Wear 263 (2007) 542–551

The critical role of fretting wear in the analysis of fretting fatigue

J.J. Madge ∗ , S.B. Leen, P.H. Shipway
School of Mechanical, Materials and Manufacturing Engineering, University Technology Centre in Gas Turbine Transmission Systems,
University of Nottingham, University Park, Nottingham NG7 2RD, UK
Received 15 August 2006; received in revised form 26 October 2006; accepted 1 November 2006
Available online 26 March 2007

Abstract
A finite element based approach is used to predict the effect of fretting wear on fretting fatigue life in the commonly used ‘rounded edge punch
on flat’ laboratory geometry. The method predicts and hence explains experimentally observed phenomena, such as cracking location under the
contact and dependence of fretting fatigue life on slip amplitude, which were not previously predictable. The results from the model are compared
against available test data across a range of slip amplitudes. The model predicts a significant effect of slip amplitude on fretting fatigue life and,
in particular, the well-known phenomenon of a critical range of slip amplitude for minimum life. The sensitivity of life and cracking location to
the assumed value of wear coefficient on life and cracking location is studied for a partial slip case and it is shown (i) that the location of cracking
switches from the contact edge to the stick–slip interface with increasing wear coefficient and (ii) that there is an associated critical value of wear
coefficient which gives a maximum fretting fatigue life. For gross sliding cases, failure is always predicted at the contact edge. The work shows
that inclusion of the effects of wear on fatigue damage accumulation is necessary for prediction of fretting fatigue life and cracking location.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Fretting wear; Fretting fatigue; Rounded punch on flat; Critical plane; Contact slip

1. Introduction life prediction methods will aid further optimisation of the

Many engineering assemblies contain contacting compo-
nents which are nominally fixed with respect to each other, yet
in reality experience micro slip due to the loads imposed across
the contact interface. The surface damage that results from this
‘fretting’ condition is known to be a key factor in the fatigue
behaviour of these components, and can cause dramatic reduc-
tions in fatigue life. The fretting process has been found to be
dependent on a large number of variables [1]. Dobromirski [2]
concluded that the variables of coefficient of friction (COF), nor-
mal load and slip amplitude are of primary importance. However,
even these ‘primary’ variables are found to be interdependent.
For example, the COF is known to affect the slip distribution
[3], whilst COF is found to be dependent on normal load; fur-
thermore, COF is found to change with the number of fretting
cycles, e.g. [4,5].
Spline couplings and turbine root dovetails are notable exam-
ples of aeroengine components where fretting occurs. Accurate
∗ Corresponding author. Tel.: +44 115 8467685.
E-mail address: eaxjm1@nottingham.ac.uk (J.J. Madge).
design of such components. Leen et al. [6] studied the per-
formance of a laboratory-scale spline coupling under different
combinations of torque, axial and rotating bending moment
loading. While plain fatigue failure occurred under torque over-
load, fretting fatigue cracking occurred under large bending
moment overloads, with the cracks occurring within the length
of engagement contact region (Fig. 1).
Similar trends have been observed in ‘punch on flat’ fret-
ting fatigue tests. In studies by Jin and Mall [7] and Nakazawa
et al. [8], failure was observed well within the contact region
under certain loading conditions. Similarly, Hertzian contact
conditions have frequently produced cracking under the con-
tact [4]; this can occur at either the contact edge or the stick–slip
interface. The effect of slip amplitude has been shown to be
significant in experimental studies, e.g. [4,9]. A second key
phenomenon in fretting fatigue is the occurrence of a mini-
mum fatigue life corresponding to a certain critical range of
slip amplitudes, normally associated with the transition from
partial slip to small amplitude gross sliding. Both of these phe-
nomena, viz. failure within the contact region away from the
contact edge, and dependence of fatigue life on slip ampli-
tude, are difficult to explain and capture using existing fretting

0043-1648/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.wear.2006.11.021
 J.J. Madge et al. / Wear 263 (2007) 542–551 543

Nomenclature
b fatigue strength exponent; also specimen width
c fatigue ductility exponent
E Young’s modulus
h wear depth
h incremental wear depth
H hardness
k bulk wear coefficient
kl local wear coefficient
kcrit critical wear coefficient value
K nondimensional wear coefficient
ni number of cycles at load cycle i
N number of fretting cycles
N cycle jump size
N* number of fretting cycles analysed within FE
Nf number of cycles to failure
Nfi number of cycles to failure at load cycle i
Fig. 1. Fretting fatigue cracking location for a laboratory scaled spline coupling
P normal load applied to fretting pad under combined torque, axial and bending moment overload [6]. (N.B. z = axial
R fatigue stress load ratio position, al = total axial length of engagement).
S sliding distance
t time gave good experimental agreement over the range of stresses
V wear volume tested.
W accumulated damage fraction In this paper, the multiaxial fatigue life prediction approach
x position across contact face of Madge et al. [10] with wear effects included, is applied to the
y depth into specimen surface from contact surface ‘rounded-edge punch-on-flat’ geometry. The method is based on
a modified version of the Archard wear equation which is used to
Greek letters predict the effect of material removal due to wear, by incremen-
δ relative slip at contact surface over one fretting tally updating the finite element mesh geometry. A critical plane
cycle implementation of the Smith–Watson–Topper (SWT) parame-
δapp maximum relative displacement applied to pad in ter [12] is combined with a damage accumulation approach to
one fretting cycle predict fatigue damage during material removal and hence fail-
δapp (t) applied displacement at time t ure. The results are validated against available test data across
εf fatigue ductility coefficient a range of slip amplitudes. The effect of slip amplitude and the
ε strain range over one cycle (transformed to plane) effect of changes in wear coefficient are investigated in order to
σf fatigue strength coefficient explain the aforementioned experimental phenomena.
σ max maximum applied fatigue stress
σ min minimum applied fatigue stress
2. FE model description
σ(t) applied fatigue stress at time t
τ time during one cycle jump
Fig. 3 shows the specimen geometry modelled, based on that
used by Jin and Mall [7]. A pair of flat fretting pads with a nom-
inal contact width of 4.45 mm and edge blend radii of 2.54 mm
analysis approaches. Conventional approaches have not been are held in contact with a flat, uniaxially loaded fatigue speci-
able to predict the effect of slip amplitude, particularly the
beneficial effect of gross sliding (Fig. 2). These approaches
also generally predict failure at the trailing contact edge. How-
ever, Madge et al. [10] have recently presented a critical plane
methodology for fretting fatigue life prediction that accounts
for the material removal effect of wear by use of the Archard
wear model. When applied to the Hertzian contact geometry,
the approach was found to predict the measured dependence
of fretting fatigue life on slip amplitude more accurately than
using a similar critical plane technique which neglects wear
(Fig. 2). Hattori and Watanabe [11] have developed a stress
intensity factor based fretting fatigue lifing methodology cou- Fig. 2. Fretting fatigue life prediction by Madge et al. [10] compared to exper-
pled with the Archard model to account for wear, this also imental values by Jin and Mall [4] (Hertzian geometry).
544 J.J. Madge et al. / Wear 263 (2007) 542–551

Fig. 5. Loading history applied to the FE model.

edges of contact to improve the stress resolution in this region

Fig. 3. (a) Experimental schematic; (b) fretting fatigue specimen; (c) fretting
pad; after [7].
men. The material used in the experiments by Jin and Mall [7]
is a dual phase Ti-6Al-4V alloy which consists of α (hexagonal
close packed) and β body centred cubic phases). The material
was solution heat treated at 935 ◦ C for 1.75 h, cooled in air and
annealed at 700 ◦ C for 2 h in vacuum and cooled in argon. The
material is modelled as linear elastic with a Young’s modulus of
126 GPa and a Poisson’s ratio of 0.32. Table 1 shows the fatigue
properties used in the fatigue lifing model.
A symmetry plane parallel to the specimen axis is employed
in the FE model so that only one pad needs to be modelled
(Fig. 4). The FE mesh created has refinement graded towards the
Table 1
Fatigue properties for Ti-6Al-4V
σf (MPa) b εf
2030 −0.104 0.841
The above data corresponds to a definition of crack initiation as the development
of a 1 mm crack from Dowling [25].
as very high stress gradients can be expected in this region. The
smallest element in the model is around 14 ␮m × 9 ␮m.
Due to the conforming nature of the contact and the relatively
abrupt edge of contact, the influence of pad misalignment due
to rig compliance was found to be an important factor. In order
to represent the flexibility of the experimental rig, the pad is
modelled to a height of 11 mm away from the contact surface.
Following the work of Sabelkin and Mall [13], the effect of rig
compliance has been incorporated by setting the elastic modulus
of the extended section of the pad to a value of 55 GPa. The top
surface of the pad is prescribed to remain plane. The ‘contact
pairs’ option [14] was used in conjunction with the Lagrange
friction algorithm to strictly enforce exact sticking behaviour
when the resultant shear stress is less than the critical shear
stress. A COF of 1.05 is taken as representative, based on the
approximate stabilised dry sliding COF measured by Jin and
Mall [7].
c In the first analysis step a normal load P of 208 N/mm (equiv-
alent to 1334 N for a 6.4 mm wide specimen) is applied to the
−0.688
fretting pad. In the next step the specimen is loaded by a cyclic
fatigue load σ(t) with a maximum value σ max of 550 MPa and a
stress ratio R of 0.03. The pad is also loaded with a prescribed
maximum displacement δapp , which is varied for each simula-
tion to cover a range of slip amplitudes including partial slip
through to gross sliding conditions. δapp (t) and σ(t) are then
applied cyclically in the FE model. Fig. 5 shows the load cycle
history for the first two load cycles.

3. Wear modelling

3.1. Wear model

The wear simulation technique used here is based on that

Fig. 4. The FE mesh used for analyses. Shaded section shows ‘compliant sec-
tion’ used to represent rig compliance.
implemented and validated by McColl et al. [5] for Super-
CMV high strength steel fretting wear tests. The specific method
employed here is that described in Madge et al. [10] for Ti-
6Al-4V in a Hertzian fretting fatigue configuration, which
demonstrated successful validation against fatigue life test data
across a range of slip amplitudes. The material removal approach
is based on the well-known Archard equation:
V P
=K (1)
S H
 J.J. Madge et al. / Wear 263 (2007) 542–551
where V is the volume wear, S is the sliding distance, K is the
wear coefficient, P is the normal load and H is the hardness of
the material.
At each point along the contact surface, Eq. (1) can be written
as
h
= kl p
S
where h is the wear depth (mm), kl is the dimensional local wear
coefficient defined as the wear per unit local slip per unit local
contact pressure, and p is the contact pressure. Therefore, the
incremental wear depth for a node experiencing an incremental
slip δ and pressure p at time t can be expressed as:
Δh(x, t) = kl p(x, t)δ(x, t)
where h(x,t), p(x,t) and δ(x,t) are the incremental wear depth,
contact pressure and relative slip at point x and time t, respec-
tively.
Direct calculation of the desired wear coefficient, kl , requires
knowledge of the local contact pressures and slips. As these
are not readily measurable, a modified form of Eq. (3) is used
to determine a bulk wear coefficient, k, for which a value of
2.75 × 10−8 MPa−1 has been estimated as representative of the
Ti-6Al-4V fretting contact pair of the present work (see [10] for
details), using measured wear scar data from Magaziner et al.
[15].
3.2. Variation of k
For computational efficiency, a cycle jump technique is
employed, as described in [10], where the assumption is made
that wear is constant over a certain number of cycles. By multi-
plying the incremental wear by a cycle jumping factor N, one
FE cycle simulation is used to model the effect of N actual
wear cycles. Eq. (3) thus becomes:
h(x, τ) = kNp(x, τ)δ(x, τ)
where τ is the time within one ‘cycle’ corresponding to N wear
cycles and the bulk wear coefficient k is employed in place of
kl . The modified Archard equation given in Eq. (4) shows that
since k and N appear as a product on the right-hand side, this
product can be represented as one constant kref which can be
treated as a reference wear coefficient, as follows:
kN ≡ kref
so that Eq. (4) can now be written as:
h(x, τ) = kref p(x, τ)δ(x, τ)
Subject to the assumption that the solution is approximately
independent of the cycle jump size N (which has been found
to be valid by the authors for the range of cases considered in
this paper), the effect of varying the wear coefficient k, within
a reasonable range of values, can be investigated by varying the
cycle jump size, as a post-processing exercise to a simulation
using a fixed value of kref . The implementation of Eq. (6) within
ABAQUS has been described previously by Madge et al. [10].
545
(2)
Fig. 6. Schematic showing the wear modelling dealt with in the ABAQUS
FE code, stress–strain results are then fed into the post processing software
to calculate critical plane SWT values and accumulated damage.
(3) In summary, Eq. (6) is programmed into the UMESHMOTION
subroutine within ABAQUS to effect incremental modification
of the contact geometry to model the material removal effect of
wear. Each FE fretting cycle was analysed in 100 increments,
resulting in the geometry being updated 100 times during each
N cycle jump.
Fig. 6 summarises the approach employed. The evolution of
contact pressures, slips, stresses and damage parameters can then
be presented in terms of a ‘scalable’ number of cycles, N* , cor-
responding to the reference wear coefficient kref , which can be
easily converted to specific numbers of cycles, N corresponding
to arbitrary wear coefficients, k, as follows:
kref ∗
N= N (7)
k
In the present study, a kref value of 4 × 10−5 MPa−1 has been
employed to expedite the wear simulations and thus reduce com-
putational time.
4. Fatigue damage model
4.1. Multiaxial fatigue methodology
(4)
The Smith–Watson–Topper parameter [12] is based upon a
combined high- and low-cycle strain–life equation, with a maxi-
mum stress term to account for the mean stress effect, as follows:
ε σ2
SWT = σmax = f (2Nf )2b + σf εf (2Nf )b+c (8)
2 E
where σ max is the peak normal stress on a plane, ε the maxi-
(5) mum normal strain range on the same plane within one fatigue
cycle, σf is the fatigue strength coefficient, E is Young’s mod-
ulus, Nf is the number of cycles to crack initiation, b is the
(6) fatigue strength exponent, εf is the fatigue ductility coefficient
and c is the fatigue ductility exponent. For most fatigue life sit-
uations, Fatemi and Socie [16] have proposed the use of one
of two multiaxial fatigue damage criteria, one for shear crack-
ing failure modes and one for tensile cracking failure modes.
Such parameters should take account of both crack initiation
and propagation. For shear cracking failure modes, the param-
eter should include shear strain amplitude for crack initiation,
and maximum normal stress on the plane of maximum shear
strain amplitude for crack propagation. For tensile cracking, the
546 J.J. Madge et al. / Wear 263 (2007) 542–551
parameter should include maximum principal strain amplitude
for initiation, and maximum normal stress on the plane of max-
imum principal strain amplitude for propagation. Thus, Socie
[17] suggests that knowledge of the cracking behaviour for the
material and loading state of interest allows selection of the most
appropriate fatigue criterion. On this basis, Araujo and Nowell
[18] have suggested that the SWT parameter is appropriate for
fretting fatigue prediction of Ti-6Al-4V.
The critical plane approach [19] is based on physical obser-
vations that fatigue cracks initiate and grow within a material on
certain planes, where the growth and orientation depends on the
normal stresses and strains on these planes. Dominant fatigue
damage occurs on these planes, which are called critical planes.
The approach has evolved from the work of Brown and Miller
on multiaxial fatigue [20] who first suggested the combined use
of maximum shear strain range and tensile strain normal to the
plane of maximum shear and is favoured due to its basis on phys-
ical interpretation of the mechanisms of fatigue crack growth.
The approach of Fatemi and Socie [16], i.e. using normal stress
instead of strain as the second parameter, overcomes limita-
tions of the Brown and Miller [20] approach with respect to the
increased damage associated with non-proportional or out-of-
phase loading attributed to the extra cyclic hardening developed
due to rotation of the principal axes. Practically, the critical plane
approach amounts to identification of the orientation of the plane
which maximises the relevant fatigue damage parameter, at each
spatial location of interest, covering the complete loading cycle.
Szolwinski and Farris [19] were the first to suggest the criti-
cal plane method (with the SWT parameter) as a technique to
deal with the multiaxial stress and strain cycles in fretting con-
tacts. Numerous other authors have since used the critical plane
SWT parameter for fretting fatigue of Ti-6Al-4V, e.g. [20,21].
A study by Lykins et al. [22] which compared the performance
of SWT against a maximum shear stress range (MSSR) damage
parameter for prediction of Ti-6Al-4V fretting fatigue life, crack
location and crack orientation, without including the effects of
wear, concluded that both parameters were equally accurate for
life and crack location prediction, but that MSSR was superior
for crack orientation prediction. Although SWT is employed in
this paper, primarily as a vehicle to illustrate the impact of fret-
ting wear modelling on life predictions, it would not be difficult
to supplant SWT with MSSR in the present methodology, for
example.
The FE implementation of the critical plane SWT parame-
ter here follows the method described by Sum et al. [23]. In
practice this is achieved by transforming the time histories of
element centroidal stresses and strain ranges onto planes at 5◦
intervals over a 180◦ range using the two-dimensional trans-
formation (Mohr’s circle) equations for stress and strain. The
maximum normal stress σ max with respect to time, and the cor-
responding strain range ε are determined for each of the 36
planes in each element. ε is the difference between the max-
imum and minimum values of strain normal to the candidate
plane over the complete loading cycle. Thus an array of SWT
values is obtained for each candidate plane in each element for
each fretting cycle analysed. These values are then employed to
establish the maximum SWT with respect to plane orientation
and spatial position (element), which in turn is used with Eq. (8)
to furnish a number of cycles to failure, Nf for each fretting cycle.
The FE-implementation (without wear effects) was successfully
validated by comparing the predicted lives for a Hertzian fretting
contact against corresponding FE-based predictions by Sum et
al. [23], and Araujo and Nowell [18] for partial slip conditions.
4.2. Damage accumulation model
A cumulative damage approach is required to take account of
the evolution of stress and strain cycles with increasing number
of fretting cycles due to material removal. Damage accumulation
is generally a non-linear interaction concerning both the nature
of each individual loading cycle and the order in which dif-
ferent cycles are applied [24]. The most commonly used fatigue
damage accumulation model is the Miner–Palmgren (M-P) rule,
which assumes a simple linear accumulation of damage, ignor-
ing the effect of order of the loading cycles, as follows:
N
 ni
W= (9)
Nfi
i=1
where W is the accumulated damage, ni is the number of cycles
experienced at loading cycle i, Nfi is the critical plane SWT
predicted number of cycles to failure at loading cycle i (found
using Eq. (8)) and N is the number of different load cycles expe-
rienced. In the case of fretting wear, each cycle differs from
the next, so that the cumulative damage for each individual
fretting cycle should be considered. When the cycle jumping
technique described above is combined with Eq. (9), the damage
rule becomes:
N
t /N
N
W= (10)
Nfi
i=1
Given the adoption of a critical plane approach to calculate
fatigue damage, it is necessary to make an assumption about how
damage on a given plane interacts with damage on other planes
from cycle to cycle. One approach is to calculate the critical
plane SWT for each wear increment and simply accumulate
this value from wear increment to increment, irrespective of the
fact that the associated orientations of the critical plane may
change with wear, i.e. assume the critical plane SWT damage
to have an isotropic damaging effect. An alternative approach is
to calculate SWT values for all plane orientations for each wear
increment and to accumulate the incremental damage on each
plane. The latter approach is significantly more computationally
intensive, requiring storage of damage on 36 different planes
at each element centroid (note that a rectangular grid of 1700
elements is monitored for damage here) for each wear increment
(typically about N* = 100 wear increments). In this study, the
former assumption is adopted, so that damage is treated as a
scalar quantity for each element. Failure is assumed to have
occurred when W reaches a value of 1 at any point.
Due to the use of adaptive meshing to update the mesh for
wear-induced removal of material, specific elements and nodes
are no longer linked uniquely to actual material points through-
out the analysis. Consequently, a ‘material point mesh’ (MPM)
 J.J. Madge et al. / Wear 263 (2007) 542–551
is created as the global reference for damage accumulation; the
nodes of the MPM have fixed coordinates throughout the analy-
sis. Cyclic damage is calculated at the centroid of each element,
and linearly interpolated back to the MPM for accumulation. In
this way, nodes on the MPM corresponding to removed material
(due to wear) do not accumulate any further fatigue damage.
5. Results and discussion
5.1. Effect of slip amplitude
The predicted effect of slip amplitude on fretting fatigue life
is shown in Fig. 7, along with experimental test data for the
same conditions as Jin and Mall [7]. The predictions show a
reasonable correlation with the relatively sparse experimental
results. The effect of slip amplitude on fatigue life is similar
to that reported by Madge et al. [10] for the Hertzian fretting
fatigue contact and generally accepted as representative of fret-
ting fatigue [3]. Specifically, a minimum in life is predicted at
the critical range of δapp amplitudes corresponding to the transi-
tion from partial slip to gross sliding conditions (approximately
50 ␮m ≤ δapp ≤ 85 ␮m), with comparatively long lives predicted
as δapp is either reduced or increased outside of this critical range.
At low (partial) slip amplitudes, the shear traction has a negligi-
ble effect so that comparatively long lives are predicted. As δapp
is increased from zero to the critical range, i.e. for δapp < 50 ␮m,
the predicted life reduces as the shear force, and in turn the
local fatigue stresses increase. At the onset of gross sliding, the
shear tractions have saturated and a minimum in life is observed.
However, as δapp is increased still further the effect of wear is to
increase fatigue life, as discussed in more detail below.
5.2. Effect of rig compliance
The presence of a complaint section in the model creates an
appreciable amount of load transfer throughout the fretting cycle
such that the pad is able to rotate under shear loading. At larger
slip amplitudes, the shear force is large enough to cause suffi-
cient deflection for the trailing edge of the pad to lift free, whilst
the leading edge is forced further into contact. This is illustrated
in Fig. 8, which shows the pressure distributions at maximum
and minimum substrate fatigue stress for the δapp = 40 ␮m par-
Fig. 7. Prediction of fretting fatigue life at different slip amplitudes as compared
to experimental results of Jin and Mall [7].
547
Fig. 8. Contact pressure distributions for points of maximum and minimum
fatigue stress (δapp = 40 ␮m).
tial slip case. Whether or not the experimental rig displays this
behaviour is not clear; however, it is interesting to note that the
pad rotates in this manner, even though it is modelled with an
encastre boundary condition and a compliant section modulus
(55 GPa) comparable to some engineering metals. This suggests
that reducing the system stiffness has a detrimental effect on life
by compromising contact alignment.
5.3. Gross sliding results
Fig. 9 shows the effect of wear on the contact pressure dis-
tribution for a gross sliding case (δapp = 90 ␮m). The two salient
predicted effects of wear in this case are (i) the significant reduc-
tion in the edge of contact pressure peak and (ii) the significant
increase in the contact area (and hence the load distribution
area) by approximately 50%. Fig. 10 shows the evolution of
critical plane SWT distribution across the contact surface with
increasing numbers of fretting cycles. For the unworn case and
for early numbers of cycles, e.g. N* = 10 (which corresponds to
N = 60,000 cycles for k = 2.75 × 10−8 MPa−1 ), there is a pro-
nounced peak at the left-hand edge of contact (viz. the trailing
edge at the instant of maximum substrate fatigue stress). How-
ever, it is clear that the predicted effect of wear here is to cause a
Fig. 9. Evolution of contact pressure distribution for the δapp = 90 ␮m gross
sliding case. Note: for a typical value of k = 2.75 × 10−8 MPa−1 , N* = 100 cor-
responds to N = 600,000 cycles.
548 J.J. Madge et al. / Wear 263 (2007) 542–551
Fig. 10. Evolution of SWT distribution across contact surface for the
δapp = 90 ␮m gross sliding case. Note: for a typical value of k = 2.75 ×
10−8 MPa−1 , N* = 70 corresponds to N = 420,000 cycles.
dramatic reduction in the SWT value at this location. The asso-
ciated rapid drop in cyclic fatigue damage is responsible for the
increase in life observed at higher slip amplitudes in Fig. 7. Fail-
ure is predicted at the trailing edge of contact for all gross sliding
cases.
5.4. Partial slip results
Fig. 11 shows the predicted evolution of contact pressure
distribution with increasing number of wear cycles for the
δapp = 40 ␮m partial slip case. As for the gross sliding case,
wear is predicted to result in an initial beneficial effect in
reducing the N* = 0 edge of contact pressure peak. Thus, for
k = 2.75 × 10−8 MPa−1 , for example, Fig. 11 shows that the
peak pressure value has dropped to about one quarter of its
initial value after about 132,000 cycles (N* = 22). For higher
wear coefficients, this number of cycles will reduce proportion-
ally, e.g. for k = 2.75 × 10−7 MPa−1 , the same attenuation will
have occurred within about 13,200 cycles. However, a critical
phenomenon of the partial slip case, as compared to the gross
sliding case, is the development of a pressure peak well away
from the contact edges as wear advances. It can be seen by
Fig. 11. Evolution of contact pressure distribution with wear for the δapp = 40 ␮m
partial slip case. Note: for a typical value of k = 2.75 × 10−8 MPa−1 , N* = 100
corresponds to 600,000 cycles.
Fig. 12. The N* = 100 slip distribution for the δapp = 40 ␮m partial slip case.
comparing Figs. 11 and 12 (which shows the slip distribution
for the δapp = 40 ␮m case) that this pressure peak in fact devel-
ops at the stick–slip interface. The discontinuity between wear
in the slip region and no wear in the stick region generates a
geometric and loading discontinuity which leads to a stress con-
centration, with load being gradually concentrated on the stick
region.
Fig. 13 shows the predicted evolution of the SWT param-
eter across the contact surface for the latter partial slip case.
Again, consistent with the attenuation in edge-of-contact pres-
sure peak (and with the gross sliding case) during the early
cycles, an initial attenuation in edge-of-contact SWT value is
predicted. However, the subsequent predicted development of
the pressure peak at the stick–slip interface leads directly to
an associated predicted increase in SWT value. The implica-
tion of this is that the issue of cracking location is predicted
to become more complex for partial slip cases than for gross
sliding; specifically, for partial slip cases, it is predicted that
failure can occur at either the contact edge or the stick–slip
interface, depending on the competition between rate of fatigue
damage accumulation and wear-induced evolution of fretting
stresses across the contact, as discussed below. A similar phe-
Fig. 13. Evolution of SWT distribution across the contact surface for a partial
slip case, δapp = 40 ␮m. Note: for a typical value of k = 2.75 × 10−8 MPa−1 ,
N* = 100 relates to 600,000 cycles.
 J.J. Madge et al. / Wear 263 (2007) 542–551
nomenon has been previously predicted for the Hertzian fretting
problem [10]. Even though wear is predicted to give an atten-
uation of edge of contact SWT peak for partial slip cases (as
for gross sliding), this is offset by an accentuation of SWT at
the stick–slip interface, which cannot be predicted by models
which do not include the effects of material removal. However,
the wear effect on partial slip life is somewhat secondary to the
effect of the shear traction increase, so that ‘no-wear’ models
can achieve acceptable life prediction accuracy for partial slip
cases.
5.5. Effect of wear coefficient on fretting fatigue behaviour
As mentioned above, it is of interest to study the predicted
interaction between wear and fatigue life and this can be eas-
ily achieved using the present methodology. For example, if
we consider the evolution of contact pressure and SWT dis-
tributions of Figs. 11 and 13, it can be seen that the time (N* )
at which the pressure peak and the associated SWT damage
peak at the stick–slip interface starts to develop is dependent
on the wear coefficient. If wear is retarded (i.e. by use of a
low wear coefficient), the initial geometry and its associated
pressure distribution (pressure and SWT peaks at either edge
of contact) will dominate fatigue damage accumulation so that
failure is predicted to occur at the contact edges. However, if
wear evolves more rapidly (i.e. high wear coefficient), the pres-
sure and SWT peaks at the edge of contact are reduced, but the
stick–slip interface SWT peaks then dominate damage accumu-
lation and hence failure is predicted at the stick–slip interface.
The effect on predicted fatigue life of varying the wear coeffi-
cient is presented here for the partial slip case of δapp = 40 ␮m.
Figs. 14 and 15 show the effect of k on predicted failure posi-
tion and life, respectively. The failure position is found to switch
from the edge of contact for k < kcrit ≈ 5.0 × 10−8 MPa−1 to the
stick–slip interface for k ≥ kcrit ≈ 5.0 × 10−8 MPa−1 . Referring
to Fig. 15, k is predicted to have relatively little effect on life
for k < 0.5 × kcrit ; the life is predicted to suddenly increase by a
factor of about 5 as k approaches kcrit . Finally, as k is increased
further, for k > kcrit , the life is found to gradually decay back
to approximately the same values as for k < kcrit . Jin and Mall
[7] have presented a wear scar for the partial slip case of the
punch on flat geometry analysed here, which was observed to
Fig. 14. Effect of wear coefficient on predicted failure position (δapp = 40 ␮m).
549
Fig. 15. Effect of wear coefficient on predicted fretting fatigue life (for the
δapp = 40 ␮m partial slip case).
give stick–slip interface cracking. A specific estimation of the
wear coefficient of k ≈8 × 10−7 MPa−1 has been obtained for
this case, following the method outlined by McColl et al. [5].
The present analyses predict failure to occur at the stick–slip
interface under partial slip conditions for this k value. It is
worth pointing out that the value of kcrit will change for dif-
ferent fretting conditions, e.g. applied displacement, normal
load.
These results show that the interaction between fatigue and
fretting wear is critical to the prediction of failure cause, location
and number of cycles. Essentially, if the component can sur-
vive the initially severe loading conditions, viz. edge-of-contact
pressure peaks and associated contact stress concentrations, so
that wear-induced material removal can alter the pressure dis-
tribution, failure will occur at the stick–slip boundary, due to a
completely different stress distribution to that given by the initial
geometry.
The reason for this peak in life at kcrit (Fig. 15) lies in
the material utilisation. Fig. 16 shows the δapp = 40 ␮m par-
tial slip damage contour plots corresponding to failure (W = 1)
for three values of wear coefficient, namely k = 0.55 × kcrit
(k = 2.75 × 10−8 MPa−1 ), k ≈ kcrit (k = 5.5 × 10−8 MPa−1 ) and
k = 1.6 × kcrit (k = 8 × 10−8 MPa−1 ). In the k = 0.55 × kcrit case
(Fig. 16a) damage is highly localised at the trailing edge of
contact. Conversely, in the k = 1.6 × kcrit case (Fig. 16c), dam-
age is highly localised to the stick/slip boundary. However,
in the k ≈ kcrit case (Fig. 16b), two regions of high damage
are found, one near the edge of contact, and a second at the
stick/slip interface. The material utilisation is high in this case
because a greater amount of fatigue damage has been sustained
by spreading it over a larger area. Fig. 15 indicates that there is an
optimum wear coefficient (kcrit ) for maximum life under given
fretting conditions. These results furthermore indicate that the
present wear-fatigue methodology has the capability to model,
and therefore ultimately design for, commonly-observed phe-
nomena from industrial practice such as ‘bedding-in’, whereby
beneficial effects can be derived from using a mixed material
combination with one wear-‘soft’ material and one wear-‘hard’
material.
550 J.J. Madge et al. / Wear 263 (2007) 542–551
Fig. 16. Showing the damage contours at failure (W = 1) for different
values of wear coefficient: (a) k = 2.75 × 10−8 MPa−1 < kcrit ; (b) k = 5.5 ×
10−8 MPa−1 ≈ kcrit ; (c) k = 8 × 10−8 MPa−1 > kcrit .
6. Conclusions
In this paper, a finite element based method has been applied
to predict and characterise the role of fretting wear on fretting
fatigue for a rounded punch-on-flat Ti-6Al-4V fretting test con-
figuration, with comparisons against published test data from
Mall and co-workers. The key findings are as follows:
• A significant effect of slip amplitude on fretting fatigue life
was predicted, viz. a critical range of slip amplitude for mini-
mum life corresponding broadly to the transition from partial
slip to gross sliding, which is consistent with (i) a generally
accepted phenomenon of fretting fatigue and (ii) correspond-
ing FE-based life predictions (with wear) [10] and test data
for a Hertzian fretting test configuration.
• The life predictions were shown to exhibit good quantitative
agreement with available test data in the region of minimum
fatigue life.
• Prediction of failure position: It was shown that for gross
sliding cases, failure is invariably predicted at the edge of
contact, consistent with experimental observations. However,
for partial slip cases, it was shown that due to the competi-
tion between wear rate and fatigue damage accumulation and
the effects of edge-of-contact versus stick–slip interface SWT
concentrations, failure location depended on the wear coef-
ficient (for assumed constant fatigue properties). Thus, for
low wear coefficients, partial slip failure is predicted at the
edge-of-contact and for high wear coefficients at the stick–slip
interface, with the sudden shift in position predicted at a single
(critical) value of wear coefficient.
• The same critical value of wear coefficient was shown to
give an optimum (maximum) fatigue life, for a given set of
fatigue constants and fretting conditions. This method pro-
vides a framework for designing for optimum fretting fatigue
life.
• The work has demonstrated that prediction of fretting wear
is a critical aspect for prediction of fretting fatigue and can
have a major effect on the predicted evolution of stresses
and damage parameters (e.g. SWT), and hence on predicted
fatigue lives and failure position. This fundamental and impor-
tant effect for fretting fatigue cannot be predicted by models
which do not include the effects of material removal due to
wear.
Acknowledgements
The authors wish to thank Rolls-Royce plc, Aerospace Group,
for their financial support of the research, which was carried
out at the University Technology Centre in Gas Turbine Trans-
mission Systems at the University of Nottingham. The views
expressed in this paper are those of the authors and not neces-
sarily those of Rolls-Royce plc, Aerospace Group.
References
[1] J.A. Collins, S.M. Macro, The effect of stress direction during fretting
on subsequent fatigue life, Proc. Am. Soc. Test. Mater. 64 (1964) 547–
560.
[2] J.M. Dobromirski, in: R.B. Waterhouse, M.H. Attia (Eds.), Variables in
the Fretting Process: are There 50 of Them? Standardisation of Fretting
Fatigue Test Methods and Equipment, ASTM, 1992, pp. 60–66.
[3] D.A. Hills, D. Nowell, The Mechanics of Fretting Fatigue, Dordrecht,
Kluwer, 1994.
[4] O. Jin, S. Mall, Effects of slip on fretting behaviour: experiments and
analyses, Wear 256 (2004) 671–684.
[5] I.R. McColl, K. Ding, S.B. Leen, Finite element simulation and exper-
imental validation of fretting wear, Wear 256 (11–12) (2003) 1114–
1127.
[6] S.B. Leen, C.H.H. Ratsimba, I.R. McColl, E.J. Williams, T.R. Hyde, An
investigation of the fatigue and fretting performance of a representative
aeroengine splined coupling, J. Strain Anal. 37 (6) (2002) 565–585 (special
issue on fretting fatigue).
[7] O. Jin, S. Mall, Influence of contact configuration on fretting fatigue behav-
ior of Ti-6Al-4V under independent pad displacment condition, Int. J.
Fatigue 24 (2002) 1243–1253.
[8] K. Nakazawa, N. Maruyama, T. Hanawa, Effect of contact pressure on
fretting fatigue of austenitic stainless steel, Tribol. Int. 36 (2003) 79–
85.
[9] O. Vingsbo, S. Soderberg, On fretting wears, Wear 126 (1988) 131–147.
[10] J.M. 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.
[11] T. Hattori, T. Watanabe, Fretting fatigue strength estimation considering
the fretting wear process, Tribol. Int. 39 (2006) 1100–1105.
[12] K.N. Smith, P. Watson, T.H. Topper, A stress-strain function for the fatigue
of metals, J. Mater. 15 (1970) 767–778.
[13] V. Sabelkin, S. Mall, Investigation into relative slip during fretting fatigue
under partial slip contact condition, Fatigue Fracture Eng. Mater. Struct.
28 (2005) 809–824.
[14] ABAQUS User’s and Theory Manuals, Version 6.5, ABAQUS Inc., Paw-
tucket, Rhode Island, 2005.
 J.J. Madge et al. / Wear 263 (2007) 542–551
[15] R. Magaziner, O. Jin, S. Mall, Slip regime explanation of observed size
effects in fretting, Wear 257 (2004) 190–197.
[16] A. Fatemi, D.F. Socie, A critical plane approach to multiaxial fatigue dam-
age including out-of-phase loading, Fatigue Fracture Eng. Mater. Struct.
11 (3) (1988) 149–165.
[17] D. Socie, Multiaxial fatigue damage models, J. Eng. Mater. Technol. Trans.
ASME 109 (1987) 293–298.
[18] J.A. Araujo, D. Nowell, The effect of rapidly varying contact stress fields
on fretting fatigue, Int. J. Fatigue 24 (2002) 763–775.
[19] M.P. Szolwinski, T.N. Farris, Mechanics of fretting fatigue crack formation,
Wear 198 (1996) 93–107.
[20] M.W. Brown, K.J. Miller, A theory for fatigue failure under multiaxial
stress-strain conditions, Proc. Inst. Mech. Eng. 187 (1973) 745–755.
551
[21] V. Fridrici, S. Fouvry, P. Kapsa, P. Peruchaut, Prediction of cracking in
Ti-6Al-4V alloy under fretting-wear: use of the SWT criterion, Wear 259
(1–6) (2005) 300–308.
[22] C.D. Lykins, S. Mall, V.K. Jain, Combined experimental-numerical investi-
gation of fretting fatigue crack initiation, Int. J. Fatigue 23 (2001) 703–711.
[23] W.S. Sum, E.J. Williams, S.B. Leen, Finite element, critical plane, fatigue
life prediction of simple and complex contact configurations, Int. J. Fatigue
27 (2005) 403–416.
[24] S.M. Marko, W.L. Starkey, A concept of fatigue damage, Trans. ASME 76
(4) (1954) 627–632.
[25] N.E. Dowling, Mechanical Behaviour of Materials: Engineering Methods
for Deformation, in: Fracture and Fatigue, second ed., Prentice Hall, NJ,
1998.