See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/313823583

Generation of Wear Particles and Running-In in

Mixed Lubricated Sliding Contacts

Article in Tribology International February 2017

DOI: 10.1016/j.triboint.2017.02.019

CITATIONS READS

0 65

3 authors:

Aydar Akchurin Rob Bosman

University of Twente University of Twente
16 PUBLICATIONS 29 CITATIONS 26 PUBLICATIONS 121 CITATIONS

SEE PROFILE SEE PROFILE

Piet Lugt
SKF
89 PUBLICATIONS 1,071 CITATIONS

SEE PROFILE

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

Thin layer flow: Elasto-Hydrodynamic Lubrication View project

Wear Particles Generation in Lubricated Contacts View project

All content following this page was uploaded by Aydar Akchurin on 06 April 2017.

The user has requested enhancement of the downloaded file.

 Tribology International 110 (2017) 201208

Contents lists available at ScienceDirect

Tribology International
journal homepage: www.elsevier.com/locate/triboint

Generation of wear particles and running-in in mixed lubricated sliding MARK

contacts

Aydar Akchurina,b, , Rob Bosmanb, Piet M. Lugtb,c
a
Materials innovation institute (M2i), P.O. Box 5008, 2600 GA Delft, The Netherlands
b
Department of Engineering Technology, Laboratory for Surface Technology and Tribology, University of Twente, P.O. Box 217, 7500 AE Enschede, The
Netherlands
c
SKF Engineering & Research Centre, Kelvinbaan 16, 3439 MT Nieuwegein, The Netherlands

A R T I C L E I N F O A BS T RAC T

Keywords: A new model was developed for the simulation of wear particles formation in mixed lubricated sliding contacts.
Wear Particles Size The simulations based on a previously developed half-space algorithm coupled with a numerical elasto-
Mixed Lubrication hydrodynamic lubrication solver utilizing the load-sharing concept. A particle removal criterion based on a
Simulation critical Von Mises stress and a geometrical boundary condition was successfully implemented. The resulting
Running-in
model allows for the determination of the size of the particles generated in dierent conditions. The model was
applied to the simulation of running-in and validated using experimental measurements of the initial and run-in
friction coecient in a mixed lubricated contact. Additionally, dynamic light scattering (DLS) measurements of
the wear particles size were performed and compared to the simulation data.

1. Introduction
The wear process in tribological contacts results in the generation of
particles of various size and shape. Besides the mechanical damage, the
particles may inuence the performance of the tribological system in
other ways. In grease lubricated systems particles can have an impact
on the grease life since thickener and base oil oxidation is accelerated
by metal wear particles [1,2]. An autoimmune reaction was found to be
dependent on the size and number of formed wear UHMWPE
fragments in articial joint replacements [3,4]. Additionally, environ-
mental eects may also be important [5].
Since the wear process involves various complex phenomena,
development of a generic wear model is problematic. Meng and
Ludema [6] identied 182 equations for dierent types of wear
developed over the years. Among them were empirical relations,
contact mechanics-based approaches, and equations based on material
failure mechanisms.
The latter models are based on the contact stresses and strains and
are typically either fully numerical or semi-analytical. Fully numerical
models are commonly based on the Finite Element Method (FEM) [7
9] and are applied to simplied cases due to the high computational
costs. Semi-analytical codes (based on boundary element method,
BEM), on the other hand, assume the surfaces to be semi-innite
solids [10,11]. The higher accuracy compared to the analytical models
(applicable only in specic cases) and the signicantly lower computa-
tional costs compared to FEM, made this approach widely used for
contact, friction and wear simulations, see for example [1214]. Here,
one of the major drawbacks is the assumption of homogeneous elastic
material behaviour, which has been tackled only for uniform layers
[15,16].
To calculate the local wear volume, BEM may be combined with a
local variant of Archard's wear law [17,18]. However, to model wear
particles generation rather than wear volume, it is necessary to dene a
material rupture criterion. A number of material rupture criteria can be
found in the literature, including the critical accumulated dissipated
energy [19], critical accumulated plastic strain [20], critical accumu-
lated damage [21], critical Von Mises stress [22] and their variations
[2326].
A number of models were developed to simulate wear particles
generation in dry or boundary lubricated contacts. Nelias et al. [27]
proposed a wear model based on the accumulated plastic strain
criterion to nd regions of potential wear. Here it should be noted,
that the simulation of wear particle generation requires a dense mesh
and the calculation time can be very high, especially since sub-surface
plasticity is considered. The application of a stress based criterion is
computationally more ecient compared to strain based approach and
is therefore preferred for practical applications. In the stress based
approach, the particle is formed if a critical stress is exceeded and yield

Corresponding author at: Materials innovation institute (M2i), P.O. Box 5008, 2600 GA Delft, The Netherlands.
E-mail address: a.akchurin@utwente.nl (A. Akchurin).

http://dx.doi.org/10.1016/j.triboint.2017.02.019
Received 5 October 2016; Received in revised form 3 February 2017; Accepted 15 February 2017
Available online 17 February 2017
0301-679X/ 2017 Elsevier Ltd. All rights reserved.
A. Akchurin et al.
Nomenclature
fCb friction coecient in boundary lubrication, [ ]
R radius of the cylinder, [m]
FT total applied load, [N ]
FC load carried by surface contacts, [N ]
FH load carried by the hydrodynamic lm, [N ]
Fsh lubricant shear force, [N ]
dynamic viscosity of the lubricant, [Pas]
1 K ,S
pressure-viscosity coecient, [ GPa ]
h (x , y ) hydrodynamic lm thickness, [m]
h0 hydrodynamic approach, [m]
PH (x, y ) hydrodynamic pressure, [Pa]
p (x , y ) contact pressure, [Pa]
u (x , y ) deection due to pressure p(x, y ), [m]
H hardness of the substrate, [Pa]
stress is frequently used as the critical value. Bosman and Schipper
[22] used a Von Mises stress based criterion (with a yield stress as a
critical value) to simulate the wear particles generation and to predict
the severity of adhesive wear. Recently, Aghababaei et al. [28] used
Molecular Dynamic (MD) simulations to discriminate two wear particle
generation regimes: by fracture or atom-by-atom. They found, that the
yield stress can be used in a simple material rupture criterion to t the
MD simulations. An excellent agreement of the model results with the
Atomic Force Microscopy (AFM) experiments was reported. Akchurin
et al. [29] applied a Von Mises stress based criterion to calculate the
wear particles size and shape for boundary lubricated contacts and a
good agreement with the experimental data obtained using DLS and
AFM was observed.
In mixed lubricated conditions a fraction of the load is carried by
the lubricant. The lubrication conditions therefore have an impact on
the shear stress on the surface and therefore on the subsurface stress
distribution. This has an impact on the size and shape of the generated
wear particles. There are two general approaches for the calculation of
the load distribution in mixed lubricated contacts: rough elasto-
hydrodynamic lubrication (EHL) theory based methods and load
sharing concept models [30,31]. In the rst group, the governing
lubricant ow and surface deformation equations are solved numeri-
cally using a discretized rough surface in both the Reynolds and the
solid deection equations [32,33]. Although more accurate, these
models encounter certain problems in the thin lm lubrication
regime, such as convergence, accuracy and mesh-dependence
[34,35]. Also the exact value of the minimum lm thickness at which
the lm collapses is still a point of discussion [36]. Next to this, an
accurate estimation of wear particles requires dense meshes and the
computation eort of full numerical solutions may therefore become
very high. On the other hand, load sharing models, rst introduced by
Johnson [37], oer advantages of robustness and a relatively simple
methodology, without the need to determine an (arbitrary) chosen lm
Fig. 1. The Von Mises stress distribution (Pa) and the particle generation criterion. The yield stress is 0.4 GPa.
202
Tribology International 110 (2017) 201208
y yield stress, [Pa]
E1, E2 Youngs modulus of the cylinder and the substrate, [Pa]
1, 2 Poissons ratio of the cylinder and the substrate, [ ]
2 12 1 2
E composite Youngs modulus, [Pa] E = E 1 + E 2
1 2
s (x , y ) measured surface topography, [m]
hs(x, y ) separation distance between the bodies, [m]
w (x , y ) wear depth, [m]
Ac direct contact area, [m 2]
inuence matrices for the surface deection due to normal
m
and tangential surface stresses, [ Pa ]
ij , VM subsurface and von Mises stresses, [Pa]
FNij ,FSij inuence matrices for subsurface stress calculation due to
normal and tangential surface stresses, [ ]
U average sliding speed, [m/s]
collapse parameter. In these models the problem is split up into two
separate problems: a smooth surface hydrodynamic lubrication pro-
blem and a dry rough contact problem, for details see [38,39]. The load
sharing concept was used by Bartel et al. [40] along with an energy
threshold based wear model to predict the volumetric wear loss in
mixed lubrication. Prediction of the friction coecient in mixed
lubrication based on load-sharing approach was reported in [41].
Morales-Espejel et al. [42,43] combined a cumulative damage
model and a local Archard wear equation to simulate mild wear and
micropitting accumulation. They found a good qualitative agreement
with the experimental data in the prediction of the degree of micro-
pitting [42] and the topography evolution due to mild wear [44]. This
approach is required for the simulation of the evolution of surfaces as a
function of time. However, in the current paper, the objective is to
predict the particle size formed during running-in and to estimate the
ultimate surface roughness (process roughness) after running-in. The
running-in time is therefore not predicted, which simplies the
approach.
The major goal of the current work is to predict the size of the wear
particles generated in mixed lubricated contacts during running-in and
the main interest is on metallic materials and running conditions
encountered in rolling element cage contacts in bearings. An
engineering algorithm based the on Von Mises stress was used to
model the wear particles generation, as presented in [29]. There are
two conditions to be met for a wear particle to form in a certain
volume: rstly the Von Mises stress should exceed the yield stress and
secondly the volume should touch the surface, see Fig. 1. This wear
model was coupled with a load-sharing based friction model developed
by the authors earlier [41]. Validation of the proposed model was done
using experimental measurements of the initial and run-in friction
coecients and the particle size measurement.
A. Akchurin et al. Tribology International 110 (2017) 201208

Table 1
Properties of the materials.

Property Cylinder, steel Disk, steel

E , GPa 210 210

0.3 0.3
H , GPa 6 1.08
y , GPa 2.14 0.39
R , mm 2
Rq , nm 55 130

Table 2
Oil Properties.

Oil , 1/GPa dynamic viscosity (at 25 C), Pa s

Mineral (Shell Vitrea) 25 0.36

Here the tangential stress is fCb p . The corresponding inuence

functions can be found in reference [13]. The systems (1) (EHL) and
(2) (dry) are solved iteratively, until the load balance condition is met:
Fig. 2. The layout of the combined wear particle formation model.
FT = FC + FH . (4)
2. Problem formulation and solution methods The mean friction coecient is then calculated according to [41]:

2.1. Friction model in mixed lubrication fC =(fCb F + Fsh )/ FT , (5)

C

In mixed lubricated conditions, a part of the load is carried by the
solid contact (contacting asperities) and a part is carried by a lubricant
lm. To predict the formation of a wear particle, it is necessary to
obtain the stress state in the loaded body. In the current work, this was
accomplished using the load sharing concept rstly introduced by
Johnson [37] and the current algorithm is based on the same physical
approximation. The half-space approximation based dry contact model
and a numerical elasto-hydrodynamic (EHL) model for smooth sur-
faces were implemented, which was shown to signicantly improve the
accuracy of the original approach used by Johnson [41]. This model
has been extensively described in [45]. However, to make this paper
easier to read, the major steps are repeated below.
The lubrication model consists of the classical system of smooth
line contact (1 dimensional) EHL equations:
h3 PH
h
U x =0
x 12 x
where fCb
has to be obtained experimentally in the boundary lubrication
regime and Fsh is obtained by integrating the hydrodynamic shear
stress. In Eq. (5), it is assumed that fCb characterizes the local shear
strength of the solid-to-solid contacts and equals the measured
coecient of friction in boundary lubrication as frequently assumed
in mixed lubrication models [38,45].
It should also be mentioned, that in the current work a contact of
two rough surfaces is considered and that a lateral interaction (forces)
of the asperities takes place. In the model, this interaction alters the
local pressures according to Eq. (3). The lateral deection, on the other
hand, was neglected. Measured roughness proles were used in the
simulations, which will be discussed later.
2.2. Wear model
Once the contact pressure is known, the subsurface stresses can be
calculated using the half-space approximation. The subsurface stress

calculation was performed assuming elastic behaviour.

FH = PH (x )dx , The Von Mises stress was calculated from [46]:

h x x2
4
ij (x, y, z ) = (FNij + FSij )p ,
( ) = h0 + PH (x)ln (x x)dx (6)
2R E
(1)
The Von Mises stress based wear particle formation criterion was
where U =(U1 + U2 )/2 , U1 and U2 are the velocities of the moving surfaces. used, following Bosman et al. [22] and validated for boundary
This system is solved numerically assuming smooth surfaces to obtain lubricated conditions in reference [29]. According to this criterion, a
lm thickness h , hydrodynamic pressure PH and hydrodynamic load FH . wear particle is formed if in a certain volume the Von Mises stress
From the calculated lm thickness h , the separating distance hs exceeds the yield stress and if this volume touches the surface. The
between the two rough surfaces is calculated using the basic iterative details of the algorithm are given in [29], but, the major steps of the
algorithm of mass conservation as described by Johnson et al. [37]. The algorithm are also listed here:
separating distance determines the contact force FC according to the
following system of equations: 1) Calculate the equivalent Von Mises stress VM .
u(x, y ) = s(x, y ) hs(x, y ), x, y Ac 2) Find the potential volumes for the wear particle formation:
Vpp = {(x, y, z ): VM > y}.
p(x, y )>0, x, y Ac
, 3) Within Vpp nd the volumes reaching the
p (x , y ) H

FC = p(x, y )dxdy surface:Vp = (x, y, z )V p : z = 0.
p
(2)
4) Within Vp , nd the wear depth w(x, y ) by locating the maximum
where the deection u(x, y ) is related to the pressure according to half- depth at which the yield stress is exceeded: (x, y )Vp ,
space approximation [10]: w(x, y ) = maxpz , (x, y ) Vp,w(x, y )=0 .
z Vp
u(x, y )=(K + fCb S )p . (3)

203
A. Akchurin et al. Tribology International 110 (2017) 201208

2.4. Experimental verication

The particle-by-particle removal model takes into account the

evolution of the surface roughness due to wear, which is closely related
to the running-in process [47]. The combined model was applied to
simulate running-in and was validated using experimentally measured
initial and run in friction coecients, e.g. the transition from a
running-in to a stable system.
The experiments were carried out using cylinder on disk sliding
tests with mineral oil as a lubricant. The cylinder (length of 6 mm ) was
made of AISI 52100 and was sliding against a bearing cage steel disk.
Roughness measurements of the cylinder and disk were performed
using a Keyence VK9700 laser scanning microscope, before and after
Fig. 3. Schematic diagram of the particle isolation procedure (reprinted from [1]). testing. The hardness measurements were performed using a LECO
micro-hardness tester LM100AT on the wear tracks before and after
Table 3 the test. It was found, that the hardness did not change noticeably after
Specification of the wear particle's sizes. the wear test. The material properties are given in Table 1.
A mineral oil (Shell Vitrea) was used, see Table 2. The tests were
Property Mean Value, nm Standard Deviation, nm
performed at room temperature 25 C. The coecient of friction was
Radius 200 145 measured at various sliding velocities at a xed normal load of 5.5 N
Length 726 600 (nominal contact pressure 102 MPa), resulting in a representative
Width 316 280
Stribeck curve. To obtain this curve the friction was measured 3 times
Thickness 405 285
and then averaged at each sliding speed. The friction coecient was
measured for 4 h and in all cases it stabilized by the end of the test. For
5) Identify wear particles based on w(x, y ) and nd the volume (and the comparison with the theoretical model, two levels of friction were
size) of each particle. used: initial and stabilized (after run-in).
The size of the wear particles was analysed using DLS technique as
It should be noted that it was shown recently that the formation of part of the validation of the presented model. After a test at 0.01 m/s
adhesive wear particles is related to the contact junction size [28]. In sliding velocity, the wear particles were separated from the oil. They
the current study, however, only the abrasive wear mechanism is were suspended in deionized water using a centrifuging and sonication
addressed. procedure and analysed using DLS. The particle collection and analysis
procedure is schematically shown in Fig. 3. For details of the experi-
mental procedure the interested reader is referred to [48].
The wear particle size simulation results were presented in the form
2.3. Combined model of equivalent radii of the particle to accommodate a direct comparison
between theory and experiments (DLS determines only equivalent
The wear and friction models were combined as shown in Fig. 2. radii). This was calculated using:
First, the parts of the total load carried by the solid contacts and by the
lubricant lm were calculated using the load sharing concept (also the V 3 1/3
friction coecient was calculated here). Further, based on the load Ri = i ,
4 (7)
carried by the asperities and calculated contact pressure, the Von Mises
stress was obtained and the wear particle formation criterion was
where Ri is the equivalent radius of the particle, and Vi is its volume
checked. Finally, the surfaces were updated to account for wear. This
obtained from the simulation. Actually, the particles are not spherical
process continued until the surface roughness reached the steady state.
as indicated in Table 3 and as it was reported earlier [29].
The computational burden for the calculation of the contact
pressure and subsurface stresses was optimized by the widely applied
discrete fast Fourier transformation technique (DC-FFT) [13].

Fig. 4. Measured surface roughness proles of the cylinder and disk. Spatial resolution is 90 nm.

204
A. Akchurin et al. Tribology International 110 (2017) 201208

Fig. 5. Friction coecient at 0.05 m/s sliding speed a) measured curve b) calculated curve.

Nx Ny Nz=200 200 50 data points with the horizontal spacing

ax = ay=90 nm and the vertical spacing az=90 nm. The computational
grid of the hydrodynamic solver was the same in all simulations and
contained N h=2000 data points with a h=1.1 m (covering a length of
2.2 mm). A typical computational time needed to perform 200 itera-
tions with the considered grids was approximately 4 h with most of the
time taken by the hydrodynamic solver. Simulations were performed
on a 2.5 GHz Intel Core i7 quad-core processor.
It should be noted, that the contact grid covers only a small part of
the grid of the hydrodynamic solver. The grids were centred at the
origin of the coordinate system and the results from the hydrodynamic
solver were linearly interpolated to the contact grid. To reduce the
computational burden only a fraction of the cylinder geometry in the
direction perpendicular to sliding (y direction) was considered.
Periodic boundary conditions were applied in this direction to mimic
the full cylinder geometry. The measured roughness proles were used
in simulations and are shown in Fig. 4.
Fig. 6. Initial and run-in friction coecient curves, experiment vs. calculation.

2.5. Simulation details 3. Results and discussion

In the simulations, a cylinder-on-disk conguration was consid- 3.1. Experimental validation of the model
ered, corresponding to the experimental conditions. The wear of the
cylinder was considered to be negligible, since it was signicantly A typical experimental friction curve is shown in Fig. 5a. Initially,
harder than the counter surface. The value of the boundary friction the friction coecient was high since the surfaces have many irregula-
coecient fCb =0.15 (obtained experimentally) was used as an input to rities, carrying most of the load. With time, these irregularities were
the model and considered to be constant. worn out, and a larger portion of the total load was carried by the
In most of the simulations, the computational domain of the lubricant. This results in a drop in friction coecient towards a steady-
contact model was a parallelepiped which contained state value. The curves were used to nd the initial friction by taking

Fig. 7. Evolution of the a) wear particles size and b) of the direct contact load share (ratio of contact load and the total load) at 0.01 m/s sliding speed.

205
A. Akchurin et al.
Fig. 8. Evolution of surface roughness of the disk at various sliding speeds, a) simulation results, b) experimental data.
Fig. 9. Equilibrium wear particles size as a function of sliding speed.
Fig. 10. Distribution of particles radius at 0.01 m/s sliding speed.
the average of the rst 1% of the data points and the run-in friction
coecient by taking the average of the last 5% of the data points.
A typical friction curve obtained using the model is shown in
Fig. 5b. Here the wear simulations were performed for 200 iterations at
each sliding speed, which was found to be enough to get a stable
friction coecient. The iterations on the x-axis do not refer to time and
these curves were therefore only used to determine the initial friction
(the rst calculated friction value) and the run-in friction coecient (by
taking the average of the 10 last data points). The Stribeck curves were
subsequently built using the initial and run-in friction coecients for
both the experimental and theoretical data, see Fig. 6. Note that all
206
Tribology International 110 (2017) 201208
three lubrication regimes are present in the curves: boundary (BL),
mixed (ML) and full lm lubrication (EHL). Prediction of the initial
friction coecient by the model is in a reasonable agreement with the
experimental data for both the initial and run-in friction coecient. It
can be noticed that for the sliding speeds of 0.03 and 0.05 m/s, the
model underestimates the friction. This is ascribed to the roughening of
the surface due to abrasive wear of particles re-entering the contact, an
eect that is not included in the model.
The largest change in the mean coecient of friction due to
running-in is achieved in the mixed lubrication part of the Stribeck
curve (0.030.05 m/s).
Obviously, the eect of running-in was absent in the EHL regime,
0.20.4 m/s. In boundary lubrication the eect of running-in was less
pronounced. After all, the fraction of the contact separated by the
lubricant is already very small and will not change much by wear.
3.2. Evolution of the Wear Particles Size, Surface Roughness and
Pressure
In this section, the evolution of some important parameters with
respect to the wear process are addressed. The wear model for
boundary lubrication was validated in an earlier paper [29]. Here the
validation will be done for mixed lubrication and therefore a sliding
speed of 0.01 m/s was chosen.
At rst, the evolution of the wear particles size was calculated, as
shown in Fig. 7a. In the beginning, small wear particles were formed,
but their size increased and stabilized. This behaviour, however, is not
general and will depend on the initial surface roughness and the
hardness of the materials. The evolution of the fraction of the load
carried by solid-to-solid contacts is shown in Fig. 7b. The gure shows
that the contact load decreases, which results in a decrease of the
friction coecient during the running-in.
The evolution of the surface roughness (Rq ) with wear at various
sliding speeds is shown in Fig. 8a. Due to larger wear particles formed
at 0.005 m/s, the nal surface was rougher than the surfaces formed at
higher speeds. At 0.03 m/s, after a short period of wear, a full lm
lubrication regime was established and no wear took place. Fig. 8b
shows the measured surface roughness Rq for comparison. It can be
seen, that the surfaces were smoothened after the wear test, which is in
agreement with the model prediction. However, the degree of smooth-
ening predicted by the model is higher. This behaviour can be ascribed
to the presence of the wear particles in the oil. These particle may re-
enter the contact, create dents and scratches and generate rougher
surfaces. Obviouslty, this eect is not described by the model. This
eect will also be addressed below.
A. Akchurin et al.
Fig. 11. a) Theoretical run-in friction coecient and b) wear particles radius as function of hardness at 0.01 m/s sliding speed.
3.3. Wear particles size
In Fig. 9 the equilibrium wear particle size and the corresponding
hydrodynamic lm thickness (central value) are shown as a function of
the sliding speed. Clearly, with the increase of the sliding velocity the
wear particles size decrease. At speeds larger than 0.03 m/s, after
running-in, full lm lubrication was obtained resulting in no wear
particles.
At sliding speeds lower than 0.01 m/s, the particles are much larger
than the separating hydrodynamic lm. As it was mentioned above, the
model assumes that the wear particles do not re-enter the contact.
Since the predicted and measured run-in friction coecient match (see
Fig. 6), the assumption that wear particles will not enter the contact in
boundary lubrication (and close to it) may be valid. At sliding speeds
higher than 0.02 m/s, the central lm thickness becomes larger than
the average diameter of the wear particle. In this region (up to 0.05 m/
s), signicant deviation of the model with the experiments is observed.
It indicates that the wear particles may re-enter the contact since they
are comparable in size with the separating lm thickness. Once re-
entered, they may create additional dents and scratches and increase
the roughness of the surface in contact. Since the model neglects this
eect, it underestimates the friction in this region. At higher speeds,
the separating lm becomes much larger than the mean wear particles
size and the eect of the particles re-entering the contact on the
developed friction coecient is less pronounced.
To validate the wear particles size prediction, DLS measurements of
the wear particles collected after test at 0.01 m/s were performed. The
size of the wear particles according to the model is 200 nm, which is
about the same as those obtained from the experiment (the DLS results
gave 223 13 nm). The wear particles dimensions are given in the
Table 3.
The particles radius distribution is given in Fig. 10. It can be seen,
that the distribution is not Gaussian, but dominated by smaller
particles. A similarly shaped distribution was reported earlier based
on AFM measurements [48].
3.4. Inuence of hardness on the running-in behaviour
To investigate the inuence of the hardness on the severity of
running-in the hardness of the disk and load were varied. The sliding
speed was xed at 0.01 m/s and the hardness of the disk varied from
700 MPa to 5 GPa. The result is shown in Fig. 11a.
It can be seen that there was a clear minimum in the run-in friction
coecient. This result can be explained by the following consideration.
If the hardness and therefore the yield stress is high enough, the
contact is close to an elastic state. Very small particles are formed in
this case as shown in Fig. 11b and the surface changes only slightly.
207
Tribology International 110 (2017) 201208
Hence, there is only minor running-in taking place and the friction
stays high. On the other hand, if the hardness is very low, very large
wear particles form as can be seen in the Fig. 11b. Large wear particles
detachment creates deep pits on the surface. These pits, in turn, retain
large volumes of the lubricant and bring the surfaces closer to each
other and more irregularities come into direct contact. This results in
an increase of the friction coecient. Therefore, there exists an
optimum value of hardness (and wear particle size), which gives a
minimum coecient of friction after running-in. The eect is similar to
the eect of surface texturing: there exists an optimum depth of the
texture feature to provide the maximum lm thickness in the contact
[49]. If the size of the texture is larger or lower than the optimum, the
lm thickness is decreased. The optimum values range will depend on
the applied conditions and the surface roughness of the harder body,
but not the initial roughness of the softer body. Minor inuence of the
initial roughness of the softer body on the running-in is also docu-
mented in literature [50]. As discussed in the reference [29], the initial
roughness of the softer body does not change the size of wear particles
in stationary running-in conditions and according to the theory
presented here cannot aect the run-in friction coecient. It should
also be noted, that the eect of directionality of the surface roughness is
not taken into account in the presented hydrodynamic model. This can
be partly improved by incorporation of ow factors into the lubrication
model [51,52].
4. Conclusion
The generation of the wear particles was eciently modelled using
a half-space approximation approach. The load-sharing concept was
employed to identify the fraction of the total load carried by the
lubricant lm and that of the solid contacts. This was then used to
calculate the shear stresses (friction), which, in combination with the
normal stresses, give the total stress eld. A wear particle is then
generated if the Von Mises stress exceeds the yield stress and if the
volume reaches up to the free surface.
The model was validated using experimental measurements of
friction and its evolution during the running-in process. A reasonable
agreement was observed, proving the possibility to simulate the
running-in due to the surface roughness wear using the presented
wear model. Since the surface roughness evolution is correlated to the
wear particles size, the agreement in the predicted evolution of friction
due to roughness, may indirectly indicate the validity of the particles
size estimation. In addition, DLS measurement of the wear particles
size was performed and compared to the simulations and previously
reported results and a reasonable agreement was observed.
For the particular combination of materials and lubricant consid-
ered in this paper, the surface roughness evolution was the main
A. Akchurin et al.
component of the running-in process. Simulations showed the ex-
istence of an optimum hardness range, which ensures the minimum
friction coecient after the running-in.
Acknowledgements
This research was carried out under project number M21.1.11450
in the framework of the Research Program of the Materials innovation
institute M2i. The authors would like to thank SKF Engineering &
Research Centre, Nieuwegein and Bosch Transmission Technology BV,
Tilburg, The Netherlands for providing technical and nancial support.
References
[1] Hurley S, Cann P, M, Spikes H,A. Lubrication and reow properties of thermally
aged greases. Tribology Trans 2008;43:2218.
[2] Jin X. The Eect of contamination particles on lithium grease deterioration. Lubr
Sci 1995:7.
[3] MacQuarrie RA, Chen YF, Coles C, Anderson GI. Wear-particleinduced osteoclast
osteolysis: the role of particulates and mechanical strain. J Biomed Mater Res Part
B, Appl Biomater 2004;69B:10412.
[4] Green TR, Fisher J, Stone M, Wroblewski BM, Ingham E. Polyethylene particles of a
critical size are necessary for the induction of cytokines by macrophages in vitro.
Biomaterials 1998;19:2297302.
[5] Olofsson U, Olander L, Jansson A. Towards a model for the number of airborne
particles generated from a sliding contact. Wear 2009;267:22526.
[6] Meng HC, Ludema KC. Wear models and predictive equations: their form and
content. Wear 1995;181-183:44357.
[7] Holmberg K, Laukkanen A, Ghabchi A, Rombouts M, Turunen E, Waudby R, et al.
Computational modelling based wear resistance analysis of thick composite coat-
ings. Tribology Int 2014;72:1330.
[8] Molinari JF, Ortiz M, Radovitzky R, Repetto EA. Finite-element modeling of dry
sliding wear in metals. Eng Comput 2001;18:592609.
[9] Salib J, Kligerman Y, Etsion I. A model for potential adhesive wear particle at
sliding Inception of a spherical contact. Tribology Lett 2008;30, [225-3].
[10] Timoshenko SP, Goodier JN. Theory of Elasticity, 3-rd edition. New York: McGraw-
Hill; 1970.
[11] Tian X, Bhushan B. A numerical three-dimensional model for the contact of rough
surfaces by variational principle. J Tribology 1996;118:3342.
[12] Bosman R, de Rooij MB. Transient thermal eects and heat partition in sliding
contacts. J Tribology 2010:132.
[13] Liu S. Thermomechanical contact analysis of rough bodies [PhD thesis].
Northwestern University; 2001.
[14] Polonsky IA, Keer LM. A numerical method for solving rough contact problems
based on the multi-level multi-summation and conjugate gradient techniques.
Wear 1999;231:20619.
[15] OSullivan TC, King RB. Sliding contact stress eld due to a spherical indenter on a
layered elastic half-space. J Tribology 1988;110:23540.
[16] Liu S, Wang Q. Studying contact stress elds caused by surface tractions with a
discrete convolution and fast Fourier transform algorithm. J Tribology
2002;124:3645.
[17] Bosman R, Schipper DJ. On the Transition from Mild to Severe Wear of Lubricated,
Concentrated Contacts: the IRG (OECD) Transition Diagram. Wear, 269 2010;
2010. p. 5819.
[18] Jerbi H, Nelias D, Baietto M-C. Semi-analytical model for coated contacts. fretting
wear simulation. Leeds-Lyon Symposium on Tribology. Lyon, France; 2015.
[19] De Moerlooze K, Al-Bender F, Van Brussel H. A novel energy-based generic wear
model at the asperity level. Wear 2011;270:76070.
[20] Boher C, Barrau O, Gras R, Rezai-Aria F. A wear model based on cumulative cyclic
plastic straining. Wear 2009;267:108794.
[21] Beheshti A, Khonsari MM. An engineering approach for the prediction of wear in
mixed lubricated contacts. Wear 2013;308:12131.
[22] Bosman R, Schipper D, J. Transition from mild to severe wear including running in
208
View publication stats
Tribology International 110 (2017) 201208
eects. Wear 2011;270:4728.
[23] Huq M, Z, Celis J-P. Expressing wear rate in sliding contacts based on dissipated
energy. Wear 2002;252:37583.
[24] Aghdam AB, Khonsari MM. On the correlation between wear and entropy in dry
sliding contact. Wear 2011;270:78190.
[25] Ajayi OO, Lorenzo-Martin C, Erck RA, Fenske GR. Analytical predictive modeling
of scung initiation in metallic materials in sliding contact. Wear 2013;301(s 1
2):5761.
[26] Nlias D, Boucly V, Brunet M. Elastic-plastic contact between rough surfaces:
proposal for a wear or running-in model. J Tribol 2005;128(2):23644.
[27] Nelias D, Boucly V, Brunet M. Elastic-plastic contact between rough surfaces:
proposal for a wear or running-inrunning-in model. J Tribol 2006;128:23644.
[28] Aghababaei R, Warner DH, Molinari J-F. Critical length scale controls adhesive
wear mechanisms. Nat Commun 2016:7.
[29] Akchurin A, Bosman R, Lugt PM. A Stress-criterion-based model for the prediction
of the size of wear particles in boundary lubricated contacts. Tribol Lett; 2016.
[30] Lugt PM, Morales-Espejel GE. A review of elasto-hydrodynamic lubrication theory.
Tribol Trans 2011;54:47096.
[31] Spikes HA. Mixed lubrication - an overview. Lubr Sci 1997;9:22153.
[32] Evans HP, Snidle RW, Sharif KJ. Deterministic mixed lubrication modelling using
roughness measurements in gear applications. Tribol Int 2009;42:140617.
[33] Zhu D, Hu Y-Z. A computer program package for the prediction of EHL and mixed
lubrication characteristics, friction, subsurface stresses and ash temperatures
based on measured 3-D surface roughness. Tribol Trans 2008;44:38390.
[34] Holmes MJA, Evans HP, Snidle RW. Discussion on elastohydrodynamic lubrication
in extended Parameter ranges, Part II: load eect. Tribol Trans 2003;46:2828.
[35] Zhu D, Liu Y, Wang Q. On the numerical accuracy of rough surface EHL solution.
Tribol Trans 2014;57:57080.
[36] Holmes MJA, Evans HP, Snidle RW. Miscellany: discussion. Tribol Trans
2003;46:2828.
[37] Johnson KL, Greenwood JA, Poon SY. A simple theory of asperity contact in elasto-
hydrodynamic lubrication. Wear 1972;19:91108.
[38] Morales-Espejel GE, Wemekamp AW, Felix-Quinonez A. Micro-geometry eects on
the sliding friction transition in elastohydrodynamic lubrication. J Eng Tribology
2010;224:62137.
[39] Bobach L, Beilicke R, Bartel D, Deters L. Thermal elastohydrodynamic simulation
of involute spur gears incorporating mixed friction. Tribol Int 2012;48:191206.
[40] Bartel D, Bobach L, Illner T, Deters L. Simulating transient wear characteristics of
journal bearings subjected to mixed friction. Proc IMechE Part: J Eng Tribol
2012;226:1095108.
[41] Akchurin A, Bosman R, Lugt PM, van Drogen M. On a model for the prediction of
the friction coecient in mixed lubrication based on a load-sharing concept. Tribol
Lett 2015;59:1930.
[42] Morales-Espejel GE, Brizmer V. Micropitting Modelling in rolling-sliding contacts:
application to rolling bearings. Tribol Trans 2011;54:62543.
[43] Brizmer V, Pasaribu HR, Morales-Espejel GE. Micropitting performance of oil
additives in lubricated rolling contacts. Tribol Trans 2013;56:73948.
[44] Morales-Espejel GE, Brizmer V, Piras E. Roughness evolution in mixed lubrication
condition due to mild wear. Proc IMechE Part: J Eng Tribol 2015;229:133046.
[45] Faraon I, Schipper DJ. Stribeck curve for starved line contacts. J Tribol
2007;129:1817.
[46] Jacq C, Nelias D, Lormand G, Girodin D. Development of a thrree-dimensional
semi-analytical elastic-plastic contact code. J Tribol 2002;124:65367.
[47] Jamari J. Running-in of rolling contacts. Enschede: University of Twente; 2006.
[48] Akchurin A, Bosman R, Lugt PM. Analysis of wear particles formed in boundary
lubricated sliding contacts. Tribol Lett 2016;63:114.
[49] Bijani D, Schipper DJ, Deladi EL. The inuence of surface texturing on the lm
thickness in parallel sliding surfaces. 20-th International Colloquium Tribology
Industrial and Automotive Lubrication. Stuttgart/Ostldern, Germany; 2016.
[50] Rowe GW, Kaliszer H, Trmal G, Cotter A. Running-in of plain bearings. Wear
1975;34:114.
[51] Patir N, Cheng HS. An average ow model for determining eects of three-
dimensional roughness on Partial hydrodynamic lubrication. J Lubr Technol
1978;100:127.
[52] Sahlin F, Almqvist A, Larsson R, Glavatskih S. Rough surface ow factors in full
lm lubrication based on a homogenization technique. Tribol Int
2007;40:102534.