................................ 30 Figure 24 Pebble wear rate, blending test (ks-value). ................................................................................. 32 Figure 25 Grinding product blending test. .................................................................................................. 32 Figure 26 (A) A elongated phlogopite grain, in coulure (brown phase). (B) Same grain as in (A) but in back
scattered electron image (BSE-image). ....................................................................................................... 35 Figure 27 Mineral grade plotted against ks-value. ...................................................................................... 48 Figure 28 Back-calculation mineralogy vs. XRF assays. ........................................................................... 49 Figure 29 Galena, trends with and without outliers. ................................................................................... 61 Figure 30 Mica, trends with and without outliers. ...................................................................................... 62 Figure 31 FeOx, trends with and without outliers. ..................................................................................... 63 Figure 32 Amphibole/pyroxene, trends with and without outliers. ............................................................ 64 Figure 33 Garnets, trends with and without outliers. .................................................................................. 65
Figure 34 Pyrrhotite, trends with and without outliers. .............................................................................. 66 Figure 35 Feldspars, trend with and without outliers. ................................................................................. 67
List of Tables
Table 1 Examples of different types of crushers. From Metso, (2015); Wills & Finch, (2015) ................... 5 Table 2 Examples of low and high speed stirrer mills. Adapted from Wang & Forssberg, 2003; Wills &
Finch, 2015. ................................................................................................................................................ 10 Table 3 Operation cost in comminution (Abbey et al., 2015). .................................................................... 15 Table 4 Different ore domains from Garpenberg, Sweden, used in this study. .......................................... 18 Table 5 Different equipment used for grindability study at Boliden Pilot Plant. ........................................ 20 Table 6 Blending test series, blending content............................................................................................ 23 Table 7 Conditions used for XRF measurements. ...................................................................................... 24 Table 8 SEM settings for the different ore domains. ................................................................................. 26 Table 9 Predicted energy consumption when grinding each domain, ks-value and theoretical throughput.
.................................................................................................................................................................... 31 Table 10 Estimated energy consumption, ks-value compared and theoretical throughput against calculated
values for the blending test. ........................................................................................................................ 33 Table 11 Automated mineralogy result for each domain, fraction size by fraction size and a calculated bulk
mineralogy. ................................................................................................................................................. 42 Table 12 Calculated mineral composition of blending domain QDM and RPSG ...................................... 43 Table 13 R2 values from plotting mineral grade against Ks-value for each mineral. .................................. 47 Table 14 Unaltered version of QDM and MN-MSPG normalized mineral composition. .......................... 52 Table 15 Zn and Pb assays for Sphalerite and Galena, from automated mineralogy. ................................ 58 Table 16 Fe-assays from automated mineralogy. ....................................................................................... 59 Table 17 XRF assays for each domain and fraction. .................................................................................. 60
Acknowledgement First of all I would like to thank Boliden Mineral AB for the financial support and opportunity to finish my
Master of Science in Natural Resource Engineering with an interesting master thesis. Secondly I would like
to thank Yousef Ghorbani, my main academic supervisor for his endless support and all the time spent
discussing on how to write a proper academic report. Also I would like to thank my other supervisors at
Luleå University of Technology, Christina Wanhainen and PhD student Glacialle Tiu. Adam Mc Elroy
acted as my industry supervisor and contact person at Boliden Mineral AB and I would like to thank him
for the time and help during my master thesis.
Most of all I would like to thank my family for the support and patience during the time of my master thesis.
To my friends, Viktor Grundström-Mattsson and Maximilian Körtge that went through the same type of
process… We made it!
1
1 Introduction
The demand for metals and minerals in the world is increasing. With lower grades and more complex
mineralogy, there is a requirement for improved knowledge of how different minerals and their attributes
affect mineral processes. With lower overall ore grades, more material needs to be processed. This results
in more material being pushed through the mineral processing plants, which requires higher energy, water
and chemical costs (Bradshaw, 2014; Dominy et al., 2018). The comminution step of mineral processing
accounts for 30-70% of the total energy consumption in a mining operation (Ballantyne and Powell, 2014;
Jeswiet and Szekeres, 2016; Napier-Munn, 2015; Wills and Napier-Munn, 2006). With a better
understanding of how chemical and physical characteristics of minerals might affect the grinding energy
and throughput of an autogenous grinding mill (AG), a more effective and responsive mine production can
be achieved through a more accurate prediction of throughput and energy cost of different ore types.
Garpenberg mine is the oldest still-operating mine in Sweden and is located approximately 180 km
northwest of Stockholm within the Bergslagen ore district (Bindler et al., 2017). It is located on the limb of
the Garpenberg syncline and the largest orebody in the mine is Lappberget with 58 million tonne combined
resources and reserves at 3.42% Zn, 1.68% Pb, 0.06% Cu, 70 g/t Ag and 0.41 g/t Au (B. Tiu et al., 2021;
Jansson et al., 2011). Through the characterization process of Lappberget a PhD study was initiated by
Boliden AB. PhD student Glacialle Tiu have to this date published, three papers about Lappberget that
concerns ore mineralogy and trace element (re)distribution; sulfide chemistry and trace element
deportment; tracking silver mineralogy through the processing plant and deposit (A. Tiu et al., 2019; B. Tiu
et al., 2021; C. Tiu et al., 2021). To compliment the work done by G. Tiu, characterization of the grinding
characteristics within Lappberget was initiated by Boliden AB. With an improved understanding of the
grinding characteristics in the spatial constrained mineralogical domains classified by G. Tiu with the aim
for a comprehensive geometallurgical model. The increasing knowledge of how different domains within
Lappberget might behave within the mill is of use for the processing plant when optimizing the mill
performance and mine planning department on how to make a more flexible mine plan.
Two different objectives were given for this study. The first one was to study a possible connection between
the mineralogy and grindability in Lappberget whereas the second one was to investigate a connection
between ore blending and throughput. The first objective is studied by using a combination of Boliden AB’s
grindability test and automated mineralogy whereas the second objective is studied through three blending
tests with known quantities and mineral composition in how different blends would affect the throughput.
3
2 Literature Review
2.1 Comminution
Mineral processing is aimed to separate the commercial valuable mineral(s) from the host rock. The mineral
processing starts with the comminution, which is to reduce the ore to a particle size suitable for enrichment
processes that separate the valuable mineral(s) from gangue. The process of a standard mineral processing
operation can be seen in Figure 2. After the ore have been separated from the bedrock through blasting it is
transported to a crusher where it is decreased in size. In the last comminution step, the ore is milled and
further decreased in size until it is at an optimal size for downstream processes. After the ore have passed
through the comminution process it will enter the classification and separation step of a mineral processing
operation that will produce concentrates and a tailings product. The concentrates are later transported to a
smelter and the tailings could either be stored in a tailings dam or be reused as backfill material in some
mining operations. In an example from Gupta & Yan (2006), after blasting, the run of mine ore (ROM) is
transported and fed to the crushers with rock sizes up to one meter. The crusher reduces the size to a
maximum of 50 mm before the material is fed to the grinding mill (i.e. ball mill) (Gupta and Yan, 2006).
In Garpenberg, Boliden Mineral AB uses a 250 mm maximum for an autogenous grinding mill. The
material that is larger than the optimum size will be screened out and returned to the crushers. The statement
from Gupta & Yan (2006) may not be true for all mining operations but illustrates a closed circuit primary
crushing operation. In the grinding mill the material will be further decreased in size, until the commercial
valuable minerals have been liberated from the gangue material in the host rock and has the desired size for
downstream processes.
4
When ROM ore is crushed, the main breakage mechanism is different from the type of breakage
mechanisms the material is exposed to during grinding (Figure 3). The main breakage mechanism in
crushing is the impact or compression breakage where the material is compressed between rigid surfaces.
However, grinding is a more random process that is subject to probability laws, grinding can be subject to
several of the presented breakage mechanism according to Wills & Finch, (2015):
Impact or compression, when forces are applied normally to the particle surface;
Chipping is due to asymmetrical forces; and
Abrasion is due to parallel forces to the surface of the particle.
Figure 2 Simplified mineral processing flowsheet, adapted from Wills & Finch (2015).
5
All of these mechanisms will distort the particles until their shape is beyond certain limits determined by
their mineralogical properties which will cause them to break (Wills and Napier-Munn, 2006). Abrasion
mechanism happens in comminution when the particle grains rub against each other and produces a finer
particle size, whereas impact and compression is more suitable when producing a larger size fraction (Bis
et al., 2018).
2.1.1 Crushing
To obtain the goal of crushing, i.e. to reach a certain particle size before grinding commences, several
different crushing stages can be used in a crushing circuit (Figure 4) (Hukki, 1961). The reduction in size
in crushing is mainly by compressing the ore against rigid surfaces or by impact against surfaces in a rigidly
constrained motion (Jeswiet and Szekeres, 2016; Wills and Finch, 2015). There are several different types
of crushers. Some examples are shown in Table 1
Table 1 Examples of different types of crushers. From Metso, (2015); Wills & Finch, (2015)
Crushers Type of crusher
Gyratory Primary
Jaw Primary
Impact Primary
Cone Secondary/tertiary
Crushing is usually a dry process in several different stages depending on how many stages is required to
reach the desired particle size. The reduction ratio for crushers is small and ranges from 3 to 6 in each step
of the crushing process (Metso, 2015). Wills and Finch (2015) defines the reduction ratio as “the ratio of
maximum particle size entering to the maximum particle size leaving” the same definition is used in
Mwanga et al. (2016) and Jeswiet and Szekeres (2016). In the set up in Figure 4 below, the crushers have
Figure 3 Different breakage mechanism (a) impact or compression, (b) chipping, (c)
abrasion. Adapted from (Wills and Napier-Munn, 2006).
6
been split into primary, secondary and tertiary crushers, in combination with screens. The primary crusher
reduces the run of mine ore before the ore passes through a screen so that the undersized material bypasses
the secondary crusher. The same procedure is repeated for the tertiary crusher where only particles too large
to pass through the secondary screen are crushed in the tertiary crusher. Crushing circuits are split between
open and closed circuits. In Figure 4, a closed circuit is used for the tertiary crushers whereas for the primary
and secondary an open circuit. Primary crushers are always in open circuits and to reduce the material
crushed in the primary crushers, screens can be installed. The undersize material will pass directly to the
secondary crusher without passing through the primary crusher. The main differences between the open
and closed circuit is that in an open circuit, the material will not be screened and then recycled back to the
crusher. According to Wills and Finch (2015) the main primary crushers are jaw and gyratory crushers,
whereas the secondary/tertiary crushers are mainly cone crushers which is a modified gyratory crusher.
2.1.2 Grinding
The aim of the last stage of comminution, grinding, is to reach the particles size where the liberation of the
valuable minerals is sufficient for downstream processes to be as effective as they can be. All ores have an
optimum particle size where the mineral grains have been fully liberated from gangue. Too coarse grinding
product will lead to insufficient liberation which limits recovery. Where too finely ground ore will lead to
increasing cost of grinding to that extent that an increasing recovery is not enough to validate the cost. Over
Figure 7 3 stage crushing closed circuit, adapted from Metso, 2015. Figure 4 Three stage crushing closed circuit, adapted from Metso, 2015.
Screens
Screens
7
grinding ore can even lead to lower recovery in downstream processes, e.g. the particles becoming too small
for efficient flotation. There is a large difference in reduction ratios between grinding mills and crushers.
According to Metso (2015), the reduction ratio is dependent on what type of grinding mill, i.e. for semi-
autogenous (SAG)/autogenous grinding (AG) mills, the reduction ratio can be up to 5000 whereas for a
standard stirred mill a possible reduction ratio is 100. The difference between the grinding types can be
quite large. It can vary with what type of grinding media is used and material movement in the mill, what
particle size is produced and what type of breakage mechanism is the most dominant factor in size reduction.
A relatively new method for comminution that is both used in fine crushing and coarse grinding is the high
pressure grinding roll (HPGR). The HPGR is a dry comminution machine that utilizes two rotating rolls
that create a compressive force on the material (Figure 5). There is some evidence that the HPGR creates
micro-fractures that enable more efficient grinding and leaching (A. Ghorbani et al., 2013; B. Ghorbani et
al., 2013; Wills and Finch, 2015).
In Figure 6, Metso, (2015) has illustrated a wide array of different grinding mills in accordance with the
capacity of the feed size, throughput and what the standard product size is. It also illustrates the reduction
ratios for the different mills if the same definition is used as for the reduction ratio for crushing. SAG/AG
mills have a large working range whereas the stirred mill have a much smaller working range. Grinding
mills are used in a similar way as crushers in a circuit, e.g. an AG mill in circuit with a ball mill to produce
the desired product size. The following subsection contains a description of the different types of mill
movement and their respective grinding mills.
Figure 5 Schematic over HPGR, reduction ratio of 25, adapted from Metso 2015.
8
2.1.2.1 Tumbling mill
Tumbling is one of the main grinding motions, examples of tumbling mills are: semi-autogenous
grinding/autogenous grinding (SAG/AG), rod and ball mills. The main difference between the mentioned
mills is the grinding media, rod mills and ball mills use steel ball or steel rods as grinding media. The ball
charge is larger in a ball mill, roughly 30 % than what is used in a SAG mill (4-15% of the mill volume).
Autogenous grinding is when the grinding media is composed of the ore itself. According to Wei and Craig,
(2009) ball mills are used in almost 50 % of mineral processing plants and SAG/AG mills are used less (ca
40 %) and rod mills are used least. In a tumbling mill, the mill diameter will determine the drop height
which in turn will determine the impact and what feed size the mill can handle, whereas the length of the
mill will determine the throughput (capacity of the grinding mill). With an increasing height, the mill
capacity to produce impact breakage will increase, the main breakage mechanisms in tumbling mills are:
impact, compression and abrasion. The tumbling mill motion can be seen in Figure 7. It is a rotational
movement where the material is located on the edge of the mill and lifted along the sides of the mill with
the help of liners. In Figure 7, the different zones of a tumbling mill are presented and the goal is to have a
rotational speed of the mill so that the impact zone is on the toe and not on the mill liners, which would
increase the wear of the liner and in the end the steel consumption (if steel media is used). If the tumbling
mill reaches the critical speed, centrifuging occurs and a tumbling action no longer takes. Then a cascading
and cataracting motion does not take place and particle breakage ceases. The cascading motion will lead to
Figure 6 Different types of grinding mills, from Metso 2015.
9
a finer product size through an increase in abrasion and if the speed increases so that a cataracting motion
is obtained the particles will be subjected to impact breakage. The impact breakage will lead to a coarser
size fraction of the product with less wear on the liner than only a cascading motion. Most tumbling mills
is run at between 50-90 % of the critical speed, after 40-50 % of the critical speed there is only a small
increase in efficiency of the milling (Wills and Finch, 2015; Wu and Wang, 2014). The different liners will
be discussed further down in a separate subsection.
2.1.2.2 Stirred Mills
The aim of stirred mills is to produce fine particles. They were developed in the early 20th century to produce
fine particles for pigment. Stirred mills are used in fine to ultrafine grinding applications, from
pharmaceuticals to mining. The main differences within stirred mills are the operating stirrer speed, i.e. low
stirrer speed where gravity have an influence, or high stirrer speed with fluidized pulp. In Table 2 examples
of the different mill types are presented.
Figure 7 Tumbling mill movement with descriptive text and example of a tumbling mill. Adapted from Will & Finch
2015 and Metso, 2015.
10
Table 2 Examples of low and high speed stirrer mills. Adapted from Wang & Forssberg, 2003; Wills & Finch, 2015.
Low speed
High speed
Metso Vertimill
Metso Stirred Media Detritor (SMD)
Eirich Towermill
Xstrata IsaMill
Outotec High Intensity Grinding Mill (HIG)
The main difference between stirred and tumbling mills is how the energy is transferred to the material
grinded. As mentioned before, the breakage mechanism for tumbling mills is impact and abrasion whereas
in stirred mills the main breakage mechanism is abrasion. The mill shell for a stirred mill is stationary unlike
the rotation shell of a tumbling mill. In the production of fine to superfine particles (less than 100 µm) the
stirred mill is more efficient than the ordinary tumbling mill due to abrasion being more efficient in the
production of small particle sizes (Jeswiet and Szekeres, 2016; Wills and Finch, 2015) Figure 8 illustrates
standard stirred mill movement. The mill is vertical and stirrer rotates the material within the mill. Stirred
mills are usually found in the regrinding of material after it has first been ground in an AG mill, for example,
to decrease the size further for downstream processes or in the regrinding of rougher flotation concentrates
to increase liberation prior to the cleaning step.
11
2.1.2.3 Mill Liners
The inside of a mill drum consists of a renewable layer, liners which is a wear and impact resistant surface.
The liners help with the transportation of the material up towards the shoulder in a tumbling mill. What
type of liners used in the mill will have an effect on what type of breakage mechanism will affect the
grinding media. In Figure 9, several different types of liners are illustrated. A smooth liner will increase the
abrasion of the particles and, therefore, result in a finer grinding size but the increase in abrasion will also
result in higher wear on the liners. Where a liner that lifts the material with the rotation of the mill, i.e. rib
liner will result in more of an impact breakage due to a higher amount of material cataracting. Liners can
be made of hard metal alloys (e.g. Ni-alloy), rubber or white cast iron. There are several efforts being made
to increase the mill liner life to help decrease the time of shutdown for re-lining the mill drum. The trend
according to Wills & Finch 2015 is to choose liners that decrease downtime through faster re-lining time
and best service life. By having shorter and fewer shutdowns for re-lining, the cost associated with
downtime would be reduced. Rubber liners have some advantages in that they are longer lasting, easier and
faster to install (decreases the downtime therefore the cost associated to re-lining) and also a significant
reduction in noise level associated with grinding. However, rubber liners are thicker and thus will decrease
the throughput and are not suitable in primary grinding where the wear on the liners will be too high due to
higher grinding forces (Wills and Finch, 2015). A special solution in protecting the mill liners is the concept
of magnetic liners where the magnetic material within the grinding material creates a protective barrier on
liners that replenishes continuously. Magnetic liners have, for smaller ball mills, an increasing grinding
efficiency, longer liner life and a more consistent performance (Wu and Wang, 2014).
Figure 8) Stirred mills. Adapted from Metso 2015.
Stirring
12
2.1.2.4 Grinding Circuits
In the mineral industry grinding is usually a wet practice, where the material is a slurry. However, there is
some applications for dry grinding where water supply is of an issue. Dry grinding leads to a large
production of dust and is one of the biggest issues. According to Will and Finch (2015), wet grinding has
several advantages: lower energy per tonne, higher throughput, wet screening and size control and
elimination of dust. Grinding circuits are divided in the same manner as for crusher circuits, open or closed.
The first two mills in a circuit are referred to as the primary and secondary mill. Circuits can be defined in
terms of primary circuit to differentiate between potential re-grinding after separation processes. In a closed
circuit, the effort is being made to remove the material as soon as it reaches the desired particle size, whereas
in an open circuit, the goal is to decrease the size in one pass. This approach leads to a narrower particle
size distribution and, in the end, reduces overgrinding of valuable minerals. With the removal of particles
of the correct size for downstream processes, the particle size within the mill is shifted towards coarser
fractions. This allocates more of the grinding energy into reducing the size of the coarser particles, a more
energy efficient grinding is obtained. Figure 10 illustrates the effects of a closed circuit, that a higher
throughput with the same power draw as the open circuit can be achieved
Figure 9 Different types of mill liners, adapted from Wills & Finch 2015
13
The two most common circuits configurations according to Wei and Craig, (2009) are (i) one-stage closed
circuit and (ii) two stage, open then closed milling circuit. In Figure 11, two different SAG/AG circuits can
be observed. In both circuits, the primary mill is in an open circuit and the secondary mill is closed with a
classifier. The differences between the two circuits presented in the figure is that in circuit-A all material
before secondary grinding is classified. This is done to decrease the overgrinding of certain minerals and to
decrease the total energy requirement for grinding. The set-up in circuit-B is that a classifier have been
placed after secondary grinding, this result in more material than needed is milled for a second time.
Figure 10 Increasing grinding capacity through applying a closed circuit. Green circle,
circles the more energy efficient solution. Adapted from Wills & Finch 2015.
14
2.1.3 Energy consumption
Mining operations is one of the most energy intensive industrial processes. The comminution step of a
mining operation can consume from 30-70% of energy used on a mine site and is estimated to consume
between 2-4 % of the global energy consumption (Jeswiet and Szekeres, 2016; Napier-Munn, 2015; Wills
and Napier-Munn, 2006). In 2025, the total energy requirement for copper mining is predicted to amount
to 41.1 terawatt (TWh), which is a 95.5% increase from the reported values in 2013, where new plants are
estimated to consume 36.2% of the total copper mining energy consumption (Jeswiet and Szekeres, 2016).
Energy consumption and the cost per tonne of ore comminuted increases from blasting to grinding (Table
3), therefore, grinding is the operation in the comminution process that has the highest cost. Cleary and
Morrison (2016) defined wasted energy as the total energy from the collision energy that is less than the
elastic threshold, the wasted energy which does not contribute to any particle damage were estimated to be
Figure 11 Two different AG/SAG mill circuit, adapted from Gupta & Yan, 2006.
B
A Classifier
Classifier
15
above 30%. The wasted energy is dissipated as thermal energy and the high energy wastage is one reason
that the standard tumbling mill is inefficient. Less than 3% of the total energy used in the milling
configuration used in Cleary and Morrison (2016) is consumed in rock breakage. This number correlates
well to the presented energy number in Napier-Munn (2015) that 1-3% of the total energy consumed
produces new particle surfaces. The operation cost reported in Abbey et al., (2015) correlates well for
grinding with the energy requirement observed in Hukki (1961) (Figure 12).
Table 3 Operation cost in comminution (Abbey et al., 2015).
Operation Energy (kWh/t)
Blasting 0,43
Crushing 3,23
Grinding 10,0
Energy requirement increases with decreasing particle size, which can be observed in Figure 12. Hukki,
(1961) described the energy curve as a hyperbolic function and the well know theories of Rittinger, Kick
and Bond is represented as tangents to the curve. The red circle illustrates in what energy and size range
the Boliden grindability test works within.
Figure 12 An imaginary example of the different reduction characteristics plotted on a logarithmic paper, adapted from Hukki,
1961.
16
Energy consumption is closely linked to ore grade and production. Decreasing ore grades and increasing
demand for metals energy consumption has increased by 46 % during the last decade whereas the
production rate has increased by 30 % (Calvo et al., 2016). In the same study time, ore grade for selected
mines have decreased by 25 %. The increase in the total amount of material processed with lower ore grades
could be an explanation to the difference between energy consumption and tonnes copper produced (Figure
13). The high energy requirement for comminution has influenced sustainability work to focus on how to
decrease the energy requirement. An example is the establishment of Coalition for Eco-efficient
Comminution (CEEC, (Coalition for Eco-efficient Comminution, n.d.)). According to Wills & Finch
(2015), one approach to decrease the energy consumption is to ask if all ore needs to reach a fine grain size
where the majority of the energy is consumed.
2.1.4 Grindability
Grindability is defined as the materials ability to be made smaller by grinding according to Wills and
Napier-Munn, (2006), whereas grindability is defined by Neikov et al., (2009), as the characterization of a
grinding techniques’ efficiency. The first definition focusing on the material ability to decrease in size
regardless of what technique is used, whereas the latter utilizing the techniques’ ability to make a material
smaller. In this study, the definition of grindability by Wills & Napier-Munn (2006) is used. Both books
emphasize the importance of a set method when measuring the grindability. The most commonly used test
for grindability is Bond’s grindability test. Another well-known grindability test is the SAG mill
Figure 13 Total amount of energy and amount Cu produced over time. The orange arrow indicates that
the total amount of energy have diverged from the total tonnes Cu produces, adapted from Calvo et al.,
2016.
17
comminution test (SMC test) which is well described in Morrell, (2006). Bond’s grindability test is quite
time consuming and a lot of effort has been made to produce a faster test to measure a materials grindability.
Several different adaptations on the standard grindability test have been developed (Armstrong, 1986;
Heiskari et al., 2019; Mwanga et al., 2015). Mwanga et al., (2015) took it one step further trying to test and
define what is required for a grindability test to pass in a geometallurgical aspect. He defined eight different
requirements for a comminution test to pass as a geometallurgical characterization tool, as listed below:
1. The test should be relatively simple and use instruments available in common analytical and mineral
processing laboratories.
2. The test should be repeatable and not dependent on person.
3. The test should be easy to execute so that technicians with basic skills in sample preparation should be
able to do it with short training.
4. The test should be fast (max 1 hour) and inexpensive.
5. The amount of sample per test should be less than 0.5 kg; preferentially the test could use assay rejects.
6. The test, or rather a combination of tests, should give measured values on both crushability and
grindability.
7. It should be possible to use the parameters derived from the test directly in the modeling and simulation
of a comminution circuit.
8. It should be easy to extend the test to include mineral liberation information.
2.2 Mineral Characterization
The aim of mineral characterization is to separate different minerals apart from other minerals. The most
common way of separating different minerals from each other is by different physical properties such as
density, luster, hardness, colure, cleavage, crystal form and fracture. There are other properties used to
differentiate between minerals such as abrasion, grain size and shape and chemical properties (Kulu et al.,
2009; Díaz et al., 2018). Some of the physical properties can easily be characterized with the naked eye or
by using tests as the Mohs hardness scale, whereas others require more advanced tests and tools. The need
for a detailed characterization of mineral’s properties have developed the technology used into advanced
instruments such as SEM and x-ray diffraction (XRD). The goal is to understand the mineral characteristics
and how these might influence the different processes that the ore is passing through from mine to mill and
is more thoroughly presented in Baum, (2014); Dehaine et al., (2021); Dominy et al., (2018); Lishchuk et
al., (2020); Lotter et al., (2018); Michaux and O’Connor, (2020).
18
3 Experimental Work
3.1 Sample Selection
Drill-core samples from four different ore domains were collected on September, 2019 at the Garpenberg
mine site in Bergslagen, Sweden and sent to the Boliden pilot plant, located in Boliden, Sweden. Four ore
domains were selected from the ten different ore domains in the mine based on the scope of the master
study, which is part of an ongoing PhD study (B. Tiu et al., 2021) (Table 4). The six other domains were
not used in this study due to insufficient amount of drill cores for a grindability study, budget, and the
amount of time for an MSc study.
Table 4 Different ore domains from Garpenberg, Sweden, used in this study.
Massive Ores Domain name Ore domain classification
RPSG Remobilized pyrite-sphalerite-galena Remobilized pyrite-rich sulphide ore
MN-MSPG Manganese-rich massive sphalerite-pyrite-
galena
Massive sulphide ore hosted by Mn-rich
rocks
Disseminated/Vein
Type Ore Domain name Ore domain classification
BCA Biotite chalcopyrite gold Biotite schist-hosted ore (Cu-Au bearing)
QDM Quartz diopside magnetite Quartz-Diopside-Magnetite ore
3.2 Sample Preparation
3.2.1 Grindability test
Standard Boliden protocol for the sample preparation for a grindability study was conducted at the pilot
plant in Boliden, Sweden. The sample preparation protocol consists of the following three steps.
1. Material selection.
- Primary wear: to reduce the initial wear in grindability test and make the material
resemble a mill charge.
- Fine crushing: to produce grinding sand.
2. Drying and sieving
- Sieving and splitting the material into three different sizes fractions.
19
3. Preparation of grindability charge.
- 4kg in total from step one, plus fine crushed material and water.
For the material selection step, the drill-cores were blended into a homogenous pile within their respective
ore domain before being split into two different batches. The blending process consisted of collecting all
materials of an individual domain into one large bucket, blending the material back and forth between three
different buckets with a minimum of 50 back and forth blends before being splitted into two different
batches. The first batch was crushed in a regular laboratory jaw crusher with a spacing of one to three cm
open side setting in between the jaws. The material was then used to construct a primary wear charge. An
autogenous primary grind was performed in a Sala mill (Figure 14) with a constant grind time, 10 hours,
for each ore domain. In step 2, the material was dried and sieved into different fractions1 before the last step
(step 3), where the grindability charge was prepared with a total weight of 4kg consisting of primary wear
material plus finely crushed material, and water. The material in batch 2 was finely crushed in the same
laboratory jaw crusher with a spacing of 1mm (closed setting) before sieving with a 3mm sieve. The
oversized material was repeatedly crushed until all the crushed material passed through the 3mm sieve. The
finely crushed material was then riffled down into smaller portions for later use as grinding sand.
3.2.2 Analytical methods
The sample preparation for the analytical methods consist of different steps and at different times after the
grindability test was conducted. Below is the chronological list of the different sample preparation steps for
analytical methods.
X-ray fluorescence (XRF).
- Riffled grinding material and placed into sample holders for XRF.
Optical microscopy.
- Polished epoxy samples containing grinding material.
Automated SEM
- Carbon coating of the polished epoxy samples.
XRF and optical microscopy preparation were done at the Boliden pilot plant. The epoxy samples were
prepared by a Boliden technician. Carbon coating was done at the Lulea University of Technology.
20
3.3 Equipment
In step 1 that introduced in section 3.2.1 Grindability test a laboratory jaw crusher was used to reduce the
sample size quickly both for primary charge material and to produce grinding sand. The grinding sand was
then sieved with a mechanical sieve until a passing size of 100 % less than 3 mm was reached. An
autogenous Sala mill was used to decrease the initial loss of mass from the drill-core by chipping away the
edges and through abrasion to create smoother edges. This is to create a more linear expression (see Section
4.1. Grindability results). When performing the grindability test, a laboratory mill (Figure 15) is used with
a selection of sieves and a laboratory density measurement machine is used (Figure 16). A list of the
different equipment and their characteristics used for this grindability study can be found in .
Table 5 Different equipment used for grindability study at Boliden Pilot Plant.
2 With regards to non-disclosure agreement (NDA), the size fraction used in the grindability test is confidential. 2 With regards to NDA the dimension on sieves used is confidential.
Equipment Dimensions Uses
Laboratory jaw crusher 1mm and ~3 cm (open setting) Crusher
Mechanical Sieve 3 mm Sieve
Sala mill 45x30 cm Large lab tumbling mill
Laboratory AG mill 24x19 cm Small lab tumbling mill
Sieves -2 Sieve
AccuPyc II 1340 - Density measurement device
Figure 14 (A) Inside of Sala mill. (B) Sala mill at Boliden pilot plant. (C) To the left,
mechanical sieve and to the right laboratory jaw crusher.
C
A
B
21
3.4.1 Experimental procedure for a grindability study
The method used in this study to determine the grindability for autogenous grinding adapted for drill-cores
was developed by Boliden Minerals AB. Figure 17 presents an overview of how the experimental procedure
is constructed. The step “cleaning” in Figure 17 pertains to the cleaning procedure for the larger size fraction
A
B
Figure 15 (A) Inside of laboratory AG mill drum. (B) Laboratory AG mill used in grindability test.
Figure 16 AccuPyc II 1340 used for density measurement.
22
from small materials so a collection of the grindability product can occur. The test is performed with a total
runtime of 300 minutes split between three different runs (60, 90 and 150 minutes). The grindability test
produces a ks-value that represents the weight %-loss in relation to the grinding time and a size distribution
graph (see Section 4.1). As part of the grindability test, a density measurement was conducted on the 90
minute run and is later used for energy consumption calculations. From the size distribution, material was
categorized into three different size fractions as listed below:
250-90 μm
90-45 μm
45-0 μm
From each size fraction representative samples were selected for chemical analysis using XRF and
mineralogical characterization by SEM.
Figure 17 Grindability test developed by Boliden Mineral AB, repeated for three different times (60, 90 and 150 min).
23
3.4.2 Blending test procedure
A blending test series of the three tests were conducted with the same grindability methodology as
mentioned above. The ore domains used were the ones that had the largest differences in ks-value (see
section 4.1 Grindability results) and resulted in the domains RPSG and QDM. The test was split into three
different grindability tests where the sample content varied between 20-80 %.The setup for the three
blending grindability tests can be seen in Table 6.
Table 6 Blending test series, blending content.
Domain Test 1 Test 2 Test 3
RPSG (wt. %) 20 50 80
QDM (wt. %) 80 50 20
3.5 Sample characterization
Different tools and techniques were used to characterize the samples. The subsequent sections contain a
more thorough presentation of the different tools and techniques used. Figure 18 is an overview of the
process of ore sample characterization. XRF, optical microscopy and automated SEM were used to obtain
the ore sample characterization.
Figure 18 Schematic overview of the ore characterisation process.
24
3.5.1 X-Ray Fluorescence analysis
XRF measurements were carried out with a Spectro Xepos energy dispersive XRF equipped with a
palladium-cobalt rod at the Boliden Pilot plant (Figure 19). The measurements were conducted with a
calibrated method for Boliden ore, consisting of four different measurements over 160 seconds. The
conditions for XRF measurements are presented in Table 7. The four different measurement Boliden uses
is to target specific elements within each measurement. With a higher energy (keV) heavier elements can
fluorescence and be quantified whereas the lower energy is used for quantification of the lighter elements.
The Boliden calibration is focusing on precise ore element (Pb, Cu, Zn, Au and Ag) measurements.
Therefore the lighter elements might be under or over reported, see Appendix A: Assay data from XRF
and automated mineralogy for sum of concentration.
Table 7 Conditions used for XRF measurements.
Run number Energy (keV) Voltage (kV) Current (mA)
1 E>19 60 0,66
2 6<E<19 45 0,887
3 3<E<6 22,4 1,777 (He-atmosphere)
4 E<3 22,4
1,777 (He-atmosphere,
without the detection filter
that is used in run number
3)
Figure 19 XRF at Boliden pilot plant with sample holders.
25
3.5.2 Optical microscopy
The optical microscopy used in this study was an Axioscope 5/7 equipped with Zeiss Core at the Luleå
University of Technology. From the size fractions in section 3.4.1 epoxy samples containing material from
each size fraction were produced by a Boliden technician (+90; +45; -45) (Figure 20). Optical microscopy
is used as a first mineral characterization tool aimed to get knowledge of each sample i.e. what type of
mineralogy could be expected in the SEM, mineral liberation and eliminating possible errors from sample
preparation.
3.5.3 Automated SEM
In this study, a Zeiss Sigma 300 VP equipped with two Bruker XFlash 6|60 for EDS, one 16 mm solid state
diode backscattered detector and Mineralogic software. The measurements were done at the QANTMIN
SEM laboratory for automated mineralogy at the Lulea University of Technology. Table 8 shows the
settings for the different SEM samples. The step size were increased with fraction size to speed up the
process, for domain RPSG size fraction 45-90 µm the step size had to be increased compared to the other
domains. Stop criteria, determines the number of particles analyzed had to be reduced for the largest size
fraction. The fields analyzed throughout the samples in Zeiss Mineralogic were randomly selected and
20 μm
Figure 20 (A) Polished epoxy mounts, (B) Optical microscope, (C) Tarnished chalcopyrite with an Ag2S film (50 times the
microscopic magnification).
C
B A
26
calculated to contain enough particles for the stop criteria to kick in. The number of fields varied between
the different fractions with 100 fields for the larger sizes and 50 for the smallest fraction (see Figure 21).
Table 8 SEM settings for the different ore domains.
Mineralogic and Mineralogic Explorer was used for reprocessing the data from SEM to decrease
unclassified particles to a range of 1-3 % for size fraction 250-90 µm and 90-45 µm, whereas for the smallest
size fraction 45-0 µm a range 3-8 % was acceptable. To decrease the % of unclassified particles within each
sample the mineral list used needed to be altered to fit the chemical properties of the unclassified
measurements. Several different elemental ratios were used to differentiate between different minerals,
where the mineralogical formula is close e.g. pyrite and pyrrhotite. To different between pyrite and
pyrrhotite a Fe/S ratio were used, where if the ratio is higher than 1.55 the grain is classified as a pyrrhotite
whereas <1.55 it were classified as a pyrite. The automated SEM analyzed the different particle grains
within each sample and provided liberation information, modal mineralogy and mineral associations. This
study focusing on modal mineralogy.
Domain Size
fraction
(μm)
Carbon
coated
(Y/N)
Mode of Analysis Step size & stop
criteria
Magnification
RPSG 0-45 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
RPSG 45-90 Y 25kV, 60 aperture,
HV
7 μm, 10000 particles 360
RPSG 90-250 Y 25kV, 60 aperture,
HV
14 μm, 2500 particles 360
MN-
MSPG
0-45 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
MN-
MSPG
45-90 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
MN-
MSPG
90-250 Y 25kV, 60 aperture,
HV
14 μm, 2500 particles 360
BCA 0-45 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
BCA 45-90 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
BCA 90-250 Y 25kV, 60 aperture,
HV
14 μm, 2500 particles 360
QDM -45 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
QDM 45-90 Y 25kV, 60 aperture,
HV
3,5 μm, 10000
particles
360
QDM 90-250 Y 25kV, 60 aperture,
HV
14 μm, 2500 particles 360
27
3.6 Equations
Energy equation and throughput calculations is under NDA, but is based on density, ks-value, the change
in grinding product and empirical constants. This will not be presented further. Equations 1-3 is used to
predict a reasonable result for ks-value, energy and throughput for the blending test. To calculate the
relative % difference between the actual result from the blending test and the results from equations 1-3,
equation 4 is used.
Figure 21 (A) SEM for automated mineralogy at QANTMIN LTU, (B) Randomized pattern of analyzed
fields, (C and D) Galena (bright/blue phase) enclosed in enstatite (dark grey/green phase) in domain
MN-MSPG, size fraction >90 µm.
A B
C D
28
𝑘𝑠𝑎 × 𝑠𝑎 + 𝑘𝑠𝑏 × 𝑠𝑏 = 𝑘𝑠𝑐 (1)
Where;
ksi is equal to the ks-value for the individual domain,
si is equal to the blending percent (0,2; 0,5; or 0,8),
ksc equal to estimated blending ks-value based on individual domain results.
𝐸𝑎 × 𝑠𝑎 + 𝐸𝑏 × 𝑠𝑏 = 𝐸𝑐 (2)
Where;
Ei is equal to the energy consumption for the individual domain,
si is equal to the blending percentage (0,2; 0,5; or 0,8),
Ec equal to the estimated blending energy consumption based on individual domain results.
𝑇𝑎 × 𝑠𝑎 + 𝑇𝑏 × 𝑠𝑏 = 𝑇𝑐 (3)
Where;
Ti is equal to the theoretical throughput for the individual domain,
si is equal to the blending percentage (0,2; 0,5; or 0,8),
Tc is equal to the estimated blending throughput based on individual domain results.
(𝑥 − 𝑥𝑟𝑒𝑓) 𝑥𝑟𝑒𝑓⁄ = 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 % 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 (4)
Where;
x is equal to the result from the blending test,
xref is equal to the individual results from equation (1-3).
29
4 Results and Discussion
4.1 Grindability results
In Figure 22, it can be observed that the pebble wear rate follows a linear equation dependent on the time
milled. There are significant differences in wear between the different ore domains, which was expected
due to the domains having differences in texture as well as in mineralogy. The primary wear time was held
constant throughout the different samples and was set after the RPSG domain, this could be an explanation
for the higher wear in the 60 minutes run for the three other ore domains. Whereas the pebble wear rate
represents the % of the weight loss at a certain time the grinding product represents the %of material below
45 µm at a certain time. The slope of the pebble wear rate for each domain reflects the ks-value mentioned
in the chapter experimental work. RPSG domain have the highest ks-value and is, therefore, easiest to mill.
In Table 9 density, energy requirement, ks-value and theoretical throughput for each domain is presented.
It should be noted that the grinding result below is purely theoretical. In the processing plant in Garpenberg,
none of these different mineral domains would be processed alone but as a mixture of several different
domains of mineralogical content and the throughput calculations does not give any indication of possible
critical size build up in the mill, critical size is when a fraction size within the grinding mill builds up that
have a negative effect on the milling. The different mineral content present in these four domains that have
been tested highlights that the different mineral compositions should have an effect on the grinding
characteristics. This can be observed by studying the differences in throughput with large differences
0
1
2
3
4
5
6
7
8
9
10
0 20 40 60 80 100 120 140 160
PEB
BLE
WEA
R, W
EIG
HT
-%
GRINDING TIME, MINUTES
RSPG 5,52
MN-MSPG 2,66
BCA 1,47
QDM 0,84
Figure 22 Pebble wear rate, four different ore domains.
ks-value
30
between the domains. The curve for the domain; RPSG and MN-MSPG reached a plateau within the test
time of 150 minutes around a P90 for <45 µm, whereas the domains of BCA and QDM observed a more
linear production of material <45 µm and reached a P80 of 45 µm at 150 minutes of grinding (see Figure
23). To reach a plateau for the two mentioned domains an increase of the milling time would be needed but
a P80 of 45 µm for both domains was reached.
It should be noted that the difference in expected energy consumption and throughput is high, nearly a
factor of three between the domains. There is also a density difference where the harder to mill domains
have a lower density, this could be one of many factors that could affect the large difference in expected
energy and throughput. The lower density could indicate a higher silicate/carbonate content whereas the
two domains that have a density above 3.7 g/cm3 have a higher sulfide mineral content (see section 4.2
Mineral characterization). Since Lappberget is a polymetallic sulfide orebody, several high-density
minerals are present, such as galena, iron sulfides and sphalerite that will affect the density of the domains.
Other mineralogical characteristics that could affect the energy and throughput are, e.g. grain size, porosity
and hardness.
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100 120 140 160
WEI
GH
T-%
<4
5 Μ
M
GRINDING TIME, MINUTES
RSPG MN-MSPG BCA QDM
.
Figure 23 Grinding product for different ore domains.
31
Table 9 Predicted energy consumption when grinding each domain, ks-value and theoretical throughput.
4.1.1 Blending ore types and the effect on energy consumption and pebble wear rate
Blending tests were conducted to observe how the energy consumption would be when blending the least
energy intensive ore domain (RPSG) with the domain that required the most energy (QDM). In Figure 24
it can be observed that the pebble wear rate was increased for the endpoint of the test (80/20 and 20/80)
compared with when the domain were tested as an individual domain. The highest increase in pebble wear
were observed in the 20/80 test, where an increase in pebble wear rate was 264 % higher against the
individual domain result. A higher pebble wear rate was observed in the 80/20 test (not as large as in the
20/80 test) with an increase of 9 %. A hypothesis is that the QDM domain that was quite energy intensive
to mill is acting as grinding media for the softer RPSG domain. Further studies need to be done in exploring
the increase in pebble wear rate, firstly a size by size analysis on the mineralogical composition of the
different tests. No automated mineralogy was done on the blending test due to time constraints, but a bulk
mineral composition was calculated from the bulk mineral composition of the RPSG and QDM domains.
The bulk mineralogy is presented in the mineral characterization section. In Figure 25 the grinding product
for the blending test is presented. In the 20/80 (RPSG/QDM) test the grinding product does not follow the
same trend as other experiments (i.e. the individual experiment presented in the same figure). This deviation
from the trend is unexpected and can be due to sampling handling errors. In Table 10 the energy requirement
for the blending test, ks-values, theoretical throughput and their original domain values is presented. Table
10 also shows the % between the estimated energy consumption of the blending test and a calculated
weighted energy from the original domain. The relationship between the observed and theoretical pebble
wear is also presented in Table 10. The hypothesis for the lower energy consumption in the blending test is
that the QDM domain acts as a grinding media and the increasing pebble wear rate in relationship to the
calculated value indicates that a higher rate of new particle surfaces have been achieved when blending
50 % soft material with 50 % hard material. The key point from the table below is that the relationship
between the actual energy and pebble wear to the calculated value from the domains used when blending
Domain Density
(g/cm3) Energy consumption (kWh/t)
ks-
value
Throughput
(t/h)
RPSG 3,7 9,8 5,52 624
MN-MSPG 3,9 15,0 2,66 406
BCA 2,9 22,2 1,47 275
QDM 2,9 26,5 0,84 230
32
is hard to predict and unknown parameters have influenced the increasing pebble wear and decreasing
energy consumption. The Boliden lab test cannot predict problems with critical size build up or whether or
not blending the samples improves or worsens this tendency
Figure 25 Grinding product blending test.
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100 120 140
WEI
GH
T %
<4
5Μ
M
GRINDING TIME
RPSG QDM Blend 50/50 RPSG/QDM
Blend 80/20 RPSG/QDM Blend 20/80 RPSG/QDM
0
1
2
3
4
5
6
7
8
9
10
0 20 40 60 80 100 120 140 160
PEB
BLE
WEA
R, W
EIG
HT
-%
GRINDING TIME, MINUTES
RPSG 5,52 QDM 0,83
Blend 50/50 RPSG/QDM 3,2 Blend 80/20 RPSG/QDM 6,04
Blend 20/80 RPSG/QDM 2,19
Figure 24 Pebble wear rate, blending test (ks-value).
33
Table 10 Estimated energy consumption, ks-value compared and theoretical throughput against calculated values for the blending
test.
4.2 Mineral characterization
In this section results from the automated mineralogy for each individual domain is presented. A bulk
mineral composition has been calculated through mass-balancing. The lowest total mineral composition %
measured is in the MN-MSPG domain, 91,3 %. The low % is related to the % of unclassified minerals and
the agglomeration of small grains. In this study, the main reason for low analysis % is agglomeration of
small grains and boundary phase issues. The calculated bulk mineral composition of the different blending
samples is presented, it is based on the original mineral composition of the domains present in the sample.
All measurements presented below is in wt. % and have been normalized to 100%. Amphibole/pyroxene is
represented in the classification tables as amph/pyro and sphalerite is represented as sph.
In summary, the softer sulfide minerals (sphalerite and galena) and carbonates seem to increase in grade
with a lower fraction size, whereas pyrite, which has a higher hardness value in the Mohs scale, is in the
larger size fraction. One possibility that could explain this is due to the harder Mohs value of pyrite, silicates
i.e. the less wear could be expected from a grindability test where abrasion is one of the main breakage
mechanism. The above statement agrees with the conclusion from Kulu et al. (2009) that the hardness of a
mineral is well correlated to the abrasivity and to the grindability. Another hypothesis for the accumulation
of harder minerals within the larger size fraction is that of larger grain size before grinding and not enough
stress were applied to fracture the particle into smaller pieces. The RPSG domain is a good example where
this hypothesis is applicable, where the main component of the >90 µm fraction is pyrite. This theory
correlates well to the grinding technique used, autogenous grinding. In the BCA domain, it can be observed
that the domain contains a significantly larger quantity of micas compared to the other domains and in
combination with the large quartz content. It will have an important effect on the ks-value due to the micas
laminar structure and high resistance to reducing in size through breakage mechanisms (Díaz et al., 2018).
Micas usually form elongated mineral grains, and it might therefore be hard to differentiate between the
Domain
Energy
consumption
(kWh/t)
Relative % difference ks-value Relative %
difference
Throughput
(t/h)
Relative %
difference
RPSG 9,77
5,52 624
QDM 26,52
0,84 230
20/80 (RPSG/QDM) 17,40 -22,0 2,19 23,3 333 7,8
50/50 (RPSG/QDM) 14,42 -29,8 3,21 0,9 423 -1,0
80/20 (RPSG/QDM) 9,93 -19,2 6,04 31,8 651 19,3
34
different fractions for this mineral. It can be elongated so that it passes through a smaller sieve than the
largest diameter is and this phenomenon might be the reason that the mica content is higher in the 45-90
fraction. Figure 26 illustrates the above mentioned problem with elongated mica grains, the (A) images is
an elongated phlogopite grain with mineral mapping whereas (B) images is a back scattered electron image
(BSE-image) of the same grain.
In Table 11 the result from the four different mineralogical domains is presented. The mineralogical result
from each fraction is presented with the wt% for each fraction and a mass balanced bulk result for each
domain. It should be noted that a possible sampling error have occurred in the domains QDM and MN-
MSPG. Mislabelling in the 45-90 µm fraction and is further discussed in more detail in section 4.6.
35
A
B
Figure 26 (A) A elongated phlogopite grain, in coulure (brown phase). (B) Same grain as in (A) but in
back scattered electron image (BSE-image).
42
Table 11 Automated mineralogy result for each domain, fraction size by fraction size and a calculated bulk mineralogy.
RPSG Pyrite Quartz Sph Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars wt%
>90 87,9 3,2 2,7 0,2 1,7 1,3 1,8 0,8 0,0 0,2 0,1 14
45-90 54,1 5,9 15,9 0,7 2,3 13,8 3,8 2,4 0,1 0,4 0,5 19
<45 20,5 4,6 40,0 2,8 1,4 23,5 3,9 2,5 0,1 0,2 0,6 68
Bulk 36,0 4,6 30,4 2,1 1,6 18,6 3,6 2,3 0,1 0,2 0,5 100
BCA Pyrite Quartz Sph Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars wt%
>90 12,9 72,5 3,5 1,3 7,2 0,1 0,5 0,3 0,2 0,8 0,7 35
45-90 5,5 72,5 8,3 2,5 9,4 0,2 0,2 0,5 0,0 0,4 0,6 16
<45 2,5 60,0 18,9 9,0 7,1 0,3 0,2 0,8 0,1 0,4 0,7 48
Bulk 6,7 66,5 11,7 5,2 7,5 0,2 0,3 0,6 0,1 0,5 0,7 100
MN-MSPG Pyrite Quartz Sph Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars wt%
>90 34,0 11,8 8,6 4,4 1,9 4,9 9,5 11,8 10,7 2,2 0,1 17
45-90 31,3 13,3 7,5 3,4 1,2 4,0 8,5 16,0 11,6 3,1 0,1 4
<45 9,1 8,0 25,6 16,9 0,4 10,2 6,4 15,6 6,5 1,2 0,0 79
Bulk 14,2 8,9 22,0 14,3 0,7 9,0 7,0 15,0 7,4 1,5 0,1 100
QDM Pyrite Quartz Sph Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars wt%
>90 18,1 62,4 0,2 0,2 0,1 1,1 16,1 0,6 0,2 1,0 0,2 37
45-90 14,2 62,2 0,7 0,3 0,1 1,3 18,5 1,5 0,1 1,1 0,1 17
<45 7,5 65,9 2,4 0,8 0,1 2,5 17,1 2,4 0,3 1,0 0,1 46
Bulk 12,5 64,0 1,3 0,5 0,1 1,8 16,9 1,6 0,2 1,0 0,1 100
43
4.2.1 Blending of two different mineralogical domains.
The blending mineral composition was calculated from the measured composition of the QDM and RPSG domain, presented in Table 11, and
introduced in Table 12. It can be observed that the bulk mineralogy for the different blending test goes from a high quartz, amph/pyro and low pyrite,
galena, mica and carbonates in the 20/80 blend with 20% soft (RPSG) and 80% hard domain (QDM). To the opposite mineralogy in the softer
blending test with an 80/20 blend with 80% RPSG domain and 20% QDM.
Table 12 Calculated mineral composition of blending domain QDM and RPSG
Pyrite Quartz Sph Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars
Bulk
20/80 17,21 52,11 7,11 0,80 0,41 5,13 14,27 1,73 0,18 0,85 0,21
Bulk
50/50 24,25 34,30 15,84 1,27 0,86 10,20 10,26 1,93 0,14 0,62 0,32
Bulk
80/20 31,30 16,50 24,56 1,75 1,31 15,27 6,25 2,14 0,10 0,39 0,44
44
4.3 Relationship between mineralogy, ks-value, and energy
In this section, the relationship between the mineralogy, ks-value and estimated energy consumption is discussed. The sample dataset has been
extended with the calculated bulk mineral composition. This increases the dataset by 75% but also increases the uncertainty of the dataset due to the
bulk mineral composition of the blended samples being calculated. As mentioned in chapter 3 Experimental Work, the ks-value represents the pebble
wear rate in the experiment over time, the mass % loss/time. In Figure 27 below, five different minerals were picked to demonstrate the trend from
plotting the bulk mineralogy against the ks-value. According to Table 13 and Figure 27 carbonates and pyrite have the best R2-value and indicate a
positive trend. The R2-value measure how much the variance in the data is explained by the regression model and varies between 0 and 1. Carbonates,
pyrite and sphalerite is linearly correlated between ks-value and wt% whereas the trend is inverted for quartz, which is inversely correlated between
ks-value and increasing wt. %. Quartz is, according to the Mohs hardness scale, a hard mineral (7 in Mohs hardness scale) and should be a mineral
less prone to abrasion. This might be one explanation to why the ks-value decreases with increasing grade of quartz. Galena has a high variance in
the dataset and therefore lack any clear linear trend. Galena’s´ lack of trend might be a results of outliers and the other minerals that is missing a
trend can be the result of outliers. The possibility of outliers is further discussed in
45
Appendix B. The number of data points in this study is too few to be conclusive. By increasing the number
of test of each domain or expanding the data set by more mineralogical domains would create a solid data
set. All R2-values and P-values for each mineral are presented in Table 13. When using Boliden’s own
energy estimation calculations and plotting the different minerals against the energy consumption, similar
trends and R2-values can be observed as in the mineral grade vs ks-value. In Table 13 the R2-values for the
different minerals are presented.
In the table below it can be observed that carbonates, pyrite, sphalerite, and quartz have R2-values that
indicate the trend fits a linear model. Linear regression was done for each mineral against ks-value, but also
the estimated energy consumption at 95% confidence. From the R2-values in Table 13 it can be observed
that the four mentioned minerals above are the only minerals that have a clear linear trend. The purpose of
linear regression is to determine if some minerals from the dataset have any significant effect on
grindability. If any mineral have a P-value below 0.05 the null hypothesis can be rejected and it can be
concluded that the observed trend in Figure 27 is with 95 % certainty not due to random chance. From the
P-values listed below in Table 13 it can be observed that the same minerals that have a high R2-value, have
a P-value below 0.05. It should be noted that the minerals with a low R2-value have a high P-value, it is not
possible to rule out the other minerals from having a significance on the ks-value and energy. The R2-value
might be due to the earlier discussed different mineralogical population within the data set. It is interesting
to consider what mineralogical attributes that could affect the outcomes of the results discussed above and
should be further studied.
47
Table 13 R2 values from plotting mineral grade against Ks-value for each mineral.
Pyrite Quartz Sphalerite Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars
R²-value ks 0,86 0,65 0,76 0,02 0,04 0,90 0,17 0,00 0,01 0,35 0,11
R²-value
Energy
0,73 0,78 0,81 0,01 0,06 0,87 0,17 0,03 0,01 0,16 0,04
P-value ks 0,00 0,03 0,01 0,79 0,69 0,00 0,36 0,71 0,66 0,36 0,33
P-value
Energy
0,01 0,01 0,01 0,88 0,61 0,00 0,36 0,93 0,38 0,64 0,27
48
y = 4,9842x + 4,6836R² = 0,8553
0
5
10
15
20
25
30
35
40
0 2 4 6 8
Pyr
ite
(wt%
)
ks (mass % loss/time)
y = -10,689x + 68,763R² = 0,6536
0
10
20
30
40
50
60
70
0 2 4 6 8
Qu
artz
(w
t%)
ks (mass % loss/time)
y = 4,5317x + 1,9394R² = 0,7594
0
5
10
15
20
25
30
35
0 2 4 6 8
Sph
aler
ite
(wt%
)
ks (mass % loss/time)
y = 3,2726x - 1,6509R² = 0,8986
0
2
4
6
8
10
12
14
16
18
20
0 2 4 6 8
Car
bo
nat
e (w
t%)
ks (mass % loss/time)
y = -0,307x + 4,6523R² = 0,0151
0
2
4
6
8
10
12
14
16
0 2 4 6 8
Gal
ena
(wt%
)
ks (mass % loss/time)
Figure 27 Mineral grade plotted against ks-value.
49
4.4 Back-calculations vs XRF data
Back-calculating the mineralogy results for each size fraction with assay data from the automated
mineralogy result and plotting the result against the XRF data for Pb, Zn and Fe gives an indication on the
validity of each analytical method. In Figure 28 the results are presented and it can be observed that the Zn
and Pb shows a good fit between the back-calculated mineralogy and the XRF-assay data. It should be
noted that the slope of the linear fitted model indicates that for Zn the assay data from the SEM is higher
than the Zn assays from the XRF with a relationship 1:1 (SEM:XRF). For Pb the slope indicates a 1.8:1
relationship between the assay from each method, this could be because the simple stoichiometry of the
galena whereas for sphalerite the stoichiometry is more complex with substitutions of Zn and Fe with Mn,
Cd and Hg (B. Tiu et al., 2021). No clear relationship for Fe could be found with the back-calculations. A
possible reason is that Fe is present in many different minerals. If the back calculated chemical data vs the
xrf data is fitted to a linear model the result indicate a relationship between XRF results and the back-
calculations is 1:2. In Appendix A: Assay data from XRF and automated mineralogy the different
assays from both the automated mineralogy and XRF measurements is presented and discussed.
4.5 Source of errors
In this subsection, different sources of error are listed for the different steps of the study and discussed in
more detail.
4.5.1 Sample preparation errors
In this subsection several different possible errors are discussed and how they might have influenced the
results of the study.
Homogeneity not reached in step one where the material within the domains were blended. This
could lead to an unrepresentative distribution of sample between the different fractions;
Figure 28 Back-calculation mineralogy vs. XRF assays.
50
Not enough primary wear, insufficient primary wear would lead to an unrealistic increase in initial
pebble wear rate. This in turn would affect the grindability result and in the end the estimated
energy, with a higher ks-value and lower energy consumption;
When constructing the grinding charge, introducing bias towards different minerals through
preferential sampling when selecting material. Reduced through randomly selecting material and
allowing a 10% variance in the mill charge. The 10 % variance will decrease the amount of picking
material out or adding material to the grindability charge. However, the variance might change the
grinding characteristics to a higher or lower wear; and
In the preparation of polished epoxy samples, the technician used a spoon to select the material
used. This could have lead to an unrepresentative sample where the majority of the material might
be from the top or bottom of the sample holder. That would affect the mineral composition of the
finished sample, where an increase of lighter or heavier minerals would have occurred and the size
distribution of the mineral grains will also be affected.
When comparing the criteria defined by Mwanga et al., (2016) to the method used for measuring
grindability in this study, the methods used do not pass as a geometallurgical test for grindability
because the test used is time consuming, material heavy, and introduces bias through preferential
sampling when compiling the grindability charge and therefore not reproducible between different
persons.
4.5.2 Analytical methods errors
Under this subsection different errors will be discussed that could have occurred during the analytical
method part of the study.
The grinding product used in the epoxy samples had a lot of fines below 45 µm. The high amount
of fines produces an error in the sample preparation of the epoxy samples. An agglomeration of
fines occurred. The agglomerated particles were classified as separate particles or similar
agglomeration around hard mineral grains. This lead to unclassified particles and boundary phase
problems;
Not enough material within a sample to make a good XRF analysis. This leads to over or
underestimations of elemental assays;
High uncertainty in light elemental analysis;
Not enough particles analyzed in the automated mineralogy, which would lead to unrepresentative
results; and
51
Mislabeling of the product have occurred between the 45-90 fraction of the QDM and MN-MSPG
domains. This resulted in a trend of the minerals to be unrealistic and not logical i.e. the quartz content
in the largest size fraction to be around 12 wt% in MN-MSPG domain and in the middle fraction to be
around 63 wt% and then down to 7 wt% in the smallest size fraction. The opposite occurred in the
QDM domain and a similar pattern could be observed for the other mineral presented. In Table 14
below the unaltered version of the QDM and MN-MSPG domain is presented, the fraction that have
been mislabeled is marked in yellow.
52
Table 14 Unaltered version of QDM and MN-MSPG normalized mineral composition.
MN-
MSPG
Pyrite Quartz Sphalerite Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars
>90 27,34 9,47 6,90 3,56 1,53 3,97 7,60 9,47 8,60 1,78 0,10
45-90 13,84 60,80 0,65 0,25 0,10 1,31 18,07 1,42 0,14 1,06 0,14
<45 7,35 6,52 20,77 13,71 0,35 8,22 5,15 12,67 5,22 1,00 0,03
Bulk 15,52 15,85 12,75 8,04 0,74 5,65 8,00 9,82 5,67 1,29 0,07
QDM Pyrite Quartz Sphalerite Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars
>90 17,45 60,26 0,19 0,23 0,11 1,03 15,53 0,56 0,17 0,93 0,16
45-90 30,35 12,91 7,27 3,32 1,21 3,86 8,23 15,47 11,29 2,98 0,08
<45 7,09 62,67 2,30 0,72 0,09 2,33 16,22 2,33 0,25 0,95 0,09
Bulk 14,84 53,31 2,37 0,98 0,29 2,11 14,61 3,92 2,10 1,29 0,12
53
5 Conclusions and recommendations
This study has answered the objectives noted in the introduction for the Lappberget mineralization at
Garpenberg mine:
1. Study a possible connection between mineralogy and grindability; and
2. Investigate a connection between ore blending and throughput.
The connection between mineralogy and grindability is best shown through throughput where each
individual domain has clear distinct throughput in combination with the four different minerals or mineral
group that have a clear influence on the pebble wear rate. Each of the three minerals and the mineral group
carbonates (pyrite, quartz, sphalerite and carbonate) trends have been validated with a linear regression
analysis and the null hypothesis have been rejected. With 95 % certainty the trend is not due to a random
pattern.
By blending soft and hard ore domains, an increase in throughput can be achieved but there is a problem
when predicting the effects that an ore blend can have on throughput from individual grindability tests. One
aspect that affects the throughput is the critical size build up, which the Boliden Lab test can not measure.
More research needs to be done to understand when the tendency for critical size build up is improving or
worsen, so a more precise model for predicting blending throughput can be achieved. Micas’ tend to
produce elongated mineral grains in autogenous grinding and would be of note to research micas’ influence
on milling. The laminar structure of micas’ is of interest and might be one of the reason for the elongation
of mineral grains. Another mineral that tend to produce elongated mineral grains is amphiboles where the
crystal structure is monoclinic. Further research should be to investigate the importance of different crystal
structures when milling.
It should be noted that the sample database in this study is not large enough to make any definite conclusion,
but the study indicates the importance of a better knowledge for how the mineralogy and blending will
affect the energy and throughput in comminution. A recommendation for future studies would be to increase
the dataset with different ore lenses at the Garpenberg mine to validate the trends observed in this study
and introduce several different blending experiments in combination with mineral characterization. If the
trends are validated, a multivariate analysis (PCA, principal component analysis) could be used to obtain a
general equation for the influence of mineralogy on grindability in Garpenberg mine. A summary of the
recommendations for future research is presented in the bulleting below:
54
Continue to investigate the connection between mineralogy and grindability by expanding the data
set, i.e. include new spatial constrained domains and expand into new ore bodies.
Investigate the critical size build up in the Boliden grindability test.
Investigate into a prediction model between mineralogy and grindability.
Investigate and improve the prediction of the grindability result and energy consumption when
blending.
Investigate Micas’ behaviour in AG.
Investigate how different crystal structures behave during milling.
Investigate on how to optimize mill performance with better understanding on spatial constrained
mineralogy and mill performance.
55
References
A. Ghorbani, Y., Petersen, J., Becker, M., Mainza, A.N., Franzidis, J.P., 2013. Investigation and modelling
of the progression of zinc leaching from large sphalerite ore particles. Hydrometallurgy 131–132, 8–
23. https://doi.org/10.1016/j.hydromet.2012.10.004
A. Tiu, G., Jansson, N., Wanhainen, C., Ghorbani, Y., 2019. Sulfide chemistry and trace element
deportment at the metamorphosed Lappberget Zn-Pb-Ag-(Cu-Au) ore body, Sweden : Implications
for Mineral Processing. 15th SGA Bienn. Meet. 27-30 August 2019, Glas. Scotl. 4, 1486–1489.
Abbey, C.E., Bansah, K.J., Galecki, G., Boateng, K.A., 2015. Rock Characteristics and Ball Mill Energy
Requirements at Goldfields Ghana Limited, Tarkwa Gold Mine (TGM). Ghana Min. J. 15, 50–57.
Armstrong, D.G., 1986. An alternative grindability test. An improvement of the Bond procedure. Int. J.
Miner. Process. 16, 195–208. https://doi.org/10.1016/0301-7516(86)90031-1
B. Ghorbani, Y., Becker, M., Petersen, J., Mainza, A.N., Franzidis, J.P., 2013. Investigation of the effect of
mineralogy as rate-limiting factors in large particle leaching. Miner. Eng. 52, 38–51.
https://doi.org/10.1016/j.mineng.2013.03.006
B. Tiu, G., Jansson, N., Wanhainen, C., Ghorbani, Y., Lilja, L., 2021. Ore mineralogy and trace element
(re)distribution at the metamorphosed Lappberget Zn-Pb-Ag-(Cu-Au) deposit, Garpenberg, Sweden.
Ore Geol. Rev. 135, 104223. https://doi.org/10.1016/j.oregeorev.2021.104223
Ballantyne, G.R., Powell, M.S., 2014. Benchmarking comminution energy consumption for the processing
of copper and gold ores. Miner. Eng. 65, 109–114. https://doi.org/10.1016/j.mineng.2014.05.017
Baum, W., 2014. Ore characterization, process mineralogy and lab automation a roadmap for future mining.
Miner. Eng. 60, 69–73. https://doi.org/10.1016/j.mineng.2013.11.008
Bindler, R., Karlsson, J., Rydberg, J., Karlsson, B., Berg Nilsson, L., Biester, H., Segerström, U., 2017.
Copper-ore mining in Sweden since the pre-Roman Iron Age: lake-sediment evidence of human
activities at the Garpenberg ore field since 375 BCE. J. Archaeol. Sci. Reports 12, 99–108.
https://doi.org/10.1016/j.jasrep.2017.01.019
Bis, K., Leveinen, J., Ngan-Tillard, Dominique Lottermoser, B., Kangas, L.S., 2018. Geometallurgical
characterization of the Kittilä gold ore deposit. Master thesis.
Bradshaw, D., 2014. The role of “process mineralogy” in improving the process performance of complex
sulphide ores, in: IMPC 2014 - 27th International Mineral Processing Congress.
C. Tiu, G., Ghorbani, Y., Jansson, N., Wanhainen, C., 2021. Tracking silver in the Lappberget Zn-Pb-Ag-
(Cu-Au) deposit, Garpenberg mine, Sweden: Towards a geometallurgical approach. Miner. Eng. 167,
106889. https://doi.org/10.1016/j.mineng.2021.106889
Calvo, G., Mudd, G., Valero, Alicia, Valero, Antonio, 2016. Decreasing ore grades in global metallic
mining: A theoretical issue or a global reality? Resources 5, 1–14.
https://doi.org/10.3390/resources5040036
Cleary, P.W., Morrison, R.D., 2016. Comminution mechanisms, particle shape evolution and collision
energy partitioning in tumbling mills. Miner. Eng. 86, 75–95.
https://doi.org/10.1016/j.mineng.2015.12.006
Dehaine, Q., Tijsseling, L.T., Glass, H.J., Törmänen, T., Butcher, A.R., 2021. Geometallurgy of cobalt ores:
A review. Miner. Eng. 160. https://doi.org/10.1016/j.mineng.2020.106656
56
Díaz, E., Voisin, L., Kracht, W., Montenegro, V., 2018. Using advanced mineral characterisation
techniques to estimate grinding media consumption at laboratory scale. Miner. Eng. 121, 180–188.
https://doi.org/10.1016/j.mineng.2018.03.015
Dominy, S.C., O’connor, L., Parbhakar-Fox, A., Glass, H.J., Purevgerel, S., 2018. Geometallurgy—A route
to more resilient mine operations. Minerals 8, 1–33. https://doi.org/10.3390/min8120560
Gupta, A., Yan, D., 2006. Mineral Processing Design and Introduction to Mineral Processing Design and
Operation 0–444.
Heiskari, H., Kurki, P., Luukkanen, S., Sinche Gonzalez, M., Lehto, H., Liipo, J., 2019. Development of a
comminution test method for small ore samples. Miner. Eng. 130, 5–11.
https://doi.org/10.1016/j.mineng.2018.10.005
Hukki, R.T., 1961. Proposal for a solomnic settlement between the theories of von Rittinger, Kick and
Bond. Trans. SME/AIME 220, 403–408.
Jansson, N.F., Allen, R.L., Allen, R.L., 2011. Timing of volcanism, hydrothermal alteration and ore
formation at Garpenberg, Bergslagen, Sweden. Gff 133, 3–18.
https://doi.org/10.1080/11035897.2010.547597
Jeswiet, J., Szekeres, A., 2016. Energy Consumption in Mining Comminution. Procedia CIRP 48, 140–
145. https://doi.org/10.1016/j.procir.2016.03.250
Kulu, P., Tarbe, R., Käerdi, H., Goljandin, D., 2009. Abrasivity and grindability study of mineral ores.
Wear 267, 1832–1837. https://doi.org/10.1016/j.wear.2009.02.025
Lishchuk, V., Koch, P.H., Ghorbani, Y., Butcher, A.R., 2020. Towards integrated geometallurgical
approach: Critical review of current practices and future trends. Miner. Eng.
https://doi.org/10.1016/j.mineng.2019.106072
Lotter, N.O., Baum, W., Reeves, S., Arrué, C., Bradshaw, D.J., 2018. The business value of best practice
process mineralogy. Miner. Eng. https://doi.org/10.1016/j.mineng.2017.05.008
Metso, 2015. Basics in minerals processing 354.
Michaux, S., O’Connor, L., 2020. How to Set Up and Develop a Geometallurgical Program.
Morrell, S., 2006. Design of AG/SAG Mill circuits using the SMC test. Sag.
Mwanga, A., Lamberg, P., Rosenkranz, J., 2016. Development of a Geometallurgical Testing Framework
for ore Grinding and Liberation Properties. Luleå Univ. Technol.
Mwanga, A., Lamberg, P., Rosenkranz, J., 2015. Comminution test method using small drill core samples.
Miner. Eng. 72, 129–139. https://doi.org/10.1016/j.mineng.2014.12.009
Napier-Munn, T., 2015. Is progress in energy-efficient comminution doomed? Miner. Eng. 73, 1–6.
https://doi.org/10.1016/j.mineng.2014.06.009
Neikov, O.D., Naboychenko, S.S., Murashova, I., Gopienko, V., Frishberg, I., Lotsko, D., Dowson, G.,
Williams, B., Tokhtuyev, V., Nelkov, I., Latysheva, L., Vasilieva, G., 2009. Handbook of non-ferrous
metal powders, First. ed.
Wei, D., Craig, I.K., 2009. Grinding mill circuits - A survey of control and economic concerns. Int. J. Miner.
Process. 90, 56–66. https://doi.org/10.1016/j.minpro.2008.10.009
Wills, B.A., Finch, J.A., 2015. Wills’ Mineral Processing Technology An Introduction to the Practical
57
Aspects of Ore Treatment and Mineral Recovery, 8th ed. https://doi.org/10.1115/DSCC2013-3715
Wills, B.A., Napier-Munn, T.J., 2006. Wills’ mineral processing technology: An introduction to the
practical aspects of ore treatment and mineral recovery, 7th editio. ed, Wills’ Mineral Processing
Technology: An Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery.
Butterworth-Heinemann.
Wu, M., Wang, V., 2014. Modeling ball impact on the wet mill liners and its application in predicting mill
magnetic liner performance. Miner. Eng. 61, 126–132. https://doi.org/10.1016/j.mineng.2013.09.019
58
Appendices In this thesis there will be two different appendixes (A and B) that will presented in this chapter.
Appendix A: Assay data from XRF and automated mineralogy In Appendix A the different assays used to back calculate and plot Fe, Zn and Pb against the XRF assays.
For each domain and fraction the average Fe, Zn and Pb assays for each of the 11 selected minerals were
exported from Mineralogic Explorer is presented in Table 15 and in Table 16.
Table 17 presents the XRF assays for each domain and fraction. It should be noted that the sum of
concentration for the XRF values have a high variance and that is due to the calibrations for the XRF
measurements. The XRF data have not been normalized when comparing to the assay data from automated
mineralogy. This as mentioned in section 3.5.1, the calibration used for XRF measurements at Boliden is
for a precise measurement for the different ore minerals (Pb, Zn, Cu, Au and Ag). If the assays are
normalized the results for the different ore elements will be under or over estimated.
Table 15 Zn and Pb assays for Sphalerite and Galena, from automated mineralogy.
Domain Size fraction Sphalerite Galena
RPSG
>90 58,6 89,4
45-90 58,6 94,3
<45 59,1 92,3
BCA
>90 57,4 96,4
45-90 57,4 93,0
<45 56,9 93,7
MN-
MSPG
>90 59,9 96,2
45-90 57,2 93,8
<45 60,2 97,2
QDM
>90 59,3 95,4
45-90 60,0 94,2
<45 58,9 97,3
59
Table 16 Fe-assays from automated mineralogy.
Domain Size fraction Pyrite Quartz Sphalerite Galena Mica Carbonates Amph/pyro Pyrrhotite FeOx Garnets Feldspars
RPSG
>90 46,8 0,0 8,2 0,0 1,5 0,0 1,0 58,3 74,5 2,8 0,2
45-90 53,2 0,1 8,2 0,0 1,7 0,0 1,0 60,4 74,7 3,9 0,1
<45 53,0 0,2 8,3 0,0 2,4 0,2 1,3 61,0 81,2 3,7 0,1
BCA
>90 46,8 0,0 7,4 0,0 2,0 0,5 6,2 60,4 72,4 0,8 0,1
45-90 46,8 0,0 7,3 0,0 2,7 0,3 5,8 61,0 75,5 2,1 0,5
<45 46,8 0,0 7,0 0,0 3,5 0,2 5,2 61,2 73,5 5,7 0,5
MN-MSPG
>90 53,2 0,0 8,6 0,0 4,8 0,8 6,9 61,8 76,2 7,9 0,5
45-90 53,1 0,1 8,7 0,0 10,9 3,3 5,7 61,3 76,6 6,2 0,0
<45 46,9 0,2 9,7 0,0 6,4 1,2 8,3 61,5 76,9 8,3 1,0
QDM
>90 46,9 0,0 7,3 0,0 17,1 4,0 5,7 60,6 69,6 6,5 0,0
45-90 46,9 0,0 8,9 0,0 4,9 1,0 7,2 61,6 76,2 7,6 0,3
<45 46,9 0,1 7,9 0,0 14,5 1,5 6,3 61,2 77,5 6,8 0,0
60
Table 17 XRF assays for each domain and fraction.
Na₂O MgO Al₂O₃ SiO₂ P S Cl K₂O CaO Ti Mn Fe Cu Zn As Sb Ba Pb Sum of Conc.
Domain Size fraction % % % % % % % % % % % % % % % % % % %
RPSG
>90 0,0 10,3 11,2 36,4 0,0 23,0 0,1 4,5 3,8 0,1 0,4 21,5 0,0 2,6 0,1 0,0 0,0 0,2 114,2
90-45 0,0 5,4 2,3 14,3 0,0 18,2 0,0 0,9 9,7 0,0 0,6 25,8 0,1 9,8 0,0 0,0 0,0 0,4 87,5
<45 0,0 5,2 0,7 7,3 0,0 23,7 0,0 0,3 11,2 0,0 1,0 30,6 0,0 24,3 0,0 0,1 0,0 2,1 106,7
BCA
>90 0,0 3,7 0,7 26,4 0,0 12,6 0,0 0,5 3,3 0,0 1,6 61,6 0,1 7,4 0,0 0,0 0,0 1,7 119,9
90-45 0,0 1,5 0,1 9,7 0,0 17,0 0,0 0,1 4,5 0,0 1,6 52,6 0,1 13,6 0,0 0,1 0,0 8,9 110,2
<45 0,0 6,8 1,8 24,8 0,0 4,4 0,0 1,5 1,4 0,0 0,7 26,8 0,0 2,5 0,0 0,0 0,1 1,0 72,0
MN-MSPG
>90 0,0 1,1 8,4 68,7 0,0 1,0 0,0 4,0 0,1 0,1 0,1 2,4 0,0 0,4 0,0 0,0 0,0 0,2 86,6
90-45 0,0 1,0 5,2 71,9 0,0 2,8 0,0 3,1 0,2 0,1 0,0 2,8 0,0 1,7 0,0 0,0 0,0 0,4 89,4
<45 0,0 1,6 6,1 76,7 0,0 6,3 0,0 2,2 0,5 0,1 0,1 4,1 0,2 7,5 0,0 0,0 0,0 3,7 109,2
QDM
>90 0,0 0,8 0,0 42,7 0,0 3,1 0,0 0,0 1,1 0,0 0,4 14,1 0,0 0,3 0,0 0,0 0,0 0,1 62,6
90-45 0,0 0,9 0,0 27,5 0,0 3,0 0,0 0,0 1,9 0,0 0,7 16,5 0,0 1,5 0,0 0,0 0,0 0,8 53,0
<45 0,0 0,8 0,0 54,4 0,0 4,1 0,0 0,1 1,0 0,0 0,4 12,0 0,0 0,1 0,0 0,0 0,0 0,1 72,9
61
Appendix B: Mineral grade vs. ks-value When plotting the wt% of each selected mineral/mineral group against the ks-value for each domain and
blending test. It can be observed that some of the mineral and mineral groups have a linear trend between
ks and wt% whereas others seems to have one or two outliers. In the main text the plots for Sphalerite,
Quartz, Carbonates, Pyrite and Galena is presented and discussed. In Appendix B the rest of the 11 different
mineral and mineral groups will be presented and discussed shortly. The main problem with Galena in the
main text is that a clear outlier, possible two outliers were present and no linear trend could be observed. In
Figure 29, Figure 30, Figure 31, Figure 32, Figure 33, Figure 34Figure 35 and Figure 35 the result of Galena,
Pyrrhotite, Garnets, Mica, FeOx, Amphibole/pyroxenes and Feldspars. When removing one potential
outlier for the aforementioned minerals and mineral groups indicative trends can be observed for all mineral
and mineral groups except Galena. To be able to observe a linear trend for Galena a second possible outlier
had to be removed. However when removing the second outlier 50% of the original data (domains) have
been removed. It should be noted that the removed outliers is different for the different mineral/mineral
groups. Mica R2-value (0,922) indicates that when removing the outlier in the Mica dataset the data points
fit well to a linear relationship between wt% and ks-value.
y = -0,307x + 4,6523R² = 0,0151
y = -0,0569x + 2,1089R² = 0,0051
y = 0,2853x + 0,2559R² = 0,9264
0,0
2,0
4,0
6,0
8,0
10,0
12,0
14,0
16,0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
Gal
ena
(wt%
)
ks (mass % loss/time)
Galena
Galena without 1 outlier
Galena without 2 outliers
Figure 29 Galena, trends with and without outliers.
62
y = -0,2451x + 2,5549R² = 0,0352
y = 0,2683x - 0,0806R² = 0,922
0,0
1,0
2,0
3,0
4,0
5,0
6,0
7,0
8,0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
Mic
a (w
t%)
ks (mass % loss/time)
Mica Mica without outlier
Figure 30 Mica, trends with and without outliers.
63
y = -0,1663x + 1,6949R² = 0,0143
y = -0,0196x + 0,2029R² = 0,6726
0,0
1,0
2,0
3,0
4,0
5,0
6,0
7,0
8,0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
FeO
x (w
t%)
ks (Mass % loss/time)
FeOx
FeOx without outlier
Figure 31 FeOx, trends with and without outliers.
64
y = -1,2287x + 12,218R² = 0,1703
y = -2,2059x + 17,234R² = 0,7476
0,0
2,0
4,0
6,0
8,0
10,0
12,0
14,0
16,0
18,0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
Am
ph
ibo
le/p
yro
xen
e (w
t%)
ks (mass % loss/time)
Amph/pyro
Amph/pyro without outlier
Figure 32 Amphibole/pyroxene, trends with and without outliers.
65
y = -0,1247x + 1,1172R² = 0,3452
y = -0,1083x + 0,9499R² = 0,673
0,0
0,2
0,4
0,6
0,8
1,0
1,2
1,4
1,6
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
Gar
net
s (w
t%)
ks (mass % loss/time)
Garnets Garnets without outlier
Figure 33 Garnets, trends with and without outliers.
66
y = -0,0653x + 3,8126R² = 0,0006
y = 0,2071x + 1,0418R² = 0,5307
0,0
2,0
4,0
6,0
8,0
10,0
12,0
14,0
16,0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
Pyr
rho
tite
(w
t%)
ks (Mass % loss/time)
Pyrrhotite
Pyrrhotite without outlier
Figure 34 Pyrrhotite, trends with and without outliers.