Wear 289 (2012) 26–38

Contents lists available at SciVerse ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Finite element modeling and experimental validation of fretting

wear scars in thin steel wires
A. Cruzado a,n, M.A. Urchegui b, X. Gómez a
a
Mondragon Goi Eskola Politeknikoa, Mondragon Unibertsitatea, Loramendi 4, 20500 Arrasate-Mondragon, Spain
b
ORONA eic, Polı́gono Industrial Lastaola s/n, 20120 Hernani, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Fretting wear is one of the main degradation mechanism produced in steel wire ropes. The experimental
Received 3 September 2011 characterization of this phenomenon is economically costly and too much time consuming. With the aim to
Received in revised form reduce this disadvantage, an optimized methodology for wear scars simulation has been developed. This
13 April 2012
paper presents the two relevant aspects of this methodology, which consist of the reduction of
Accepted 19 April 2012
computational time and the validation of the wear scar with respect to the experimental data. This
Available online 19 May 2012
methodology has been implemented via the commercial FE code ABAQUS in a 901 crossed-cylinders
Keywords: configuration. With respect to the reduction in computational time the effect of mesh size, simulation wear
Fretting wear increments per fretting cycle and cycle jump technique are studied. An optimum mesh size for minimal
Wires
computational time that should be in the range from 3% to 4% of the final longitudinal wear width is
FEM
defined. On the other hand, an optimum value of 40 increments per fretting cycle and 100 fretting cycles
Crossing angle
Parameters optimization with the corresponding cycle jump is enough to simulate the wear scar in any conditions with differences
in dimensions less than 4%, in comparison with more refined simulations. The results of the optimized
model are validated with respect to the experimental data obtained by the tests carried out in a small
sliding amplitudes tribometer for the same crossed cylinder configuration. The validation consists of three
steps: wear scar dimensions, wear scar depth and wear scar volume. This procedure demonstrates that
within an error of 10% this optimized method predicts accurately the wear scars.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction The experimental analysis of the different parameters involved

Steel wire ropes are used in many industrial applications, like
structural elements (reinforcement for tires, bridges, barges, etc.)
or as elements for transporting purposes (cranes, lifts, funicular
railway, ski lifts) due to their high axial strength and bending
flexibility. These mechanical properties are largely dependent on
their construction and the properties of the wires itself, because
the wire ropes are composed of individual wires wound into
strands, which are then wound to form the final rope (Fig. 1).
Nevertheless in the case of ropes running over sheaves, which
it is the main traction system used in lift industry, an oscillatory
motion between the wires is produced and as a result fretting
wear is prone to occur [1,2]. Urchegui et al. [3] studied this effect
into a 7  19 stranded rope, where the fretting problem is
presented in point contacts, typically between the outermost
wires of strands and core or between the outermost wires of
adjacent strands. As a consequence, high level of wear exists in
wires and the failure may result from increased stresses at worn
cross-sections.
n
Corresponding author. Tel.: þ34 943 25 33 63; fax: þ 34 943 79 15 36.
E-mail address: acruzado@mondragon.edu (A. Cruzado).
in fretting wear, require high number of tests, with high time
consuming and the cost associated to these tests. For this reason
and due to the complexity of the wire rope construction and the
difficulty of the detection of wear in point contacts during their
life, in this work, a methodology for the simulation fretting wear
in thin steel wires has been developed and validated. This
methodology is presented as a powerful tool for the design of
new wire ropes, taking into account the reduction in the cross
section that cause the failure of the wires as a consequence
of fretting wear. In this point the wear simulation using FEM
(finite element methodology) has been chosen.
The wear simulation with FEM has been presented by many
authors with the aim to improve the knowledge of the evolution of
fretting and predict trends. Johanson [4] was the first author who
incorporates in a FE model a local implementation of Archard’s wear
law to evaluate the contact pressure evolution due to change of
contact geometry. More recently McColl et al. [5] developed an
incremental wear approach for the simulation of fretting wear in
cylinder on flat configuration. Following this work, Ding et al. [6]
presented a finite element based approach to simulate the effects of
debris, Madge et al. [7] analyzed both fretting wear and fretting
fatigue and Mohd Tobi et al. [8] presented the influence of cyclic
plasticity. On the other hand Fouvry et al. [9] presented and

0043-1648/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.wear.2012.04.018
 A. Cruzado et al. / Wear 289 (2012) 26–38
Fig. 1. Wire rope used in electric elevators for people transporting.
validated the wear simulation in a cylinder plane configuration
using a local wear model approach based on the dissipated friction
energy. Following these works this paper presents a methodology
for simulation fretting wear in 901 crossed cylinders 3D problem
combining different strategies for computational time reduction and
using the local Archard’s wear law.
The wear simulation therefore, is based on an iterative process, in
which the domain of space is divided into small increments of slip
where both normal and tangential problem are solved to apply a
local wear law as mentioned previously. This implies a high
computational cost, since the resolution of contact must be repeated
as many times as iterations of wear have been defined. In the case of
steel wires the problem is much more complex because, firstly it is a
3D problem that involves a high computational cost due to the
significant increase of the degrees of freedom and secondly is
presented as a multiple contact problem with the associate com-
plexity for the analysis of the different parameters that play when
the wear is simulated. Taking into account these problems, this
paper has limited the problem to a crossed cylinders system
subjected to fretting that represent two wires in contact, in order
to study the wear behavior of different geometric variables (inter-
section angles, contact radius, wire geometry) and operational
(contact pressure, stroke). Hence and optimized methodology for
simulation both fretting and reciprocating sliding wear in 3D time
consuming problems is presented. Following the experimental work
carried out by Cruzado et al. [10], where the effect of contact
pressure in thin steel wires for a 901 crossed cylinder configuration
was analyzed, a FE model for the same configuration has been
developed. This study has been done as a previous work to extend to
any configuration presented in wire ropes.
This work has been carried out using the ABAQUS user
subroutine UMESHMOTION, which has been validated by many
authors, i.e., Madge et al. [11] has validated this subroutine in
both partial and gross sliding regimes, and Hegadekatte et al. [12]
for sliding and slipping wear in pin-on-disc and twin-disc tribo-
systems. In both cases the authors have simplified the model to a
two dimensional problem.
From the results provided by this tool, an increase in the
knowledge of the evolution of the degradation of steel wires has
been presented and criteria for increasing safety and durability of
these elements could be defined
2. Experimental details
2.1. Tribological testing
The extended description of the tribological testing and results
is presented in [10]. It is briefly outlined here for completeness.
The tests were carried out with a 901 crossed cylinders config-
uration. The wires used for the experiments are wires which are
27
Fig. 2. Schematic of a wire rope composed of different strands of wound
steel wires.
usually fabricated into a 7  19 rope (7 strands with 19 individual
wires in each one) as presented in Fig. 2. For the tests wires with a
diameter of 0.45 mm were used. They are cold-drawn from
eutectoid carbon steel with 0.8% C with a tensile strength greater
than 2800 MPa and a hardness of 659 781 HV0.05. The surface
average roughness (Ra) of these wires along its axis was 0.35 and
0.70 mm in perpendicular direction. These values were obtained
from the measurements carried out at an unworn wire using a
confocal imaging profiler (Plm, Sensofar).
Fretting wear tests were carried out on specific tribometer for
small sliding amplitudes (10–200 mm) and small loads (0.1–10 N).
The tribometer consists of a top arm with an integrated load
cell for the measurement of the friction force and a bottom
vibrating horizontal table. The movement of this table is given
by a frequency-controlled DC motor with an eccentric adjustment
[13]. Special holders were designed for the correct fixing of the
wires as shown in Fig. 3. In these holders the wires represent a
toroid with a curvature radius of 10 mm. The upper holder
(stationary sample), is fixed at the end of a horizontal arm, and
the bottom holder is fixed to the vibrating horizontal table in the
tribometer, with the wire axis parallel to the direction of motion.
During each test the real stroke (Dxreal), friction force (Ff) and
total linear wear (Wl,tot) were measured. Relative humidity (RH)
and temperature (T) were kept constant during the test. All
measured quantities were recorded on-line multiple times. After
each test friction and wear quantities are derived from the stored
values. The planimetric wear (Wq) and the volumetric wear (Wv)
subsequently determined from wear scar profilometry (Hommel-
tester, Fa. Hommel, Homburg, Germany).
The tests were carried out in a range of average contact
pressures that varies from 1960 MPa to 3560 MPa, two strokes
65 mm and 130 mm in a range of cycles that ranges from 20,000 to
200,000. The specific details of these tests are shown in Table 1.
2.2. Tribological quantities
For the wear simulation by FEM, there are two representative
tribological quantities. The first one corresponds to the friction
behavior which is described by the coefficient of friction.
The coefficient of friction f in each cycle is obtained by dividing
the friction force Ff by the normal load Fn according to Eq. (1) and
thus stored 500 times per test. However the mean coefficient of
friction used in this research fav represents the average value
calculated from the data of the last half of the experiment.
The friction force as a function of the displacement is mea-
sured 120 times during a closed motion cycle. The hysteresis
shows a trapezoidal shape with an inclination in the flanks due to
the stiffness of the system including the fixation of the wires and
28 A. Cruzado et al. / Wear 289 (2012) 26–38

Fig. 3. Wire support holders for bottom (left) and top (right) wires.

Table 1
Fretting test conditions for different mean pressures and 901 crossing angle.

Parameter Symbol Value Tribosystem

Normal load (N) Fn 0.5; 1; 2; 3

Average contact pressure (MPa) Pav 1960; 2470; 3100; 3560
Maximum contact pressure (MPa) Pmax 2920; 3690; 4650; 5400
Stroke (mm) Dx 65 75; 130 7 5
Frequency (Hz) f 10
Number of cycles (103) n 20; 50; 100; 200
Lubricant None
Temperature (1C) T 25 71
Atmosphere Laboratory air
Relative humidity (%) RH 507 5

Δxpeak-peak the tribological contact (Fig. 4).

F f ,av
4 f av ¼ ð1Þ
Fn
3 where Ff,av is the average friction force; fav is average friction
2 coefficient and Fn is normal force.
Friction force Ff (N)

The second one corresponds to the wear behavior which is

1 described by the coefficient of wear presented in Eq. (2). The total
0 volumetric wear is the sum of the wear volumes at both specimens.
Wv Wv
-1 k¼ ¼ ð2Þ
sF n 2Dxreal nF n
-2
The volumetric wear is obtained by confocal microscopy (Fig. 5),
-3 creating the entire topography of the wear scar, using a method
described by Urchegui et al. [14].
-4
-40 -30 -20 -10 0 10 20 30 40
Displacement x (m) 3. FEM methodology

Δxreal
3.1. FE model

Fig. 4. Hysteresis loop of friction force Ff during one cycle of displacement equal The FE model used for the simulation of fretting wear in thin
to 2Dx. steel wires is shown in Fig. 6. In this study a 901 crossed cylinders

configuration, the same as the specimens in the tests, with a
diameter of 0.45 mm has been chosen. Moreover the curvature
radius that acquires the wire in the fixation support (10 mm) is
included. The meshes are carried out according to a partition
methodology, in which a more refine mesh is used in the contact
zone, to capture precisely the variation of contact pressures and
geometry. Moreover the meshes in the contact zone have a square
shape. A study of the optimum mesh size for minimal computa-
tional time will be presented in point 4.1.
Moreover the parametrization of the different parameters of
the model (geometry, partitions and meshes) are included with
the aim to reduce the cost that implies new models with different
geometrical parameters. In this case the great advantage is that
the contact partitions shown in Fig. 6(b) and (c) can be modeled
easily for different wear scars dimension, with the aim to reduce
the computational time depending on the intensity of the wear.
Three-dimensional, eight nodes lineal brick elements are used in
the analysis because of second order elements can give fluctuating
pressure distribution in frictional contact problems [18].
The contact surface interaction between both cylinders is
defined via the finite sliding contact pair algorithm, which uses
the master–slave algorithm to enforce the contact constraints.
40
20
0
z m
-20
-40
-60
400
350
300
250
y  200 150
200
m 150
100 100
50 50
0 0
Fig. 5. Confocal image treated in Matlab of the bottom wear scars.
Fig. 6. FE model and mesh details of 901 crossed cylinders: (a) full model, (b) contact region detail for top cylinder and (c) contact region detail for bottom cylinder.
A. Cruzado et al. / Wear 289 (2012) 26–38 29
In this study the top surface is chosen as slave surface and the
bottom one as master. This configuration was chosen because
after doing a previous study was seen that the other configura-
tion, i.e. the top surface as master and the bottom one as slave,
produces convergence problems in the resolution of the contact
problem. The Lagrange multiplier contact algorithm was used to
ensure the exact stick constrain when the shear stress is less than
the critical value according to the Coulomb friction law. A COF of
0.7 is chosen according to the experimental tests carried out in
thin steel wires [11].
The load and the boundary conditions are applied in the free
surfaces of the model through a surface-based coupling, which
couples the motion of all free surfaces to the motion of a reference
node. Moreover a kinematic coupling is used, so as the rigid body
motion is defined by the reference node. It allows in the upper
wire, where a normal force is applied in the reference point, only
the vertical movement and in the bottom wire, where an alter-
native displacement condition is imposed, only the horizontal
movement. This is because the wire maintains the rigid body
movement of the holder in the testing machine. The elastic
modulus and Poisson’s ratio of both cylinders are taken as
210 GPa and 0.33, respectively.
In order to validate the model the contact pressures obtained
by FEM has been compared with the well know contact pressure
distribution obtained according to the analytical solution for the
general cases given by Hertz, which were described by Johnson [15].
Therefore the contact pressure distribution is given by Eq. (3)
rffiffiffiffiffiffiffiffiffiffiffiffiffi
x2
500 pðxÞ ¼ pmax 1 2 ð3Þ
450 a
400
350
300 where the maximum contact pressure is given by Eq. (4):
250
m
x
Pmax ¼ 3F n =ð2p abÞ ð4Þ
where Fn is the applied normal load and a and b are the semi-axes
of the elliptical contact. To obtain both semi-axes Eqs. (5) and (6)
30 A. Cruzado et al. / Wear 289 (2012) 26–38

are solved using the bisection method: where Dh(x,t), p(x,t) and s(x,t) are the incremental wear depth, the
2 contact pressure and the relative slip for a specific point x (node) at
ða=bÞ EðeÞKðeÞ
Rrel ¼ ð5Þ specific time t, respectively. On the other hand kl represents the local
KðeÞEðeÞ
wear coefficient. This coefficient in principle could not be the same
  hn o as the volumetric coefficient of wear, because the last one represents
3F n Re 4 i1=2
ðabÞ3=2 ¼ ðb=aÞ3=2 ða=bÞ2 EðeÞKðeÞ KðeÞEðeÞ an average value across the complete wear scar. As there is not any
4Ee pe 2
method for estimating the local wear coefficient kl, in this study and
ð6Þ
with the aim to simplify the problem it is assumed that the
where E(e) and K(e) are complete elliptic integrals of argument volumetric coefficient of wear k is the same as the local coefficient
e¼(1 b2/a2)1/2, boa. Re is the equivalent radius of curvature and of wear kl. Moreover some authors like McColl et al. [5] or Fouvry
Rrel is the ratio of the relative curvatures R0 and R00 . Both et al. [16] has justified the close relation existed between the local
parameters are obtained according to Eqs. (7) and (8) respec- wear coefficient and the volumetric wear coefficient.
tively: Taking into account Eq. (14), one fretting cycle is divided in
small slip increments; it implies that the FE wear simulation is
Re ¼ 12ðABÞ1=2 ð7Þ
the product between these increments and the total number of
 0 cycles. As a consequence, high computational time is needed for
B R
Rrel ¼ ¼ ð8Þ the simulation of one fretting test. In this point cycle jump
A R00
technique proposed by different authors like McColl et al. [5]
where A and B are positive constants and are obtained solving and Mary et al. [17] is used, where it is made the assumption that
Eqs. (9) and (10): the contact pressure and the slip distribution remain constant in
  each finite element analysis over the next DN cycles. In this way
1 1 1 1 1
ðAþ BÞ ¼ 0 þ 00 þ 0 þ 00 ð9Þ Eq. (14) is multiplied by DN to obtain Eq. (15), such that the wear
2 R1 R1 R2 R2
corresponding to DN cycles can be simulated in one fretting cycle.
( 2  2    )1=2
1 1 1 1 1 1 1 1 1 Dhðx,tÞ ¼ DNkl pðx,tÞsðx,tÞ ð15Þ
9BA9 ¼  þ  þ2  0  00 cos 2a
2 R01 R001 R02 R002 R01 R001 R2 R2
ð10Þ
3.3. Wear simulation method
R01 , R001 , R02 , R002 are the principal radius of curvature of upper and
bottom wire and a is the crossing angle between both bodies. Ee is
Wear simulation consists of an iterative process called wear
the composite modulus of the two contacting bodies and is
simulation algorithm which is presented in Fig. 7 and it is imple-
obtained according to Eq. (11):
mented via the commercial FE code ABAQUS (Version 6.9) and the
1n21 1n22 user-defined subroutine UMESHMOTION. This subroutine allows to
Ee ¼ þ ð11Þ
E1 E2 define the motion of nodes inside an adaptative mesh constraint
where E1, E2 are the elastic modulus and v1, v2 are Poisson’s ratios node set an amount equal to the local wear obtained by the Archard
of upper and bottom wires, respectively. The resolution of these modified wear equation. The nodes are moved in the local normal
equations is done with a specific program developed in Matlabs. direction [18].
For this specific study R0 represents the radius of the cylinder The wear algorithm begins with the construction of the FE model
(0.225 mm) and R00 represents the curvature radius of specimen in described previously. Once the equilibrium equations for one sliding
the holder. increment converge, the adaptative meshing algorithm of Abaqus
applies the local wear in two steps in an Eulerian analysis. First, the
surface nodes are moved an amount equal to the local wear creating
3.2. Wear local model a new mesh through a process called sweeping. The main objective
of the mesh sweeping is the reduction of the element distortion.
For the wear simulation in any tribological system it is Second, the material quantities from the old mesh to the new mesh
necessary to develop a local wear model. In this study the Archard are remapping with a process called advection by solving the
modified equation developed by McColl et al. [5] for fretting wear advection equations using a second order numerical method called
problems is used. With this equation it is assumed that fretting the Lax–Wendroff method. Nevertheless the advection of the
wear can be evaluated by applying Archard’s equation to local material quantities will generally result in loss of equilibrium, which
contact conditions along the entire contact interface. is corrected by solving the last time increment of the contact
The local wear model begins with the Archard model presented in problem. This procedure is repeated in many sliding increments as
Eq. (12), where the volumetric wear Wv is proportional to the normal needed to complete the entire fretting cycles.
force Fn and the sliding distance s and inversely proportional to the On the other hand UMESHMOTION subroutine only gives the
hardness of the material. K is the dimensionless wear coefficient of contact results for one of the contacting surfaces (slave surface),
the material. i.e., in situations that can be modeled as deformable rigid contact
Fns problems, where it is enough to simulate the wear in one of the
Wv ¼ K ð12Þ
H contacting surfaces because the wear in the other surface is
negligible [19]. As in this work, according to the experimental
If Eq. (12) is considered in each nodal contact point dA of the
results, the wear in both contacting bodies is significant, the study
FE model contact surface over an incremental local slip ds, the
of different interpolation methods were carried out. It lies in
incremental nodal wear dh can be expressed according to Eq. (13).
interpolating the contact pressures and the slip distribution from
In the case of fretting wear problems as described by McColl et al.
slave nodes to the master nodes, taking into account the nodal
[5], Eq. (13) results in Eq. (14).
coordinates of each contact surface.
dh ¼ kl pðxÞds ð13Þ The studied interpolation methods are: bivariate interpolation,
triangle-based linear interpolation (baricentric interpolation) and
Dhðx,tÞ ¼ kl pðx,tÞsðx,tÞ ð14Þ nearest interpolation. And extended description of the bivariate
 A. Cruzado et al. / Wear 289 (2012) 26–38
Fig. 7. Wear simulation algorithm including UMESMOTION subroutine.
method is found in the work developed by Akima [20]. In this
method the plane is divided into triangles which maximize the
minimum interior angle of the triangles cells according to the
nodal coordinates and then the desired z value is obtained by
interpolating in the closest x and y position triangle with a
bivariate fifth-degree polynomial as presented in Eq. (16). Where
q are the 21 coefficients of each local polygon, x and y are
coordinate distances relative to some local origin, and j and k
are dummy variables. Moreover Akima’s method can also be used
for extrapolation, though the accuracy decreases for such an
application. The coefficients of the local polynomials are obtained
once the triangles are formed, such that the partial derivatives of
the surface are continuous across the sides of the triangles and at
the vertices.
X 5j
5 X
zðx,yÞ ¼ qjk xj yk
j¼0k¼0
The second interpolation method consists of triangle-based
linear interpolation, described in detail by Watson [21]. In this
case the desired z value is obtained by Eq. (17), where z(i)
represents the z value of one node of the closest triangle in the
position (x, y) and bc represents the baricentric coordinates of the
triangle, which can be obtained following Eqs. (18)–(21).
X
3
zðx,yÞ ¼ zðiÞbcðiÞ
i¼1
del ¼ ½ðxð2Þxð1ÞÞðyð3Þyð1ÞÞ½ðxð3Þxð1ÞÞðyð2Þyð1ÞÞ
½ðxð1ÞxiÞðyð2ÞyiÞðxð2ÞxiÞðyð1ÞyiÞ
bcð3Þ ¼
del
½ðxð3ÞxiÞðyð1ÞyiÞðxð1ÞxiÞðyð3ÞyiÞ
bcð2Þ ¼
del
½ðxð2ÞxiÞðyð3ÞyiÞðxð3ÞxiÞðyð2ÞyiÞ
bcð1Þ ¼
del
In the case of the nearest interpolation it is not needed any
triangulation of the surface, because it consists of obtaining the
desired z value according to de z value of the nearest point npi to
the desired x and y position as described in Eq. (22), where npi is
obtained according to the node point position that meets with the
condition of the minimum distance described in Eq. (23).
31
The distance between the desired x and y position (master node)
and each slave node xi and yi is obtained according to the Eq. (24),
where ndp represents the number of data points in the slave
surface, which cover the interval of numbers between the first
slave node and the last number of slave points nsp.
zðx,yÞ ¼ zðnpiÞ ð22Þ
npi-ndpðminðdistÞÞ ð23Þ
dist ¼ ðxðndpÞxiÞ2 þ ðyðndpÞyiÞ2 -ndp½1,nsp ð24Þ
3.4. FE validation procedure
The validation of the FE wear simulation methodology has
been carried out following three different procedures. The first
one corresponds to the validation of wear scars dimension,
comparing both dimensions of each wear scars obtained by FEM
with the dimensions obtained by Scanning Electronic Microscopy
(SEM) as presented in Fig. 8(a) and (c). The second one corre-
sponds to validate the wear scar depth with the profile obtained
by FEM comparing the longitudinal profile of the wire in the
middle of the wear scar, where the maximum wear depth is
found, as is shown in Fig. 8(b) and (d).
The last validation corresponds to the comparison of the wear
scar volume obtained experimentally with the volume obtained
by FEM. The volumetric wear (Wv) of the wear scar was calculated
by numerical integration (trapezoidal rule): the planimetric wear
(Wp) of each profile obtained in the longitudinal direction, as it is
shown in Fig. 9, is multiplied by the mesh size along the
transversal direction of the specimen. This procedure allows to
validate accurately the shape of the entire wear scar in both top
and bottom specimens.
4. Results and discussions
ð16Þ 4.1. Wear modeling and computational time optimization
As has been mentioned previously, FE wear simulation method
consists of an iterative process which requires too much time for
one wear scar simulation. In the case of 3D problems or in cases
where high number of fretting cycles is analyzed, this time
increase drastically because of the increase in the freedom
degrees and in the number of iterations, respectively.
To reduce the computational time, in this section, two differ-
ð17Þ ent studies are presented. In the first one the analysis of the three
interpolation methods presented previously will be carried out.
In the second one the study of the principal parameters involved
ð18Þ in the FE wear method will be studied: contact zone mesh size
refinement, FE wear increments per fretting cycle and the cycle
jump technique DN. After this study the relation between the
ð19Þ
error committed in the final wear scar and the reduction in
computational time will be obtained.
ð20Þ It should be pointed that for this study the same wear
coefficient k¼30  10  6 mm3/Nm in both specimens, a normal
force of Fn ¼2 N and stroke of Dx ¼65 mm has been used.
ð21Þ
4.1.1. Interpolation method study
This section presents the analysis of the three interpolation
methods (bivariate, baricentric and nearest) presented previously.
For this, the differences obtained in the results of longitudinal
wear scar with the different interpolation methods in each
specimen are compared.
32 A. Cruzado et al. / Wear 289 (2012) 26–38

X = 219m 40
Y = 214m Original
Experimental

Wear depth [μm]

30

20

10

0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
Longitudinal contact widht [mm]

X = 322m
Y = 220m 0

Wear depth [μm]

-10

-20
Original
-30
Experimental
-40
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
Longitudinal contact width [mm]

Fig. 8. Wear scars of top and bottom: (a) top wear scar obtained by SEM, (b) top longitudinal profile obtained by confocal microscopy, (c) bottom wear scar obtained by
SEM and (d) bottom longitudinal profile obtained by confocal microscopy.

0.04 As the nearest method is presented as the fastest method, now

Original the validation of this methodology is analyzed. For this, the
MEF bottom surface is considered as the slave surface and the inter-
0.03
Wear depth [mm]

polation is done in the top surface. After this the interpolated

0.02
0.01
0 tion is made in the top surface, problems with respect to the
-0.3 -0.2 -0.1 0 0.1 0.2
Longitudinal contact widht x [mm]
Fig. 9. Planimetric wear of one profile of top wear scar.
Fig. 10 shows the longitudinal wear scars of top (Fig.10(a)) and
bottom (Fig.10(c)) specimens. In the slave surface the difference
in the wear between the three methods scar is depreciable, as it
can be seen in the magnification of the profile (Fig.10(b)).
However, in the master surface (Fig. 10(d)) the different inter-
polation methods present different wear depth; nevertheless the
wear scar width is the same in all cases. The baricentric method
produce the shallowest wear scar, the nearest method produce
the deepest one and the bivariate method is in the middle of both
methods. Analyzing Fig. 10(d) it is obtained that the difference
between baricentric and nearest method is around 9%. Never-
theless the nearest method is three times faster than the other
two methods, because it is not needed any triangulation method.
If only the contact pressure is interpolated and the relative
sliding of each node is considered the same as the imposed
displacement, Fig. 10(d) shows that the difference due to the
interpolation is depreciable. After this it can be concluded that the
difference presented previously between the different methods is
due to the interpolation the relative slips of each node.
wear scars can be compared directly with the wear scars obtained
by ABAQUS without interpolating. When the interpolation is
made in the top surface Fig. 11(a) the difference is less than 1%
and when the interpolation is made in the bottom surface
Fig. 11(b) the difference is less than 2%, being in both interpola-
tion the wear scar width the same. Nevertheless if the interpola-
0.3 contact problem stability appears and a little bit more time is
needed for interpolation, because more points enter in the nearest
method for each node interpolated, as a consequence of the
greater number of nodes that are in the bottom surface.
Taking into account that the nearest interpolation is the less
time consuming method and that the differences in comparison
with ABAQUS results is less than 2%, the wear FE methodology for
thin steel wires it is carried out with nearest interpolation.
4.1.2. FE wear parameters optimization
This section presents the analysis of the three principal
parameters involved in wear simulation (mesh size, increments
per fretting cycle and cycle jump). For this analysis the long-
itudinal profile of the bottom wear scar is studied, because it is
the most critical. On the one hand it is the wear scar obtained
with the contact pressures and slip distribution interpolated from
the top surface (slave surface) with the nearest interpolated
method mentioned in point 4.1.1 and on the other hand this
profile presents distortions in geometry faster than the top one.
The first and the most important parameter to reduce the
computational time is the mesh size in the contact zone. For this
analysis the other two parameters and the number of cycles have
been fixed: a cycle jump of DN ¼200, 40 wear simulation
 A. Cruzado et al. / Wear 289 (2012) 26–38 33

Original
0.014 Original 0.0129 Nearest
Nearest Bivariate
0.012 Bivariate 0.0128 Baricentric
Baricentric Near: no slip
0.01
Wear depth [mm]

Wear depth [mm]

Near: no slip Bivar: no slip
0.0127
Bivar: no slip Bari: no slip
0.008 Bari: no slip
0.0126
0.006
0.0125
0.004

0.002 0.0124

0
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 -0.01 -0.005 0 0.005 0.01
Longitudinal contact widht y [mm] Longitudinal contact widht y [mm]

0 -0.0075

-0.002 -0.0077
Wear depth [mm]

Wear depth [mm]

-0.004
-0.0079
-0.006
Original
Original Nearest
-0.008 Nearest -0.0081
Vibariate
Vibariate Baricentric
-0.01 Baricentric
-0.0083 Near: no slip
Near: no slip Vibar: no slip
-0.012 Vibar: no slip Bari: no slip
Bari: no slip
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 -0.024 -0.014 -0.004 0.006 0.016 0.024
Longitudinal contact widht y [mm] Longitudinal contact widht y [mm]

Fig. 10. Longitudinal wear scar comparison with different interpolation methods: (a) top wear scar, (b) magnification of (a), (c) bottom wear scar and
(d) magnification of (c).

0.014 0
Original
0.012 Abaqus -0.002
Interpolated
Wear depth [mm]

0.01 -0.004
Wear depth [mm]

0.008 -0.006

0.006 -0.008

0.004 -0.01 Original

Interpolated
0.002 -0.012 Abaqus
0
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.15 -0.1 -0.05 0 0.05 0.1 0.15
Longitudinal contact widht y [mm] Longitudinal contact widht y [mm]

Fig. 11. Longitudinal wear scar with and without interpolation: (a) top wear scar and (b) bottom wear scar.

increments per fretting cycle and two different numbers of cycles
1000 and 20,000. The range of mesh size studied goes from 3 mm
to 24 mm.
In Fig. 12(a) it is shown that the mesh size has a great
influence in the initial contact pressure distribution in compar-
ison with the Hertzian contact pressure. While small mesh size
(3 mm) adjusts properly with the Hertzian contact pressure
distribution, greater mesh size (10 mm) increase both contact
pressure and contact size. Nevertheless as the wear happens it
can be seen in Fig. 12(b) that despite the contact pressure in the
initial cycles is greater, this difference is reduced considerably
during the exponential decrease of the contact pressure and
stabilized during the stable period of the contact pressure, being
the error during this period around 2%. Although this study
presents that the mesh size has not a great influence in the
contact pressure once the wear reach the stable period, it is
necessary to study the profile of the wear scar to validate an
optimum mesh size to reduce the computational time.
For this, in the next paragraphs the influence of the mesh size
in the dimension and the shape of the wear scar profile is studied.
Fig. 13(a) and (b) shows the wear scars produced as a conse-
quence of different mesh size for 1000 and 20,000 cycles respec-
tively. It can be seen that the increase from 6 mm to 24 mm
(Fig. 13(b)) of mesh size produce a trapezoidal wear scar profile
34 A. Cruzado et al. / Wear 289 (2012) 26–38

7000
6000 3 μm 3 μm
6000

Max contact Pressure [MPa]

6 μm 6 μm
Contact Pressure [MPa]

5000
10 μm 5000 10 μm
4000 HERTZ
4000
3000 3000
2000 2000

1000 1000

0 0
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0 100 200 300 400 500 600 700 800 900 1000
y [mm] Number of cycles

Fig. 12. Hertzian contact pressure vs FEM contact pressure for different mesh: (a) contact pressure distribution without wear and (b) maximum contact pressure evolution
in the first wear 100 cycles.

0 0

-0.0004 Wear depth [mm] -0.002

Wear depth [mm]

-0.0008 -0.004

-0.0012 -0.006
original original
3 μm-1000 6 μm-20000
-0.0016 -0.008
10 μm-1000 12 μm-20000
15 μm-1000 24 μm-20000

-0.1 -0.05 0 0.05 0.1 -0.15 -0.1 -0.05 0 0.05 0.1 0.15
Longitudinal contact widht [mm] Longitudinal contact widht [mm]

Fig. 13. Longitudinal wear scar in the bottom specimen for different mesh sizes: (a) 1000 number of cycles and (b) 20,000 number of cycles.

due to the poor discretización, nevertheless the error in contact produce convergence problems during the contact problem reso-
width is less than 6% and in wear depth is less than 5%. lution and greater values do not present significant differences in
After this it is proposed that the optimum mesh size without any the wear scar profile.
distortions in geometry should be in the range from 3% to 4% of the On the other hand, in Fig. 14(b) a great influence of the cycle
final longitudinal wear width. In the case of 20,000 cycles, using and jump DN in the wear scar profile is seen. The wear scar becomes
optimum mesh size of 8 mm a reduction of two times in the deeper and slender with high cycle jumps, such that with
computational time is obtained with respect to a refined mesh of excessive cycle jumps the profile is distorted. While for 1000
6 mm which is close to the Hertzian contact pressure (Fig. 12(a)). cycles there is an increase in the wear depth of 5% between a
On the other hand to obtain the approximated dimensions of the cycle jump of DN ¼100 to DN ¼200, for 20,000 cycles there is the
wear scar depth and width, before applying the relation between the same increase between DN ¼100 and DN ¼400. It means that with
mesh and the longitudinal wear width proposed previously, greater higher wear cycles greater cycle jumps can be used. For this study
meshes can be used. In this case using a greater mesh of 24 mm a it is proposed that the optimum value for the study of different
reduction of 8 times in the computational time with respect to a cases along different number of cycles without any distortions in
refined mesh of 6 mm is achieved. profile and with an error in depth less than 3% should be 100
The second and the third parameter to reduce the computa- fretting FE cycles and the corresponding DN, i.e., in the case of
tional time are based in the determination of the maximum wear 20,000 cycles DN ¼200 or in the case of 50,000 cycles DN ¼ 500.
increment according to Eq. (15). For this, two different studies Taking into account that each fretting cycle is discretized in 40
were carried out: the first one corresponds to the contact wear increments, any wear simulation, regardless the conditions used,
simulation increments per fretting cycle, and the second one is modeled with 4000 FE wear increments.
consists of the study of the cycle jump DN technique. In these
studies, an optimum mesh size of 6 mm for the case of 1000 cycles 4.2. Wear modeling validation
and 8 mm for the case of 20,000 cycles is used, following the
relation proposed previously between the longitudinal wear scar In this point, the wear scars (Fig. 15) obtained by the optimized
width and the mesh size. wear methodology presented in point 4.1 are validated using the
On the one hand, Fig. 14(a) shows that the influence of the experimental wear scars obtained in tribometer under different
increments discretization per cycle in the wear scar is depreci- tests conditions of: load, stroke and number of cycles with the
able. For this study 40 increments/cycle is selected as the aim to present a global methodology for simulation any experi-
optimum discretization, because smaller increments per cycle mental cases. As has been mentioned previously the validation
 A. Cruzado et al. / Wear 289 (2012) 26–38 35

0 0

-0.0004 -0.002
Wear depth [mm]

Wear depth [mm]

1000 Cycles original
-0.0008 -0.004 20000 Cycles 100
200
500
-0.0012 original -0.006
1000
40 inc
100
60 inc
-0.0016 -0.008 200
100 inc
500
200 inc
1250

-0.1 -0.05 0 0.05 0.1 -0.2 -0.1 0 0.1 0.2

Longitudinal contact widht y [mm] Longitudinal contact widht y [mm]

Fig. 14. Longitudinal wear scar in the bottom specimen: (a) for different increments per cycle and 1000 number of cycles and (b) for different cycle jumps for 1000 and
20,000 number of cycles.

Fig. 15. FEM wear scars for a 1N-130 mm-50000 cycles: (a) top wear scar and (b) bottom wear scar.

Table 2
FEM vs experimental wear scar dimension analysis for both top and bottom specimens.

Top wear Error (%) Bottom wear Error (%) Coefficient of wear
scar (lm) scar (lm) (10  8 mm3/Nmm)

1 N-65 mm Experimental 151/166 3–1 161/200 9–6 2.59–3.65

20,000 cycles FEM 146/164 146/212
2 N-65 mm Experimental 189/190 6–1 190/230 7–4 2.65–3.46
20,000 cycles FEM 178/192 176/240
3 N-65 mm Experimental 200/200 8–2 200/250 9–2 2.31–2.76
20,000 cycles FEM 184/204 182/245
3 N-65 mm Experimental 230/230 3–6 239/280 6–2 2.10–1.91
50,000 cycles FEM 224/245 224/286
1 N-130 mm Experimental 180/180 2–2 190/280 4–2 3.01–4.08
20,000 cycles FEM 183/184 183/286
1 N-130 mm Experimental 200/210 3–5 210/310 5–3 2.32–2.34
50,000 cycles FEM 196/221 200/319
1 N-130 mm Experimental 250/250 8–3 260/359 8–3 1.54–0.69
200,000 cycles FEM 230/257 240/372

methodology consists of three steps, which will be analyzed in detail
in the next paragraphs. For the validation of all cases the specific
wear coefficients obtained in each specimen and test has been
chosen.
The first validation method consists of the determination of
the wear scar dimensions in top and bottom wear scars, taking
into account that each wear scar is composed by two principal
dimensions: the longitudinal and the transversal direction with
respect to the wire. Table 2 shows the values obtained experi-
mentally and by FEM for different load, stroke and number of
cycle cases, where the first value represent the wear scar width in
the transversal direction and the second one represents the wear
scar width in the longitudinal direction to the wire. It can be seen
that the error in the transversal direction is greater than in the
longitudinal direction, nevertheless the error is less than 10%,
acceptable in this type of tests taking into account that the
experimental results could present dispersions greater than 15%
[22]. In the range of loads, strokes and number of cycles studied
the dimensions of the wear scar could be predicted with an error
that goes from 1% to 9%.
The second validation method consists of the comparison of
the maximum wear depth profile in the longitudinal direction of
the specimen in top and bottom wear scar with the profile
obtained by FEM, as it is shown in Figs. 16 and 17 respectively
for different number of cycles.
On one hand the top wear scar profile obtained by FEM adjust
precisely to the profile obtained experimentally (Fig. 16(d)–(f)) with
an error in the maximum wear depth less than 6% as it is shown in
36 A. Cruzado et al. / Wear 289 (2012) 26–38

X = 198m X = 219m X = 265m

Y = 189m Y = 214m Y = 261m

0.04 0.04 0.04

Original
Original Original
MEF
MEF MEF
Wear depth [mm]
Wear depth [mm]

Wear depth [mm]

0.03 0.03 0.03 Experimental
Experimental Experimental

0.02 0.02 0.02

0.01 0.01 0.01

0 0 0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.3
Longitudinal contact widht x [mm] Longitudinal contact widht x [mm] Longitudinal contact widht x [mm]
Fig. 16. Wear scars of top specimen for 1N-130 mm: (a) 20,000 cycles SEM, (b) 50,000 cycles SEM, (c) 200,000 cycles SEM, (d) 20,000 cycles MEF vs experimental,
(e) 50,000 cycles MEF vs experimental and (f) 200,000 cycles MEF vs experimental.

X = 295m X = 322m X = 389m

Y = 193m Y = 220m Y = 268m

0 0 0
Wear depth [mm]
Wear depth [mm]

-0.01
Wear depth [mm]

-0.01 -0.01

-0.02 -0.02 -0.02

Original Original Original
-0.03 MEF -0.03 MEF -0.03 MEF
Experimental Experimental Experimental
-0.04 -0.04 -0.04
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
Longitudinal contact widht y [mm] Longitudinal contact widht y [mm] Longitudinal contact widht y [mm]

Fig. 17. Wear scars of bottom specimen for 1N-130 mm: (a) 20,000 cycles SEM, (b) 50,000 cycles SEM, (c) 200,000 cycles SEM, (d) 20,000 cycles MEF vs experimental,
(e) 50,000 cycles MEF vs experimental and (f) 200,000 cycles MEF vs experimental.

Table 3. On the other hand the bottom wear scar (Fig. 17(d)–(f)),
obtained by FEM presents more uniform shape than the profile
obtained experimentally. In this case the error in the maximum
wear depth goes up to 14%. The reason of this error could be due to
two different errors: the first one corresponds to the error com-
mitted due to the error assumed in the bottom wear scar as a
consequence of the optimized wear simulation methodology, and
the second one corresponds to an error that is committed in this
type of tests as a consequence of the bending of the wire in the
holder, because longest is the wear scar greater is the part of the
bending that is covered, for this reason if the radius of the curvature
is not exactly 10 mm or the wires are not perfectly centered, the
 A. Cruzado et al. / Wear 289 (2012) 26–38 37

Table 3 develop an optimized methodology for the fretting wear scar

FEM vs experimental wear scar depth analysis for both top and bottom specimens. simulation and validation procedure in a simple 901 crossing
angle configuration. This work will be extended to more realistic
Top wear Error Bottom Error
scar (%) wear (%) crossing angles arrangements with the aim to study the influence
(lm) scar (lm) of wear scars in the fatigue life reduction in future publications.
The wear scars produced during the fretting wear process are
1 N–130 mm Experimental 12.7 2.3 9.0 13 the main stress concentration points for the initiation and
20,000 cycles FEM 13 7.9
1 N–130 mm Experimental 18.3 1.1 10.6 13
propagation of cracks due to fatigue. This tool will help in
50,000 cycles FEM 18.5 9.17 understanding the severity of the wear scar under different
1 N–130 mm Experimental 32.8 5.2 10.07 14 parameters either constructive (such as diameter, curvature,
200,000 FEM 31.1 8.65 crossing angle) or operative (such us contact pressure, stroke).
cycles
On the one hand, the developed simulation-experimental meth-
odology could be used to obtain the reduction in the cross section
area of the wire which produces the increase in the stress fields
Table 4 and the rupture of the wire. On the other hand as the wear
FEM vs experimental wear scar volume analysis for both top and bottom simulation methodology has been developed in a FEM framework
specimens. it could be calculated the stress concentration factors produced
by the different wear scars. To understand better the influence of
Top wear Error Bottom Error
scar (%) wear (%)
each constructive and operative parameter a DoE (Design of
(lm3) scar (lm3) Experiments) could be carried out, obtaining a mathematical
model which describes the severity of the wear scars in a range
1 N–65 mm Experimental 66,000 0.5 93,000 1.2 of design and operative conditions. Summarizing, this tool pre-
20,000 cycles FEM 65,644 91,992
tends to help to the designers to understand better the severity of
2 N–65 mm Experimental 135,000 1.2 177,000 2.5
20,000 cycles FEM 133,430 172,710 the fretting wear scars in the construction of new wire ropes, with
3 N-65 mm Experimental 172,000 3 206,000 4 the aim to minimize the effect of fretting wear and improve the
20,000 cycles FEM 166,930 197,660 life of these components.
3 N–65 mm Experimental 404,000 3.6 367,000 5
50,000 cycles FEM 389,380 348,930
1 N–130 mm Experimental 157,000 2.3 212,720 1.7
20,000 cycles FEM 153,470 209,130 Summary and conclusions
1 N–130 mm Experimental 305,250 3.1 307,590 3.1
50,000 cycles FEM 295,790 297,920 A FEM methodology for the prediction wear scars in 901
1 N–130 mm Experimental 820,000 8 365,770 10
crossed cylinders thin steel wires arrangement has been devel-
200,000 FEM 753,450 328,520
cycles oped. Different strategies for computational time optimization
have been analyzed in this work and the results have been
validated in comparison with experimental data obtained in
laboratory fretting tests.
error is committed both in the experimental tests or in the The main conclusions obtained in this work are:
measurement of the profile. Nevertheless the error committed
during the simulation is assumable for this type of tests.  The nearest interpolation method is presented as the optimum
Taking into account the good correlation existed in the wear method for the transition of the contact pressure and slip
scar dimensions and in the wear scar profile in the maximum distribution from the slave surface to the master surface.
wear depth point, the final validation procedure consists of the In comparison with the other two methods studied in this
comparison of the volume of both wear scars as can be seen in paper (bivariate method and baricentric method), the differ-
Table 4. In all cases the wear volume predicted by FEM is smaller ences obtained in the resulting final wear scar are depreciable,
than the one obtained experimentally. Nevertheless this error is nevertheless the computational time needed for the resolution
smaller than 10%. of the wear simulation problem is considerably inferior, three
This validation has demonstrated that this optimized metho- times faster than the other two methods.
dology for simulation fretting wear in crossed cylinder arrange-  In the analyzed range, the mesh size in the contact zone has
ments predict within an mean error less than 10% the entire not a great influence in the final wear scar dimensions and
shape of wear scar in both top and bottom specimens depth. Nevertheless with high mesh sizes the profile presents
a trapezoidal wear scar profile, for this, an optimum mesh size
4.3. General guidelines about the application of this methodology that could be in the range from 3% to 4% of the final long-
for the improvement the wire ropes design itudinal wear width has been define. This proposal achieves
the best compromise between the wear scar shape and the
Although the aim of this paper is not to study directly the computational time spending.
influence of the wear scar in wire rope systems, this section  The discretization of the FE wear simulation increments per
pretends to outline the general guidelines about how this wear fretting cycle doesn’t present influence in the final wear scar
simulation tool can be applied in the design and in the life profile. An optimum value of 40 increments/cycle is found,
improvement of wire rope systems. since below this value convergence problems appear in the
As it has been mentioned previously, the wire ropes analyzed contact problem resolution and above this value the differ-
are constructed by small thin steel wires wound into strands, ences in the wear profile are negligible. On the other hand, the
which are then wound to form the rope. Due to this complex cycle jump technique is very sensitive to the wear profile. After
structure, fretting wear problem is presented between wires that this study, it is proposed that an optimum simulation without
are crossed at different angles such us the outermost wires of any distortions in profile and with an error in depth less than
strands and core or between the outermost wires of adjacent 3% should be carried out with 100 fretting FE cycles, 40
strands. This is the first paper which describes the steps to increments/cycle and the corresponding DN.
38 A. Cruzado et al. / Wear 289 (2012) 26–38
 The optimized simulation methodology presents very good
correlation with respect to the experimental data, which
implies that the entire wear scar can be predicted with errors
less than 10%. Acceptable in this tribological system, where the
experimental data presents dispersion greater than 15%.
Acknowledgments
The authors would like to thank Ion Lizarralde for his work in
the parametrization of the FE models, Hodei Usabiaga for the
technical assistance in the interpolation methods and the Basque
Government for the financial support under the Universidad
Empresa program in the frame of the SIVICA project (Ref.
UE2009-5).
References
[1] K.K. Schrems, Wear related fatigue in a wire rope failure, Journal of Testing
& Evaluation 22 (1994) 490–499.
[2] R.B. Waterhouse, Fretting in steel ropes and cable—a review, ASTN Special
Technical Publication 1425 (2002) 3–14.
[3] M.A. Urchegui, W. Tato, X. Gómez, Wear evolution in a stranded rope
subjected to cyclic bending, Journal of Materials Engineering and Perfor-
mance 17 (2007) 550–560.
[4] L. Johanson, Numerical simulation of contact pressure evolution in fretting,
Journal of Tribology 116 (1994) 247–254.
[5] I.R. McColl, J. Ding, S.B. Leen, Finite element simulation and experimental
validation of Fretting wear, Wear 256 (2004) 1114–1127.
[6] J. Ding, I.R. McColl, S.B. Leen, P.H. Shipway, A finite element based approach to
simulating the effects of debris on fretting wear, Wear 263 (2007) 481–491.
[7] J.J. Madge, S.B. Leen, P.H. Shipway, A combined wear and crack nucleation-
propagation methodology for fretting fatigue prediction, International
Journal of Fatigue 30 (2008) 1509–1528.
[8] A.L. Mohd Tobi, J. Ding, G. Bandak, S.B. Leen, P.H. Shipway, A study on the
interaction between fretting wear and cyclic plasticity for Ti–6Al–4V, Wear
267 (2009) 270–282.
[9] S. Fouvry, C. Paulin, C. Meunier, Finite element modelling of fretting wear
surface evolution: application to a Ti–6Al–4V contact, Wear 264 (2008)
26–36.
[10] A. Cruzado, M. Hartelt, R. Wäsche, M.A. Urchegui, X. Gómez, Fretting wear of
thin steel wires. Part 1: influence of contact pressure, Wear 268 (2010)
1409–1416.
[11] J.J. Madge, S.B. Leen, I.R. McColl, P.H. Shipway, Contact-evolution based
prediction of fretting fatigue life: effect of slip amplitude, Wear 262 (2007)
1159–1170.
[12] V. Hegadekatte, S. Kurzenhäuser, N. Huber, O. Kraft, A predictive scheme for
wear in tribometers, Tribology International 41 (2008) 1020–1031.
[13] D. Klaffke, On the influence of test parameters on friction and wear of
ceramics in oscillating sliding contacts, TriboTest 1 (1995) 311–320.
[14] M.A. Urchegui, X. Gómez, W. Tato, A method for evaluating fretting wear
scars in thin steel roping wires based on confocal imaging profilometry,
Journal of Testing and Evaluation 35 (2007) 357–363.
[15] K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge,
1985.
[16] S. Fouvry, T. Liskiewicz, C. Paulin, A global–local wear approach to quantify
the contact endurance under reciprocating-fretting sliding conditions, Wear
263 (2007) 518–531.
[17] C. Mary, S. Fouvry, Numerical prediction of fretting contact durability using
energy wear approach: optimisation of finite-element model, Wear 263
(2007) 444–450.
[18] Abaqus User’s Manual,Version 6.9, Hibbit, Karlsson & Sorensen, Inc., USA,
2009.
[19] B. Kanavalli. Application of user defined subroutine UMESHMOTION in
ABAQUS to simulate dry rolling/sliding wear. Master thesis. Stockholm,
Sweden: Royal Institute of Technology (KTH); 2006.
[20] H. Akima, A method of bivariate interpolation and smooth surface fitting for
values given at irregularly distributed points, Journal ACM Transactions on
Mathematical Software 4 (1978) 148–164.
[21] David F. Watson, Contouring: A Guide to the Analysis and Display of Spacial
Data, Pergamon, 1994.
[22] H. Czichos, S. Becker, J. Lexow, International multilaboratory sliding wear
tests with ceramics and steel, Wear 135 (1989) 171–191.