Libro DDD

Journal Pre-proof
Advanced comminution modelling: part 1 – crushers
Paul W. Cleary , Gary W. Delaney , Matt D. Sinnott ,

Sharen J. Cummins , Rob D. Morrison
PII: S0307-904X(20)30324-3
DOI: https://doi.org/10.1016/j.apm.2020.06.049
Reference: APM 13475
To appear in: Applied Mathematical Modelling
Received date: 11 August 2019

Revised date: 21 June 2020
Accepted date: 24 June 2020
Please cite this article as: Paul W. Cleary , Gary W. Delaney , Matt D. Sinnott , Sharen J. Cummins ,
Rob D. Morrison , Advanced comminution modelling: part 1 – crushers, Applied Mathematical Mod-
elling (2020), doi: https://doi.org/10.1016/j.apm.2020.06.049
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2020 Published by Elsevier Inc.

Prepared for Applied Math. Model.
Highlights
 Drivers for growth and challenges for the mineral processing industry are
reviewed.
 DEM models including breakage are capable of simulating industrial scale
crushers.
 Compression and impact crushers can be investigated in detail using these models.
 Breakage and flow are analysed in twin roll and cone crushers
 Crushing behaviour in a VSI involves multiple mechanisms of impact breakage.
1/58
Advanced comminution modelling: part 1 – crushers
Paul W. Cleary1, Gary W. Delaney1, Matt D. Sinnott1, Sharen J. Cummins1, and Rob D.
Morrison2
1
CSIRO Data61, Private Bag 10, Clayton South, 3168, Australia
2
Retired
* Corresponding author: [email protected]
Abstract
Particle scale modelling of comminution processes can provide significant insight into the
flow of particles, their breakage, the effect of slurry, wear and energy utilisation within
these machines. The ability to use such models to assist in faster and lower cost design of
new comminution devices and in the improvement of existing ones will be critical to the
ability of industry to respond to the substantive challenges facing mineral processing in
the next decade. These challenges are reviewed and drivers for change are discussed.
Understanding individual unit process performance needs to be in the context of the
flowsheets in which they are used so this is also reviewed. Advances in particle based
comminution modelling are presented with this work divided into two parts. This first part
focuses on recent advances in particle based modelling of crushing. Three crusher types
are used to demonstrate these capabilities:
1. Twin roll crusher

2. Cone crusher
3. Vertical Shaft Impactor (VSI)
These show the nature and level of fidelity that is now possible to include in particle scale
crusher models including breakage of non-spherical particles and prediction of the product
size distribution and throughput.
Keywords
Comminution; DEM; breakage; crusher; flowsheet; mineral processing challenges
2/58
1 Introduction
1.1 Comminution and industry drivers
Comminution is the class of processes relating to the deliberate reduction in the size of
particles. This is a critical and very energy intensive step in production of base metals,
light metals, precious metals and iron ore which form the basis of our industrial and
consumer economy. Continuing access to such metals is a critical determinant of standard
of living (particularly in development of infrastructure and technology). The total value of
the base metal component (Al, Cu, Ni and Zn) of such product is in excess of US$800
billion p.a. [1]. Large scale comminution of rock involves huge capital costs (with typical
new green fields mineral processing developments as of 2019 being of order US$1-5
billion) and huge energy usage. Energy consumption is typically reported as being of order
25-50% of total energy used in metal production. La Nauze and Temos [2] reported that
WMC operations direct energy consumption in comminution was around 25-30% of total
usage based on measured greenhouse gas (GHG) emissions. However, energy utilized in
manufacture of wear parts for comminution is often of a similar order suggesting that the
comminution share of total energy is considerably higher than is usually reported [3].
Ballantyne and Powell [4], drawing on a range of sources, reported that between 25-50%
of total energy consumption was used in comminution. Curry et al. [5] reported ranges for
comminution energy of 30-50% for a range of sites but observed that late in mine
operating life this could increase to as high as 80-90% as ore grades declined. The general
pattern across different sources then is consistently that around 30-50% of direct energy
usage is attributed to comminution.
Reliable estimates of the total direct energy consumption in comminution for mineral
processing of metalliferous and iron ores are difficult to obtain. A DOE study [6] found
that around 3% of worldwide energy consumption could be attributed to these usages, but
in the period since then ore grades have decreased and production volume has increased so
it is likely higher now. In contrast, Napier-Munn [7] estimated that these usages result in
consumption of order 0.5% of world energy supplies, rising to 1.8% when
cement/aggregate and coal grinding are included. When the energy involved in creating
grinding media and mill liners is also included, which is about 7kWh/kg [3], or about
150% of the energy required for the reduction of ore to metal, then the percentage
increases sharply. Rankin [8] estimates total embodied energy per kilogram for steel at 6.3
kWh but because so much steel is produced, a greenhouse gas contribution of 7% of
global production was found in 2007. Cement manufacture also requires a great deal of
comminution energy and even though it only embodies 1.5 kWh/kg, so much is
manufactured that it is estimated to produce an additional 8% of global greenhouse gasses.
3/58
Collectively these sources indicate that comminution is conservatively responsible for

more than 50% of mineral and cement processing energy usage and world total usage is
likely to be in the range of 7-8% with matching contributions to GHG generation.
The commercial imperatives for mineral processing operators are to:
1. Increase throughput,
2. Improve recoveries,
3. Reduce operating costs, and
4. Reduce energy consumption,
typically in the order listed. The first two considerations dominate because the capital cost
of the processing plant is the dominant financial consideration. The more product tonnes
that can be produced from a plant the lower is the capital cost per unit output and the
better is the economic return on investment. Operating costs are an influence whose
importance is more cyclical with strong focus during downturns and with less importance
during times of strong commodity prices. Energy consumption, which impacts via
operating costs, is not overly controllable once a plant is constructed and is rarely a
decision driver for either plant design or operation unless energy costs increase
substantially.
There are several very strong long term trends which present specific and very substantial
challenges to this industry:
 Decreasing ore grade: high grade deposits are largely depleted, and new deposit
developments are broadly prioritised by their ore grade with many more recently
started mines typically being of quite low grade but of very large scale. This means
that the ratio of waste or gangue to ore is increasing and that for the same unit of
product ever increasing volumes of rock must be ground to liberate the required
mineral fractions, [9]. The effect of decreasing grade on comminution cost per
product tonne for copper is shown clearly by Ballantyne and Powell [4] in their
Figure 10. The larger proportions of gangue, which is a consequence of decreasing
ore grades, sometimes present opportunities for early rejection of gangue (by
sorting or more traditional beneficiation processes) if it can be liberated at much
coarser sizes than is required for liberation of valuable minerals. Such
opportunities are determined by the nature of the ore mineralogy. A broad range of
such opportunities has been reviewed by Bearman [10]. The consequence of such
4/58
opportunities is greater variation and complexity in the grinding circuits than are
required for mineral liberation.
 Increasingly finer grinding is required for liberation: as ore grade decreases and the
granularity of the valuable minerals decreases, this requires the rock to be ground
to increasingly finer sizes to allow commercially useful separation of the valuable
ore components from the waste. The energy cost of grinding finer rises rapidly
with decreasing grind size, as shown in Napier-Munn et al. [11] which based on
the easiest such version by Hukki [12]. Not only is more energy required to
generate greater surface area (halving particle size generates eight times more
particles per unit mass with twice the surface area), but particles also become more
difficult to crush as fluid forces increasingly tend to carry them around colliding
media as they become finer. Hence, more recently developed devices such as the
IsaMill [13] can use less energy than the Hukki curve suggests at very fine sizes.
 Increasing volumes of final metals are required as the world population increases
and as an increasing fraction of that population has an industrialized/consumer
standard of living. This is a very strong double driver for increasing demand for
industrial metals. It has and will continue to drive significant demand for
increasing metal volumes.
The combination of these primary drivers means that energy usage in comminution for
mineral processing has been rising, and can be expected to continue to rise, extremely
rapidly into the foreseeable future. These trends are clearly demonstrated by historical data
presented by Valero and Mudd [14] (Figure 2) with copper energy usage doubling
between 1990 and 2004 and increasing 11-fold over the last 50 years. Similar behaviours
are demonstrated for nickel, lead, zinc, silver and gold. Doubling every 1-2 decades into
the future, from the current relatively high usage base, is unlikely to be sustainable.
Secondary trends include drivers such as
 Electrification of cars where electric cars typically require three times more copper
metal in construction than internal combustion cars (typically 83 kg for battery
electric vehicles compared to 30 kg for mid-range internal combustion vehicles)
(https://www.copper.org/publications/pub_list/pdf/A6191-ElectricVehicles-
Factsheet.pdf) with an estimated 1.7 million tonnes required per annum for electric
vehicles by 2027.
 Needs for newer metal products such as lithium (for batteries for electric cars or
large volume energy storage of renewable generated energy) and rare earth metals
for electric and magnetic systems.
will add further to the base line demand growth.

5/58
1.2 Comminution machines as part of grinding circuits
The outcome required from comminution is typically a fairly narrow fine size range of
particles that are suitable for downstream separation processes such as screens or flotation.
The rock starts in the mine after blasting with typical top sizes as large as 1 m or more. To
get from the ROM (run-of-mine) size to the required product size usually requires several
incremental steps that progressively reduce the particle size distribution. The coarser
stages are termed “crushing” while the finer ones are termed “grinding”, largely as a result
of the nature of the processes involved at each stage. Usually between 1 and 4 stages of
crushing is used. Crushing is usually more energy efficient, but this becomes less clear for
very fine crushing in tertiary or later stages. Commonly used machines for large scale
crushing include gyratory and twin roll crushers (for primary crushing) and high pressure
grinding rolls (HPGR) and cone crushers (for secondary crushing). Vertical shaft
impactors (VSI) are also used extensively in the quarry industry.
Post crushing (when top sizes are typically between 100 and 200 mm), the feed material
then proceeds through one or more stages of milling. The most broadly adopted primary
mills are tumbling mills (large rapidly rotating vessels that both throw and shear particles).
When used with grinding media (“balls”) they are called SAG (semi-autogenous grinding)
and when only rock is present they are called AG (autogenous grinding) mills. The most
common secondary mill is a ball mill in which relatively fine feed particles (0.1-10.0 mm
top size) are ground in a charge consisting predominantly of steel balls. Mills can be
operated in open or closed modes (which refers to whether the product is size separated in
some way with the oversize fed back into the mill). Substitution of some crushing and
milling steps by machines such as HPGR and VRM (Vertical Roller Mills) that have
attributes shared by both crushers and mills has been a trend over the last decade or so in
response to the complexities of newer ore types and in the search for increased energy
efficiency for new plants.
The sequence of steps and processes for comminution is usually termed the “comminution
circuit” or sometimes the “grinding circuit”. The nature of the circuit needed and the
specific choices of machines within this and how they are ordered and linked is a
consequence of both the ore mineralogy and the volume of rock planned for processing.
At the start of the 20th century, mines were producing and treating very high-grade ore
from typically small deposits. Hence, circuit capacities of one to a few hundred tonnes per
hour were common practice. Because of high ore grade, these operations could produce
enough metal to underpin profitable operation. As shown in Figure 2a, a large jaw crusher
was used to reduce ore from the mine to an 80% passing size (P80) of 100-150 mm. A
cone crusher was then used in closed circuit with a screen producing a 20-30 mm product
6/58
which could then be fed with some added water into a tumbling mill filled with heavy
steel rods (called a rod mill). Rod mill product might go directly into a tumbling mill filled
with steel balls (a ball mill) or be classified after the addition of a lot more water, in a
gravity device such as a rake or screw classifier. The oversize product from the classifier
would be directed back into the ball mill and the fine product would then be separated by
gravity (according to mineral density) or by the then very new process of flotation.
Flotation employs reagents which are selectively attracted by different surface
characteristics of exposed mineral surfaces and then to air bubbles giving the process its
name.
As ore grades decreased, much larger tonnages of ore had to be processed to generate
sufficient product for a mine to be profitable. The process remained much the same, but
the scale of process equipment was increased substantially. Figure 2b shows a large
gyratory crusher processing run of mine ore to 100-150 mm followed by two stages of
cone crushing to generate a P80 of 6-12 mm which could be fed into a large ball mill
operating in closed circuit with hydrocyclones. By using multiple process lines, these
circuits could process up to 100,000 tonnes of ore per day. These circuits became common
in the 1960s and 70s. Bougainville Copper [15] and Pinto Valley [16] are typical
examples.
Demand for simplicity and reduced operating costs drove the development of ever larger
equipment. Large autogenous (ore grinding ore) and semi-autogenous (balls plus ore)
mills followed by very large ball mills (see Figure 2c for a typical such circuit) took over
in the 1990s and have dominated ever since although clusters of tower/Verti-mills are
beginning to challenge dominance of ball mills [17]. The Mawby Memorial volumes [15,
18, 19, 20] provide a useful record of the evolution of large SABC circuits as they gained
acceptance from the early adoption of very large ball mills at Bougainville Copper to the
rise to dominance of SAG mills and to the uptake of very large (20 MW) SAG and ball
mills at Cadia and other sites.
The underlying driver has been the scale of equipment. Doubling the capacity of a piece of
equipment typically only increases its cost by about 40%. Hence, the heuristic of the
“square of two factor” for the cost increase required for double the capacity. As a single
piece of equipment requires about as much operator attention no matter how big it is,
operating costs are also decreased by increases in scale. A single SAG/ball mill line can
often process as much as 100,000 t/day, at a reasonable investment cost level, and have
sufficiently low operating costs to enable profitable operation. However, these huge
increases in capacity bring their own challenges. The main challenges are then to discover
sufficiently large and uniform enough deposits to support multi-billion dollar investments
which require decades of operation to pay them back. The other challenge is to have the
7/58
increasingly large unit machines operate efficiently and robustly. The use of sorting to
preferentially remove gangue material (waste) can modify the circuits and selectively
change the economics of production for specific ores. These drivers and the inherent large
variation in the mineral structure between deposits have driven the development of a wide
range of variations of the more standard grinding circuits (such as Figure 2c).
Hence, while increasing scale has served the industry well for more than a century, the
next century will require more process equipment innovation to satisfy demands for much
reduced energy and water usage as well as minimisation (or elimination) of long term
effects such as acid mine drainage.
1.3 Challenges for developing new machines and improving existing ones
A dominating issue is that the energy efficiency of the current main work horse mill type
(the SAG mill) is of order 1%. Beke [21], as quoted by Lynch [22], estimated that the
“theoretical” energy requirement based on the energy cost of creating new surface area
when compared with actual energy use was only 0.6%. When comparing the energy to
create new surface to the energy used in physically breaking rocks by impact or
compression one at a time the efficiency is better at 40% [3]. This means that the
inefficiency inherent in the SAG process in breaking the wrong particles to make the
wrong progeny far outweighs the energy inefficiencies in any individual breakage event.
This is highlighted by the DEM predictions, using incremental breakage [23, 24] which
showed that energy consumption used in producing body breakage of particles in a pilot
SAG mill was between 0.5 and 2.0% depending on conditions and rock brittleness. This
means that a significant minority fraction of the world’s energy consumption is wasted in
the production of heat, sound and ultra-fine particles (in the form of slimes) that form part
of the tailings. Such slimes can create significant waste storage, leaching and remediation
challenges. These inefficiencies contribute to increased greenhouse gas generation both in
the comminution process and in the manufacture of the grinding media used [25].
The cost and time for development of new comminution machines based on the last 50
years activity appears to typically be at least USD100 million and a 30 year time frame for
each new generation of machines. The time from patent to industry adoption is an
indicator of how long to complete development and to gain user acceptance. For example,
AG and SAG mills were introduced in the 1950s and took till the 1990s for widespread
adoption. Another example is the Schonert patent for high pressure crushing which was
granted in 1982 [22] and licensed to two manufacturers by 1985 [26]. The McIvor review
states that high pressure grinding rolls came into use in the cement industry during the
1990s, and also found acceptance in the diamond industry. However, tests on base metals
experienced serious wear problems and introduction into mineral processing required the
8/58
development of roll surfaces with very high wear resistance. However, as the cement
industry had long used roller crushers, at least some manufacturing capability already
existed. Following the same theme, tower mills continue to experience a development
cycle of more than half a century. They became commercially available in 1955 after
development from small scale mills used for pigment and other industrial fine grinding
applications, but the first mineral industry application quoted as a regrind mill was in
1980. Kubota Tower Mill developed larger units as regrind mills over the following
decade. MPSI which later changed to become METSO acquired the technology in the
early 1980s [27] and has developed ever larger units ever since as Verti-Mills. Large
Verti-Mills are now coming into use for secondary grinding in high capacity processing
plants [28]. The cost of more than 50 years of development across several companies
would be difficult to quantify with any accuracy but must have cost many, many tens of
millions in today’s dollars.
In general, a new device must work well enough in an existing application to displace the
standard approach and earn enough to fund continuing development. Such cases are quite
rare. For high pressure grinding rolls, the energy cost of cement grinding was the driving
factor.
Neither the cost nor the timescale are compatible with the medium to long term future
needs of the mineral processing industry to be able to respond to the primary trends that it
is subject to and they particularly make rapid adaption in response to sharp changes in
energy cost or GHG regulation problematic. The core reasons for this are:
 the lack of understanding of breakage mechanisms for particles in comminution

devices and how these are created by the motions generated inside the various
machine designs, and
 that a conventional design process starts with a laboratory prototype which is
scaled up through several step sizes to the required industrial throughput size.
Unfortunately mill performance involves multiple mechanisms (most importantly
particle breakage and particle transport/mixing) but these scale differently with
size. This means that performance can vary substantially with scale and often
means that little of what is learned at smaller scale can be usefully used at larger
scale. The cost of operating test mills at near industrial scale is significant with
substantial problems arising with supply of the required amount of feed at the
required feed rate for tests and then in disposing of ground product. These often
necessitate testing at remote sites which requires operators to be prepared to
support this, which further increases costs and reduces opportunities for testing.
9/58
The relative differences in the behaviour of breakage and transport mechanisms with
increasing physical scale of a comminution device also presents significant difficulties in
delivering working machines of increasing scale in response to the ongoing economically
driven demand for increasing mill sizes. Although the cost economics favour using larger
mills in response to the increasing rock volumes, this does not necessarily make the larger
machine operate as expected (in accordance with what was observed at the smaller scales
observed previously).
Furthermore, physical experimentation can provide little useful information about what is
happening inside comminution machines. Measurement is usually restricted to inputs and
outputs which are correlated and used to construct scale-up process models that developers
hope will assist in the design process – usually with only limited success. Measurement
inside the mill of detailed charge motion is possible with PEPT (Positron Emission
Particle Tracking) [29, 30] Govendar et al. 2011) but this is restricted to laboratory scale
mills.
1.4 Modelling and its role in comminution process design and optimisation
Particle scale modelling can provide detailed predictions of the internal flow behaviour (of
both solid and liquid components) as well as breakage, wear and energy consumption.
Particle scale models can provide fundamental understanding of each mill design at the
particle level where breakage occurs for all physical scales including large industrial scale
machines. They are able to cost-effectively provide critical performance information so
that testing can be performed on virtual machines of hypothetical design that do not yet
exist. They also allow scale-up to be performed virtually (quickly and at low cost)
providing clear information about how different mechanisms and processes change with
increasing physical scale and facilitating much more informed adaption and optimisation
of design concepts.
The Discrete Element Method (DEM) was originally developed by Cundall and Strack
[31]. Its usage for comminution was first introduced by Mishra and Rajamani [32, 33] as a
way to model tumbling mills. Rajamani and Mishra [34] then applied it to modeling of a
SAG mill. This was extended to explore the effect of operating parameters on mill
performance and wear prediction [35, 36, 37, 38, 39]. At the start, these DEM models
were two dimensional. Over the following decade or so, it was common practice to
consider thin axial slices through the middle of the mill (well away from end wall effects).
3D modelling of tumbling mills was introduced by Cleary [40, 41] and Morrison and
Cleary [42]. Since then there has been increasing adoption of the method for analysis of
dry tumbling mill performance [43, 44, 45, 46, 47, 48, 49, 50]. Reviews by Weerasekara et
al. [51] and Tavares [52] discuss the state of the art.
10/58
Crusher modelling using DEM is comparatively more recent because of the requirement
for direct resolution of breakage of particles in the DEM model in order for the particles to
actually transit the machine from feed to discharge. For compression-based crushers,
which squeeze particles in progressively narrowing spaces between converging surfaces,
DEM particles are unable to physically pass through the machine unless they actually
break up into smaller particles. Since this type of breakage capability is less common,
there is correspondingly less reported DEM modelling of crushers compared to mills.
A particle replacement method to enable this type of modelling was proposed by Cleary
[37]. Early DEM simulation of crushers includes lab-scale cone crushers [53, 54], single
particle breakage in jaw crushers [55, 56] and VSI [57] which all used a form of bonded
particle model. The Discrete Grain Breakage (DBG) [53] used bonded tetrahedral
elements rather than spheres. This was based on earlier work by Potapov and Campbell
[58] who considered detailed fracture predictions of small numbers of rocks modelled as
large polyhedral assemblies. Cleary [41] presented DEM models of an industrial scale jaw
crusher and a VSI using the particle replacement model and spherical particles. Lichter et
al. [59] introduced a Fast Breakage (FB) model which uses polygonal DEM elements and
a microscale Population Balance Model (PBM) to control particle replacement following a
fracture event. da Cunha et al. [60] used EDEM with spherical particles to predict
behaviour in a pilot scale VSI. Li et al. [61] presented a PFC model of coarse size
spherical feed in a batch, lab-scale, cone crusher with a simplified geometry which also
used a very simplified particle replacement model for the breakage.
Cleary and Sinnott [62, 63] presented studies of compression and impact crushers
covering broad ranges of available machine types at industrial scale. These DEM models
used the replacement method of Cleary [37] and spherical particles. A non-spherical
particle replacement model based on super-quadric particle shapes [40] using packing
methods described in Delaney et al. [64, 65] and t10 based progeny generation [11] was
used by Delaney [66] to perform realistic DEM simulation of an industrial scale cone
crusher. This method was used to predict cone crusher performance as a function of
changes in a wide range of crusher operating variables [67]. The polyhedral particle
breakage DEM model from [58] was also demonstrated for simulating a cone crusher [68].
Quist and Evertsson [54] and Johansson et al. [69] used an alternative approach in EDEM
with a particle breakage model for simulating a laboratory scale cone crusher. They
modelled the rocks as bonded assemblies of spheres with internal bond forces given in
[70] which was then used to predict breakage. Barrios and Tavares [71] used a particle
replacement model in EDEM with a very simple progeny generation model to simulate a
laboratory scale HPGR.
11/58
This paper is organised in two parts, with the first focused on crushing and the second on
milling, and reflects the status of comminution modelling by our group as presented in a
keynote talk at the 2nd Computational Particle Technology works which was held in
conjunction with the 13th International Conference on CFD in the Minerals and Process
Industries, in whose virtual special issue they appear. This first part focuses on
demonstrating advanced DEM/breakage prediction in crushers. Predictions for two
different types of compression crusher, being a twin roll crusher and a cone crusher, and
for one type of impact crusher, specifically a VSI (Vertical Shaft Impactor) are presented.
In contrast to the earlier DEM modelling of the twin roll crusher and VSI [62, 63] which
used spherical particles and a geometric sub-division model for progeny generation and
packing, the results presented here use super-quadric shaped particles, progeny generation
based on t10 theory and compression and incremental damage breakage methodologies.
Advances in such modelling and challenges remaining are discussed.
2 Summary of DEM method
The DEM method is now well known so will not be explained in detail. The
implementation used here was developed in-house by CSIRO and has been extensively
applied for the study of comminution [35, 72, 36, 37, 38, 73, 40, 42].
Particle shape is critical to flow behaviour, including the effective yield stress leading to
failure and flow, its porosity and permeability (see [41] for a discussion about shape
effects in granular flows). Here the particles are represented as super-quadrics, which in
their principle reference frame, are:
m m m
 x  y z
      1 (1)
a b c
The particle aspect ratios are given by the ratios of the semi-major axes b/a and c/a. The
exponent m determines the blockiness of the particle. A spherical particle has exponent m
= 2 and unit aspect ratios. As the aspect ratios increase from unity, they become more
elongated. As m increases, they become progressively more cubical with increasingly
sharp edges and corners. The contact detection method is based on that given in [74]. This
type of particle allows for a good representation for both naturally occurring particles
(grains and rocks) and non-round manufactured particles. Since particles in a crushing
operation are typically sized by some sort of vibrating screen or sieve, the intermediate
semi-major axis is used to define the particle size which is S=2×b.
The use of a non-spherical particle shape is critical for DEM prediction of realistic flow of
rocks and grains. Spherical particles flow too easily and lead to substantive errors in
12/58
several critical aspects of flow and granular assembly structure. These are detailed at
length in [41] and include:
1. Assembly failure strength is too low (with the most common symptom being an
excessively low angle of repose, as demonstrated in [75, 76].
2. Incorrect prediction of porosity due to the very different packing behaviour.
3. Substantial differences in the permeability of granular assemblies which strongly
affects the flow of interstitial gas and fluid and leads to much lower inter-phase
forces which in turn affects mechanical strength and flow behaviour of the granular
material [77].
4. Unique decomposition of the linear and rotational motion of particles in terms of
the normal and tangential forces from the contact model (for example rotation is
generated only the friction model) whereas any other particle shape has strong
coupling of both contact model components to linear and rotational motion.
5. Very different granular thermodynamic behaviour.
The use of non-spherical particle shapes is critical to producing realistic flow and
assembly failure behaviour. The methods broadly adopted by different schools of thought
include
 clusters of spherical primitives (common in PFC and EDEM and many in-house
DEM codes),
 polyhedral particles as originally developed by Hopkins [78, 79] and Potapov and
Campbell [58] and used in Rocky, and many other codes since, and
 super-quadrics (which are favoured by the authors of this paper).
Each approach has strengths and weaknesses which will not be explored here. The critical
point is that a suitable non-spherical shape needs to be used.
For super-quadric particles, very strong quantitative agreement has been obtained in
comparing to experiments both for dynamic flow and for slope failure behaviour (see for
example [40, 41, 76, 80, 81]). These studies show that the super-quadric shape descriptors
(two aspect ratios and a blockiness measures) are sufficient to reproduce realistic
behaviour for such granular flow. For crushers, there is an additional factor of the
interaction of the detailed shape of the particle with the loading of the particle and its
subsequent failure. With all shape models there is a level of detail resolution that is
computationally affordable, and which is measurable for real materials. It is not currently
clear whether these under-resolved finer scale shape features materially affect the process
level outcomes for the breakage behaviour of the particles. With significant practical
difficulties in measuring the required data (both input and output) and in computationally
13/58
resolving all these features and the resulting internal stresses and fractures within particles
this is an important question for the future. One possible approach is to use very detailed
numerical models that directly predict complete stress fields and fracture such as
demonstrated for elastic-brittle fracture of particles and structures [82, 83].
The contact model used is a linear spring-dashpot. Details are contained in the earlier
references and in [84]. Particles are allowed to overlap with the overlap denoted by x and
with dissipation controlled by the normal relative velocity vn. The normal force is:
Fn  knx  Cnvn . (2)
The linear spring provides a repulsive force while the dashpot provides the necessary
dissipation with the normal damping coefficient Cn chosen to give the required coefficient
of restitution. The force in equation (2) is restricted to being positive to prevent unphysical
attractive forces at the end of collisions when the first term is small and the latter term is
larger and negative.
The tangential force is controlled by the relative tangential velocity vt using an

incremental spring:
Ft  min  Fn ,  ktv t t  Ctv t , (3)
where the vector force Ft and velocity vt are defined in the tangent plane at the contact
point. The summation term is an incremental spring that stores energy from the relative
tangential motion and corresponds to the elastic tangential deformation of the contacting
surfaces. The dashpot produces dissipation of energy in the tangential direction and
represents plastic deformation of the contact in the tangential direction. The total
tangential force Ft is limited to the Coulomb friction Fn after which sliding occurs at the
contact point.
The resulting force from each contact is calculated at each timestep. The torque generated
by these contact forces are also calculated based on the force and the vector from the
particle center to the contact point. The forces and torques applied to each particle are
summed to give the net force and torque at that timestep. The governing equations are just
Newton’s equation for the acceleration of each particle based on the net force and
corresponding kinematic equation for linear motion. Similar governing equations apply for
the spin and orientation of the particle based on the net torque at each timestep. These
equations are integrated using a second order predictor-corrector method. The timestep
required for stability of the integration scheme and to give an accurate prediction of the
force and energy dissipation is 1/15th of the time required for a complete collision of a pair
14/58
of the lowest mass particles. Details of the timestep calculation can be found in [84]. The
search method for contacts is based on hierarchical grids and is described in [85].
The power consumption by a crusher is calculated by summing all the energy dissipation
components from the contact model in Equations (2) and (3) for all collisions (both
between particles and between particles and the crusher liner). When the load within the
crusher and the flow are at steady state the rate of power draw by the crusher and the rate
of energy dissipation are equal.
The breakage method is a replacement method, of the type originally proposed by Cleary
(2001b) but modified for non-round particles and using DWT data as described in Delaney
et al. (2013, 2015). The breakage model includes rules that control when a particle breaks,
how severely it breaks, the progeny distribution generated, the DEM particles created for
these progenies, and finally their packing into the original parent volume. The model used
in this work is motivated by Vogel and Peukert [86] and is described in substantial detail
in Cleary et al. (2017). The key components of the breakage model are
1. Determination of when a particle breaks
For compression breakage this uses the elastic energy stored in particle contacts (as
given in [67]. This elastic strain energy at a contact is given by:
(4)
It is the amount of elastic energy stored in the contact due to the deformation xi,j
(as indicated by the particle overlap) at that time. The elastic strain energy i,j at
each contact j of particle i is compared to the threshold elastic strain energy (which
we term E0) that is required to initiate cracking. The breakage outcome function is:
{ , (5)
where an outcome of 1 indicates breakage. Here ξi denotes the material of particle

i. When this threshold is exceeded (ζ > 0) then the particle breaks and is replaced
by the required set of progeny particles. While the elastic strain energy remains
below this threshold the particles do not accumulate any damage and do not break.
For impact breakage the energy absorbed by a particle in collisions above an

elastic threshold E0 is used. The probability of a particle being broken by collision j
with energy dissipation Ej, from [24], is:
15/58
- ( ( ) ), (6)
where En is a material constant that determines how intensively damage occurs to a

particle per unit of input energy. Note that this is the same method as used for
incremental breakage of particles in mills. In terms of earlier published versions of
the incremental breakage model, En = 1/b′. For this application, b′ is determined
from tests to characterise incremental breakage [87]. The form of the equation
follows the standard t10 equation (7).
2. Determination of the severity of breakage and the progeny size distribution
The degree of breakage of a particle is given by the JKMRC breakage map [88] as
represented by t10 (which indicates the fraction of the original parent mass that is
in fragments smaller than 1/10th of the parent size):
( ( )), (7)
where A and b are material dependent parameters that characterise the breakage of
the rock [11] and En = 1/b.
For compression breakage the energy controlling breakage intensity is the specific
breakage energy parameter Eb (in kJ/kg) which depends on the material (as is the
case for E0) and which needs to be calibrated to match measured size distributions
from some test data. This is used in the traditional t10 model with E = Eb.
For impact breakage the breakage intensity E is given by E which is the total
damage energy absorbed by the particle during all of its preceding collisions which
have contributed to damaging the particle up to the point of failure:
∑ ( ). (8)
3. Generation of the daughter particles
This step is the same for compression and impact breakage and is as described in
detail in [67].
4. Packing of the daughters within the parent particle
This step involves packing the specific progeny sequence generated in the previous
step into the volume of the parent particle. Each packing calculation is performed
on the fly in the DEM code since no two particle breakage events are identical and
16/58
so use of pre-computed or simplified packings creates a significant risk of biasing

the product size distributions predicted by the model. The dynamic packing
technique is described in [64, 65] and [67].
5. Replacement of the parent with the packed assembly of daughter particles
This replacement process is performed as an instantaneous change to the DEM

particles and reflects the fact that the timescale for the fracturing process is
significantly faster than the DEM contact model timescale.
6. Unresolved progeny and predicting fine product
A key advantage of this DEM breakage model which uses the empirical t10
approach to estimating progeny sizes is that progeny smaller than the minimum
resolved size of DEM particles are also known. This fine product particle size
distribution information is tracked and allows prediction of the crusher product
below the sizes included explicitly in the model.
Pre-processing, data analysis and visualisation for the simulations shown is performed in
software developed using Workspace which is a Software Workflow System (SWS) [89].
3 Need and challenges in validation of industrial scale DEM

crusher models
The DEM method is well validated for tumbling and other forms of mills. The DEM
solver used in this paper has been specifically validated using both photographic and
PEPT measurements [90, 76, 91]. However, validation for crushers is much more difficult
because of the lack of availability in the public domain of well controlled, suitably
instrumented and sufficiently complete data sets. At laboratory scale validation, such as in
[53], is possible, but at industrial scale suitable data is rarely collected and/or is not made
publicly available. The data required includes power draw, feed rate, feed composition and
hardness, crusher resident load, crusher control settings, throughput and product size
distribution and the shape change of the working surfaces of the crusher (captured by laser
scan) all measured over the full life cycle of the crusher liner. Difficulties include
challenges in instrumenting all these quantities in an operating plant and the very high cost
associated with doing so. However, the development of such crusher data sets suitable for
validation would be very valuable for progressing this type of micro-mechanical
modelling.
17/58
4 Double roll crusher
4.1 Geometry, model parameters and operating conditions
A double roll crusher consists of a pair of oppositely rotating rolls with large teeth for
gripping and compressing rocks. Figure 3 shows a schematic of such a crusher. Particles
are drawn down from a hopper above and are gripped between opposite sets of teeth on
the sides of the rolls leading to strong compression. In this case, the rolls have teeth of two
sizes. The larger ones have steep moderately curved leading faces and a relatively sharp
tip at the outer end with trailing faces that are highly curved. This shape enables the
opposite pairs of teeth to grip quite large particles at locations above the two rolls and drag
them down between the rolls where they are compressed and fractured. The large teeth are
arranged to be offset from each other in the axial direction. Small teeth are located
between the large teeth. This type of crusher is often used for sizing large particles and can
be used as a primary crusher. Sizing is a process which ensures that all the particles are
less than a required maximum size whilst not being overly concerned about the details of
the product size distribution. This maximum is determined by the diagonal distance
between the large teeth of opposite rolls. This type of crusher can be run in batch mode (as
is the case when loaded directly from a truck) or continuously (as when loaded from a
conveyor belt). Here we model a batch operation with a single dump of feed being
processed.
The geometric and operating condition information for this crusher is given in Table 1.
The rolls are 0.5 m long and 0.45 m in diameter. The roll rotation rate is 60 rpm. The feed
size range is uniform from 100 mm to 300 mm. The rock density is 2800 kg/m3. The feed
material has a blockiness range from 2.5 – 4.0 and aspect ratio ranges for intermediate and
short axis length to long axis length of 0.9-1.0 and 0.8-1.0 respectively. The breakage
product material has broader blockiness range from 2.5–5.0 allowing more angular
progeny than found in the feed. The breakage product aspect ratio ranges for both
intermediate and short axis length to long axis are 0.6-0.9. The minimum breakage
fragment size resolved in the simulation is 15 mm. Unresolved progeny from 15 mm down
to 4 mm are also predicted as described in the method section above.
Breakage in this crusher is compressive so we use the compression breakage method

described earlier for the DEM breakage prediction. The rock strength is characterised by
energies E0 and Eb which reflect the elastic contact energy at first fracture and when
fracture is complete. It is common for the strength of rock to vary with size as a result of
the higher frequency of flaws in larger rock. If the rock is uniformly strong with size then
its specific fracture initiation energy would be constant (that is, E0 would vary with cube
of the particle size). Commonly this energy decreases with size with the simplest
18/58
dependence being inversely proportional to feed particle size. This gives a quadratic
dependence of E0 on particle size. For a specific rock, this dependence would need to be
evaluated in either laboratory breakage experiments or obtained as part of calibration of
the breakage properties from field data. Here, we simply wish to demonstrate the
operation of DEM with compression breakage using a representative variation of material
strengths. Neither the DEM nor breakage models depend on the assumed form of the E0
dependence – this is purely a matter of ore characterisation for specific ores. For the
representative example here, the rock has E0 = 250 J for the smallest resolved particles in
the simulation of 15 mm with this value increasing quadratically based on the particle size
with E0 = 1250 J for the largest feed of 300 mm. We consider a softer ore with Eb = 1100
J/kg and a harder ore with Eb = 1500 J/kg.
The coefficients of restitution are 0.6 and 0.3 for rock-crusher and rock-rock contacts
respectively. The friction coefficients are 0.5 for rock-crusher and 0.6 for rock-rock
contacts. The normal spring stiffness is 106 N/m.
4.2 Flow through the crusher
The crusher contains an initial batch weight of 322 kg of material and the simulation was
run until all material was broken and the crusher was emptied. Figure 4 shows the flow
and breakage of material through the crusher for the softer ore with an Eb = 1100 J/kg. The
feed material is large and so is coloured from red to yellow to green for the smallest feed.
The counter-rotation of the rolls predisposes particles to being trapped between the rolls
with loads applied by teeth from either side. The weight of the particles above limits their
ability to rebound from the teeth and avoid being loaded. When the elastic energy at any
one or more of the contact points on any of the particles exceeds its elastic breakage
threshold E0, then the particle breaks.
In this case, one particle which is near the front of the crusher breaks shortly after the time
shown in Figure 4a. This is the orange particle directly above the teeth between the rolls.
In Figure 4b, this has been replaced by an assembly of smaller light and dark blue progeny
around a larger green colored fragment. Finer (blue) material from previous breakage
events can be seen passing below the rolls discharging from the crusher. The fragments
from the breakage event shown in Figure 4b are all sufficiently small to be able to fit
between the interlocking rows of teeth on the opposite rolls. In Figure 4c, the intermediate
green particle is now passing between the rolls and for a time obstructs the flow of finer
blue material above it (which would otherwise be able to travel unimpeded and more
rapidly through the crusher). The light blue particle to the right now experiences
significant compression between the rolls. In Figure 4d, some of the finer blue material
remains trapped above the intermediate sized green particle while a moderate amount of
19/58
the finer material from this breakage event is now exiting the crusher. The light blue
particle on the right that had been strongly loaded has now broken with its progeny now
able to discharge from the crusher. The sequence of loading of larger particles, their
breakage, subsequent clearance of finer product, and re-loading of any still over-sized
particles continues until all particles are broken and the crusher is empty.
4.3 Product size distribution for different Eb
Figure 5 shows the product size distribution for both the softer ore with Eb = 1100 J/kg
and a harder ore with Eb = 1500 J/kg. The crusher takes approximately 7 s to completely
empty the batch load of 322 kg and the top size predicted is consistent with the typical
design product size of 150 mm for this crusher configuration. The harder ore produces
moderately finer size distribution, in particular for the intermediate size ranges from 16
mm up to 90 mm. The two size distributions converge at both the top and bottom of the
size ranges. This behavior is expected. Ores which show more resistance to breakage by
impact in a drop weight test also experience a higher value of maximum breakage “A”.
This suggests that hard ores can absorb more elastic energy before they fail resulting in
finer progeny as that energy is dissipated in the fracture process.
5 Cone crusher
A cone crusher is a machine that breaks rocks by squeezing them between an eccentrically
gyrating cone (also called a mantle) and a concave [11]. Particles enter the cone crusher
from the top, travel down as a packed bed until they are near the working surfaces where
they become trapped between the mantle and the concave and are in due course broken
once or more by the large compressive forces applied. Larger rocks are broken first with
their progeny falling to lower positions in the crusher where they are typically broken
multiple times until they are sufficiently small to exit from the bottom of the crusher.
Traditionally modelling of a cone crusher is performed using population balance models
(see Napier-Munn et al., [11] for an introduction). Bengtsson and Evertsson [92] also
presented a new stochastic approach for predicting particle size distribution and energy
consumption using a first-order differential equation to predict how the variance depends
on the compression length and combined this with a bimodal Weibull distribution to
predict the cumulative size distribution and power draw. DEM with a compression
breakage model (see examples in [53, 54, 62, 66, 67, 69] provides a detailed micro-
mechanical approach to predicting particle breakage coupled to granular flow for any type
of liner configuration. This is the approach presented here. For a broad introduction to
cone crushers and understanding the effect of cone crusher geometry and how to measure
these, see [93].
20/58
A cone crusher consists of a mantle or cone, which is a precessing conical shaped structure
in the center of the machine, and a concave which is the surrounding outer structure.
Particles flow, are trapped, compressed and then broken in the space between the mantle
and the concave. The output of a cone crusher is controlled by the Closed Side Setting
(CSS) which is the minimum approach of the mantle and concave at any height. This
determines the peak strain that can be applied to a particle of a specific size and therefore
whether it will break and into what fragments. The Open Side Setting (OSS) on the
opposite side to the CSS is the maximum constricted separation of the mantle and concave
and essentially controls the vertical flow of material as the gap opens and particles become
mobile and move downwards ([94] and Chapter 6 of [11]).
Figure 6 shows a schematic of a 7’ Symons cone crusher. This is the same configuration
as used by [66, 67]. The mantle is the central structure in the lower half of the figure,
while the concave is the surrounding outer structure. The feed arrangement is also shown
at the top. The mantle has a diameter of 2.134 m at the bottom and a diameter at the top
(just below the feed spreader) of 0.505 m. The mantle is 1.642 m high (including the feed
spreader plate). The side angle of the mantle is 50 degrees to the horizontal. The concave
has a height of 1.538 m, a bottom internal diameter of 2.47 m and a top internal diameter
of 1.79 m. The mantle precesses with its axis rotating clockwise (when viewed from
above) at constant speed. The axis is inclined at a tilt angle of 1.06 degrees to the vertical.
The tilted axis of the mantle and the vertical axis of the concave intersect at a fixed pivot
point located at the centre of the top surface of the feed spreader plate on the mantle. The
feed table at the top of the mantle spreads feed around the crusher (nominally uniformly
with circumferential position). The cone tilt gives an open side setting (OSS) for the
crusher of 58 mm, and a closed side setting (CSS) of 11 mm. In broad terms, the CSS
controls the compression and breakage of the particles, whilst the OSS controls the rate of
vertical flow of material through the crusher. The angular speed of the mantle precession
is 255 rpm. The design objective of the cone crusher is to achieve as close to pure
compression as is possible while minimising shear loading of the crusher (which is
required to give acceptable wear life). The geometric and operating condition information
for the cone crusher is summarised in Table 2.
A generic competent ROM (run-of-mine) is used here. These material properties match
the base case values that were used in [67]. The feed top size is 38 mm and the bottom size
is 6.7 mm. The feed size distribution is given in Table 3 and is based on data by Andersen
[95]. Progeny in the simulation are resolved down to 6.7 mm. Unresolved progeny from
6.7 mm down to 0.5 mm are also predicted as described in Section 2 above. The rock
density is 2800 kg/m3. Feed and breakage product particles have a blockiness range of
21/58
2.5–5.0. The aspect ratio ranges of the other side lengths to long axis length are 0.5-0.8
respectively. The friction coefficient is 0.55 (for both rock-rock and rock-liner). The
coefficient of restitution is 0.5 for rock-rock and rock-steel liner.
The normal spring stiffness used in the contact model is k = 107 N/m. This crusher is
operated as choke fed with a continuous stream of particles entering the top of the crusher.
The feed rate is 330 tonnes/hr. The typical model size was around 200,000 particles after
10 s of operation.
Breakage in this crusher is compression generated so the DEM model of this crusher again
uses the compression breakage method. The material breakage parameters for the DEM
breakage model are the elastic energy threshold E0 at a contact for breakage to occur and
the breakage intensity Eb which represents the energy absorbed at the end of breakage and
controls the progeny size distribution generated for that breakage event. Here we use E0 =
100 J and Eb = 2.05 kJ/kg. These parameters correspond to a moderate to high strength
competent rock.
5.2 Flow within the crusher
Figure 7 shows the flow within the cone crusher. The particles are shown three times with
different colouring to enable the flow, breakage and stresses to be analysed. The tilt of the
cone means that on one side of the code the gap is maximally open and on the other it is at
a minimum (indicated by the OSS and CSS respectively). These locations precess around
the crusher as the cone gyrates (or nutates) around its pivot point at the top. In this figure
the closed side is on the left and the open side is on the right.
Figure 7a shows the particles coloured by their speed. In general, the particle speeds are
higher on the open side in the lower half of the crusher as these particles are relatively
unconstrained and able to accelerate into the space produced by the opening of the gap and
downwards. This material is typically green trending to yellow. Once particles exit the
bottom of the crusher they accelerate under gravity and are coloured orange and then red.
Above the freely flowing material on the open side is a large dark blue region which is the
slowly moving densely packed bed in the upper half of the crusher that results from the
use of a choke feed mode. On the left side of the crusher (in Figure 7a) the closing of the
gap constrains the particles which are slowed by friction. These particles are coloured light
and dark blue reflecting the much slower particle speeds in the closed side of the crusher.
In the middle front of the crusher, where the gaps is intermediate size but closing there is a
mottled colouration with flow slowing as particles jam and some areas of red where
particle speeds are high following fracture of particles in these locations.
22/58
In Figure 7b, the particles are coloured by their total compression force which shows
which parts of the particle bed are strongly loaded. The red colour indicates loading forces
that are approaching those required for breakage and so show where the loading and
breakage occur. This high load region is in the front left which is the part of the crusher
where the cone and mantle are approaching each other and where compression is
strongest. This region is narrower higher up on the cone and becomes wider but more
dilute in the lower part of the cone. The lower section is further to the left which reflects
the high degree of approach of the cone and mantle surfaces that are needed to compress
the smaller particles that occur lower down the cone. Elsewhere the loads on the particles
are comparatively small with particles being coloured dark blue. This high compression
region precesses around the crusher moderately in advance of the CSS location.
Figure 7c shows the particle bed coloured by size with red being large, green intermediate
and blue smaller. Different particle sizes are reasonably uniformly distributed down to the
level of the compression zone where the cone and mantle surfaces sharply approach each
other. The larger and intermediate size particles (red through green) rapidly disappear
from around halfway down the cone as they are broken by the high compression forces
seen in Figure 7b. In the lowest section of the cone/mantle all intermediate particle sizes
(green) are broken and only fine product (mostly dark blue with a small amount of light
blue coloured material) remains, which is then discharged as product.
In order to better understand the detailed micromechanical interactions leading to

breakage, we show the particle bed in a cross-section of the right side of the crusher when
this is alternatively fully closed (Figure 8a) and fully open (Figure 8b) with particles
coloured by their speed. In each case, some inset close-up panels are included for the
lowest parallel section of the compression region and at the top where flow into the
compression regions occurs. When the crusher position is closed the particles in the
parallel final breakage section (which is intended to ensure that no particles larger than the
CSS can survive) are slowed (coloured light blue with a few dark blue). Most of the
particles are at least slightly smaller than the gap so can still flow downwards, albeit at a
relatively slow speed due to frictional sliding of this nearly mono-layer of particles along
the steeply inclined surface. Particles that are at the CSS size or which were larger and
have just broken typically are nearly stationary (dark blue) reflecting their currently or just
previously jammed state. The particle speeds above are much lower with much of the bed
coming to rest due to the high degree of convergence of the feeding zone at the top which
vastly exceeds the capacity below for particles to move through.
Once the gap opens fully (Figure 8b), there is significant space available throughout most
of the compression region. Particles quickly become only loosely packed and are no
longer constrained to slide in a narrow gap, so they accelerate rapidly. The maximum
23/58
speeds attained, and the distances travelled downward are controlled by the time period
between the open and closed phases (which is determined by the gyration rate of the
cone). The particles at the bottom of the parallel section reach speeds of 1.5 m/s or more.
The accelerating flow becomes quite dilute (as seen in the inset close-up) and is essentially
a gravitationally driven granular chute flow. Higher in the crusher where the surfaces
converge to create the compression zone the particle flow has accelerated with large areas
reaching a light green (around 1.2 m/s). This reflects material feeding down into the top of
the compression zone. The shape of the mantle strongly affects this feeding flow with the
angular inflexion point producing strong resistance to flow. Once this corner is rounded by
wear of the mantle, then flow down from the converging region may be more efficient
leading to an increase in the volumes of material filling the compression zone during the
open phase. This will strongly affect the following compression loading cycle and the
degree of breakage. This demonstrates a strong mechanistic dependency of the breakage
on the shape of the crusher liner surfaces.
6 Vertical Shaft Impactor (VSI)
The Vertical Shaft Impactor (VSI) is a popular class of impact crusher. The machine
accelerates particles to high speed and projects them into solid obstacles, leading to high
energy collisions that commonly lead to fracture of these particles.
Figure 9 shows the design of one type of VSI which is studied in this paper. This is very
similar to the same one used in the previous generation of this model using spherical
particles [63]. A continuous stream of particles is centrally fed from a hopper down into
the center of the rotor which rotates at high speed. The feed stream is dense and a crusher
operated in this way is termed “choke” fed. The particles in the feed stream form a dense
granular mass when they encounter the gentle conical surface on the impeller table and are
pushed sideways. Contact with the impeller blocks (or shoes) causes particles to accelerate
rapidly in the radial direction reaching speeds in excess of 60 m/s. The particles are then
thrown against the ring of anvils that are oriented to produce predominantly normal
impacts and to shatter the particles.
This VSI is 2.5 m in diameter and 1.85 m high. Its rotor is 1.2 m diameter, has three shoes
which are 450 mm high and is rotating at 750 rpm. The geometric and operating
specifications for this crusher are summarised in Table 4.
The feed size range has a uniform mass weighted probability distribution from a bottom
feed size of 20 mm to a top size of 70 mm. The rock density is 2700 kg/m3. The feed rate
24/58
is 400 kg/s. Feed and breakage product particle blockiness range (for the super-quadric
shape) is 2.5 – 5.0. The aspect ratio ranges for both intermediate and short axis length to
long axis lengths are 0.6-0.8 and 0.5-0.7 respectively.
Breakage in this crusher is due to high energy impacts so we use the incremental breakage
method described earlier for the DEM breakage prediction. The rock strength is
characterised by energies E0 (the elastic threshold for damage to begin) and En which
reflect ability of the rock to withstand fracture. Since the impacts in the VSI are very
energetic the value of E0 is comparatively small and has little effect on overall particle
damage. A value of E0 = 20 J/kg is used, corresponding to a strong competent rock. The
parameter En controls the fineness of the progeny generated from each fracture event so
this is the dominant material parameter in this type of simulation.Values of En = 100 and
3000 J/kg are considered here to evaluate the effect of the material strength on the
breakage behaviour. The minimum breakage fragment size resolved in the simulation is
4 mm. Unresolved progeny from 4 mm down to 0.2 mm are also predicted as described in
the method section above.
The coefficient of restitution was 0.5 and 0.3 for rock-crusher and rock-rock contacts
respectively. The friction coefficient was 0.5 for all contacts. These are taken from the
earlier modelling of this VSI using spherical particles [63]. The normal spring stiffness
used was 4000 MN/m.
6.2 Steady state flow structure in the VSI
Figure 10 shows the steady state particle distribution within the VSI with the particles
coloured by speed. The central choke fed material is all coloured mid-blue since it is
moving relatively slowly. A coherent curved stream of high speed particles (coloured red)
can be seen flowing off the end of the front face of each shoe of the rotor. This curves
behind the shoe, reflecting the anti-clockwise rotation of the impeller. In Figure 10a (at 2.8
s) the shoe on the left is approaching and the layer of high speed feed particles flowing
outward across the shoe surface is well defined. This stream is reasonably dense but has a
thickness only of one large particle. The material in the three streams is thrown outward
where it predominantly strikes the ring of fixed anvils and typically fractures. The finer
product particles rebound with intermediate speeds (coloured light blue to green) and
create a moderately dense granular gas within the grinding chamber. There are substantial
amounts of very slow moving material trapped above and behind the anvils. Gravity is
relatively weak in comparison to the turbulent like fluctuations which support the granular
gas cloud leading to quite slow settling of particles. This leads to perhaps surprisingly
large residence times and the system takes around 2 s to reach steady state. The settling
flow is strongest at the outside of the chamber where collisions with the stationary
25/58
structures reduce the granular temperature and allow a denser flow of fractured particles
down along the chamber walls. In Figure 10b, which shows the crusher load at 3.0 s, the
details of the fractured particle cloud within the crusher including the higher densities and
slower speeds at the top of the chamber and around the walls vary but the general structure
is similar (as should be the case for steady state flow).
Figure 11 shows the VSI crushing chamber from a steeper angle and with particles
coloured by their size. The feed material and coarse breakage products (70 mm down to 20
mm) are coloured green through yellow to red. Finer resolved breakage product (20 mm to
4 mm) is coloured green to light blue to dark blue. There are still coarse particles
remaining after collision with the anvils, as can be seen by the red and orange particles
present around the periphery of the crusher chamber. This is a consequence of
1. these particles having a non-negligible chance of surviving because the peak

energy dissipation gives a low but not zero chance of survival,
2. some particles being slowed by collisions with other particles before reaching the
anvils, and
3. some particle striking the edges or corners of the anvils obliquely which dissipates
much less of the incident kinetic energy than does a front on collision.
Note that the large majority of the particles making up the central granular gaseous cloud
of post-breakage fragments are smaller particles which are able to maintain a high
granular temperature for longer and so settle less quickly.
Figure 12 shows the particles in a thin horizontal slice through the mid-height of the rotor
and the anvils with particles again coloured by speed. This view very clearly shows the
thin (one large particle diameter) flow along each of the three shoe blocks of the rotor and
the backward arcing stream that flows from each of their ends. The higher speed
rebounding fragments, post breakage, can be seen behind the current impingement points
of these streams on the anvils. The steady state build-up of slow, often coarser fragments
around and behind the ring of anvils is also clearly visible.
The effect of non-spherical shape and of the probabilistic incremental damage model in
the VSI model can be seen by comparing Figures 10 and 11 to their counterparts in [63]
for the earlier generation spherical particle model. The granular cloud is denser and there
is more material, observably coarser, trapped in the upper section of the crushing chamber
above the level of the anvils. Coherent down-flow along the walls of the chamber was not
identifiable in the previous spherical model but now is what we attribute to higher energy
dissipation in collision of non-round fragments with the chamber walls resulting in higher
density and more rapid settling flow.
26/58
Figure 13 shows the distribution of breakage events and their intensity within the VSI
crushing chamber. Each particle breakage occurring within the previous 1 ms is shown by
a red circle superimposed on the standard flow picture with particles coloured again by
speed. There are around 225,000 such breakage events per second. These can be classified
into four regimes:
1. Breakage of particles travelling outward along the faces of the shoes of the rotor.
This motion is bouncing type flow not rolling or sliding (as was seen for the
spherical particle model). Impacts with the corners or edges of the non-round
particles used here cause small excursions of the particles away from the face of
the shoes. As the particles travel outward there is a large change in circumferential
speed of the rotor shoe. So as a particle moves outward and away from the shoe,
the speed of the shoe at the next point of contact with the particle is higher. So as
particles flow/bounce outwards they experience a serious of collisions which
provide the acceleration to the tip speed at the end of the shoe, but these can also
generate breakage (typically fairly weak) of the particles. This leads to some of the
breakage events at the face of the shoes that can be seen in Figure 13. This is a
weak primary breakage mechanism.
2. The dominant primary breakage mechanism is from the high intensity fracture of
the high speed outward travelling feed particles when they collide with the anvils.
This is the mechanism around which the crusher was designed. This makes the
largest contribution to breakage.
3. Breakage also occurs for particles in the granular cloud as a result of the feed
particles travelling outwards in the coherent streams from the impeller shoes. A
critically important effect of the large moderate density cloud of particles in the
steady state VSI is that the particles being ejected from the rotor are not free to
travel directly to the anvils but instead experience large numbers of collisions with
the cloud of particles. These collisions typically involve larger high speed feed
particles with smaller more slowly moving particles that have been broken
previously. The energy dissipated can be attributed equally to the two particles
because they have the same stiffness. The energy per unit mass (which controls the
probability of breaking) is then much larger for the smaller particle in the collision.
This means that the cloud of broken particles has a limited effect on the outward
streams of fresh feed material, but this stream has a very significant effect on the
finer particles in the cloud generating significant re-breakage of this previously
broken material. This is a secondary breakage mechanism which leads to a non-
negligible amount of additional re-breakage beyond what was intended with the
crusher design.
27/58
4. The cloud of finer broken fragments has reasonable fluctuating velocities (meaning
that a reasonable fraction is behaving essentially randomly). Some of these
particles randomly walk radially inwards into the path of the particle laden shoes.
Groups of fragments from direct impact of larger particles with the anvils often
rebound back towards the center of the chamber taking these sets of particles back
into the paths of the shoes as well. So, there are two different opportunities for
fragments to experience re-breakage by the rotor. A reasonable fraction of the
breakage events shown in Figure 13 at the face of the shoes is a result of this
mechanism. This is another secondary breakage mechanism that also contributes to
non-trivial re-breakage of material within the crusher.
Significant numbers of collisions occur between particles in the cloud and the outward
travelling streams from the particles which makes this an important contribution to the
comminution from rock-rock collisions (mechanism 3). There are also significant amounts
of breakage in the rotor region both with the shoes striking the outward flowing particles
sufficiently to produce at least weak breakage events (mechanism 1) and with strong
collisions between particles in the drifting cloud of particles moving back into the paths of
the shoes leading to strong impacts (mechanism 4). Since these particles are typically
much smaller than the feed the energy intensities of these secondary collisions are quite
high and lead to significant amounts of size reduction.
Only mechanism 2 was able to be predicted by the previous generation of our breakage
model since this used an average value of the energy absorbed and this value is necessarily
above that for mechanisms 1, 3 and 4 but below that of mechanism 2. The newer breakage
method used here is based on a probabilistic model for determining breakage so even
weak collisions can potentially lead to fracture but the resulting degree of breakage is
small and the resulting progeny (aside from the reduced parent particles) is very fine. This
allows different types of breakage (as generated by the very different collision
mechanisms in different parts of the crusher) to be predicted.
The power draw is 1045 kW for a feed rate of 400 kg/s giving a crusher charge (when full)
of 0.32 tonnes (corresponding to 33,000 particles). The energy dissipation is split with
80% in rock-crusher collision (mechanisms 1, 2 and part of 4) and 20% in rock-rock
collisions which is far from negligible (mechanism 3 and part of 4). This demonstrates that
re-breakage from rock-rock collisions is an important minority component of the breakage
process in the VSI. The ability to attribute breakage between different mechanisms is one
of the strengths of such micro-mechanical modelling. Experimental measurement of these
quantities is difficult and not reported in the literature.
28/58
Of the energy absorbed by the crusher, this is split with 13% absorbed by the rotor and
87% by the anvils and to a much lesser degree the crusher liner above and below the
anvils. Wear is directly correlated with the energy absorption. The energy dissipation split
shows that the anvils have a wear rate around 6 times higher than the shoes on the rotor
but that the rotor wear is not at all negligible, so wear of the rotor needs to be considered.
The throughput is necessarily 400 kg/s so as to match the feed rate (since the flow is at
steady state).
Figure 14 shows the particle size distributions (PSD) shown as % passing. The green
curve shows the feed which is nearly uniformly distributed giving a nearly linear %
passing between the feed bottom size (20 mm) and top size (70 mm). The red dashed
curve shows the size distribution of particles (both resolved and unresolved) that are
within the crusher volume. The product size, as measured by a sample plane, at the bottom
of the crusher is shown as dark blue. The resolved product is from 4 mm up to near the top
size. Since this material is extremely hard (En = 200 J/kg which is equivalent to b′ = 0.005)
up to half the feed material survives usually with small reductions in size from minor
breakage events. Note that the DEM breakage model used here allows prediction of the
product finer than the sizes resolved by the DEM model, albeit without any further re-
breakage from any further collisions. The minimum size tracked by this model is 0.2 mm
which is predominantly controlled by the measured progeny from the DWT which
measures down to 1/75th of the parent size (which is 1 mm for the feed top size) but
smaller for the finer feed and coarser products. The form of the product size distribution is
very similar to those seen in the literature for this type of crusher. This product PSD is
qualitatively very similar to experimentally measured distributions (see for example
Figure 9 from [92].
6.3 Effect of material strength En on PSD
The primary material breakage property that affects the fracturing of the particles and the
resulting particle sizes is the material strengths En. It is useful to understand how the
model responds to variations in this material parameter. The product size distributions for
four values of material strength En are shown in Figure 15. This shows that the softest
material gives the finest product size distribution (as can be seen with the blue curve for En
= 100 J/kg being consistently to the left of the other curves which indicates that it is finer).
En = 300 J/kg is a mid-range material strength (giving the somewhat coarser product
shown in the red in Figure 15). A material with En = 1000 J/kg is becoming somewhat
difficult for this VSI to break with the product given as the green curve. The coarsest
product (violet curve on the right) occurs for the hardest material with En = 3000 J/kg. For
this strength the amount of breakage is quite small, and the feed material substantially
survives giving the steep section for particles above 20 mm and a thinner tail of product at
29/58
lower sizes from the breakage. The forms of the curves are very similar for En ≤ 1000 J/kg
which converge for the finest size material. These show that behaviour ranging from
significant breakage to only modest levels of breakage can be predicted by the model.
The product size distribution can be characterised by the size at which the % passing curve
reaches specific percentage levels. The P20 (which measures the size below which there is
20% of the product) is a useful measure of the fine end of the product distribution. The
P50 (the size at which half the product is smaller) is a good measure of the medium part of
the size range. The P80 is often used to characterise the coarse end of the product and
sometimes as a proxy for the product top size. This is preferred since the actual top size
has very few particles and is heavily subject to sample noise and is typically hard to
measure reliably. The variation of the P20, P50 and P80 with material strength En is
shown in Figure 16 with logarithmic spacing for the independent axis En. The P50 and
P80 increase approximately linearly with increasing material strength reflecting the
increasing difficulty of breaking particles. By contrast, the P20 varies little for softer rocks
but then varies more quickly as the rock becomes harder, with larger increases from En =
1000 to 3000 J/kg than were found for P50 and P80. From a process perspective these are
realistic variations in product size.
Figure 17 shows the corresponding variation in the VSI power draw with the hardness of
the rock (again with a logarithmic variation in En). The net power draw (excluding
frictional losses in the impellor and its drive) is around 1.0 MW for soft rock and increases
with ore hardness but with smaller fractional changes as it becomes very hard.
These results demonstrate that the DEM breakage model for the VSI is producing
qualitatively sensible predictions for the product size and that these vary sensibly with the
hardness of the rock material being broken. Future work on the VSI will look at
quantitative validation of these predictions for specific measured rocks.
7 Conclusions
The challenges confronting comminution for the mineral processing industry around rising
energy consumption, decreasing efficiency, the need to grind finer driven by ore depletion,
the need to use lower grade and more mineralogically complex ores and rising worldwide
demand for industrial metals were explored in detail. This creates strong need to better
understand, improve and optimize existing machine performance and to be able to design
and develop new machines quickly and efficiently. Particle scale modelling of
comminution processes can provide significant insight into the flow of particles, their
breakage, the effect of slurry, wear and energy utilisation within these machines.
Comminution performance also requires changes and adaption in flowsheet design.
30/58
Simulation capability for comminution unit operations needs to be performed in the

context of these evolving flowsheets. These changes were also reviewed in this paper.
This first part paper is concerned with the particle scale modelling of crushers. The
foundational method is DEM using efficient shape descriptions and coupled to breakage
models that allow the fracturing of particles from both high speed impact and compression
to be predicted. Simulation at full industrial scale for reasonable durations with realistic
particle sizes (for both feed and product) is now quite feasible for any type of crusher.
Information that can now be routinely generated include product size distribution,
throughput, wear on liners, power draw, stresses on working surfaces and feed segregation
effects. Such simulations can be used parametrically to understand sensitivity of flow,
breakage and performance to changes in equipment design, operational parameters and to
material property variation.
Three crusher types are used to demonstrate these capabilities:
1. Twin roll crusher

2. Cone crusher
3. Vertical Shaft Impactor
Results presented for the twin roll crusher demonstrate how these models can be used to
understand the flow features within the crusher and how variations in material properties
affect the output PSD. For the particular twin roll crusher examined, it was found that
harder materials (higher Eb value) produce finer size distribution (as should be expected
based on DWT data and experience).
For the cone crusher, we have demonstrated how understanding of the detailed flow
features inside the crusher can be obtained and highlighted the effects of geometric
variations along the vertical profile in the liners. The mantle shape in the cone crusher
strongly affects the feeding flow and the degree of breakage, so wear of the mantle will
significantly impact the performance of the crusher. These detailed analyses can be used in
design of improved liner geometries to optimize crusher behavior. Perhaps the most
valuable opportunity is to maintain crusher performance for as long as possible as the
liners wear.
For the VSI, the use of incremental breakage for predicting impact based fracture of
particles was demonstrated. Breakage not only occurs because of the direct impact of
particles on the anvils but also from secondary breakage as the moderately dense cloud of
previously broken particles that fills the crushing chamber are either struck by the
coherent outward spiral streams that extend from the shoes of the impeller or because
these particles travel back into the high speed impeller. The dependence of breakage and
31/58
of the product size distribution on particle hardness was explored with particle breakage
ranging from strong for soft ores to only modest for very hard ores where the specific
energies needed to break are near the upper end of the range of energies that the machine
can generate.
These models show the nature and level of fidelity that is now possible to include in
particle scale crusher models including breakage of non-spherical particles and prediction
of the product size distribution and throughput. However, they do not take into account –
as yet – mechanical deformation of the machine structures which is likely to occur in
heavily loaded comminution equipment and which can strongly affect gap sizes and
therefore loading of particles particularly in compression based crushers.
Although now quite advanced it is important to appreciate that there are still many
challenges remaining for such particle scale modelling of crushers. Such developments
include:
 Better linkage of material characterisation data to laboratory measurable particle

properties taking account of rock texture, variability and particle size.
 Taking account of the geometric details of the loading condition in the criteria for
when breakage occurs and in the determination of the progeny size and shape
distribution.
 Continual reduction in the resolved particle sizes used so that increasing fractions
of product can be explicitly included in the model
 Selective breakage and liberation to be included within the computational
framework
 Inclusion of air flow which for some crushers is very important to the classification
performance and also to dust generation.
 Understanding the effect of machine liner wear on life cycle process performance.
 Validation both for flow and breakage within laboratory scale crushers and for
prediction of industrial scale machines using quality integrated data sets (which
regrettably do not currently exist in the public domain and whose absence is
limiting specific validation for individual crushers).
 Application to optimization of design and wear operation of existing equipment.
 Last and definitely not least, the Virtual Comminution Machine approach offers
much more rapid development of proposed devices at a substantial reduction in
development cost as well as calendar time.
32/58
8 References
[1] A., Harrison, Base Metal Fundamentals: an overview of 2018-2019. Metal

Bulletin, Shanghai Derivatives Market Forum, (2018).
[2] R. D., La Nauze, and J., Temos, Technologies for sustainable operations, CMMI
Congress, Cairns, Australia, 27-28th May 2002, (2002),
[3] F., Musa, and R. D., Morrison, A more sustainable approach to assessing
comminution efficiency, Minerals Engineering, 22, (2009), 593-601.
[4] G. R., Ballantyne, and M. S., Powell, Benchmarking comminution energy

consumption for processing copper and gold ores, Minerals Engineering, 65, (2014), 109-
114.
[5] J. A., Curry, M. J. L., Ismay, and G. J., Jameson, Mine operating costs and the
potential impacts of energy and grinding, 56, (2014), 70-80.
[6] DOE, (2007), Mineral industry energy bandwidth study. U.S. Department of
Energy - Industrial Technologies Program.
[7] T., Napier-Munn, Is progress in energy-efficient comminution doomed?, Minerals

Engineering, 73, (2015), 1-6.
[8] W. J., Rankin, Australasian mining and metallurgical practices. the Sir Maurice
Mawby Memorial Volume, (ED), AusIMM Monograph #28, (2013).
[9] G. M. Mudd, and D.V., Boger, The ever growing case for paste and thickened
tailings – towards more sustainable mine waste management, Aus IMM Bulletin, No 2
April, 56-59, (2013).
[10] R. A., Bearman, Step change in the context of comminution, Minerals

Engineering, 43/44, (2013), 2-11.
[11] T.J., Napier-Munn, S., Morrell, R.D., Morrison, T., Kojovic, Mineral comminution
circuits: their operation and optimisation, JKMRC, 1996.
[12] R.T., Hukki, Proposal for a Solomonic settlement between the theories of von
Rittinger, Kick and Bond., Trans SME/AIME, 220, (1961), 403-408.
[13] F., Shi, R. Morrison, A. Cerellin, F. Burns and F. Musa, Comparison of energy
efficiency between ball mills and stirred mills in coarse grinding, Minerals Engineering,
22, (2009), 673-680.
33/58
[14] A. Valero, and G. M., Mudd, Exergy – A Useful Indicator for the Sustainability of
Mineral Resources and Mining, SDIMI conference, Gold Coast, Australia, 6-8 July,
(2009).
[15] J. T., Woodcock, Australasian mining and metallurgical practices. the Sir Maurice
[16] W. D., Gould, Pinto Valley Concentrator Grinding with Large Diameter Ball
Mills, SME Transactions, 260, AIME, 260, (1976), 264-274.
[17] D. B., Mazzinghy, J. F. C., Russo, J., Lichter, C. L., Schneider, J., Sepúlveda and
A., Videla, The grinding efficiency of the currently largest vertimill installation on the
world. SAG17, September, Vancouver, (2015).
[18] J. T., Woodcock, Mining and metallurgical practices in Australasia, The Sir
Maurice Memorial Volume. Monograph Series No 10. Published by the AusIMM., 1980.
[19] J. T., Woodcock, and K., Hamilton, Australasian mining and metallurgical
practices, the Sir Maurice Mawby Memorial Volume. (ED), AusIMM Monograph #19.,
(1993).
[20] W. J., Rankin, Australasian mining and metallurgical practices. the Sir Maurice
[21] B., Beke, Principles of comminution. Maclaren, London, (1964).
[22] A. J. Lynch, and C.A., Rowland, The history of grinding, Published SME,
Littleton, Colorado, (2005).
[23] P. W., Cleary, and R. D., Morrison, Comminution mechanisms, particle shape
evolution and collision energy partitioning in tumbling mills, Minerals Engineering, 86,
(2016), 75–95.
[24] P. W., Cleary, R. D., Morrison, and M., D., Sinnott, and G. W., Delaney,
Incremental damage and particle size reduction in a pilot SAG mill: DEM breakage
method extension and validation, Minerals Engineering, 128, (2018), 56-68.
[25] W. J. Rankin, Minerals, Metals and Sustainability: Meeting Future Material Needs,
Chapter 9, CSIRO PUBLISHING, (2011).
[26] R., McIvor, High pressure grinding rolls – a review. Chapter 13 of Advances in
comminution. Published SME, Denver, (1997).
34/58
[27] F., Pena, Recent developments and application of the tower mill. SME pre-print
86-380, (1986).
[28] D. B., Mazzinghy, J., Lichter, C. L., Schneider, R., Galéry, and J. F., Russo,
Vertical stirred mill scale-up and simulation: Model validation by industrial samplings
results. Minerals Engineering, 103–104, (2017), 127–133.
[29] L. S. Bbosa, I. Govender, A. N. Mainza, and M. S., Powell, Power draw

estimations in experimental tumbling mills using PEPT, Minerals Engineering, 24, (2011),
319-324.
[30] I., Govender, N., Mangesana, A. N., Mainza, and J.-P., Franzidis, Measurement of
shear rates in a laboratory tumbling mill, Minerals Engineering, 24, (2011), 225-229.
[31] P. A., Cundall and O. D. L., Strack, A discrete numerical model for granular
assemblies, Geotechnique, 29, (1979), 47-65.
[32] B. K., Mishra, and R. J., Rajamani, The discrete element method for the simulation
of ball mills, App. Math. Modelling, 16, (1992), 598-604.
[33] B. K., Mishra, and R. K., Rajamani, Simulation of charge motion in ball mills. Part
1: experimental verifications, Int. Journal Mineral Processing, 40, (1994), 171-186.
[34] R. K., Rajamani, and B. K., Mishra, Dynamics of ball and rock charge in SAG
mills, Proc. SAG 1996, University of British Columbia, (1996).
[35] P. W., Cleary, Predicting charge motion, power draw, segregation, wear and
particle breakage in ball mills using discrete element methods, Minerals Engineering, 11,
(1998), 1061-1080.
[36] P. W., Cleary, Charge behaviour and power consumption in ball mills: Sensitivity
to mill operating conditions, liner geometry and charge composition, Int. J. Min.
Processing, 63, (2001), 79-114.
[37] P. W., Cleary, Recent advances in DEM modelling of tumbling mills', Minerals
Engineering, 14, (2001), 1295-1319.
[38] P. W., Cleary, Modelling Comminution Devices using DEM, Int. J. for Numer.
Anal. Meth. Geomechan., 25, (2001),83-105.
[39] J. A., Herbst, and L., Nordell, Optimization of the design of sag mill internals
using high fidelity simulation, Proc. of the SAG conference, Eds. Barratt, Allan and
Mular, IV, 150 -164, (2001).
35/58
[40] P. W., Cleary, Large scale industrial DEM modelling, Engineering Computations,
21, (2004), 169-204.
[41] P. W., Cleary, Industrial particle flow modelling using DEM, Engineering
Computations, 26(6), (2009), 698-743.
[42] R. D., Morrison, and P. W., Cleary, Towards a Virtual Comminution Machine,
Minerals Engineering, 21, (2008), 770-781.
[43] O., Hlungwani, J., Rikhotso, H., Dong, and M. H., Moyes, Further validation of
DEM modelling of milling effects of liner profile and mill speed, Minerals Engineering,
16, (2003), 993-998.
[44] J. T., Kalala, M. M. Bwalya, and M. H., Moys, Discrete element method (DEM)
modelling of evolving mill liner profiles due to wear. Part I. DEM validation, Minerals
Engineering, 18, (2005), 1386–1391.
[45] J. T., Kalala, M.M. Bwalya, and M.H., Moys, Discrete element method (DEM)
modelling of evolving mill liner profiles due to wear. Part II. Industrial case study,
Minerals Engineering, 18, (2005), 1392–1397.
[46] N. Djordevic, Discrete element modelling of the influence of lifters on power draw
of tumbling mills, Minerals Engineering, 16, (2003), 331-336.
[47] N., Djordjevic, R. Morrison, B., Loveday, and P. W., Cleary, Modelling
comminution patterns within a pilot scale AG/SAG Mill, Minerals Engineering, 19,
(2006), 1505-1516.
[48] M.S., Powell, I. Smit, P., Radziszewski, P., Cleary, B., Rattray, K., Eriksson, L.
and Schaeffer, The Selection and Design of Mill Liners. In Advances in Comminution,
Ed. S.K. Kawatra. ISBN-13: 978-0-87335-246-8, Society for Mining, metallurgy, and
exploration, Inc., Colorado, USA. pp. 331-376, (2006).
[49] M. S., Powell, N. S. Weerasekara, S. Cole, R. D. LaRoche, and J., Favier, DEM
modelling of liner evolution and its influence on grinding rate in ball mills, Minerals
Engineering, 24, (2011), 341-351.
[50] R. M. de Carvalho and L. M. Tavares, Predicting the effect of operating and design
variables on breakage rates using the mechanistic ball mill model, Minerals Engineering,
43-44, 2013, 91-101.
36/58
[51] N. S., Weerasekara, M. S., Powell, P. W., Cleary, L. M., Tavares, M., Evertsson,
R. D., Morrison, J., Quist, and R. M. Carvalho, The contribution of DEM to the science of
comminution, Powder Technology, 248, (2013), 3-24.
[52] L. M., Tavares, A Review of Advanced Ball Mill Modelling, KONA Powder and
Particle Journal, 34, (2017), 106-124.
[53] J. A. Herbst and A. V. Potapov, Making a Discrete Grain Breakage model practical
for comminution equipment performance simulation, Powder Technology, 143-144,
(2004), 144-150.
[54] J., Quist. and C. M., Evertsson, Cone crusher modelling and simulation using
DEM, Minerals Engineering, 85, (2016), 92-105.
[55] D., Legendre, and R. Zevenhoven, Assessing the energy efficiency of a jaw
crusher, Energy, 74, (2014), 119-130.
[56] A., Refahi, J., Aghazadeh Mohandesi, and B. Rezai, Discrete element modeling for
predicting breakage behavior and fracture energy of a single particle in a jaw crusher,
International Journal of Mineral Processing, 94, (2010), 83-91.
[57] N., Djordjevic, F. N., Shi, and R. D., Morrison, Applying discrete element
modelling to vertical and horizontal shaft impact crushers, Minerals Engineering, 16,
(2003), 983-991.
[58] A. V., Potapov, and C. S., Campbell, A three-dimensional simulation of brittle

solid fracture, Int. J. of Modern Physics C, 7, (1996), 717-729.
[59] J., Lichter, K., Lim, A., Potapov, and D. Kaja, New developments in cone crusher
performance optimization, Minerals Engineering, 22(7), (2009), 613-617.
[60] E. R., da Cunha, R. M., de Carvalho, and L. M., Tavares, Simulation of solids flow
and energy transfer in a vertical shaft impact crusher using DEM, Minerals Engineering,
43–44, (2013), 85–90.
[61] H., Li, G., McDowell, and I. Lowndes, Discrete element modelling of a rock cone
crusher, Powder Technology, 263, (2014), 151-158.
[62] P. W., Cleary, and M., Sinnott, Simulation Of Particle Flows And Breakage in
Crushers Using DEM: Part 1 - Compression crushers, Minerals Engineering, 74, (2015),
178-197.
37/58
[63] M., Sinnott, and P. W., Cleary, Simulation Of Particle Flows And Breakage in
Crushers Using DEM: Part 2 - Impact crushers, Minerals Engineering, 74, (2015), 163-
177.
[64] G., Delaney, P. W., Cleary, M. D., Sinnott, and R. D., Morrison, Novel application
of DEM to modelling comminution processes, Proc. 9th World Congress on
Computational Mechanics, 19th July, Sydney, IOP Conf. Series: Materials Science and
Engineering, 10, 012099, (2010).
[65] G., Delaney, and P. W., Cleary, The packing properties of superellipsoids,
Europhysics Letters, 89, 34002, (2010).
[66] G. W, Delaney, R. D, Morrison, M. D. Sinnott, and S., Cummins, and P. W,

Cleary, DEM modelling of non-spherical particle breakage and flow in an industrial scale
cone crusher, Minerals Engineering, 74, (2015), 112–122.
[67] P. W, Cleary, G. W, Delaney, M. D., Sinnott, R. D, Morrison, and S., Cummins,

Analysis of cone crusher performance with changes in material properties and operating
conditions using DEM, Minerals Engineering, 100, (2017), 49-70.
[68] X., Qiu. DEM Simulations in Mining and Mineral Processing, 7th Int. Conference
on Discrete Element Methods, Paper Number: G010111, 1st-4th August 2016, Dalian,
China, (2016).
[69] M., Johansson, J., Quist, M., Evertsson, and E., Hulthén, Cone crusher
performance evaluation using dem simulations and laboratory experiments for model
validation, Minerals Engineering, 103-104, (2017), 93-101.
[70] D. O., Potyondy, and P. A., Cundall, A bonded-particle model for rock, Int. J.
Rock Mech. Min. Sci., 41, (2004), 1329-1364.
[71] G. K. P., Barrios and L. M., Tavares, A preliminary model of high pressure roll
grinding using the discrete element method and multi-body dynamics coupling.
International Journal of Mineral Processing, 156, (2016), 32-42.
[72] P. W., Cleary, The Filling of Dragline Buckets, Mathematical Engineering in

Industry, 7, (1998), 1-24.
[73] R., Morrison, and P. W., Cleary, Using DEM To Model Ore Breakage Within A
Pilot Scale Sag Mill, Minerals Engineering, 17, (2004), 1117-1124.
38/58
[74] P. W., Cleary, A. N., Stokes, and J., Hurley, Efficient collision detection for three
dimensional super-ellipsoidal particles, Proc. 8th Int. Computational Techniques and
Applications Conference, pp. 139-144, 29th Sept – 1st Oct, Adelaide, Australia, (1997).
[75] P. W., Cleary, G., Metcalfe, and K., Liffman, How well do discrete element
granular flow models capture the essentials of mixing and segregation processes?, Applied
Mathematical Modelling, 22, (1998), 995-1008.
[76] P. W., Cleary, R., Morrison, and S., Morrell, Comparison of DEM and experiment
for a scale model SAG mill, Int. J. Min. Processing, 68, (2003), 129-165.
[77] J. E., Hilton., L., Mason, and P. W., Cleary, Dynamics of gas–solid fluidised beds
with non-spherical particle geometry, Chemical Engineering Science, 65, (2010), 1584-
1596.
[78] M. A., Hopkins, W. D., Hibler, and G. M., Flato, On the numerical simulation of
the sea ice ridging process, Journal of Geophysical Research, 96, (1991), 4809–4820.
[79] M. A., Hopkins, Onshore ice pile-up: a comparison between experiments and
simulations, Cold Regions Science and Technology, 26, (1997), 205-214.
[80] P. J., Owen, P. W., Cleary, and C., Meriaux, Quasi-static fall of planar granular
columns: comparison of 2D and 3D discrete element modelling with experiment,
Geomechanics and Geoengineering: An International Journal, 4, (2009), 55-77.
[81] W., McBride, and P. W., Cleary, An investigation and optimization of the 'OLDS'
elevator using Discrete Element Modeling, Powder Technology, 193, (2009), 216-234.
[82] R., Das, and P. W., Cleary, Effect of rock shapes on brittle fracture using
Smoothed Particle Hydrodynamics, J. Theoretical and Applied Fracture Mechanics, 53,
(2010), 47-60.
[83] R., Das, and P. W., Cleary, Application of a mesh-free method to modelling brittle
fracture and fragmentation of a concrete column during projectile impact, Computers and
Concrete, 16, (2015), 933-962.
[84] C., Thornton, S., Cummins, and P. W., Cleary, An investigation of the comparative
behaviour of alternative contact force models during inelastic collisions, Powder
Technology, 233, (2013), 30-46.
[85] P. W, Cleary, Discrete Element Method, US Patent 9,633,142, (2017),
39/58
[86] L., Vogel, and W., Peukert, Determination of Material Properties relevant to
grinding by practicable labscale milling tests, International Journal of Minerals
Processing, 74S: S329-S338, (2004).
[87] R. D., Morrison, F., Shi and R., Whyte, Modelling of incremental rock breakage
by impact - For use in DEM models, Minerals Engineering, 20, (2007), 303-309.
[88] S. S., Narayan, and W. J., Whiten, Breakage characteristics for ores in ball mill
modelling, Proc. AusIMM 286, (1983), 31-39.
[89] P. W., Cleary, D., Thomas, L., Hetherton, M., Bolger, J. E., Hilton, and D.,
Watkins, Workspace: a workflow platform for supporting development and deployment of
modelling and simulation, Mathematics and Computers in Simulation, 175, (2020), 25-61.
[90] P. W., Cleary, and D., Hoyer, Centrifugal mill charge motion: comparison of DEM
predictions with experiment, Int. J. Min. Proc., 59, (2000), 131-148.
[91] I., Govender, P. W., Cleary, and A., Mainza, Comparisons of PEPT derived charge
features in wet milling environments with a friction-adjusted DEM model, Chemical
Engineering Science, 97, (2013), 162-175.
[92] M., Bengtsson, and M., Evertsson, Modelling of output and power consumption in
vertical shaft impact crushers, Int. J. Miner. Process., 88, (2008), 18–23.
[93] P. W., Cleary, and R. D., Morrison, A more general approach to measurement and
analysis of cone crusher geometry, to appear: Minerals Engineering, (2020).
[94] J. S. Andersen, and T. J., Napier-Munn, The influence of liner condition on cone
crusher performance, Minerals Engineering, 3, (1990), 105-116.
[95] J. S., Andersen, Development of a cone crusher model, University of Queensland,

(1988).
40/58
Tables
Diameter of rolls 450 mm

Length of rolls 500 mm
Height of primary (p) roll teeth above roll surface 75 mm
Width of primary roll teeth 70 mm
Height of secondary (s) roll teeth above roll surface 75 mm
Width of secondary roll teeth 50 mm
Height of tertiary (t) roll teeth above roll surface 35 mm
Width of tertiary roll teeth 30 mm
Number of roll teeth 9 (axial) x 12 (angular)
Minimum distance between the teeth of the two rolls 80 mm
Feed top size 300 mm
Angular speed of roll 60 rpm
Product top size 150 mm
Table 1: Geometric and operating specification for a double roll crusher.
Concave height 1.538 m

Concave internal diameter (top) 1.79 m
Concave internal diameter (bottom) 2.47 m
Mantle diameter (at top) 0.505 m
Mantle diameter (at bottom) 2.134 m
Mantle height 1.642 m
Mantle side angle (from horizontal) 50 deg
Feed top size 38 mm
Closed side setting 11 mm
Open side setting 58 mm
Inclination of mantle for nutating motion 1.06 deg
Angular speed of mantle 255 rpm
Table 2: Geometric and operating specification for the cone crusher.
41/58
Size (mm) % Passing

38.0 100.0
25.4 86.6
19.0 72.5
13.2 45.8
9.5 11.4
6.7 0.0
Table 3: Size distribution of feed material used for the cone crusher.
Feed opening diameter 400 mm

Feed top size 70 mm
Feed rate 400 kg/s (1440 tph)
Discharge opening diameter at bottom 1650 mm
Height of mill 1850 mm
Outer diameter of mill 2500 mm
Height of crushing chamber 1000 mm
Diameter of impeller table 1200 mm
Height of table central distribution cone 70 mm
Distance from chamber roof to cone top 530 mm
Number of shoes on table 3
Height of shoe 325 mm
Angle of inclination of shoe 30 deg
Number of anvil rings 20
Angle of inclination of a single anvil 30 deg
Height of a single anvil 500 mm
Width of single anvil 350 mm
Angular speed of rotating table 750 rpm
Product top size 30 mm
Table 4: Geometric and operating specification for the VSI.
42/58
Figures
10 6
Energy Consumed (kWh/ tonne)
10 5
10 4 Conv entional
3
Ritt inger Grinding Range
10 (slope=-1. 0)
10 2
lit tle known range
10 1
Bond Kick
0 (slope=-0. 5)
10 (slope=0)
10 -1
10 -4 10 -2 10 0 10 2 10 4 10 6
Size (µm)
Figure 1: Energy cost of comminution as a function of the fineness of the ground product.
43/58
a)
b)
c)
Flowsheet symbol nomenclature
Figure 2: Common comminution circuits, a) early in the 20th century, b) larger volume
scaled up circuits typical of the 1960s and 1970s, and c) large SABC circuits from the
1990s. (The flowsheets were drawn using IES-integrated Extraction Simulator).
44/58
a)
Housing
Roll
b)
Figure 3: Double roll crusher viewed a) obliquely from above and b) from in-front. The
two crushing rolls, with alternating large and small downward directed teeth, are shaded
grey. The feed hopper and crusher shell is shown as orange. The front wall has been
omitted for visibility.
45/58
a) b)
Figure 4: Particle flow and breakage in a double roll crusher for breakage of a softer ore
(Eb = 1100 J/kg). The view is oblique and from above so that the behaviour in the
breakage region above and between the rolls can be seen. The frames are at closely spaced
times and show the fracture of a medium sized particle between and the teeth. The
particles are coloured by size with red being large (250-300 mm), green being
intermediate (around 150 mm) and blue being fine (15-30 mm). The finer breakage
product particles are dark blue.
c) d)
46/58
Figure 5: Product size distribution variation with the strength Eb of the ore.
47/58
a)
b)
Figure 6: Cone crusher with cone (mantle) shown green, bowl (concave) shown brown,
and feed table shown purple. a) entire crusher, and b) close-up of the cross-section of the
crushing region.
48/58
a)
b)
c)
Figure 7: Steady state particle flow within the cone crusher. The particles when the left
side of the cone crusher is closed coloured by: a) speed (with red being 3 m/s, green being
1.5 m/s and dark blue being stationary), b) total force on the particles (with red being 20
kN, green being 10 kN, light blue being less than 5 kN), and c) by particle size with red
being large (>30 mm), green intermediate and blue small (<10 mm).
49/58
a)
b)
Figure 8: Charge distribution for successive closed and open positions of the cone crusher.
The colour range is the same as for Figure 7a.
50/58
a)
b)
Figure 9: View of the VSI with central rotating impeller table and the peripheral anvil
ring.
51/58
a)
b)
Figure 10: Steady state particle distribution in the VSI from a shallow viewing angle at
two times coloured by speed with red being fast and dark blue being slow.
52/58
Figure 11: Steady state particle distribution in the VSI from a steeper viewing angle
coloured by particle size with red being large (> 65 mm) and dark blue being small (<8
mm). The bottom size of the feed is green. The dense high speed flow across the rotor
shoe at the front right creates a coherent stream that arcs outwards towards the anvils.
53/58
Figure 12: Particle flow in the VSI shown in a thin horizontal slice coloured by speed with
red being fast and dark blue being slow. Particles trapped or moving very slowly form a
moderate density autogenous layer around and behind the anvils.
54/58
Figure 13: Distribution of breakage events and their intensity within the VSI crushing
chamber. Each red circle shows the breakage of one particle within the previous 1 ms. The
particles are coloured by speed with the same colour bar as in Figure 12.
55/58
Figure 14: Particle size distributions shown as % passing. The green curve shows the feed
which is nearly uniformly distributed giving a nearly linear % passing between the feed
bottom size and top size. The red dashed curve shows the size distribution of particles
(both resolved and unresolved) that are within the crusher volume. The product size, as
measured by a sample plane, at the bottom of the crusher is shown as dark blue.
56/58
100
90
En = 100
80
En =300
70
En=1000
60
% passing
En=3000
50
40
30
20
10
0
0.1 1 10
Product particle size (mm)
Figure 15: Product size distribution for the VSI for three material strengths En.
60
P20
50
Product size measures (mm)
P50
P80
40
30
20
10
0
100 1000
En (J/kg)
Figure 16: Variation of the P20, P50 and P80 with material strengths En for the VSI.
57/58
1.4
1.2
1.0
Power Draw (MW)
0.8
0.6
0.4
0.2
0.0
100 1000
En (J/kg)
Figure 17: Variation of VSI power draw with material strengths En.
58/58

Libro DDD

Uploaded by

Copyright:

Available Formats

Libro DDD

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Libro DDD

Uploaded by

Copyright:

Available Formats

Journal Pre-proof

Advanced comminution modelling: part 1 – crushers

Paul W. Cleary , Gary W. Delaney , Matt D. Sinnott ,

To appear in: Applied Mathematical Modelling

Received date: 11 August 2019

© 2020 Published by Elsevier Inc.

Advanced comminution modelling: part 1 – crushers

* Corresponding author: [email protected]

1. Twin roll crusher

Comminution; DEM; breakage; crusher; flowsheet; mineral processing challenges

1.1 Comminution and industry drivers

Collectively these sources indicate that comminution is conservatively responsible for

The commercial imperatives for mineral processing operators are to:

3. Reduce operating costs, and

4. Reduce energy consumption,

Secondary trends include drivers such as

will add further to the base line demand growth.

1.2 Comminution machines as part of grinding circuits

 the lack of understanding of breakage mechanisms for particles in comminution

2 Summary of DEM method

Fn  knx  Cnvn . (2)

The tangential force is controlled by the relative tangential velocity vt using an

Ft  min  Fn ,  ktv t t  Ctv t , (3)

1. Determination of when a particle breaks

where an outcome of 1 indicates breakage. Here ξi denotes the material of particle

For impact breakage the energy absorbed by a particle in collisions above an

where En is a material constant that determines how intensively damage occurs to a

2. Determination of the severity of breakage and the progeny size distribution

3. Generation of the daughter particles

4. Packing of the daughters within the parent particle

so use of pre-computed or simplified packings creates a significant risk of biasing

5. Replacement of the parent with the packed assembly of daughter particles

This replacement process is performed as an instantaneous change to the DEM

6. Unresolved progeny and predicting fine product

3 Need and challenges in validation of industrial scale DEM

4 Double roll crusher

4.1 Geometry, model parameters and operating conditions

Breakage in this crusher is compressive so we use the compression breakage method

4.2 Flow through the crusher

4.3 Product size distribution for different Eb

5.1 Geometry, model parameters and operating conditions

5.2 Flow within the crusher

In order to better understand the detailed micromechanical interactions leading to

6 Vertical Shaft Impactor (VSI)

6.1 Geometry, model parameters and operating conditions

6.2 Steady state flow structure in the VSI

1. these particles having a non-negligible chance of surviving because the peak

6.3 Effect of material strength En on PSD

Simulation capability for comminution unit operations needs to be performed in the

Three crusher types are used to demonstrate these capabilities:

1. Twin roll crusher

 Better linkage of material characterisation data to laboratory measurable particle

[1] A., Harrison, Base Metal Fundamentals: an overview of 2018-2019. Metal

[4] G. R., Ballantyne, and M. S., Powell, Benchmarking comminution energy

[7] T., Napier-Munn, Is progress in energy-efficient comminution doomed?, Minerals

[10] R. A., Bearman, Step change in the context of comminution, Minerals

[21] B., Beke, Principles of comminution. Maclaren, London, (1964).

[29] L. S. Bbosa, I. Govender, A. N. Mainza, and M. S., Powell, Power draw

[58] A. V., Potapov, and C. S., Campbell, A three-dimensional simulation of brittle