minerals
Article
Semi-Autogenous Wet Grinding Modeling with CFD-DEM
Vladislav Lvov * and Leonid Chitalov






Citation: Lvov, V.; Chitalov, L.
Semi-Autogenous Wet Grinding
Modeling with CFD-DEM. Minerals
2021, 11, 485. https://doi.org/
10.3390/min11050485
Academic Editors: Nadezhda N.
Nikolaeva, Tatyana Nikolayevna
Aleksandrova and Thomas Mütze
Received: 26 February 2021
Accepted: 27 April 2021
Published: 1 May 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Minerals 2021, 11, 485. https://doi.org/10.3390/min11050485
Mineral Processing Department, Saint Petersburg Mining University, 199106 Saint Petersburg, Russia;
opiopi@spmi.ru
* Correspondence: lvov_vv@pers.spmi.ru; Tel.: +7-(812)-328-82-85
Abstract: The paper highlights the features of constructing a model of a wet semi-autogenous
grinding mill based on the discrete element method and computational fluid dynamics. The model
was built using Rocky DEM (v. 4.4.2, ESSS, Brazil) and Ansys Fluent (v. 2020 R2, Ansys, Inc., United
States) software. A list of assumptions and boundary conditions necessary for modeling the process
of wet semi-autogenous grinding by the finite element method is presented. The created model
makes it possible to determine the energy-coarseness ratios of the semi-autogenous grinding (SAG)
process under given conditions. To create the model in Rocky DEM the following models were used:
The Linear Spring Rolling Limit rolling model, the Hysteretic Linear Spring model of the normal
interaction forces and the Linear Spring Coulomb Limit for tangential forces. When constructing
multiphase in Ansys Fluent, the Euler model was used with the primary phase in the form of a pulp
with a given viscosity and density, and secondary phases in the form of air, crushing bodies and ore
particles. The resistance of the solid phase to air and water was described by the Schiller–Naumann
model, and viscosity by the realizable k-epsilon model with a dispersed multiphase turbulence model.
The results of the work methods for material interaction coefficients determination were developed.
A method for calculating the efficiency of the semi-autogenous grinding process based on the results
of numerical simulation by the discrete element method is proposed.
Keywords: AG; SAG; FEM; DEM; CFD; Rocky DEM; Ansys Fluent; comminution; modeling; energy-
coarseness ratios; energy spectra; grinding efficiency
1. Introduction
The Finite Element Method (FEM) and computational fluid dynamics (CFD) are
branches of fundamental numerical modeling associated with the calculation of all contact
interactions that occur between conditionally indivisible objects. Applied to the analysis
of comminution processes, FEM began to be used with the development of computer
technology at the end of the 20th century [1]. This method is especially widely used to
optimize the design of comminution equipment [2,3].
The Discrete Element Method (DEM) is a numerical technique for predicting the
behavior of bulk solids. DEM is a mesh-free method and does not solve the continuum
equations of motion; this is how DEM differs from FEM. DEM modeling of tumbling
mills reveals the characteristics of the mill media motion [4], predicting the wear of lifters,
lining [5,6], grinding media [7] and energy consumption under various conditions [8,9].
However, the number of particles involved even in grinding processes largely limits the
application of this method today. With the development of computer technology and
software, the FEM finds more and more application in the field of comminution processes
modeling and begins to compete with and complement software for steady-state modeling
of comminution processes (JKSimMet, USIM PAC, etc.) [10–12].
The main advantage of fundamental modeling over empirical modeling is the abil-
ity to effectively analyze any factors affecting the comminution process. The combined
use of these two directions of modeling allows engineers, designers, and scientists to
https://www.mdpi.com/journal/minerals
Minerals 2021, 11, x FOR PEER REVIEW 2 of 17

Minerals 2021, 11, 485 2 of 17

financial and labor resources required to justify the adopted technical and technological
solutions [13–15].
reduceIn the financialdecades,
the coming and labor resources
subject to therequired to justify
preservation of thethe adoptedoftechnical
dynamics and
the develop-
technological
ment of computer solutions [13–15].FEM modeling [16,17] will increasingly displace empirical
technology,
In the coming decades, subject to the preservation of the dynamics of the develop-
modeling due to its universality.
ment of computer technology, FEM modeling [16,17] will increasingly displace empirical
The paper presents the concept of an approach to the design of a combined DEM-
modeling due to its universality.
CFD model of semi-autogenous grinding, and the processing of the results obtained. The
The paper presents the concept of an approach to the design of a combined DEM-CFD
main motivation for this work is the need to study various technical and technological
model of semi-autogenous grinding, and the processing of the results obtained. The main
parameters, the influence of which on the SAG process cannot be investigated in static
motivation for this work is the need to study various technical and technological parame-
modeling.
ters, the influence of which on the SAG process cannot be investigated in static modeling.
2. Materials
2. Materials and
and Methods
Methods
The object
The object of
of the
thestudy
studyisisaasemi-autogenous
semi-autogenousgrinding
grindingmillmillSAG
SAG3434x17
× 17 (the diameter
(the diameter
of the mill is 34 feet, the length is 17 feet), which processes charges of copper–nickel ores.
of the mill is 34 feet, the length is 17 feet), which processes charges of copper–nickel ores.
The ores
The ores that
that make
make up up the
the charge
charge are
are divided
divided into
into 33 conventional
conventional types.
types. For
For each
each type
type of
of
ore, the
ore, the strength
strength characteristics
characteristics werewere determined
determined using
using the
the drop
drop weight
weight test
test methods
methods (JK
(JK
DWT [11,17,18],
DWT [11,17,18], SMC
SMC Testing
Testing [19],
[19], Australia)
Australia) and
and F.F. Bond
Bond test
test [20]
[20]for
forball
ballgrinding.
grinding. The
The
results are presented in Table
results are presented in Table 1. 1.

Table 1.
Table 1. Strength Strength characteristics
characteristics of ore types.of ore types.

Ore Type A SCSE[21],

[21], 3
Ore Type A b Abb Ab ta ta SCSE kWh/t SG,
SG, t/m
t/m 3 BWi, kWh/t
BWi, kWh/t
kWh/t
M1 76.00 2.81 213.60 1.23 5.02 4.51 11.61
М1 76.00 2.81
M2 68.30 213.60
2.39 163.24 1.23
1.01 5.02
5.52 4.51 11.61
4.1712.34
М2 68.30 2.39
T1 69.70 163.24
0.67 46.70 1.010.43 5.52
10.05 4.17 12.34
3.5418.60
Т1 69.70 0.67 46.70 0.43 10.05 3.54 18.60
The test
The test results
results show
show that
that the
the studied
studied ores
ores M1
M1 and
and M2
M2 based
based on
on their
their strength
strength prop-
prop-
erties belong to very soft types of ores, and ore T1 to medium ones. Thatthe
erties belong to very soft types of ores, and ore T1 to medium ones. That is, is,ores
the com-
ores
posing the charge
composing are contrasting
the charge in their
are contrasting strength
in their properties.
strength properties.
The mass
The mass fractions
fractionsofofeach
eachtype
typeofofore
ore
inin
thethe charge
charge processed
processed by by
the the plant
plant in 2019
in 2019 are
are shown in Figure
shown in Figure 1. 1.

45

40
Mass fraction, %

35

30

25

20
1 2 3 4 5 6 7 8 9 10 11 12
М1 М2 Т1 Month

Figure 1.
Figure 1. The
The composition
composition of
of the
the processed
processed charge
charge by
by type
type of
of ore
ore in
in 2019.
2019.

The number of the SAG unit samplings for one one year
year showed
showed that
that the
the feed
feed size
size of
of the
the
F80 mill
F80 millvaried
variedinin
thethe range
range fromfrom 47,603
47,603 to 103,993
to 103,993 μm,forand
µm, and the for
P80the
from P80
920from 920µm.
to 5267 to
5267 The
μm.contrast in the strength properties of the processed ores and the instability of the
massThe contrast
fraction in thetype
of each strength
of oreproperties of thehave
in the charge processed ores and
a negative theoninstability
effect of the
the process of
mass fraction of each
semi-autogenous type of ore in the charge have a negative effect on the process of
grinding:
•semi-autogenous grinding: of soft types of ores occurs, leading to sliming and a decrease
In the mill, overgrinding
• In the
in theefficiency
mill, overgrinding
of subsequentof soft types of ores
beneficiation occurs,[22–24];
processes leading to sliming and a de-
• Accumulation of criticaloffraction
crease in the efficiency ore inbeneficiation
subsequent the mill is observed;
processes [22–24];
Minerals 2021, 11, x FOR PEER REVIEW 3 of 17

Minerals 2021, 11, 485 3 of 17

• Accumulation of critical fraction ore in the mill is observed;
• The instability of the composition (Figure 1) and the ore size lead to the instability of
• the size
The of the mill
instability product.
of the The mass
composition fractions
(Figure 1) andof
thevarious
ore sizetypes
lead of
to ores are selected
the instability of
based on the useful component content in them.
the size of the mill product. The mass fractions of various types of ores are selected
The aim
based on of
thethe workcomponent
useful was to create a stable
content FEM model of the wet semi-autogenous
in them.
grinding process and to quantify the efficiency
The aim of the work was to create a stable FEM of processing
model of copper–nickel ores under
the wet semi-autogenous
given conditions. The SAG 34x17 model was created by the discrete element method
grinding process and to quantify the efficiency of processing copper–nickel ores under and
computational fluid dynamics in Rocky DEM and Ansys Fluent software.
given conditions. The SAG 34 × 17 model was created by the discrete element method and
computational fluid dynamics in Rocky DEM and Ansys Fluent software.
2.1. Setting the Modeling Boundary Conditions
2.1. Setting the Modeling
DEM modeling Boundary
of bulk Conditions
materials is a special case of the FEM method. Ore particles
represented in such aofmodel
DEM modeling are indivisible
bulk materials objects,
is a special between
case the surfaces
of the FEM method.ofOrewhich inter-
particles
actions are calculated
represented by various
in such a model types of contact
are indivisible models.
objects, betweenWith
thean increase
surfaces of in the number
which interac-
of particles
tions participating
are calculated in the simulation,
by various and with
types of contact a decrease
models. in their
With an size,in
increase the total
the num-
number
berparticles
of of fixed contacts growsinexponentially.
participating the simulation, In and
this regard, DEM modeling
with a decrease in theirissize,
inevitably as-
the total
number of fixed contacts grows exponentially. In this regard, DEM modeling
sociated with the assumptions inherent in the FEM method [25,26]. In this study, the fol- is inevitably
associated with the assumptions
lowing assumptions were made: inherent in the FEM method [25,26]. In this study, the
following
• assumptions
The shape of the ore were made:is represented by polyhedrons with 10 vertices;
particles
•• The shape
minimum of the ore particles
particle is represented
size shown by polyhedrons
in the model with 10
is 5 mm, particle sizevertices;
distribution is
• The
shownminimum
in Figureparticle sizegrinding
2 and the shown inmedia
the model is 5 mm, particle
is represented by 100%size
ballsdistribution
Ø125 mm; is
• shown
Part of in
theFigure
solid 2fraction
and theisgrinding
modeledmedia is represented
as a suspension by 100%
in the balls
pulp of Ø125 density
a given mm;
• Part of the solid
and viscosity. fraction is modeled as a suspension in the pulp of a given density
and viscosity.
• The design area of the mill is represented by a linear section of the cylindrical part of
• The design area of the mill is represented by a linear section of the cylindrical part of
the drum 500 mm long, the end walls of which are solid surfaces with zero coeffi-
the drum 500 mm long, the end walls of which are solid surfaces with zero coefficients
cients of friction in relation to particles and pulp.
of friction in relation to particles and pulp.

100

80

60
Passing, %

40

20

0
0 10 20 30 40 50 60 70 80 90 100 110 120 130

Particle size, mm
Figure 2. Particle size distribution of the material included in the simulations.

2.2. Contact Models

2.2. Contact Models
The choice of
The choice ofcontact
contactinteraction
interactionmodels
models was
was dictated
dictated by by a decrease
a decrease in computa-
in the the com-
putational
tional power consumed during modeling. The following models were used in thein
power consumed during modeling. The following models were used the
DEM
DEM simulation module: particle rolling resistance model—not chosen (due to the use of
simulation module: particle rolling resistance model—not chosen (due to the use of poly-
polyhedrons instead of spheres); normal contacts model—Linear Spring Dashpot [27,28];
hedrons instead of spheres); normal contacts model—Linear Spring Dashpot [27,28]; ad-
adhesion model—disabled (due to the use of three-phase modeling, adhesion of particles
hesion model—disabled (due to the use of three-phase modeling, adhesion of particles
inside the pulp should not be observed); and tangential contact model—Linear Spring
inside the pulp should not be observed); and tangential contact model—Linear Spring
Coulomb Dashpot.
Coulomb Dashpot.
The following models were used in the CFD module: Euler’s multiphase model [29]
The following models were used in the CFD module: Euler’s multiphase model [29]
with three phases (the primary phase is pulp, the secondary phases are ore particles, balls
with three phases (the primary phase is pulp, the secondary phases are ore particles, balls
and air). In a pair of pulp–air phases, a symmetric interaction was used, in a pair of
particles–pulp and particle–air phases—the Schiller–Naumann interaction model [30]. The
orthogonal cell size of the computational grid was 60 mm, with a single cell for the entire
width of the computational zone. The laminar viscosity model was used. The air had a
 Minerals 2021, 11, 485 4 of 17

density of 1.225 kg/m3 and a viscosity of 1.789 × 10−5 kg/m·s (reference data) [31]; the
pulp had a density of 1992 kg/m3 and a viscosity of 1.3 kg/m·s (according to the results of
laboratory determination of viscosity with a disk viscometer and density by measuring
with the Marcy system). The method for calculating the combined pressure and velocity
fields was Phase Coupled SIMPLE [32]. Spatial discreteness of pressures was calculated by
the PRESTO! [33].

2.3. Calibration of Material Interaction Coefficients

To achieve the best convergence of DEM simulation results with experimental data, it is
necessary to carry out calibration tests [34,35]. Calibration tests are laboratory experiments
designed to find coefficients used in models of contact interaction of particles with each
other and with geometry surfaces in DEM simulations. These include the coefficients of
friction and restitution between material pairs “ore-ore”, “ore-steel” and “steel-steel”.
The essence of each calibration test is to conduct such a laboratory experiment, the
result of which depends to a greater extent on the desired interaction parameter and to a
lesser extent on other variables. The conditions and result of the laboratory test are recorded.
After that, the laboratory experiment is carried out numerically in the DEM model, while
maintaining all the conditions of the experiment. By varying the sought-for interaction
parameter, one finds such a value at which the result of numerical simulation corresponds
to a laboratory experiment. After that, the found value of the required parameter is
considered calibrated.
In this work, a representative sample of the investigated charge was prepared for
calibration tests in accordance with the arithmetic mean proportion of the studied types
of ores in it (mass fraction of ore M1—40%, M2—25%, T1—35%). Particles of the correct
shape from the size class −12.5 +10.0 mm of a representative sample of the investigated
charge were used. To determine the coefficients of friction, the surface of the particles used
was sprayed with water from a spray bottle until the visual effect of wetting with water
was achieved.

2.3.1. Restitution Coefficient (Ore–Steel)

This experiment consisted of alternately dropping 50 ore particles onto an inclined
nerals 2021, 11, x FOR PEER REVIEW
steel surface of a steel cube with sides of 145 mm, as shown in Figure 3. The output of the
experiment was the average distance traveled by particles from the point of impact on a
steel inclined surface to the point of impact on a horizontal surface.

Figure
Figure 3. Restitution
3. Restitution coefficient
coefficient "ore-steel" "ore-steel"
experiment conditions. experiment conditions.
Minerals 2021, 11, 485 5 of 17

The arithmetic mean rebound of the particles was 182 mm. Based on the results of
varying the restitution coefficient in the model of this experiment, duplicated in Rocky
DEM software, this rebound value corresponds to a recovery factor of 0.703.

2.3.2. Restitution Coefficient (Ore–Ore)

This experiment consisted of alternately dropping 50 ore particles onto an inclined
surface of a flat ore sample. The general configuration of the experiment is shown in
Figure 4. The output of the experiment was the arithmetic mean distance traveled by ore
particles from the point of impact on the inclined surface of a flat ore sample to the point of
Figure 3. Restitution coefficient "ore-steel" experiment conditions.
impact on the horizontal surface [36].

Figure
Figure 4. Restitution
4. Restitution coefficientcoefficient "ore-ore"
"ore-ore" experiment experiment
conditions. conditions.
The arithmetic mean rebound of the particles was 182 mm. Based on the results of
2.3.3.the
varying Friction Coefficient
restitution coefficient in (Ore–Ore)
the model of this experiment, duplicated in Rocky
DEM software, this rebound value corresponds to a recovery factor of 0.623.
The test was conducted to determine the angle of repose of the
particles
2.3.3. of the charge.
Friction Coefficient (Ore–Ore)The massif is formed after lifting from the h
The test was
vertically conducted
located to determine
plastic the angle
tube filled of repose
with of thematerial.
the test assemblage The
of themass o
particles of the charge. The massif is formed after lifting from the horizontal surface a
such that
vertically theplastic
located heighttubeof thewith
filled material column
the test material. Theinmass
theoftube is atmust
particles least
be four t
diameter
such [37].ofInthethe
that the height present
material columntest, the
in the particles’
tube mass
is at least four was
times 6856.3
the inner tubeg. The i
diameter [37]. In the present test, the particles’ mass was 6856.3 g. The inner pipe diameter
was 103 mm. The tube was lifted at a speed of approximately 2 cm/s, a
was 103 mm. The tube was lifted at a speed of approximately 2 cm/s, and the assemblage
ofof particles
particles was thenwas then
located under located
the tube. under the tube.
The arithmetic Theofarithmetic
mean angle mean a
repose was 39.1º.
39.1º. According
According to the results oftothethe results
numerical of theinnumerical
simulation simulation
Rocky DEM software, in ofRocky
this value
the angle of repose corresponds to a coefficient of static friction equal to 0.870. Figure 5
value the
illustrates of numerical
the angle of repose
(a) and full-scale corresponds
(b) of the experiment.to a coefficient of static fric
Figure 5 illustrates
To estimate the effect ofthe numerical
the dynamic friction(a) and full-scale
coefficient of the ore on(b)
theof the experime
dynamics of
material movement in a rotating cylinder, a series of numerical experiments was carried
out in Rocky DEM. It was found that the coefficient of dynamic friction did not significantly
affect the trajectories of material. In this regard, the coefficient of dynamic friction was
taken to be practically equal to the coefficient of static friction (0.855). Figure 6 shows
2 experiments from this series of experiments in which all the initial parameters were the
same, except for the coefficient of dynamic friction between the ore particles, equal to
0.870 and 0.570, respectively.

Minerals 2021, 11, 485
such that the height of the material column in the tube is at least four times the inner tube
diameter [37]. In the present test, the particles’ mass was 6856.3 g. The inner pipe diameter
was 103 mm. The tube was lifted at a speed of approximately 2 cm/s, and the assemblage
of particles was then located under the tube. The arithmetic mean angle of repose was
39.1º. According to the results of the numerical simulation in Rocky DEM software, 6 ofthis
17
value of the angle of repose corresponds to a coefficient of static friction equal to 0.870.
Figure 5 illustrates the numerical (a) and full-scale (b) of the experiment.

Minerals 2021, 11, x FOR PEER REVIEW 6 of 17

out in Rocky DEM. It was found that the coefficient of dynamic friction did not signifi-
cantly affect the trajectories of material. In this regard, the coefficient of dynamic friction
was taken to be practically equal to the coefficient of static friction (0.855). Figure 6 shows
2 experiments from this series of experiments in which all the initial parameters were the
same, except for the coefficient of dynamic friction between the ore particles, equal to 0.870
Figure
and 5. Friction
0.570, coefficient “ore-ore” measurement test. (a) Simulated experiment, (b) Laboratory
respectively.
Figure 5. Friction coefficient “ore-ore” measurement test. (a) Simulated experiment, (b) Laboratory experiment.
experiment.

To estimate the effect of the dynamic friction coefficient of the ore on the dynamics
of material movement in a rotating cylinder, a series of numerical experiments was carried

Figure6.6. Two
Figure Twoimages
images from
from aa series
series of
of numerical
numerical experiments
experiments totodetermine
determinethe
theinfluence
influenceofofthe
thedynamic
dynamicfriction
frictioncoefficient
coefficient
betweenore
between oreparticles
particleson
on the
the overall
overall dynamics
dynamics of
of the
the material
material in
in the
the rotating
rotating drum.
drum.

2.3.4. Friction
2.3.4. Friction Coefficient
Coefficient (Ore–Steel)
(Ore–Steel)
The test was carried
The test was carried out outtoto determine
determine thethe inclination
inclination angle
angle of steel
of the the steel surface
surface at
at which
which
the the particle
particle assemblage
assemblage slid off slid off particles
it. The it. The particles
were inwere in paper
a thick a thickring
paper ring in
50 mm 50height
mm
in height and 160 mm in diameter. The weight of the particles was 2454.0
and 160 mm in diameter. The weight of the particles was 2454.0 g, and the steel surface g, and the steel
surface
and the and the particle
particle assemblage
assemblage were wetted
were wetted with water.
with water.
The
The arithmetic mean sliding angle under thedescribed
arithmetic mean sliding angle under the describedconditions
conditionswas was28.0°.
28.0◦Accord-
. Accord-
ing to the results of numerical simulation, this value of the sliding
ing to the results of numerical simulation, this value of the sliding angle correspondsangle corresponds to
the static friction coefficient at rest, equal to 0.553. Figure 7 shows the numerical
to the static friction coefficient at rest, equal to 0.553. Figure 7 shows the numerical (a) (a) and
laboratory
and (b) conditions
laboratory of theof
(b) conditions experiment. By analogy
the experiment. with thewith
By analogy “ore-ore” friction coeffi-
the “ore-ore” friction
cients, the dynamic
coefficients, friction
the dynamic coefficient
friction was taken
coefficient as almost
was taken equalequal
as almost to thetostatic friction
the static co-
friction
efficient (0.545).
coefficient (0.545).

2.3.5. Friction Coefficient (Steel–Steel)

The test was performed to determine the angle of inclination of a steel surface, at which
an array of three steel 80 mm balls, fastened to each other, slid off. Prior to the experiment,
the balls and steel surface were wetted with water to achieve the friction conditions
 Figure 6. Two images from a series of numerical experiments to determine the influence of the dynamic friction coefficient
between ore particles on the overall dynamics of the material in the rotating drum.

2.3.4. Friction Coefficient (Ore–Steel)

Minerals 2021, 11, 485 The test was carried out to determine the inclination angle of the 7 ofsteel
17 surfac
which the particle assemblage slid off it. The particles were in a thick paper ring 50
in height and 160 mm in diameter. The weight of the particles was 2454.0 g, and the s
surface and the particle assemblage were wetted with water.
estimated for wet grinding. The average sliding angle obtained with 5 measurements was
The arithmetic mean sliding angle under the described conditions was 28.0°. Acc
22.4◦ . According to the results of numerical simulation of this experiment, carried out in
ing to the results of numerical simulation, this value of the sliding angle correspond
the Rocky DEM software, this value of the slip angle corresponds to a coefficient of static
the static friction coefficient at rest, equal to 0.553. Figure 7 shows the numerical (a)
friction of 0.405. Figure 8 displays the numerical (a) and full-scale (b) of the experiment
laboratory (b) conditions of the experiment. By analogy with the “ore-ore” friction co
conditions. By analogy with the “ore-ore” friction coefficients, the sliding friction coefficient
cients, the dynamic friction coefficient was taken as almost equal to the static friction
was taken as almost equal to the static friction coefficient (0.400).
efficient (0.545).
Minerals 2021, 11, x FOR PEER REVIEW 7

2.3.5. Friction Coefficient (Steel–Steel)

The test was performed to determine the angle of inclination of a steel surfac
which an array of three steel 80 mm balls, fastened to each other, slid off. Prior to
experiment, the balls and steel surface were wetted with water to achieve the friction
ditions estimated for wet grinding. The average sliding angle obtained with 5 meas
ments was 22.4°. According to the results of numerical simulation of this experiment
ried out in the Rocky DEM software, this value of the slip angle corresponds to a co
cient
Figure 7. Friction coefficient of staticexperiment
“ore-steel” friction ofconditions.
0.405. Figure 8 displays
(a) Simulated the numerical
experiment, (a) and full-scale
(b) Laboratory experiment.(b) o
Figure 7. Friction coefficient
experiment “ore-steel” experiment
conditions. conditions.
By analogy with the(a)“ore-ore”
Simulated experiment,
friction (b) Labo-
coefficients, the sliding
ratory experiment.tion coefficient was taken as almost equal to the static friction coefficient (0.400).

Figure 8. Friction coefficient “steel-steel“ experiment conditions. (a) Simulated experiment, (b) Laboratory experiment.
Figure 8. Friction coefficient “steel-steel“ experiment conditions. (a) Simulated experiment, (b) Labo-
ratory experiment.
The interaction coefficients obtained in calibration tests are presented in Table 2
The interaction coefficients obtained in calibration tests are presented in Table 2.
Table 2. Material interaction coefficients obtained from the calibration tests.

Table 2. Material interaction coefficients obtained from the calibration tests.

Restitution Friction Coefficient
Contact Pair
Coefficient Static Dynamic
Friction Coefficient
Contact Pair Ore–ore
Restitution Coefficient0.623 0.870 0.855
Ore–steel 0.703 Static 0.560 Dynamic 0.545
Ore–ore Steel–steel 0.623 0.900 1 0.870 0.405 0.855 0.400
Ore–steel 1 The restitution coefficient
0.703 0.560 0.545 to reference da
of the pair “steel-steel” is taken according
Steel–steel 0.900 1 0.405 0.400
1 The restitution coefficient of the pair “steel-steel” is Conditions
taken according to reference data.
2.4. Description of the Model
The geometry of the mill was simplified to the surface of its lining. The calcu
2.4. Description ofarea
the Model Conditions is shown in Figure 9.
with dimensions
The geometry of the mill was simplified to the surface of its lining. The calculated
area with dimensions is shown in Figure 9.
Table 3 shows the main characteristics of the semi-autogenous grinding process,
adopted for modeling. The duration of the simulation time was 20 s. The number of
particles in the model was 74 thousand.
 Ore–steel 0.703 0.560 0.545
Steel–steel 0.900 1 0.405 0.400
1 The restitution coefficient of the pair “steel-steel” is taken according to reference da

Minerals 2021, 11, 485 2.4. Description of the Model Conditions 8 of 17

The geometry of the mill was simplified to the surface of its lining. The calcu
area with dimensions is shown in Figure 9.

Minerals 2021, 11, x FOR PEER REVIEW 8 of 17

Table 3 shows the main characteristics of the semi-autogenous grinding process,

adopted for modeling. The duration of the simulation time was 20 s. The number of par-
ticles in the model was 74 thousand.
Figure 9. Section of the lining of the cylindrical part of the drum used in the simulation.
Figure Section
9. Initial
Table 3. of the
data lining of the cylindrical part of the drum used in the simulation.
for modeling.

Parameter
Table 3. Initial data for modeling. Value Unit of Measurement
Modeling area width 500 mm
Parameter Value Unit of Measurement
The degree of filling with
Modeling 11 %mm
grinding media area width 500
The degree of filling with grinding media 11 %
Balls diameter 125 mm
Balls diameter 125 mm
Total media mass
Total media mass 17.123 17.123 t t
Total filling degree
Total filling degree 24 24 %%
T1 total mass
Т1 total mass 4.375 4.375 t t
M1 total M1masstotal mass 5.000 5.000 t t
M2 total mass 3.123 t
M2 total mass 3.123 t
% Solid 65 %
% Solid
Rotational speed 65 9.4538 %rpm
Rotational speed 9.4538 rpm
3. Results
3. Results
The general dynamics of the material in the mill (a) and the distribution of the mass
The general dynamics of the material in the mill (a) and the distribution of the mass
fraction of the liquid phase in the volume (b) are shown in Figure 10.
fraction of the liquid phase in the volume (b) are shown in Figure 10.

Figure 10.
Figure 10. General
General dynamics
dynamics of
of the
the material
materialin
inthe
themill
mill(a)
(a)and
andthe
thedistribution
distributionofofthe
themass
massfraction
fraction
of the
of the liquid
liquid phase
phase (b).
(b).

The steady-state
The steady-state mode
mode ofof operation
operation of
of the
theSAG
SAGmill
millwas
wasachieved
achievedwhen
whenthetheweighted
weighted
average coordinates
average coordinates of
of all
all particles
particles and
and grinding
grinding media
mediaalong
alongthe
thevertical
verticaland
andhorizontal
horizontal
axes came
axes came toto conventionally
conventionally unchanged
unchanged values
values (±
(±0.15
0.15m)
m)(Figure
(Figure11).
11).The
Theanalysis
analysisofofthe
the
energy interactions
energy interactions of
of the
the material
material was
was carried
carried out
outfor
forthe
thesimulation
simulationtime
timeinterval
intervalfrom
from
10 to
10 to 20
20 s.
s.
Minerals 2021, 11, 485 9 of 17
Minerals
Minerals2021,
2021,11,
11,xxFOR
FORPEER
PEERREVIEW
REVIEW 99of
of17
17

44 0.5
0.5

m
X-coordinate, m

Y-coordinate, m
33 -0.5

X-coordinate,

Y-coordinate,
-0.5

22 -1.5
-1.5

11 -2.5
-2.5

00 -3.5
-3.5
00 22 44 66 88 10
10 12
12 14
14 16
16 18
18 20
20 22
22
Time,
Time,ss
X-coordinate
X-coordinate Y-coordinate
Y-coordinate
Figure
Figure11.
11.Average
Averageparticle
particlecoordinates
coordinatesduring
duringsimulation.
Figure 11. Average particle coordinates during simulation.
simulation.

Figure
Figure12
Figure 12shows
12 showsthe
shows thenet
the netpower
net powerconsumed
power consumedby
consumed bythe
by themill,
the mill,obtained
mill, obtainedfrom
obtained fromthe
from thesimulation
the simulation
simulation
results
results after
resultsafter reaching
afterreaching steady-state
reachingsteady-state mill
steady-statemill operation
milloperation
operation(after(after
(after1010 seconds).
10s).
seconds). These values
Thesewere
These values values were
werere-
recalculatedre-
calculated
calculated
in proportionin proportion
in proportion to the
to thein
to the difference difference
difference
mill feedin in mill feed
mill feed
weight, weight, considering
weight, considering
considering the simulationthe simulation
the ofsimulation
a 500 mm
of
ofaa500
drum 500 mm
width drum
mmfromdrum5180width
widthmm.fromThe5180
from 5180 mm.
mm.The
arithmetic The arithmetic
meanarithmetic
value was mean
mean value
7017 kW. was
value was7017
This valuekW.
7017 kW. This
This
of energy
value
consumption does not consider the imperfection in all units of a mill. To calculate theTo
value of
of energy
energy consumption
consumption does
does not
not consider
consider the
the imperfection
imperfection in
in all
allunits
units of
of aamill.
mill. To
real
calculate
calculate the
the real
real energy
energy consumption,
consumption, it
it is
is necessary
necessary to
to consider
consider the
the
energy consumption, it is necessary to consider the power losses in the engine (up to 8.5%), power
power losses
losses in
in the
the
engine
engine
main (up
(up to
bearings,to 8.5%),
8.5%), main
main
girth gear, bearings,
shaftgirth
bearings,
pinion girth
bearingsgear, pinion
gear,(uppinion shaft
shaft
to 2.5%) bearings
and gearbox(up
bearings (up
(2%)to 2.5%)
to[38].
2.5%) and
andit
Thus,
gearbox
gearbox
can (2%)
be assumed [38].
(2%) [38]. Thus,
thatThus, it can be assumed
it canpower
the total be assumed
consumed that the
that the total
by the power
totalmill
power consumed
consumed
drive is 8026 kW,by the
by the
whichmill drive
millisdrive
7.96%
is
is8026
8026kW,
different kW, which
the isactual
which
from is7.96%
7.96% different
different
power from
fromthe
consumption theactual
actual
of the power
power consumption
consumption
mill drive under the of the
themill
ofgiven mill drive
drive
grinding
under the given
under the given
conditions grinding
(8720 grinding
kW). conditions (8720
conditions (8720 kW). kW).

10000
10000
9000
9000
8000
8000
7000
7000
kW
power, kW

6000
6000
Net power,

5000
5000
4000
4000
3000
3000
Net

2000
2000
1000
1000
00
10
10 11
11 12
12 13
13 14
14 15
15 16
16 17
17 18
18 19
19 20
20
Time,
Time,ss
Figure
Figure12.
Figure 12.Net
12. Netdrive
drivepower
powerobtained
obtainedfrom
obtained fromsimulation.
from simulation.
simulation.

3.1.
3.1. SAG
3.1.SAG Energy
SAGEnergy Spectra
EnergySpectra
Spectra
The
The main
Themain challenge
mainchallenge
challengein inDEM
in DEMmodeling
DEM modelingisis
modeling isinterpreting
interpretingthe
interpreting thesimulation
the simulationresults
simulation resultsin
results ininterms
terms
terms
of
of the
the efficiency
efficiency of
of the
the grinding
grinding
of the efficiency of the grinding process. process.
process.
Due
Due
Dueto to the
tothe fact
thefact thatdirect
factthat
that directmodeling
direct modelingof
modeling ofthe
of theore
the oreparticles’
ore particles’breakage
particles’ breakagewith
breakage withthe
with theformation
the formation
formation
of
of fragments is an extremely resource-intensive calculation,
of fragments is an extremely resource-intensive calculation, the process is evaluated by
fragments is an extremely resource-intensive calculation, the
the process
process is is evaluated
evaluated byby
analyzing
analyzing the
the statistics
statistics of
of collisions
collisions of
of particles
particles with
with each
each other,
other, balls
balls
analyzing the statistics of collisions of particles with each other, balls and lining. The main and
and lining.
lining. The
The main
main
tool
tool
toolin in the
inthe analysis
analysisisis
theanalysis isaaaset
set
setofof so-called
ofso-called “energyspectra”
so-called“energy
“energy spectra”curves
spectra” curvesfor
curves forthe
for thestudied
the studiedgrades
studied grades
grades ofof
of
ore
ore size
oresize
sizeinin the
inthe mill.
themill.
mill.TheThe energy
Theenergy spectra
energyspectra is
spectraisisthethe distribution
thedistribution
distributionof of
ofthethe cumulative
thecumulative
cumulativespecific specific
specificpower power
power
(W/kg)
(W/kg) attributableto
(W/kg)attributable
attributable toto impacts
impacts
impacts ofof
of various
various
various specific
specific
specific energy
energy
energy (J/kg
(J/kg(J/kg
or or kWh/t)
orkWh/t)
kWh/t) over over narrow
overnarrow
narrow par-
par-
particle
ticle size size classes.
classes.
ticle size classes.
The
The obtained
Theobtained
obtainedenergyenergy spectrafor
energyspectra
spectra forthe
for thecharges
the chargesunder
charges understudy
under studyare
study areshown
are shownin
shown inFigure
in Figure13.
Figure 13.All
13. All
All
obtained
obtained curves
curves are
are divided
divided into
into size
size fractions
fractions
obtained curves are divided into size fractions and types of ores. and
and types
types ofof ores.
ores.
Minerals 2021, 11, 485 10 of 17
Minerals 2021, 11, x FOR PEER REVIEW 10 of 17

1000000

Cumulative specific power, kW/t

100000

10000

1000

100

10

1
0.00001 0.0001 0.001 0.01 0.1 1 10 100
Specific energy, kWh/t
T1 -5 T1 -20+5 T1 -50+20 T1 -100+50 T1 -115+100 T1-130+115
M1 -5 M1 -20+5 M1 -50+20 M1 -100+50 M1 -115+100 M1-130+115
M2 -5 M2 -20+5 M2 -50+20 M2 -100+50 M2 -115+100 M2-130+115

Figure 13.
Figure 13. Energy
Energy spectra
spectra obtained
obtained by
by modeling.
modeling.

Each narrow
Each narrow size fraction is characterized
characterized by by its
its own
own average
averagevalue
valueofofthe
thespecific
specific
minimum energy required
minimum energy required to to break a particle. The larger the particle, the lower the specific
The larger the particle, the lower the specific
impact energy
impact energy required
required to break it (hereinafter,
(hereinafter, referred
referred totoas
asthe
thenominal
nominalbreakage
breakageenergy
energy
of the particles).
of the particles). This is due
due to the natural fracturing of mineral raw materials:statistically,
to the natural fracturing of mineral raw materials: statistically,
large ore
large ore samples
samples show
show more cracks of varying degrees
degrees of of liberation
liberation than
thansmall
smallones.
ones.

Method
3.1.1. Methodfor Determining
for Determiningthe Nominal
the NominalEnergy of Particles
energy of ParticlesBreakageBreakage
To
To establish the facts of the destruction of particles of varioussizes,
establish the facts of the destruction of particles of various sizes,aamethodology
methodologyis
proposed
is proposed forfordetermining
determining thethe nominal
nominal particle-specific
particle-specific breakagebreakage energy
energyof the investigated
of the investi-
samples
gated samples in different size classes.
in different This value
size classes. This allows us to determine
value allows us to determinethe amount of energy
the amount of
accounted
energy accounted for by suchfor bycollisions
such collisions of particles that that
of particles leadlead to their
to theirfailure.
failure.ThisThismethod
methodis
intended
is intended primarily
primarily forfor
comparative
comparative analysis
analysis ofofSAG
SAG under
under various
variousconditions.
conditions.
The
The laboratory
laboratory equipment for the study study is is proposed
proposed to touse
useaadrop dropweight
weighttestertesterused
used
in
in the
the JK DWT DWT [17] [17]and
andSMC SMCtest test[19]
[19]methodologies.
methodologies. TheThe JK DWT
JK DWT test test methodology
methodology in-
includes a stage for determining the nature of damage in the
cludes a stage for determining the nature of damage in the low-energy region (parameter low-energy region (parameter
ttaa); however,
however, in inthis
thistest,
test,several
several particles
particles areareusedused simultaneously
simultaneously withoutwithout recording
recording the
the
numbernumber of destroyed
of destroyed ones.ones.
In the Inarea
the area
of highof high specific
specific energiesenergies
in theinJKthe DWT JK DWT test
test (pa-
(parameters
rameters A and A and b), b),
an an energy
energy of ≥of0.1≥ kWh/t
0.1 kWh/t always
always leadsleads to particle
to particle breakage.
breakage. In thisIn
this
regard,regard, the results
the results of theofJKtheDWT JK DWT test cannot
test cannot be processed
be processed in such inasuch
wayaasway as to obtain
to obtain from
from
themthem the value
the value of the of the nominal
nominal specificspecific
energy,energy,
which iswhich is the minimum
the minimum necessarynecessary
to destroyto
destroy
particlesparticles of a size.
of a certain certain size.case,
In this In this
thecase,
dropthe droptester
weight weight is tester is a ready-made
a ready-made tool
tool for de-
for determining
termining this value. this value.
The
The study
studyassumes
assumesthe theselection
selectionofofparticles
particlesofofnarrow
narrowsize size fractions,
fractions, which
which can bebe
can ob-
tained
obtained from intermediate
from intermediate sizesize
fractions when
fractions sampling
when for the
sampling forJKtheDWTJK DWTtest (− 75.0
test + 63.0
(−75.0 mm,
+ 63.0
−mm,53.0−53.0
+ 45.0 mm,mm,
+ 45.0 −37.5 + 31.5
−37.5 + 31.5mm mm andand−26.5
−26.5++22.422.4 mm).
mm). The average average particle
particlesizessizesinin
these
these setssets are
are 69.0,
69.0, 49.0,
49.0, 34.5 and 24.5 mm, respectively.respectively. ItIt has has been
beencalculated
calculatedthat thatthethe
masses
masses of of particles
particles in these size fractions will make
fractions will make itit possible
possible to to achieve
achievethe theminimum
minimum
specific
specificimpact
impact energy
energyfromfrom 0.005
0.005 kWh/t
kWh/t forfor 24.5
24.5 mmmm andandfrom from0.0005
0.0005kWh/t kWh/tfor for69.0
69.0mm,
mm,
provided that the
provided that the light DW light DW base (LB, 2.7 kg) is dropped at a height of 50 mm.
at a height of 50 mm. Such a short Such a short
drop
dropheight
heightwill willnotnotlead
leadtotothe thebreakage
breakage of of
particles,
particles,i.e.,i.e.,
will be be
will lower
lower than their
than minimum
their mini-
specific
mum specificenergyenergy of breakage.
of breakage. By gradually
By gradually increasing
increasing the thedrop dropheight,
height, it is necessary
it is necessaryto
achieve
to achieve particle
particlebreakage.
breakage.
Figure
Figure 14 14 shows
shows a conditional probability graph
conditional probability graph ofof particle
particlebreakage
breakagedepending
dependingon on
the
the value
value of of the
the specific
specific impact energy.
MineralsMinerals
2021, 2021,
11, x11,
FOR485 PEER REVIEW 11 of 17

Figure
Figure 14. 14.
The The dependence
dependence of the probability
of the probability of breakage of of breakage
particles of particles
on specific on specific
impact energy. im
To analyze the curves of energy spectra, it is necessary to determine the nominal
specificTo analyze
impact energythe curves
0.95
Ecs ( j)
, at which of energy
a particlespectra,
of a certain it is
sizenecessary to determine
will be broken with a th
.
cific impact
probability energy 𝐸 ( ) , at which a particle of a certain size will be broke
of 0.95.
The test methodology consists of the following algorithm of actions:
ability of 0.95.
(1) Preparation of sets of 20 particles of each size fractions: −75.0 + 63.0 mm,
−53.0 +The45.0 test
mm, methodology
−37.5 + 31.5 mm and consists
−26.5 +of22.4 themm.following
The selectionalgorithm of actions:
is performed by
analogy1) with the SMC test [20],
Preparation of sets but with a decrease
of 20 particles in theoftolerance
each size of thefractions:
mass of particles
−75.0 + 63
in the set from ± 30 to ± 5% of the average mass of a particle in the set. All particles must
45.0 mm, −37.5 + 31.5 mm and −26.5 + 22.4 mm. The selection is perform
have the correct shape. At the end of the selection, four narrow size fractions should be
with the
selected, SMC
within test
which the[20],
lightest but withdiffers
particle a decrease
in its mass infrom
the the
tolerance ofnothe
heaviest by moremass of
than 10%;
set from ± 30 to ± 5% of the average mass of a particle in the set. All parti
(2) Sequential breakage of all particles of each set in the JK DWT tester;
the2.1
correct shape. At the end of the selection, four narrow size fractions shou
The initial drop height of a light weight (LB, 2703.4 g) for all particles is 50 mm
within
from which
the top surfacethe lightest
of the particle to particle
the impact differs
surfacein of its mass from
the dropping the heaviest by
weight;
10%;2.2 The weight is dropped onto the particle, and after each drop the set height is
increased by 10 mm until the particle is broken. The height is recorded in the table as hi,j ,
2) Sequential breakage of all particles of each set in the JK DWT tester
at which the particle is broken. A particle is considered broken when its largest fragment
2.1 Thea sieve
passes through initial
withdropa cell height
size of theoflower
a light limit weight (LB, 2703.4
of the corresponding sizeg) for all part
fraction
from the top surface of the particle to the impact surface of the dropping w
or has a mass of no more than 0.8 of the average particle masses in the set;
(3) The average value of the installation heights is calculated, at which the breakage of
2.2 The weight is dropped onto the particle, and after each drop the s
particles of each of the size fractions occurred. The resulting values are recorded in Table 4;
creased by 10 mm
(4) Calculation of theuntil
average the
drop particle
height ofis thebroken.
weight at The
whichheight is recorded
the particles breakage in the
which
in the j-th the particle is broken. A particle is considered broken when its la
set occurs:
passes through a sieve withh ja=cell ∑20 1 hi,j of the lower limit of the correspondin
i =size
, m (1)
20
or has a mass of no more than 0.8 of the average particle masses in the set;
The average specific energy of particles’ breakage of each set is determined by
3) The average value of the installation heights is calculated, at whic
the equation:
of particles of each of Ethe Q · hj · G
0.5 sizew fractions occurred. The resulting (2) values a
cs( j) = 3, 6 · Q , kWh/t
Table 4; pj

where:
0.5 —specific energy of particles breakage, at which with a probability of 0.5 a
Ecs ( j)
particle in the j-th set will be broken, kWh/t;
Qw . —weight mass, kg;
Minerals 2021, 11, 485 12 of 17

hj —average drop height of the weight at which the particles breakage in the j-th set
occurs, m;
Qpj —average particle mass in the j-th set, g;
G—acceleration of gravity, m/s2 .
The nominal specific energy of particles breakage of each set is determined by
the equation:
0.95 0.5
Ecs ( j) = Ecs( j) + 2σj , kWh/t (3)

where Ecs0.95 —specific impact energy, with a probability of 0.95 leading to the j-th set particle
( j)
breakage, kWh/t; σj —standard deviation of the achieved specific energy Ecs of the j-th set
of particles breakage.

Table 4. Test report for determining the nominal specific energy of particle breakage.

Set No. j 1 2 3 4
Size fraction, mm −75.0 +63.0 −53.0 +45.0 −37.5 +31.5 −26.5 +22.4
Particle amount, i 20 20 20 20
Set mass Qnj , g Qn1 Qn2 Qn3 Qnj
Average particle mass Qpj , g Qn1 /20 Qn2 /20 Qn3 /20 Qnj /20
Weight mass QW , kg 2.7034 2.7034 2.7034 2.7034
Particle No. i Drop height, hij , leading to particle breakage, m
1 h1,1 h1,2 ... h1,j
2 h2,1 h2,2 ... h2,j
... ... ... ... ...
i = 20 hi,1 hi,2 ... hi,j
Average height, m h1 h2 ... hj
0.5 , kWh/t 0,5 0,5 0.5
Average breakage specific energy Ecs ( j) Ecs (1)
. Ecs (2)
Ecs ( j)
Standard deviation of Ecs(j) , % σ1 σ2 ... σj
Nominal breakage specific energy Ecs0.95 , kWh/t 0.95
Ecs 0.95
Ecs ... 0.95
Ecs
( j) (1) (2) ( j)

According to the method described above, for the tested samples of M1, M2 and T1
ores, the dependences of the nominal specific particle’s breakage energy on their size were
obtained. The results are shown in Table 5.
The results obtained from the analysis of the dependence of the nominal specific
particle’s breakage energy on the various size fractions of the three studied types of ores
are presented in Figure 15.
The obtained value of nominal specific particle’s breakage energy Ecs 0.95 cuts on the
( j)
energy spectra of the j-th size fraction the useful collisions, occurring in the mill, from those
which do not lead to breakage of the material, see Figure 16. Numerically, the total power
transmitted to the j-th size fraction can be divided into two components:

Ws( j) = Wl ( j) + Wu( j) , kW/t (4)

where Ws( j) —is the cumulative specific power of the j-th size fraction, kW/t; Wl ( j) —specific
cumulative power of the j-th size fraction, performed by impacts that do not lead to the
particle’s breakage, kW/t; and Wu( j) —useful cumulative specific power of the j-th size
fraction, performed by impacts leading to the particle’s breakage, kW/t.
This method assumes that the ratios of the infinitely narrow particle size fraction
destruction rate for different operating modes must be equal to the corresponding ratios
of useful energies Wu( j) for these size fractions in the investigated SAG operating modes.
Because the energy spectra curves are similar to each other (in specific particle size frac-
tions), the difference in the selected nominal specific particle’s breakage energy (within
reasonable limits) will not have a very high impact on the final useful energies’ ratio. This
 Table 5. Nominal specific particle’s breakage energy of various sizes of the studied ore types.

Minerals 2021, 11, 485 Specific Energy 𝑬𝟎.𝟓

𝒄𝒔(𝒋) Leading to the Particle Breakage, kWh/t 13 of 17
Ore Type M1 M2 Т1
Particle Size,
mm / 24.45 34.50
is the 49.00
reason why69.00the effect
24.45 of particles
34.50 weakening
49.00 69.00
[39,40]24.45
during 34.50
contacts 49.00 69.00
with specific
Particle no. energies smaller than the nominal specific breakage energy may be not considered.
1 0.0049 0.0023 0.0023 0.0009 0.0090 0.0082 0.0038 0.0016 0.0108 0.0091 0.0084 0.0032
2 Table 5. Nominal
0.0068 0.0049 specific
0.0040particle’s
0.0029breakage
0.0099energy 0.0087of various
0.0048sizes0.0033
of the studied
0.0154 ore types.
0.0116 0.0097 0.0071
3 0.0088 0.0054 0.0044 0.0029 0.0119 0.0088
Specific Energy E0.5
0.0056 0.0039 0.0159 0.0127 0.0118 0.0078
cs(j) Leading to the Particle Breakage, kWh/t
4 0.0094 0.0065 0.0047 0.0035 0.0120 0.0105 0.0064 0.0044 0.0162 0.0137 0.0121 0.0085
Ore Type M1 M2 T1
5
Particle Size, mm/0.0095 0.0075 0.0053 0.0035 0.0124 0.0113 0.0071 0.0050 0.0178 0.0140 0.0133 0.0090
6 No. 0.010124.450.0077
Particle 34.50 49.00
0.0054 69.00
0.0041 24.45
0.0125 34.50
0.0114 49.00
0.0076 69.00
0.0053 24.45
0.0179 34.50
0.0147 49.00
0.0134 69.00
0.0091
7 1 0.0049 0.0023 0.0023 0.0009 0.0090 0.0082 0.0038 0.0016
0.0104 0.0082 0.0055 0.0044 0.0129 0.0122 0.0077 0.0058 0.0186 0.0152 0.0141 0.0094 0.0108 0.0091 0.0084 0.0032
0.0068 0.0049 0.0040 0.0029 0.0099 0.0087 0.0048 0.0033 0.0154
8 2 0.0107 0.0089 0.0058 0.0046 0.0129 0.0124 0.0078 0.0058 0.0187 0.0116
0.0152 0.0097
0.0142 0.0071
0.0097
3 0.0088 0.0054 0.0044 0.0029 0.0119 0.0088 0.0056 0.0039 0.0159 0.0127 0.0118 0.0078
9 0.0115 0.0091 0.0058 0.0050 0.0136 0.0125 0.0080 0.0060 0.0189 0.0154 0.0144 0.0098
4 0.0094 0.0065 0.0047 0.0035 0.0120 0.0105 0.0064 0.0044 0.0162 0.0137 0.0121 0.0085
10 0.0117 0.0093 0.0062 0.0054 0.0140 0.0126 0.0081 0.0061 0.0191 0.0168 0.0147 0.0101
5 0.0095 0.0075 0.0053 0.0035 0.0124 0.0113 0.0071 0.0050 0.0178 0.0140 0.0133 0.0090
11 6 0.0117 0.0094
0.0101 0.0077
0.0070
0.0054
0.0059
0.0041
0.0142
0.0125
0.0127
0.0114
0.0085
0.0076
0.0061
0.0053
0.0200 0.0147
0.0179
0.0173 0.0134
0.0147 0.0091
0.0112
12 7 0.0122 0.0095
0.0104 0.0082 0.0072
0.0055 0.0059
0.0044 0.0147
0.0129 0.0127
0.0122 0.0094
0.0077 0.0061
0.0058 0.0201
0.0186 0.0174
0.0152 0.0151
0.0141 0.0124
0.0094
13 8 0.0126 0.0099
0.0107 0.0089 0.0075
0.0058 0.0061
0.0046 0.0148
0.0129 0.0134
0.0124 0.0096
0.0078 0.0075
0.0058 0.0203
0.0187 0.0178
0.0152 0.0152
0.0142 0.0125
0.0097
14 9 0.0126 0.0100
0.0115 0.0091 0.0079
0.0058 0.0062
0.0050 0.0153
0.0136 0.0137
0.0125 0.0098
0.0080 0.0076
0.0060 0.0212
0.0189 0.0187
0.0154 0.0157
0.0144 0.0126
0.0098
1510 0.0132 0.0100
0.0117 0.00930.0080
0.0062 0.0062
0.0054 0.0155
0.0140 0.01380.0126 0.0099
0.0081 0.0076
0.0061 0.0225 0.0168
0.0191 0.0195 0.0147
0.0167 0.0101
0.0127
1611 0.0136 0.0110
0.0117 0.00940.0085
0.0070 0.0063
0.0059 0.0166
0.0142 0.01390.0127 0.0099
0.0085 0.0079
0.0061 0.0228 0.0173
0.0200 0.0195 0.0147
0.0172 0.0112
0.0132
1712 0.0122
0.0139 0.00950.0085
0.0110 0.0072 0.0069
0.0059 0.0168
0.0147 0.01390.0127 0.0107
0.0094 0.0081
0.0061 0.0201
0.0229 0.0174
0.0197 0.0151
0.0177 0.0124
0.0132
18 13 0.0140 0.0116 0.0102 0.0075 0.0177 0.0150 0.0111 0.0100 0.0237 0.0202 0.0190 0.0125
0.0126 0.0099 0.0075 0.0061 0.0148 0.0134 0.0096 0.0075 0.0203 0.0178 0.0152 0.0150
14 0.0126 0.0100 0.0079 0.0062 0.0153 0.0137 0.0098 0.0076 0.0212 0.0187 0.0157 0.0126
19 0.0148 0.0117 0.0103 0.0084 0.0180 0.0151 0.0117 0.0100 0.0255 0.0204 0.0199 0.0152
15 0.0132 0.0100 0.0080 0.0062 0.0155 0.0138 0.0099 0.0076 0.0225 0.0195 0.0167 0.0127
20 0.0153 0.0119 0.0106 0.0087 0.0192 0.0155 0.0131 0.0101 0.0268 0.0226 0.0202 0.0158
16 0.0136 0.0110 0.0085 0.0063 0.0166 0.0139 0.0099 0.0079 0.0228 0.0195 0.0172 0.0132
Average 0.0114 0.0088 0.0068 0.0053 0.0142 0.0124 0.0085 0.0064 0.0198 0.0166 0.0149 0.0109
17 0.0139 0.0110 0.0085 0.0069 0.0168 0.0139 0.0107 0.0081 0.0229 0.0197 0.0177 0.0132
𝟎.𝟓
σj of 𝑬18 𝒄𝒔(𝒋) ,
0.0026 0.0025 0.0022 0.002 0.0026 0.0021 0.0023 0.0022 0.0037 0.0034 0.0031 0.0150
0.0140 0.0116 0.0102 0.0075 0.0177 0.0150 0.0111 0.0100 0.0237 0.0202 0.0190 0.0031
kWh/t 19 0.0148 0.0117 0.0103 0.0084 0.0180 0.0151 0.0117 0.0100 0.0255 0.0204 0.0199 0.0152
𝑬𝟎.𝟗𝟓
𝒄𝒔(𝒋) , kWh/t
20 0.0165 0.0138
0.0153 0.01190.0112
0.0106 0.0093
0.0087 0.0195
0.0192 0.01660.0155 0.0132
0.0131 0.0108
0.0101 0.0272 0.0226
0.0268 0.0234 0.0202
0.0211 0.0158
0.017
Average 0.0114 0.0088 0.0068 0.0053 0.0142 0.0124 0.0085 0.0064 0.0198 0.0166 0.0149 0.0109
0.5 ,
σ j ofEcs ( j)
kWh/t 0.0026 0.0025The0.0022
results obtained
0.002 from the
0.0026 analysis
0.0021 of the 0.0022
0.0023 dependence
0.0037of the nominal
0.0034 specific
0.0031 par-
0.0031
0.95 ,
Ecs kWh/t 0.0165
ticle’s
0.0138
breakage
0.0112
energy
0.0093
on the
0.0195
various
0.0166
size fractions
0.0132
of
0.0108
the three
0.0272
studied
0.0234
types
0.0211
of ores
0.017
are
( j)
presented in Figure 15.

0.030

0.025 y = −0.0095ln(x) + 0.0575

M1
R² = 0.9887
0.020

0.015 y = −0.0085ln(x) + 0.0467

R² = 0.9965 M2
0.010
y = −0.0070ln(x) + 0.0387
0.005 R² = 0.9943
T1
0.000
20 30 40 50 60 70 80
Particle size, mm
Figure 15.
Figure 15. Nominal
Nominal specific
specific particle’s
particle’s breakage
breakageenergy
energyof
ofvarious
varioussizes
sizesof
ofthe
thestudied
studiedore
oretypes.
types.

The overall efficiency of the grinding process can be determined as the weighted
average over all size fractions, and the completeness of the transfer of the total consumed
energy into the contact interaction of particles in the range of specific energies leading to
their breakage:
j Wu( j) γ j
Een.gr = ∑n=1 , % (5)
Ws( j)
 where 𝑊 ( ) —is the cumulative specific power of the j-th size fraction, kW/t; 𝑊 ( ) —spe-
cific cumulative power of the j-th size fraction, performed by impacts that do not lead to
the particle’s breakage, kW/t; and 𝑊 ( ) —useful cumulative specific power of the j-th size
fraction, performed by impacts leading to the particle’s breakage, kW/t.
This method assumes that the ratios of the infinitely narrow particle size fraction de-
Minerals 2021, 11, 485 14 of 17
struction rate for different operating modes must be equal to the corresponding ratios of
useful energies 𝑊 ( ) for these size fractions in the investigated SAG operating modes.
Because the energy spectra curves are similar to each other (in specific particle size frac-
tions), Thus,
the difference in the selected
having determined nominalspecific
the nominal specificparticle’s
particle’sbreakage
breakage energy
energy 0.95
Ecs(within
in the
( j)
reasonable limits)
entire range will notsizes
of material havebya very high impact
calculation on the
from the final useful
obtained energies’itratio.
dependencies, This
is possible
is to
the reasonthe
process why theofeffect
data energyofspectra
particles weakening
under specified[39,40] during
conditions contacts
and withthe
determine specific
overall
energies smaller than the nominal specific
efficiency of the SAG process, see Table 6. breakage energy may be not considered.

Figure 16.16.
Figure Graphic interpretation
Graphic ofofprincipal
interpretation principaldetermining
determininguseful
usefulenergy
energyon
onan
anenergy
energyspectra
spectra using
using the nominal
the nominal specific
specific energy
energy value.
value.

Table
The 6. Semi-autogenous
overall efficiency of grinding efficiency
the grinding calculations.
process can be determined as the weighted av-
erage over all size fractions, and the completeness of the transfer of the total consumed
−100.0 −115.0 −130.0
Particle Fraction,
energymminto the contact
5.0 −20.0+5.0
interaction −50.0+20.0
of particles in the range+50.0
of specific+100.0
energies leading
Ore Type +115.0 to
their breakage:
Particle Size, mm 5.0 12.5 35.0 75.0 107.5 122.5
( )
0.95 ,
Ecs kWh/t 0.0274 0.0210𝐸 . = ∑0.0138 , %0.0085 0.0060 0.0050(5)
( j) ( )

Ws( j) , kW/t Thus, having

133,279.8 27,280.6 .
determined the nominal 2839.6 1497.8 breakage
specific particle’s 1079.2
energy 𝐸1053.4
( ) in
M1 Wu( j) , kW/t 132,923.5
the entire range 27,054.3
of material sizes 2183.5
by calculation 481.6
from the obtained 131.4
dependencies, it122.8
is pos-
sible to process0.040
γi,j , % the data of energy
0.360 spectra under
1.600 specified conditions4.000
30.000 and determine
4.000the
overall efficiency of the SAG process, see Table 6.
Ei,j , % 0.040 0.357 1.230 9.646 0.487 0.466
0.95 ,
Ecs kWh/t 0.0330 0.0252 0.0165 0.0100 0.0069 0.0058
( j)
Ws( j) , kW/t 121,879.8 27,010.4 2392.5 1501.2 995.9 961.0
M2 Wu( j) , kW/t 121,629.6 26,703.0 1742.4 425.5 140.8 133.2
γi,j , % 0.025 0.225 1.000 18.750 2.500 2.500
Ei,j , % 0.025 0.222 0.728 5.314 0.353 0.347
0.95 , kWh/t
Ecs 0.0422 0.0335 0.0237 0.0165 0.0131 0.0118
( j)
Ws( j) , kW/t 101,134.3 31,596.8 3266.1 1994.4 1519.9 1436.7
T1 Wu( j) , kW/t 100,770.7 31,200.4 2301.0 631.3 233.3 196.0
γi,j , % 0.035 0.315 1.400 26.250 3.500 3.500
Ei,j , % 0.035 0.311 0.986 8.309 0.537 0.477
Charge Een.gr , % 29.87

In Table 6, it is possible to track the share of each j-th size fraction in the value of the
overall efficiency of the grinding process. In fact, the value Ei,j shows the rate of breakage
of the j-th size fraction relative to all others. For the simulation with the conditions in
which the change in the particle size distribution of material is expected: the fine fractions’
masses must decrease, and the coarse fractions’ masses must increase, according to their
Ei,j /γi,j ratios.
Minerals 2021, 11, 485 15 of 17

The obtained value of the efficiency of the semi-autogenous grinding process Een.gr can
be used as an optimization criterion when comparing various technical and technological
parameters of the SAG process in FEM-designed models. In comparative calculations with
the presented method, it is expected that the resulting value of the grinding efficiency Een.gr
will be inversely proportional to the specific power consumption of the SAG unit.

4. Discussion
With the development of computer hardware and software, fundamental discrete
element modeling is increasingly used in the field of comminution processes modeling.
This study presents the author’s approach to the construction of a multiphase model of the
SAG process and the processing of its results.
The developed integrated approach allows us to create dynamic models of AG and
SAG mills. In the future, dynamic models can be used to increase the versatility of static
models, as well as to obtain design points when creating digital twins of semi-autogenous
grinding mills.
Further directions of research in this area:
• standardization of laboratory methods for calibrating the parameters of material
interactions—this kind of research should be easy to perform and aimed at find-
ing specific parameters that govern the model of interaction of materials present in
the model;
• standardization and substantiation of laboratory methods for studying the strength of
mineral raw materials, the results of which can be used to analyze models of the SAG
process, designed with DEM;
• development of methods for processing the results of DEM modeling of tumbling
mills, aimed at predicting the productivity and grain size distribution of its product.

5. Conclusions
As a result of the work, a set of models for the interaction of ore and grinding media
was determined, which makes it possible to design a stable multiphase DEM-CFD model
of wet semi-autogenous grinding. A method for determining the coefficients required to
create a DEM-CFD model of wet semi-autogenous grinding and a method for laboratory
determination of the nominal specific energy mineral raw material particle’s breakage in
various sizes fractions using a JK DWT tester have been developed. This indicator can be
used when processing the results of DEM modeling of the AG/SAG mill.
A method for calculating the efficiency of the semi-autogenous grinding process based
on the results of numerical simulation by the discrete element method is proposed.

Author Contributions: Conceptualization, L.C. and V.L.; methodology, L.C. and V.L.; software, L.C.
validation, L.C.; formal analysis, L.C. and V.L.; investigation, L.C. and V.L.; resources, L.C.; data
curation, L.C. and V.L.; writing—original draft preparation, L.C.; writing—review and editing, V.L.;
visualization, L.C.; supervision, V.L.; project administration, V.L. All authors have read and agreed to
the published version of the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Cleary, P.W. Predicting charge motion, power draw, segregation, wear and particle breakage in ball mills using discrete element
methods. Miner. Eng. 1998, 11, 1061–1080. [CrossRef]
2. Gutiérrez, A.; Ahues, D.; González, F.; Merino, P. Simulation of Material Transport in a SAG Mill with Different Geometric Lifter
and Pulp Lifter Attributes Using DEM. Soc. Min. Metall. Explor. 2018. [CrossRef]
3. Kolahi, S.; Chegeni, M. Investigation of the effect of the number of lifters on performance of pilot-scale SAG mills using discrete
element method. J. Min. Environ. 2020, 11, 675–693. [CrossRef]
4. Bian, X.; Wang, G.; Wang, H.; Wang, S. Effect of lifters and mill speed on particle behavior, torque, and power consumption of a
tumbling ball mill: Experimental study and DEM simulation. Miner. Eng. 2017, 105, 22–35. [CrossRef]
Minerals 2021, 11, 485 16 of 17

5. Kalala, J.T.; Bwalya, M.; Moys, M. Discrete element method (DEM) modelling of evolving mill liner profiles due to wear. Part II.
Industrial case study. Miner. Eng. 2005, 18, 1392–1397. [CrossRef]
6. Powell, M.; Weerasekara, N.; Cole, S.; Laroche, R.; Favier, J. DEM modelling of liner evolution and its influence on grinding rate
in ball mills. Miner. Eng. 2011, 24, 341–351. [CrossRef]
7. Sato, A.; Kano, J.; Saito, F. Analysis of abrasion mechanism of grinding media in a planetary mill with DEM simulation. Adv.
Powder Technol. 2010, 21, 212–216. [CrossRef]
8. Cleary, P.W.; Morrison, R.; Morrell, S. Comparison of DEM and experiment for a scale model SAG mill. Int. J. Miner. Process. 2003,
68, 1–4. [CrossRef]
9. Rosenkranz, S.; Breitung-Faes, S.; Kwade, A. Experimental investigations and modelling of the ball motion in planetary ball mills.
Powder Technol. 2011, 212, 224–230. [CrossRef]
10. Delaney, G.; Cleary, P.; Morrison, R.; Cummins, S.J.; Loveday, B. Predicting breakage and the evolution of rock size and shape
distributions in Ag and SAG mills using DEM. Miner. Eng. 2013, 50, 132–139. [CrossRef]
11. Chitalov, L.S.; Lvov, V.V. New approaches in mineral raw materials comminution tests modelling. In Advances in Raw Material
Industries for Sustainable Development Goals; Litvinenko, V.S., Ed.; Taylor & Francis Group: London, UK, 2021; pp. 146–151.
ISBN 978-0-367-75881-3.
12. Klichowicz, M.; Lieberwirth, H. Grain-Based DEM for Particle Bed Comminution. Minerals 2021, 11, 306. [CrossRef]
13. Gospodarikov, A.P. Development of nonlinear mathematical models and numerical modeling of the prediction of the stress-strain
state of a rock massif. J. Min. Inst. 2016, 219, 382–386. [CrossRef]
14. Aleksandrova, T.N.; Nikolaeva, N.V.; Lvov, V.V.; Romashev, A.O. Ore processing efficiency improvements for precious metals
based on process simulations. Obogashchenie Rud 2019, 2, 8–13. [CrossRef]
15. Ushakov, E.; Aleksandrova, T.; Romashev, A. Neural network modeling methods in the analysis of the processing plant’s
indicators. In Advances in Intelligent Systems and Computing, Proceedings of the International Scientific Conference Energy Management
of Municipal Facilities and Sustainable Energy Technologies (EMMFT); Springer: New York, NY, USA, 2019. [CrossRef]
16. Guryev, A.A. Sustainable development of the ore-raw material base and processing facilities of Apatit JSC based on the best
engineering solutions. J. Min. Inst. 2017, 228, 662–673. (In Russian) [CrossRef]
17. Stark, S.; Perkins, T.; Napier-Munn, T.J. JK drop weight parameters: A statistical analysis of their accuracy and precision and
the effect on SAG mill comminution circuit simulation. In Proceedings of the Conference MetPlant 2008, Perth, WA, Australia,
18–19 August 2008.
18. Tikhonov, N.O.; Skarin, O.I. Semiautogenous grinding mills’ capacity calculation by energy indices. Gorn. Zhurnal 2014, 11, 6–10.
19. Morrell, S. Predicting the specific energy of autogenous and semi-autogenous mills from small diameter drill core samples. Miner.
Eng. 2004, 17, 447–451. [CrossRef]
20. Burke, J.M. Determining the Bond Efficiency of industrial grinding circuits. In Global Mining Standards and Guidelines Group;
2015. Available online: https://gmggroup.org/wp-content/uploads/2016/02/Guidelines_Bond-Efficiency-REV-2018.pdf
(accessed on 29 April 2021).
21. Matei, V.; Bailey, C.W.; Morrell, S. A new way of representing A and b parameters from JK Drop-Weight and SMC tests: The
“SCSE”. In Proceedings of the Conference Materials: SAG 2015, Vancouver, Canada, 20–24 September 2015; pp. 1–12.
22. Lieberwirth, H.; Popov, O.; Aleksandrova, T.; Nikolaeva, N. Scientific substantiation and practical realization of selective
comminution process of polymetallic mineral raw materials. E3S Web Conf. 2020, 192. [CrossRef]
23. Aleksandrova, T.; Nikolaeva, N.; Lieberwirth, H.; Aleksandrov, A. Selective Desintegration and Concentration: Theory and
Practice. E3S Web Conf. 2018, 56, 1–7. [CrossRef]
24. O’Connor, C.; Alexandrova, T. The geological occurrence, mineralogy, and processing by flotation of platinum group minerals
(PGMS) in south Africa and Russia. Minerals 2021, 11, 54. [CrossRef]
25. Boikov, A.V.; Savelev, R.V.; Payor, V.A.; Erokhina, O.O. Evaluation of bulk material behavior control method in technological
units using dem. Part 2. CIS Iron Steel Rev. 2020, 20, 3–6. [CrossRef]
26. Boikov, A.V.; Savelev, R.V.; Payor, V.A.; Erokhina, O.O. The control method concept of the bulk material behavior in the pelletizing
drum for improving the results of DEM-modeling. CIS Iron Steel Rev. 2019, 17, 10–13. [CrossRef]
27. Branco, A.; Mesquita, A.; Vaz, J. Application of the linear spring-dashpot model in the CFD-DEM simulation of alumina
fluidization. Eng. Térmica (Therm. Eng.) 2015, 14, 95–101. [CrossRef]
28. Cundall, P.A.; Strack, D.L. A discrete numerical model for granular assemblies. Geotechnique 1979, 29, 47–65. [CrossRef]
29. Li, C.; Zhang, M.; Wang, T.; Wang, H. Simulation and computational study of U-mold membrane components by CFD multiphase
flow Euler model. Mater. Res. Express 2019, 6, 1–10. [CrossRef]
30. Zbib, H.; Ebrahimi, M.; Ein-Mozaffari, F.; Lohi, A. Comprehensive analysis of fluid-particle and particle-particle interactions in a
liquid-solid fluidized bed via CFD-DEM coupling and tomography. Powder Technol. 2018, 340, 116–130. [CrossRef]
31. Cleary, P.; Hilton, J.E.; Sinnott, M. Modelling of industrial particle and multiphase flows. Powder Technol. 2017, 314, 232–252.
[CrossRef]
32. Guo, Y.; Wu, C.-Y.; Thornton, C. Modeling gas-particle two-phase flows with complex and moving boundaries using DEM-CFD
with an immersed boundary method. AIChE J. 2012, 59, 1075–1087. [CrossRef]
33. Kuang, S.; Zhou, M.; Yu, A. CFD-DEM modelling and simulation of pneumatic conveying: A review. Powder Technol. 2020, 365,
186–207. [CrossRef]
Minerals 2021, 11, 485 17 of 17

34. Vasilyeva, N.V.; Erokhina, O.O. Post-impact recovery coefficient calibration in DEM modeling of granular materials. Obogashchenie
Rud 2020, 4, 42–48. [CrossRef]
35. Zhou, H.; Hu, Z.; Chen, J. Calibration of DEM models for irregular particles based on experimental design method and bulk
experiments. Powder Technol. 2018, 332, 210–223. [CrossRef]
36. Jiang, Z.; Du, J.; Rieck, C.; Bück, A.; Tsotsas, E. PTV experiments and DEM simulations of the coefficient of restitution for irregular
particles impacting on horizontal substrates. Powder Technol. 2020, 360, 352–365. [CrossRef]
37. Roessler, T.; Katterfeld, A. Scaling of the angle of repose test and its influence on the calibration of DEM parameters using
upscaled particles. Powder Technol. 2018, 330, 58–66. [CrossRef]
38. Andreev, E.E.; Tihonov, O.N. Droblenie, Izmel’chenie i Podgotovka Syr’ya k Obogashcheniyu (Crushing, Grinding and Ore Pretreatment
for Benefication); Saint Petersburg Mining University: Saint Petersburg, Russia, 2007; p. 99.
39. Tavares, L.M.; Carvalho, R.M. Impact work index from continuum damage model of particle fracture. Miner. Eng. 2007, 20,
1368–1375. [CrossRef]
40. Pauw, O.G.; Mare, M.S. The determination of Optimum Impact-Breakage Routes for an Ore. Powder Technol. 1998, 54, 3–13.
[CrossRef]