Sunteți pe pagina 1din 9

Minerals Engineering 127 (2018) 81–89

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Quantifying variability of ore breakage by impact – Implications for SAG T


mill performance

Farhad Faramarzi , Vladimir Jokovic, Robert Morrison, Sarma S. Kanchibotla
Julius Kruttschnitt Mineral Research Centre, Sustainable Minerals Institute, The University of Queensland, 40 Isles Road, Indooroopilly, Qld 4068, Australia

A R T I C LE I N FO A B S T R A C T

Keywords: The standard JK drop weight test breaks single particles at five size fractions each at different energy levels by
Rock heterogeneity dropping a given weight from a certain height. The progenies of all particles at a given energy level are sieved
Ore intrinsic variability and the degree of breakage of all those particles at that energy level is presented as t10% (percent passing 1/10th
Breakage variability of original size). This method assumes that all of the particles in a single size fraction have the same mass and
Ore characterization
experience a similar degree of breakage and calculates the average breakage characteristics of an ore domain
t10
JKMRC’s breakage model
from impact loading. Hence, this standard approach does not capture the breakage variability that exists among
Comminution percentile curves ore particles. To investigate the breakage variability of ore, the drop weight testing method has been extended
Complex orebodies where the energy applied and the degree of breakage for each particle within the sample are measured sepa-
JKSimMet rately. Therefore, the degree of breakage for the ore domain is expressed as an envelope of t10% vs energy curves.
SAG mill This approach provides some insight into the intrinsic variability of the response to impact loading within an ore
domain. Several rock types were tested by the extended DWT testing approach and the implications of ore
breakage variability for a SAG mill treating these materials are simulated for a standard SAB circuit using the
JKMRC grinding models.

1. Introduction domain is not characterised and translated into likely variation in


process performance. This variation can cause unstable performance of
Heterogeneity is a common feature of rocks, but the degree of grinding circuits (Hart et al., 2001), inconsistent fineness of grind,
heterogeneity depends on many factors, such as alteration, oxidation, operation and optimisation difficulties, reduction in recovery (Putland,
stress history etc. as well as variations in grain size and concentrations 2006) and reduction in classification efficiency. Putland (2006) has
of mineral grains. Heterogeneity is a key factor in defining rock argued that variability of ore competency as a result of the degree of
breakage characteristics and introduces variability into the rock mass, alteration is important and has observed relatively consistent Bond ball
which can pose challenge in mining operations from blasting to work index despite variation in the ore competency. The impact of ore
grinding. This heterogeneous feature of rock translates into breakage variability on process performance was observed at the OK Tedi mine in
variability when particulates of a given rock type are broken at the Papua New Guinea, where the variations in feed ore hardness (Bond
same conditions. This implies that variable breakage behaviour among work index 5–16 kWh/t) and in feed size distributions caused the SAG
these particles might be representative of the variability within an ore mill throughput to fluctuate between 700 and 3000 t/h (Sloan et al.,
type. 2001). Additionally, Bye (2011) pointed out just how significant the
Size reduction of rocks and ores is a fundamental process in most impact of ore variability is on mine valuation, production schedule and
mining and mineral processing operations and the energy required to economic gains. In an example, he employed a geometallurgical model
break ore particles to a certain size fraction depends on the ore char- to show the impact of ore variability on the mine plan and identified
acteristics, type of size reduction process (or circuit), size and installed high-risk periods to optimise the schedule. Narayanan (1985) also
power of equipment (Bailey et al., 2009). pointed out the necessity of developing a prediction scheme for varia-
Despite the fact that breakage characteristics of ores vary, most of tion of the breakage function with hardness variability of ore for si-
the current ore testing methods do not capture this variability and plant mulation and optimisation of ball mill circuits, which are liable to treat
designs and process performance predictions are made using only ores of varying hardness.
average values. Hence, the inherent variability that exists within an ore This paper, proposes an approach to extend the standard JKMRC


Corresponding author.
E-mail address: f.faramarzi@uq.edu.au (F. Faramarzi).

https://doi.org/10.1016/j.mineng.2018.07.007
Received 1 June 2018; Accepted 17 July 2018
0892-6875/ © 2018 Elsevier Ltd. All rights reserved.
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

drop weight testing method to quantify the breakage variability of an


ore domain for impact breakage. In the course of experimental devel-
opment, we will address considerations necessary during the new test
work. The extended DWT breakage testing approach was conducted
over a broader range of input energies to determine breakage percentile
curves for the tested samples. One aim of this work is to investigate how
breakage variability may affect SAG mill performance through running
several simulations using the JKSimMet software. This approach has the
potential to characterise comminution characteristics of complex or-
ebodies and transition ore domains that by nature impose large un-
certainty on process design and performance.

2. JKMRC background – ore breakage characterisation and


comminution modelling

2.1. Impact breakage testing device

To investigate the patterns of energy utilisation in single particle


impact breakage studies, Narayanan (1985) developed and in- Fig. 2. Size distribution parameter tn versus breakage intensity t10 (Narayanan
strumented the twin pendulum apparatus which had been used by an and Whiten, 1988).
earlier study (Awachie, 1983). Later, to allow testing over a wider range
of input energies and particle sizes, a new larger twin pendulum device
was constructed to supplement the tests conducted on the smaller
pendulum used by Narayanan (Leung, 1988). However, due to some
issues including restrictions of the twin pendulum in terms of input
energy and particle size range, (as well as being time-consuming in
operation and imprecise calculation of breakage energies due to occa-
sional secondary motion of the rebound pendulum), the JK drop weight
test was developed as an alternative (Napier-Munn et al., 1996).
The original ore impact breakage test was invented by Narayanan
and Whiten (1983). Later, the JK drop weight tester (JKDWT) was
developed at JKMRC by Doug Brown and Steve Morrell (Brown, 1992)
for characterising impact breakage of ores for the purpose of AG/SAG
mill and crusher modelling (Fig. 1).
In the standard testing approach, 15 sets of particles selected at
random from five size fractions are broken at three energy levels. The
products of each set are collected and sized. Therefore, breaking 15 sets
of particles at estimated energy levels generate 15 average coordinates, Fig. 3. Illustration of typical t10%-Ecs graph from the standard approach.
expressible as t10% versus Ecs in a Cartesian plane (see Fig. 3). The
measured energies include a correction for the height of particle re- at a particular energy input per unit mass generally is higher than for
maining after impact. The average t10% value for coarser size fractions finer size fractions for a similar degree of relative breakage or t10%.
This means that more energy is required for smaller particles to produce
the same degree of breakage as coarse particles. Shi and Kojovic (2007)
developed a relationship to quantify this variation.

2.2. Breakage modelling

The degree of fragmentation or the breakage function of an ore is


described by ‘tn-family’ of curves. A detailed description of this ap-
proach is given by Narayanan and Whiten (Narayanan, 1985;
Narayanan and Whiten, 1988). This approach formed the basis for the
later standard JK drop weight test. In this approach, product size dis-
tribution generated at different levels of input energies is described by
relating a number of parameters called ‘tn-family’ parameters taken
from the size distribution curves. The tn-family represents the fraction
smaller than 1/nth of the parent size where common values for ‘n’ are 2,
4, 10, 25, 50 and 75. From the correlations between the t10% and tn-
family points on each vertical line, a complete size distribution can be
generated using spline regression analysis (Fig. 2). Among these, t10 is
employed as a characteristic size reduction, which is defined as the
fraction of the progeny that is smaller than 1/10th of the original size
fraction. Each progeny size distribution has a unique value of t10% and
all ‘tn’ values are related to that individual distribution (Napier-Munn
Fig. 1. A schematic of JKMRC drop weight tester. et al., 1996).

82
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

The relationship between the specific comminution (Ecs) and


breakage intensity (t10%) is described as a power-law relationship
(Napier-Munn et al., 1996) shown in Eq. (1).

t10(%) = A (1−e−bEcs ) (1)

where

t10 is the percentage of progenies that are smaller than 1/10th of


original size
Ecs is the specific comminution energy in kWh/t
A is limiting value of t10%
b parameter defines the steepness of t10%– Ecs curve

The A × b value known as impact parameter is the slope of the


curve at zero input energy and can be used to characterise the amen-
ability of the material to fragmentation by double sided impact (Napier-
Munn et al., 1996). It has to be noted that the ‘tn-family’ approach sa-
tisfactorily agrees with the experimental data and has been adopted by
various researchers in model development (Kapur et al., 1997; King and
Bourgeois, 1993; Pauw and Maré, 1988). The breakage map (Fig. 2) is
consistent across most brittle ores and the standard map is used in many
models.
The JKDWT approach is used by industry as one of the standard
approaches to design, operate and optimise SAG mill circuits. The JK
breakage model does not account for the particle size effect on the
breakage intensity ‘t10%’, and describes all particle sizes by one set of A
and b parameters. However, Banini (2000) developed a breakage model
incorporating the effect of particle size by conducting detailed drop
weight tests on eight mineralised ore types. In a different research, Shi
and Kojovic (2007) modified the original JKMRC breakage model to
accommodate size-dependency between applied energy and breakage Fig. 4. Standard JKDWT breakage testing approach.
intensity. Shi (2016) argued that the advantage of using the size-de-
pendant model over the JKMRC prior art breakage model (Eq. (1)) is domain from impact loading. The standard approach fits the breakage
that the effect of particle size on the breakage response can be directly model to the average t10% values derived from sizing the whole product
quantified. after breaking a group of particles at an average Ecs for a given size
The accuracy and precision of the A × b value of the JKDWT are of fraction, height and weight. An illustration of standard JKDWT ap-
prime importance as the impact parameter is a key input in selection proach is given in the figure below.
and design of AG/SAG mills. Furthermore, plant optimisation exercises The breakage characteristic of an ore in the standard approach is
also rely on the accuracy of A × b value to improve the performance of based on 15 averaged coordinates on the plot, namely (t10%, Ecs) and
an existing grinding circuit (Stark et al., 2008). A set of 24 JKDWT thus, disregards different sources of errors and intrinsic variability with
conducted on a homogeneous material to statistically study the source an ore domain. The potential sources of errors and variations in the
of variation in JKDWT (Stark et al., 2008). Results suggested a standard standard approach are due to the following reasons:
deviation of 5.7% for the A × b values. The question of how this
random error solely arisen from laboratory measurement of A × b af- 1. Variations in drop height and weight, which are considered
fects the predicted throughput and specific power of a comminution minimal.
circuit was investigated by a simulation exercise. A comminution pro- 2. Residual height (i.e. the height that the drop weight has gone down
cess simulation involving 100 simulations were conducted with the for each particle after breakage). In the standard approach, residual
JKSimMet software, which used 100 correctly distributed, random va- height is measured for only ten particles for each size fraction and
lues of ‘A’ and ‘b’ for a particular mill configuration. With 95% con- the mean residual height from these ten particles is used to estimate
fidence, results showed ± 7.4% variance in throughput which is the actual input energy for each size fraction.
variability that can be expected from random variation of A × b values. 3. Variation in particle mass: Particles within a certain size fraction do
Stark et al. (2008) pointed out real differences in ore characteristics not have the same mass, because there is a variation in size (and
(which can be referred as ore breakage variability) as a major source for volume) and density between the particles. The variation due to
additional variation in practice. density is an intrinsic property that depends on their mineralogical
composition. The standard approach uses the average mass for each
size fraction and therefore calculates the average specific energy
2.3. Limitations of the standard JKDWT testing approach
based on the drop weight and height.
4. Determination of t10%: In the standard approach, all the particles
The standard JK drop weight test breaks single particles at five size
within a size fraction are broken at a given input energy and then
fractions each at different levels of input energies by dropping a given
are combined and sieved to determine the average degree of
weight from a certain height. The progenies of all particles at a given
breakage (expressed as tn-curves). Therefore, this approach ignores
energy level are sieved and the breakage intensity of all those particles
the breakage variability among the particles within the same size
at that energy level is presented as t10% (percent passing 1/10th of
range.
original size) (Fig. 4). This approach assumes that all of the particles in
5. The other limitations include the DWT device itself, mechanical
a single size fraction have the same mass and similar degree of frag-
breakage limitation (in particular at higher levels of input energies)
mentation and calculates the average breakage characteristics of an ore

83
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

plot rather than a single line of mean values.


It has to be noted that the proposed approach calculates the specific
comminution energy differently than the standard JKDWT approach.

• In the standard JKDWT approach, Ecs Standard is defined as:


Mg (h−hR )
EcsStandard =
mp (3)
where hR represents an average of 10 measurements from breaking the
initial 10 particles and mp is the mean mass of particles as:
1 N
mp =
N
∑i =1 mpi (4)
th
where N is number of particles and mpi is the mass for the i particle.
Therefore, by substituting the Eq. (4) in Eq. (3), the Ecs is:
NMg (h−hR )
EcsStandard = N
∑i = 1 mpi (5)

• In the extended DWT approach, the specific comminution energy for


each particle (Ecsi ) is calculated using Eq. (2) and the average value
is used:
1 N
EcsExtended =
N
∑i =1 Ecsi (6)
Substituting Eq. (2) in Eq. (6) gives:
1 N Mg (h−hRi )
EcsExtended =
N
∑i =1 mpi (7)
By comparing Eqs. (5) and (7), it could be concluded that the values
for EcsStandard and EcsExtended are the same only in case when all particles
have the same mass and the residual height, see Eq. (8).
Fig. 5. Extended DWT breakage testing approach. N 1 N 1
N
=
N
∑i =1 mpi
if mpi = mp and hRi = hR forevery i
∑i = 1 mpi (8)
and particle shape irregularity.
As the particles’ masses are not the same, the values of EcsStandard and
EcsExtended are slightly different as shown later in this text.
The above limitations of standard JKDWT test masks the breakage
Overall, in the standard JKDWT, the Ecs value is calculated based on
variability of ore, hence the extended DWT testing approach has been
the average mass of a group of particles, whereas in the extended DWT
developed.
approach, it is average of actual Ecs values of the individual particles.

3. Extended breakage testing approach 4. Results

The extended single particle impact breakage testing approach is 4.1. Repeatability experiments
based on measuring the breakage behaviour of each particle within
each size range and energy input (Fig. 5). The extended DWT testing approach was conducted on four rock
In the extended DWT, the following measurements are made for types. For each rock type, 90 particles (30 particles per energy level)
each particle: were broken at three levels of energy such as 0.1, 0.25 and 0.5 kWh/t.
The test was repeated again with another set of samples from the same
• The mass of each particle within a size range is measured. rocks to test the repeatability of the approach. The results shown in
• Particles within each size range were broken individually in a range Fig. 6, clearly indicate that the envelope of repeat test fairly overlap the
of input energies. original data.
• The residual height after breakage is measured for each particle, Analysis of variance (ANOVA) between the two tests showed that at
thus the specific energy for each particle is calculated by using Eq. 95% confidence level, there is no difference between the two for all the
(2). rock types; hence, it is assumed that the new testing approach is re-
Mg (h−hRi ) peatable.
Ecsi =
mpi (2)
4.2. Comminution percentile curves
where M is mass of the drop weight, g is gravitational constant, h is drop
height, hRi is the residual height after impact and mpi the mass of par- The extended DWT was conducted on four rock types for a wider
ticle. range of input energies to determine the breakage percentile curves.
The progenies for each particle are sieved on a set of sieves whose The means for t10% and Ecs values are given in Fig. 7. The horizontal
apertures were chosen equivalent to t2, t5, t10 and t50 criteria. Note that error bars represent mass-generated variation and the vertical error bar
this approach still references the degree of breakage against the geo- reports the uncertainty of the t10% measurements. In these tests, 180
metric mean of the size fraction. The breakage intensity of ore is then particles of size fraction 22.4 + 19 mm were broken at six input energy
expressed as an envelope of t10% values vs specific comminution energy levels of 0.5, 1.0, 2.0, 2.5, 3.0 and 4.0 kWh/t within the new testing

84
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

Fig. 6. Variability in breakage behaviour of particles captured by the extended DWT.

framework except for Rock Type-1, which used 150 particles, broken at existence of actual competence differences among the particles for all
five input energies with a maximum amount of 3.0 kWh/t. The results the characterised rock types. If the latter is the dominant effect, this
shown in Fig. 7 clearly indicate significant scatter for all the rock types. variability may provide a way to represent breakage variability.
This scatter might indicate variability in the testing process or the The JKMRC model (Eq. (1)) is fitted to the data to characterise

Fig. 7. 95% confidence interval for the population means for four rock samples.

85
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

Table 1
Comminution percentile curves of 75% and 25% for size fraction 22.4 + 19mm.
Description A × b values Mean Range Relative % Range
A×b (100 × Range/Mean)
75% 25%

Rock Type-1 35.5 27.0 30.6 8.5 28


Rock Type-2 36.9 25.5 30.1 11.4 38
Rock Type-3 44.0 23.3 30.2 20.7 69
Rock Type-4 93.4 57.1 71.3 36.3 51

Table 2
Comminution percentile curves of 93.75% and 6.25% for size fraction
22.4 + 19mm.
Description A × b values Mean Range Relative % Range
A×b (100 × Range/Mean)
93.75% 6.25%

Rock Type-1 42.7 22.5 30.6 20.2 66 Fig. 10. Comminution percentile curves for Rock Type-3 (22.4 + 19 mm size
Rock Type-2 43.3 20.0 30.1 23.3 77 interval broken at 0.5, 1.0, 2.0, 2.5, 3.0 and 4.0 kWh/t input energies).
Rock Type-3 76.3 16.9 30.2 59.3 196
Rock Type-4 128.9 47.4 71.3 81.5 114

Fig. 11. Comminution percentile curves for Rock Type-4 (22.4 + 19 mm size
interval broken at 0.5, 1.0, 2.0, 2.5, 3.0 and 4.0 kWh/t input energies).
Fig. 8. Comminution percentile curves for Rock Type-1 (22.4 + 19 mm size
interval broken at 0.5, 1.0, 2.0, 2.5, and 3.0 kWh/t input energies).
breakage model was fitted to these subsets separately to obtain 75%
and 25% curves (Table 1). The same procedure was repeated until the
model parameters for 93.75% and 6.25% curves were obtained
(Table 2). Therefore, the best fitting definition for the JKMRC model
(Eq. (1)) splits the data. Figs. 8–11 show the 6.25%, 50% and 93.75%
percentile fitted curves and the estimates of A × b values for these
curves are presented in Table 2. The breakage percentile envelope is
defined between 93.75% and 6.25% curves. These results suggest that a
distribution of A × b values (expressible in terms of percentile curves)
rather than a mean value, might provide a useful description of ore
breakage variability.
Stark et al. (2008) established the variance in A × b due to each
stage of the JK drop weight test process, comprising crushing, sieving,
mixing, test and operator error and showed that the estimated A × b
value could take a standard deviation of 5.7%. They pointed out real
differences in ore characteristics as a major source for additional var-
iation in practice. Referring to Table 2, the extent of A × b variation
captured by the extended DWT approach is far beyond the overall error
in a JK drop weight test. Hence, it is assumed that competence differ-
Fig. 9. Comminution percentile curves for Rock Type-2 (22.4 + 19 mm size
ences among the particles is the dominant source of the measured
interval broken at 0.5, 1.0, 2.0, 2.5, 3.0 and 4.0 kWh/t input energies).
variation in the extended DWT. The results presented in Table 2 in-
dicate that the first three samples have similar mean A × b values.
breakability of each rock type over a range of input energies on According to the relative percent ranges (100 × Range/Mean), Rock
average. These parameters (the 50% fitted curve) were used to split Type-3 exhibits more variance in breakage behaviour than the other
data points into two subsets. Minitab software (Minitab, 2014) was used samples. In the case of Rock Type-3, A × b value changes between 17
to split the data points along the 50% fitted curve. In next step, the and 76 (Table 2). That means, there is 6.25% chance of treating some

86
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

Fig. 12. Product size distributions at different levels of input energies.

very competent materials with A × b < 17. This material will accu- decreases the steepness ‘A × b’.
mulate in the SAG mill load and may limit throughput. However, this The JKMRC breakage model fitted to the data with EcsStandard and
may be a quite normal condition and without a database of ore varia- EcsExtended values, separately (Table 3). For Rock Types 1 and 2, the
tion, no useful conclusions can be made. On the other hand, there is A × b value decreased down to 30, for Rock Type-3 down to 32 and for
6.25% chance of treating weaker materials with A × b > 76 which Rock Type-4 down to 73. As the extended DWT approach accounts for
might be beneficial in terms of throughput increase but, as noted above, mass of individual particles, therefore the Ecs value can be estimated
may be the normal situation for processing an ore. with a higher resolution compared to the standard approach.
The product size distributions reflect the expected behaviour re-
garding limitations inherent in mechanical breakage of rock at high
levels of input energies that restricts achieving very high reduction 5. Implications for SAG mill performance
ratios in a single impact (Fig. 12). As shown in Fig. 12, Rock Type-3
shows more amenability to produce fines when compared with Rock As a demonstration of the potential of this approach, the breakage
Types 1 and 2, which are on average as competent as Rock Type-3. Rock curves for the four rock types from the extended DWT tests were used to
Type-4 is the weakest as the product at initial input energies is finer. understand the maximum potential impact of breakage variability on
However, from the size distribution graphs, as the input energy in- the SAG mill performance. In practice, the expected variations for
creases, the product size distribution reaches and limits to an identical mixed distributions should be much smaller. The simulations were
extent for all of the tested samples. conducted using the JKMRC SAG mill model (Morrell and Morrison,
The extended DWT approach measures variance in resistance to 1989). The circuit simulated is shown schematically in Fig. 13.
breakage within each size fraction of an ore sample. If – as this initial As the SAG model in JKSimMet does not take account of the var-
suite of ore tests suggests – that the degree of variance itself varies iation in breakage rates at different mill loads, the SAG mill load for
substantially between ore types, these results may provide some insight different A and b values (Table 5) is kept the same as in the base case
into the intrinsic variability in terms of A × b value of an ore type. This (Table 4). Simulated throughput was compared with the base case and
information could be used to estimate implications of ore breakage the results are presented in Tables 6 and 7).
variability for a SAG mill performance throughout simulations. The amount of variation in the SAG throughputs in tonnage com-
pared to the base case is presented in Table 6 and the percent of these
changes is given in Table 7. Referring to Table 7, for the extremities, the
4.3. EcsStandard vs. EcsExtended impact of ore competence variability on the SAG throughput is notable
for all the rock types. In case of Rock Types 3 and 4 (the most variable
As explained (refer to Section 3), in the JK standard method, the Ecs ore types), a throughput fluctuation of –32.5% to +71.5% and 22.5%
value is calculated based on the mean mass of a group of particles, to 40% is expectable, respectively, which generates from the ore in-
whereas in the extended DWT approach, it is a mean of measured Ecs trinsic variability. On the contrary, for the less heterogeneous rock,
values of the individual particles. With a 95% confidence level, there is namely Rock Type-1, throughput variation for the extremities is limited
not a significant difference between the Ecs values obtained from the between −16.1% and 20.3%, however still pronounced.
two methods. However, as presented in Table 3, difference between the These simulations estimate the potential impact on SAG mill
two averaging methods can have a minor influence on the model-fitting throughput if all of the SAG feed was similar to the marker points on the
outcomes. The reason is that in the extended DWT approach, the fitted measured particle competence distributions.
curve slightly shifts to the right because of rather higher Ecs values that

87
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

Table 3
A comparison of Ecs values from the standard JKDWT and the extended DWT approaches and the potential impact on A × b value.
Rock Type-1 Rock Type-2 Rock Type-3 Rock Type-4

EcsStandard (kWh/t) Ecs Extended (kWh/t) EcsStandard (kWh/t) Ecs Extended (kWh/t) EcsStandard (kWh/t) Ecs Extended (kWh/t) Ecs Extended (kWh/t) EcsExtended (kWh/t)

0.506 0.545 0.509 0.533 0.508 0.559 0.508 0.553


1.004 1.073 1.013 1.089 1.011 1.090 1.007 1.102
2.037 2.185 2.026 2.172 2.038 2.247 2.027 2.308
2.511 2.641 2.518 2.702 2.517 2.681 2.511 2.676
3.028 3.171 3.026 3.157 3.023 3.181 3.023 3.429
N/A N/A 4.013 4.240 4.022 4.269 4.020 4.436

A × b = 33 A × b = 31 A × b = 33 A × b = 30 A × b = 35 A × b = 32 A × b = 79 A × b = 73

6. Discussion and conclusions Table 4


Parameters used for base case simulations.
The intrinsic heterogeneity within an ore domain affects the per- Description ta* A b BBWi (kWh/t) Density (t/m3)
formance of SAG mill. Currently, this risk is not very well understood
and the extended DWT provides some insight into the intrinsic ore Rock Type-1 0.13 100 0.31 15.4 2.86
variability and its impact on SAG mill performance. Rock Type-2 0.23 70.14 0.43 19.7 2.74
Rock Type-3 0.30 60.30 0.50 12.7 2.70
The extended DWT approach described the breakage variability as Rock Type-4 0.91 55.68 1.28 11.9 2.55
an envelope of A × b values for four different rock types. Based on the
simulation results, the potential impact of ore competence variability * Abrasion ore parameter.
upon the SAG throughput is notable. Results show that for the 6.25%
and 93.5% fitted curves, the predicted SAG throughput may decrease particle is an area of ongoing work.
by 32.5% and 22.5% for the Rock Types 3 and 4, respectively. On the The main conclusions made are summarized as follows:
other side, the soft end of these rock types can achieve an increase in
throughput up to almost 71.5% and 40%, respectively. The same dis-
cussion is valid for a narrower envelope for each rock type with 25%
• The extended DWT testing approach provides a better quantitative
insight into the intrinsic variability of an ore type, which can be
and 75% fitted curves. applied to the particle size fractions used in the standard JK drop
As demonstrated through simulations, the ore intrinsic variability weight test.
not only imposes penalties (the most competent particles below the
average curve) but also offers opportunities (the weakest particles
• As a database of these variances is accumulated, it may be that the
magnitude of such variances can be used to enhance the reliability
above the average curve). It has to be noted that the very competent- of prediction of comminution equipment performance.
end generally is a major driver in plant design and its performance and
has a pronounced impact on the project viability. Therefore, an in-
• Based on simulation results, the additional information about the
breakage variability might be used to assess its impact on the
creased knowledge of ore breakage variability can help in handling the grinding performance, but this work is still in early stage. A t10% is
associated risks within the comminution circuits. still used to express intensity of breakage and rock competence is
In the JK standard method, the Ecs value is calculated based on the described in terms of a proxy, namely A × b which is not a profi-
average mass of a group of particles, whereas in the extended DWT cient representative of ore competence.
approach, it is average of actual Ecs values of the individual particles.
The extended DWT provides an estimate of the intensity of breakage Work in progress:
and Ecs per particle, but the ‘t10 size’ used to express intensity of
breakage is based on the root mean square of the upper and lower
screen sizes. Developing an estimate of an equivalent size for each
• The extended DWT allows the progeny of the particles in each
percentile group to be tested further. This work is in progress and

Fig. 13. An example of comminution circuit used in simulations within the JKSimMet software.

88
F. Faramarzi et al. Minerals Engineering 127 (2018) 81–89

Table 5
Parameters that were used in SAB circuit simulation.
Description 93.75% Fitted Curve 75% Fitted Curve 25% Fitted Curve 6.25% Fitted Curve

A b A b A b A b

Rock Type-1 100 0.43 98.67 0.36 100 0.27 100 0.23
Rock Type-2 78.50 0.55 72.34 0.51 68.92 0.37 75.61 0.26
Rock Type-3 66.25 1.15 60.22 0.73 61.24 0.38 69.75 0.24
Rock Type-4 66.95 1.93 60.23 1.55 51.87 1.10 45.68 1.04

Table 6 circuit design. Paper presented at the Proceedings of the 10th Mill Operatirs’
Throughput variation in tons per hour (tph) compared to the base case for Conference, Adelaide, SA.
Banini, G.A., 2000. An integrated description of rock breakage in comminution machines.
different A and b values.
(PhD Thesis), University of Queensland (JKMRC), Australia.
Description 93.75% 75% Base case 25% 6.25% Brown, D., 1992. Private comminution, and JKMRC internal reports.
Fitted Curve Fitted 50% Fitted Fitted Bye, A., 2011. Case studies demonstrating value from geometallurgy initiatives. In: Paper
presented at the GeoMet 2011-1st AusIMM International Geometallurgy Conference
Curve Fitted Curve Curve
2011.
Curve
Hart, S., Valery, W., Clements, B., Reed, M., Song, M., Dunne, R., 2001. Optimisation of
the Cadia Hill SAG mill circuit.
Rock Type-1 329.0 295.5 273.5 252.5 229.5
Kapur, P.C., Pande, D., Fuerstenau, D.W., 1997. Analysis of single-particle breakage by
Rock Type-2 307.0 280.0 247.5 223.0 188.5 impact grinding. Int. J. Mineral Process. 49 (3), 223–236. https://doi.org/10.1016/
Rock Type-3 406.5 298.5 237.0 200.0 160.0 S0301-7516(96)00008-7.
Rock Type-4 469.0 393.0 335.0 292.0 259.5 King, R.P., Bourgeois, F., 1993. A new conceptual model for ball milling.
Leung, K., 1988. An energy based, ore specific model for autogenous and semi-autogenous
grinding mills.
Minitab, I., 2014. MINITAB Release 17: Statistical Software for Windows. Minitab Inc,
Table 7
USA.
Throughput variation in percentage compared to the base case for different A Morrell, S., Morrison, R., 1989. Ore charge, ball load and material flow effects on an
and b values. energy based SAG mill model. In: Proceedings Advances in Autogenous and SAG
Technology. Vancouver, pp. 697–712.
Description 93.75% Fitted 75% Fitted 25% Fitted 6.25% Fitted Napier-Munn, T.J., Morrell, S., Morrison, R.D., Kojovic, T., 1996. Mineral Comminution
Curve Curve Curve Curve Circuits: Their Operation and Optimisation (Vol. 2): Julius Kruttschnitt Mineral
Research Centre. University of Queensland.
Rock Type-1 20.3 8.0 −7.7 −16.1 Narayanan, S., 1985. Developement of a laboratory single particle breakage technique
Rock Type-2 24.0 13.1 −9.9 –23.8 and its application to ball mill modelling and scale up. (PhD). University of
Rock Type-3 71.5 25.9 −15.6 –32.5 Queensland.
Rock Type-4 40.0 17.3 −12.8 –22.5 Narayanan, S., Whiten, W., 1988. Determination of comminution characteristics from
single-particle breakage tests and its application to ball-mill scale-up. Trans. Inst.
Min. Metall. Sect. C-Mineral Process. Extractive Metall. 97, C115–C124.
Narayanan, S., Whiten, W.J., 1983. Breakage characteristics of ores for ball mill model-
will be reported in due course. ling. In: Proceedings AusIMM, vol. 286, June, pp. 31–39.
• Applying the extended DWT approach for five size fractions, each Pauw, O.G., Maré, M.S., 1988. The determination of optimum impact-breakage routes for
an ore. Powder Technol. 54 (1), 3–13. https://doi.org/10.1016/0032-5910(88)
being broken at three levels of input energies.
80043-3.
Putland, B., 2006. Comminution Circuit Selection–Key Drivers and Circuit Limitations.
Acknowledgements Department of Mining Engineering, University of British Columbia.
Shi, F., 2016. A review of the applications of the JK size-dependent breakage model: Part
1: Ore and coal breakage characterisation. Int. J. Mineral Process. 155, 118–129.
The authors would like to express thanks to The University of https://doi.org/10.1016/j.minpro.2016.08.012.
Queensland and JKTech Pty Ltd. for sponsoring this research. The au- Shi, F., Kojovic, T., 2007. Validation of a model for impact breakage incorporating par-
thors extend their sincere appreciation to Prof. Tim Napier-Munn and ticle size effect. Int. J. Mineral Process. 82 (3), 156–163. https://doi.org/10.1016/j.
minpro.2006.09.006.
Prof. Malcolm Powell for their valuable advices on this research.
Sloan, R., Parker, S., Craven, J., Schaffer, M., 2001. Expert systems on SAG circuits: three
comparative case studies. Paper Presented at the SAG 2001. University of British
References Columbia, Vancouver.
Stark, S., Perkins, T., Napier-Munn, T., 2008. JK Drop weight parameters: a statistical
analysis of their accuracy and precision and the effect on SAG mill comminution
Awachie, S.E.A., 1983. Development of Crusher Models Using Laboratory Breakage Data. circuit simulation. In: Paper Presented at the MetPlant 2008.
(PhD). University of Queensland.
Bailey, C., Lane, G., Morrell, S., Staples, P., 2009. What can go wrong in comminution

89

S-ar putea să vă placă și