See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/286447937

Prediction of Wear of Mill Lifters Using Discrete Element Method

Conference Paper · January 2001

CITATIONS READS

17 1,120

4 authors, including:

Xiangjun Qiu Alexander Potapov

Metso Minerals ESSS Rocky DEM
18 PUBLICATIONS   313 CITATIONS    30 PUBLICATIONS   884 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Particle breakage embedded in DEM View project

rolling resistance model View project

All content following this page was uploaded by Xiangjun Qiu on 10 December 2015.

The user has requested enhancement of the downloaded file.

 IV-260 IV-261

the surface of the material. Perhaps the most notable wear model is
Archard’s wear law. Archard’s wear law takes the following mathematic
form

V = K S P/H (1)

where V is the volume of material worn from the surface, S is the sliding
distance, P is the shear force, H is the hardness of the material
subjected to wear, and K is a constant. There are many other similar
wear models developed. These models only try to express K as a
Prediction of Wear of Mill Lifters Using Discrete Element Method function of specific surface conditions or load conditions. Rabinnowicz
(1965) presented a model in which K of Eq.(1) is expressed in terms of
Xiangjun Qiu*, Alexander Potapov, Ming Song, Lawrence Nordell the attack angle of abrasive particles. Zum Gahr (1988) developed a
model, which takes into account material properties of work hardening,
Conveyor Dynamics Inc., 1111 W. Holly St., Bellingham, WA 98225 ductility and fracture toughness. Finnie (1972) proposed a special wear
*e-mail: [email protected] model, which addresses the erosion of a surface by a stream of solid
particles. Although the original mathematic form of Finnie’s model looks
ABSTRACT quite different from Eq.(1), it can be proven that Finnie’s model reduces
to Eq.(1) with K dependent on particle mass and moment of inertia. Due
3D-Discrete Element Method can be employed to simulate material wear to its simplicity and generality, we only attempt to implement Archard’s
inside of mining equipment. We introduce a new algorithm, which wear model in this paper. The validation of Archard’s wear law in
modifies existing DEM computer code to implement the well-known prediction of lifter wear will be further discussed in the following section.
Archard’s constitutive wear law. We show that the modified DEM For implementation purpose, we rewrite Eq(1) in a form suitable for
program is able to predict wear behavior such as: wear shape, rate, and integration in a DEM model: i.e.
life. We further show that the DEM tool, with wear-prediction capability,
can help to better design mining comminution components. The A dh = C dw (2)
comparisons between measurements and predictions for the wear of mill
lifters are presented in the paper. where A is the surface area of a DEM element subjected to wear, and
dh is the incremental loss of depth of the element, dw is the incremental
INTRODUCTION shear work done on the surface area A, and C (equal to K/H) is defined
as the wear rate. The implementation of Eq. (2) in a DEM program is
3D-Discrete Element Method (DEM) has been successfully employed in simple and does not need further discussion. However, the selection of
simulation of SAG, AG and ball mills to predict mill charge motion, power the wearing rate C is not straightforward and does need elaboration.
and energy spectra (Herbst and Nordell, 2001). In this paper, we try to
explore a new application of 3D-DEM --- prediction of mill lifter and liner Since a typical SAG mill lifter can last for more than 10 months, it is not
wear. For this purpose, we implement Archard’s (1980) constitutive wear practical to try to simulate the actual period of wear life by using DEM. It
law in the DEM model. In the following sections, we will first discuss the is desirable that the DEM simulation time be shortened by several
theoretical background of implementing a DEM wear model. Then orders of magnitudes while the characteristics of the lifter wear are still
demonstrate the DEM capability of wear-prediction. We will show that the captured by the model. For this purpose, we choose a wear rate in the
DEM with wear-simulation capability is able to predict worn lifter shape, simulation, denoted by Cmodel, which is several orders of magnitudes
wear rate and service life. higher than the actual wear rate C of Eq.(2). Our experience shows that
the suitable C model must be determined by the following two rules:
IMPLEMENTATION OF ARCHARD’S WEAR LAW

To date, numerous models for wear have been developed. In general, 1. Cmodel must satisfy the constraint of proportionality of Eq.(2). In
these models relate the volume loss of a material to load conditions on other words, increasing the Cmodel by a factor of n and reducing
 IV-262 IV-263

the simulation time by the same factor must produce the same wear The next example (shown in Fig. 5) demonstrates the effects of
characteristics. changing model wear rate Cmodel on simulation results. From Fig. 5, it is
seen that when C model /C is changed from 5 (10)5 to 1 (10) 6 the
2. The entire simulation time must be sufficiently long that the simulated wear profiles coincide with each other very well. This means
statistical variance associated with discrete events is small in the
DEM output data. that rule 1 is obeyed. However, when C model /C is increased to 1 (10)7,
the simulated wear profile significantly deviates from the profiles with
Both rule 1 and 2 are violated when a very large C model is used in the Cmodel /C=5 (10)5 and 1 (10) 6. and therefore rule 1 is violated.
simulation. Usually, a very large Cmodel will lead to a fast change of lifter EVALUATION OF LIFTER GEOMETRY
geometry so that the mill charge motion will be strongly disturbed. The
examples shown in the next section illustrates how to choose C model for With the help of a DEM tool, which includes wear-prediction capability,
lifter wear. To distinguish it from the overall (macroscopic) lifter wear we can evaluate different types of lifter geometries in terms of lifter life
rate, we call Cmodel a microscopic wear rate. and mill comminution efficiency. In this paper, let us consider two types
of lifters that are assumed to be installed in the SAG mills with the same
VALIDATION OF WEAR MODEL operating conditions (e.g. mill diameter, number of lifters, mill speed,
and holdup). One type of lifter is symmetrical with 30O face angle
Once Archard’s (microscopic) wear law is implemented in the DEM (2x30O) and runs in a bi-directional mode, and the other type is
model, we can simulate the lifter wear in SAG, AG and ball mills. Our unsymmetrical with 30O face angle (1x30O) and runs in unidirectional
numerical results show that, in general, for most conventional types of mode. For the newly installed condition, both lifters have the same
lifers, the lifter cross-sectional area decreases almost linearly as mill metal areas (mass) in their cross sections. We define a constant
running time increases (e.g. see Fig. 7). In other words, the macroscopic number of operating hours (say 750 hours) as one stage. After
wear rate for a conventional type of lifter is constant in the simulation. simulating four stages of wearing history, we demonstrate the results in
We need to verify this by checking measured lifter wear data. In the Figs 6 to 10. Figure 6 shows the lifter profile evolution in the four stages.
measured data, any deviation from a constant lifter wear rate suggests Figure 7 shows how the cross section area changes. We can see that
that the Archard’s wear law is invalid. Figure 1 shows a typical SAG mill the 1x30O lifter wears faster than 2x30O lifter. Figure 8 shows mill power
with several lifter wear rates measurement corresponding to five different draws corresponding to the two types of lifters. From this Figure, we see
time periods. Obviously, the wear rates were not constant in these time
that the power draw corresponding to 1x30O lifter type is larger, for most
periods. However, our further study shows that the ore hardness H ore of its life cycle, than that corresponding to 2x30O lifter type. Figure 9
was not constant in the five periods (see Fig. 2). After introducing a illustrates the energy dissipated to particles in the mills with two different
correction factor, denoted by CT, for SAG operating time based on ore
b types of lifters. The SAG mill with 2x30O lifter type has higher dissipated
hardness, using an equation of the form CT=a*Hore , the wear rates energy than the SAG mill with 1x30O. Figure 10 shows the energy
become surprisingly constant (see Fig. 3). This validates our adoption of utilization in SAG mills related with the two types of lifters, where the
Archard’s model. It must be noted that the macroscopic wear rates for energy utilization is defined by the collisional energy divided by the
some non-conventional types of lifters (with novel geometry) may not be
power draw. From these results, we conclude that 2x30O lifter type is
constant especially in their earlier stages of life.
better than 1x30O lifter type.
Now let us explore what accuracy the DEM wear model can achieve.
The example shown here is the simulation of a SAG mill lifter wear This example demonstrates a specific condition. We do not suggest that
history. The results are shown in Fig. 4. The dashed lines represent the proper design of a unidirectional mill cannot achieve, or exceed, the
measured lifter profiles and the solid lines represent the DEM calculated performance of a bi-directional mill. There are many important factors of
lifter shape, comminution performance and wear life that are not
profiles. In the simulation, the ratio of Cmodel /C is 1.7 (10)6 . The entire addressed in this paper. Subtle modifications to the manufactured lifter
simulation period is less than 1.5 revolutions. The comparison between shapes can produce significant improvements to life, power draw and
the simulations and measurements indicates that the DEM wear comminution efficiency. Implementation of a DEM wear model validates
prediction is reasonably accurate.
 IV-264 IV-265

comminution performance at a fraction of the cost, time, and risks spent

on real mill modifications.
Lifter Wear Rate Before Correction
CONCLUSIONS

We demonstrate that DEM combined with Archard’s wear law can be 1.2
employed to simulate mill lifter wear. We find that the wear rate used in 29 days

Normaslized Metal Loss / Day

the DEM model can be several orders of magnitudes higher than the 1 55 days 63 days 20 days
actual wear rate and the entire wearing process can be simulated within
a few mill revolutions. Examples of lifter wear simulations show that such 0.8 54 days
a modeling scheme still keeps a reasonable accuracy in comparison with
measurements. We propose that DEM accurately replicates a mill’s 0.6
granular motions and forces by noted strong correlation between
calculated and measured lifter wear morphology. We introduce two rules 0.4
about the selection of model wear rate. Finally, we show that that the
DEM tool with wear-prediction capability can help to better design mining 0.2
comminution components.
0
ACKNOWLEDGEMENTS
1 2 3 4 5
The authors gratefully acknowledge the assistance of J.A. Herbst in the
Sequential Period
preparation of this manuscript. A portion of the work performed here was
done under the sponsorship of Svedala Optimization Services.
Fig. 1 Measured SAG Mill Lifter Wear Rate Before Correction
Period 1: 0-54days; Period 2: 55-109 days; Period 3: 110-138 days;
REFERENCES Period 4: 139-201 days; and Period 5: 202-221 days.

Archard, J. F., 1980, “Wear Theory and Mechanisms”, Wear Control Handbook,
American Society of mechanical Engineers, New York,, pp. 35-80.
Finnie, I., 1972, “Some Observations on the Erosion of Ductile Metals”, Wear, 19,
pp81-90
Herbst, J. A. and Nordell, L., 2001, “Optimization of the Design of SAG Mill
Internals Using High Fidelity Simulation”, SAG 2001, Vancouver, B.C. October
2001
Rabinowicz, E., 1965, Friction and Wear of Materials, Wiley, New York.
Zum Gahr, K. H., 1988, “Modeling of Two-Body Abrasive Wear”, Wear, 124, pp.
87-103
 IV-266 IV-267

Ore Hardness

80

60
Frequency of Time (%)

Soft
Medium
40
Hard

20

0
1 2 3 4 5
Sequential Period

Fig. 2 Ore Hardness Fig.4 Simulation and Measurement of SAG Mill Lifter Wear History

Lifter Wear Rate After Correction

New Lifter 500h; Measured
0.35 500h; Cmodel/C = 5*10^5 500h; Cmodel/C = 10^6
1.2 500h; Cmodel/C = 10^7 1310h; Cmodel/C = 5*10^5
55 days 20 days
Normailized Metal Loss/ Day

29 days 63 days 0.3 1310h; Cmodel/C = 10^6 1310h; Cmodel/C = 10^7

1 54 days
0.25

0.8 0.2
0.15

y(m)
0.6 0.1
0.4 0.05
0
0.2
-0.05
0 -0.1
-0.05 0.05 0.15 0.25 0.35 0.45 0.55
1 2 3 4 5
x (m)
Sequential Period
Fig.5 Effect of Selection of Cmodel on Simulation Accuracy
Fig. 3 Measured SAG Mill Lifter Wear Rate After Correction
 IV-268 IV-269

New Lifter Cross Section Area 2x30Deg

Stage1
1x30Deg
0.3 Stage2
1.2
Stage3

Normalized Cross Section Area

Stage4
1
0.2
0.8

0.1 0.6

0.4
0
0.2

-0.1 0
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 New Stage1 Stage2 Stage3 Stage4
O
Fig.6a SAG Mill 2x30 Lifter Wear History, (Unit of coordinates: Meter) Fig. 7a Lifter Cross-Section Areas
Normalized with respect to the Cross-Section Area of New Stage

New
Lifter Cross Section Area Change
Stage1
0.3 1
Stage2 2x30Deg

Reduction of Cross Sectional Area

0.9
Stage3 1x30Deg
0.8
0.2 Stage4
0.7
0.6
0.1 0.5
0.4
0.3
0
0.2
0.1
-0.1 0
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 Stage New to 1 Stage 1 to 2 Stage 2 to 3 Stage 3 to 4

Fig.6b SAG Mill 1x30O Lifter Wear History, (Unit of coordinates: Meter) Fig. 7b Lifter Cross-Sectional Area Reduction with respect to the Cross-
Section Area of New Stage
 IV-270 IV-271

Power
2x30Deg
Energy Utilization 2x30Deg
1x30Deg
1.02 1x30Deg
1.12E+00
1.01
1.10E+00
1
1.08E+00
0.99 1.06E+00

0.98 1.04E+00
1.02E+00
0.97
1.00E+00
0.96
9.80E-01
0.95 9.60E-01
New Stage1 Stage2 Stage3 Stage4
9.40E-01
Fig.8 SAG Mill Power Draw New Stage1 Stage2 Stage3 Stage4
Normalized with respect to that of 1x30O New Stage Fig.10 SAG Mill Energy Utilization
Normalized with respect to that of 1x30O New Stage
Particle Collision Energy 2x30Deg
1x30Deg

1.08

1.06

1.04

1.02

0.98

0.96
New Stage1 Stage2 Stage3 Stage4

Fig.9 Particle Collision Energy in SAG Mill

Normalized with respect to that of 1x30O New Stage

View publication stats