Fretting fatigue life estimations based on fretting mechanisms

Toshio Hattori
a,n
, Vu Trung Kien
b
, Minoru Yamashita
c
a
Department of Mechanical and System Engineering, Gifu University, Gifu, Japan
b
Graduate School of Engineering, Gifu University, Gifu, Japan
c
Department of Mechanical and System Engineering, Gifu University, Gifu, Japan
a r t i c l e i n f o
Article history:
Received 14 May 2010
Received in revised form
20 October 2010
Accepted 21 October 2010
Available online 14 January 2011
Keywords:
Stress singularity
High cycle fatigue
Low cycle fatigue
Critical distance theory
a b s t r a c t
Generally the fretting fatigue SN curve has two regions: one is the high cycle (lowstress) region and the
second is the low cycle (high stress) region. In a previous paper we introduced the fretting fatigue life
estimation methods in high cycle region by considering the wear process; with this estimation method
the fretting fatigue limit can be estimated to be the crack initiation limit at the contact edge. In this paper
we estimate the lowcycle fretting fatigue life based on a newcritical distance theory, modied for a high
stress region using ultimate tensile strength s
B
and fracture toughness K
IC
. The critical distance for
estimating low cycle fretting fatigue strength was calculated by interpolation of the critical distance on
the fretting fatigue limit (estimatedfroms
w0
andDK
th
) withcritical distance onstatic strength(estimated
from s
B
and K
IC
). By unifying this low cycle fretting fatigue life estimation method with the high cycle
fretting fatigue life estimation method, which was presented in the previous paper, we can estimate the
total fretting life easily. And to conrmthe availability of this estimation method we performthe fretting
fatigue test using NiMoV steel.
& 2010 Elsevier Ltd. All rights reserved.
1. Introduction
Frettingcanoccur whena pair of structural elements are incontact
under a normal load while cyclic stress and relative displacement are
forced along the contact surface. This condition can be seen in bolted
or rivetedjoints [1,2], inshrink-ttedshafts [3,4], inthe blade dovetail
region of turbo machinery [5,6], etc. During fretting the fatigue
strengthdecreases toless thanone-thirdof that without fretting [7,8].
The strength is reduced because of concentrations of contact stresses,
such as contact pressure and tangential stress at the contact edge,
where fretting fatigue cracks initiate and propagate.
This concentration of stress can be calculated using the nite
element method [9] or the boundary element method. Methods of
estimation of the strength of fretting fatigue have been developed,
which use the values of this stress concentration on a contact surface
[3,5]. However, thestress elds near thecontact edges showsingularity
behavior, where the stresses at contact edges are innite. Thus,
maximumstresses cannot be usedtoevaluate fretting fatigue strength.
So, in our previous papers we presented the fretting crack
initiation estimation method, using stress singularity parameters at
contact edges [10,11,13], and fretting fatigue limit or life estimation
methods using fracture mechanics [7,12,13]. Using these fretting
fatigue strength or life estimation methods we could not predict
the super-high-cycle fretting fatigue problems in the industrial eld.
For instance, a 660 MW turbo-generator rotor failed in England
during the 1970s as a result of fretting fatigue cracking as shown
in Fig. 1 [14]. In this case the number of loading cycles in just one year
was about 1.610
9
, and this problemwas observed after many years
of operation. This very-high-cycle fatigue life cannot be explained
using only initial stress analysis results. Then in a previous paper we
presented the very-high-cycle fatigue life estimation method con-
sidering fretting wear [15] and change in the contact surface [16,17].
In the case of designing rotational machinery such as turbo
machinery, we must estimate the low cycle fatigue strength or the
life of the blade/disk that connects structures in the start/stop process
(as shown in Fig. 2). Here, in this paper we introduced the low cycle
fatigue life method using the critical distance theory, and nally we
canestimatetheSNcurveof theentireprocess fromlowcyclefatigue
toveryhighcyclefatigue. Toconrmtheavailabilityof this estimation
method we perform the fretting fatigue test using NiMoV steel.
2. Fretting fatigue process
In our previous paper [16,17] we presented a fretting fatigue
process model as illustrated in Fig. 3. Cracking due to fretting
fatigue starts very early in fretting fatigue life. We used stress
singularityparameters at the contact edge toestimate the initiation
of these cracks [10,11,13]. During this early period, fretting fatigue
cracks tend to close and propagate very slowly especially in a low
stress amplitude range due to the high contact pressure acting near
this contact edge. But wear on the contact surface reduces the
contact pressure near the contact edge, and cracks gradually start
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/triboint
Tribology International
0301-679X/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.triboint.2010.10.020
n
Corresponding author. Tel.: +58 293 2503.
E-mail address: hattori@gifu-u.ac.jp (T. Hattori).
Tribology International 44 (2011) 13891393
to propagate. Hence, the fretting fatigue life in the low stress
amplitude range will be dominated by the propagation of these
small cracks that initiate at the contact edge. To estimate the
fretting fatigue strength or life in this lowstress region, the precise
estimation of the fretting wear progress is indispensable. The
propagation life of a long crack length region can be estimated
using fracture mechanics. In our previous paper [16,17] we
discussed the method of estimation of wear extension on contact
surfaces near the contact edge, and presented the fretting fatigue
crackpropagationestimationmethodconsidering the frettingwear
extension process.
On the other hand, in the case when the stress range is high, the
crack initiationwill leadtofailure easily without wear. Inthis paper
we estimate the fretting fatigue life in this high stress range using
the critical distance theory.
3. Fretting fatigue life analysis considering fretting wear in
high cycle region
In Fig. 4 the owof fretting fatigue life analysis, considering the
extension of fretting wear, is shown. Firstly the fretting wear
amount is estimated using contact pressure and relative slippage
under each loading condition [16,17], and then the shapes of
contact surfaces are modied following the fretting wear amount.
This fretting wear amount is estimated using the classic Archards
equation as follows [15]:
W KPS 1
where W is the wear depth, K the wear coefcient, P is the contact
pressure and S the accumulative slippage.
In this equation the slippage S was calculated from the
deformation of each node on the contact surface, and the frictional
coefcient on this contact surface was set at 0.7.
Then, the fretting fatigue life under each loading condition,
considering the wear process, can be estimated by comparing the
operating stress intensity factor range DK with the threshold stress
intensity factor range DK
th
. If the operating DK is higher than the
Fig. 1. Fretting fatigue failure example inturbo-generator rotor. After Lindley and Nix
(1991) [14].
Fig. 2. Assembled gas-turbine compressor rotor and blade dovetail joint.
0
0
0
0
0
0
K
K
K
Pressure
Pressure
Pressure
Stress
Stress
Stress
Crack initiation
Wear extension
Crack propagation
Stress singularity
parameters H,
Archards eq.
W=A x p x s
Fracture mechanics
da/dN=C (K)
m
Wear
K

Fig. 3. Fretting fatigue mechanisms of various processes.

Fretting fatigue strength and life
Structural and load condition
Surface condition
Stress analysis
Fracture mechanics analysis
Wear analysis
K > K
TH
Surface modification
Fig. 4. Flow chart of fretting fatigue life analysis.
T. Hattori et al. / Tribology International 44 (2011) 13891393 1390
threshold stress intensity factor range DK
th
we can estimate this
load cycle to be a fretting life, and if the operating DK is lower than
the threshold stress intensity factor range DK
th
the fretting wear
amount will be calculated using a new contact pressure and a new
relative slippage; these processes will be repeated until operating
DK reaches the threshold stress intensity factor range DK
th
. In these
comparisons, the threshold stress intensity factor ranges were
estimated considering the crack length and stress ratio, derived in
the previous paper [7,12,13]. By connecting these fretting
threshold conditions we can estimate the fretting fatigue SN
curve considering the wear process.
4. Fatigue life analysis using critical distance theory in low
cycle region
Even under the fretting conditions, the fretting wear will be
neglected in a high stress region. In these cases the fatigue life will
be estimated to be the crack initiation condition. In this paper we
also estimated the fretting fatigue life using the critical distance
stress theory (point method and line method). In this method the
fatigue strength limit can be obtained from typical material
strength parameters, such as the fatigue limit of smooth specimens
s
w0
and the threshold stress intensity factor range DK
th
of the
cracked specimens as shown in Figs. 5 and 6. Inthe case of the point
method, the fatigue failure is supposed to occur when the stress
range at specic length r
P
from the maximum stress point reaches
Ds
w0
, and in the case of line method the fatigue failure is supposed
to occur when the mean stress range between the maximumstress
point and specic lengthpoint r
L
reaches Ds
w0
(s
w0
2); r
P
and r
L
can be derived as follows:
For point method, r
P
DK
th
=Ds
wo

2
=2p 2
And for line method, r
L
2DK
th
=Ds
wo

2
=p 3
In this paper we extended this method to the low cycle fatigue
regions. First the critical distance in the low cycle fatigue region is
derived through the interpolation between the critical distance in
the fatigue limit and the critical distance in static strength. This
static strength critical distance can be derived using ultimate
strength of smooth specimen s
B
and the fracture toughness K
IC
of the cracked specimen as shown in Eqs. (4) and (5):
For point method, r
P
u K
IC
=s
B

2
=2p 4
And for line method, r
L
u 2K
IC
=s
B

2
=p 5
2r
K
th
(r) =

w0
r
P r
crack
Fig. 5. Derivation of critical distance r
P
.
2r

th
(r) =
S
2
S
1
r
L
crack
r

Fig. 6. Derivation of critical distance r
L
(point method and line method).
S
t
r
e
s
s

a
m
p
l
i
t
u
d
e

=

2

a
C
r
a
c
k

p
r
o
p
a
g
a
t
i
o
n

r
a
t
e


d
a
/
d
N
10
0
10
2
10
4
10
7
Stress intensity factor range K

w0

B
K
IC
K
IC
K
IC

Estimated cycles to failure

Crack propagation rate
Number of cycles to failure N
f

w0

Specific distance and stress

Distancer
2
(
w0
= 2
w0
)
S-N Curve of pane specimen
Stress distributions
of objectives
K
th
K
th
.. 1
K
th

w0 2
2
1
K
th

w0
2
2
1
K
IC
B
2
2
1
K
IC

B
2
2

w0
K
th
Fig. 7. Derivation of specic distance in low cycle fatigue region and estimation of low cycle fatigue life.
T. Hattori et al. / Tribology International 44 (2011) 13891393 1391
In this section we illustrate the approach using the point
method. The critical distance at each stress level is calculated by
interpolation of critical distance on fretting fatigue limit (r
P
,
estimated from s
w0
and DK
th
) with critical distance on static
strength (r
P
0
, estimated froms
B
and K
IC
) as shown by chain line in
Fig. 7(right). The critical distance under objective conditions
(structure, load) can be estimated by reecting the stress distribu-
tions of the objective structure as shown by dotted line in
Fig. 7(right). The low cycle fatigue life in this objective condition
can be estimated by applying this reference stress s at critical
distance r on SNcurve of smooth specimens as shown in Fig. 7(left
upper).
5. Application on low cycle fretting fatigue life analysis
The extended critical distance theory will then be applied in the
fretting fatigue life prediction. In Fig. 8(left upper side) the SN
curve of the NiMoV steel smooth specimen under complete
reversed loading conditions (R 1), and in Fig. 8(left lower) the
crack propagation characteristics of the cracked specimen are
shown. From these material characteristics we found that the
critical distance r
P
is 0.011 mm and r
P
0
is 2.13 mm as shown in
Fig. 8(right). The stress distributions under the fretting conditions
were calculated using the FEM model as shown in Fig. 9. The
calculated example of stress distribution near the contact edge is
shown in Fig. 10. The mean contact pressure s
p
and mean axial
stress s
a
in this case are 200 and 100 MPa, respectively.
The critical distance r on each loading condition can be
estimated by reecting these stress distributions in Fig. 8(right)
as shown by the dotted line. As an example we can estimate the
critical distance r just 2 times higher than that of loading condition
shown in Fig. 10. As a cross point of stress distribution (dotted line)
with solid line, r and s can be derived as 0.12 mm and 490 MPa,
respectively. The low cycle fretting fatigue life in the loading
condition, with nominal axial stress s
a
200 MPa, can then be
estimated by applying the stress (s490 MPa) on the SN curve of
smooth specimens as shown in Fig. 8(left upper). By connecting
these fretting fatigue life at each stress level we can estimate the
fretting fatigue SN curve as shown in solid line in Fig. 12.
S
t
r
e
s
s

(
M
p
a
)
Distance r (mm)

a
=200MPa

W0
K
th
K
th
K
IC
K
IC

B
S-N Curve of plane specimen
Crack propagation rate
Specific distance and stress

w0
Stress distributions
in fretting model
Ni-Mo-V Steel
10
3
10
5
10
6
10
7
10
8
10
4
100
500
1000
Number of cycles to failure N
f
S
t
r
e
s
s

a
m
p
l
i
t
u
d
e

a

(
M
p
a
)

B
=705MPa

W0
=360MPa
Estimated cycle to failure
2
(
w0
= 2
w0
)
2
2
K
th

w0

2
2
K
IC

Fig. 8. Derivation of specic distance in low cycle fatigue region and estimation of low cycle fretting fatigue life.
Fig. 9. FEM fretting model.
T. Hattori et al. / Tribology International 44 (2011) 13891393 1392
6. Comparison with the experimental results
To conrm the validity of this estimation method we compare
the estimated results with the experimental results. The fretting
fatigue test apparatus is shown in Fig. 11. The specimen material is
NiMoVsteel. The contact pressure betweenspecimenandpads is
set to 200 MPa by screw. The contact pressure and crack initiation
at the contact edges are monitored by the strain gages A and B,
respectively, with calibration.
The experimental results of the fretting fatigue tests are shown
in Fig. 12, denoted by symbols. The estimated results of low cycle
fretting fatigue life using the critical distance theory are shown by
the dottedline inFig. 12. The estimatedresults of highcycle fretting
fatigue life considering fretting wear process, which was presented
inprevious paper [16,17], is shownby the dashedline inFig. 12. The
estimated fretting fatigue limit (142 MPa) without considering
fretting wear, which was presented in previous paper [12,13], is
shownbytwopoints onthe dotdashline inFig. 12. We cansee that
these three kinds of fretting fatigue strength and life prediction
results coincided well with the experimental results at each stress
and life level, and we can conrm the validity of these fretting
fatigue strength and life estimation methods.
7. Conclusions
1. Low cycle fretting fatigue strength was estimated using critical
distance theory.
2. These fretting fatigue strength and life estimated results coin-
cidedwell withthe fretting fatigue test results at eachstress and
life level, and we can conrm the availability of these fretting
fatigue strength and life estimation methods through the
standardized fretting fatigue SN curve estimation method.
References
[1] Gassner E. The value of surface-protective media against fretting corrosion on
the basis of fatigue strengthtests. Laboratoriumfur Betriebsfestigkeit TM19/67
1967.
[2] Buch A. Fatigue and fretting of pin-lug joints with and without interference t.
Wear 1977;43:9.
[3] Hattori T, Kawai S, Okamoto N, Sonobe T. Torsional fatigue strength of a shrink-
tted shaft. Bulletin of the JSME 1981;24(197):1893.
[4] Cornelius EA, Contag D. Die Festigkeits-minderung von Wellen unter dem
Einu b von Wellen-Naben- Verbindungen durch Lotung, Nut und Pa b feder,
Kerbverzahnungen und Keilprole bei wechselnder Drehung. Konstruktion
1962;14(9):337.
[5] Hattori T, Sakata S, Ohnishi H. Slipping behavior and fretting fatigue inthe disk/
blade dovetail region. in: Proceedings of the 1983 Tokyo international gas
turbine congress, 1984. p. 945.
[6] Johnson RL, Bill RC. Fretting in aircraft turbine engines. NASA TM X-71606
1974.
[7] Hattori T, Nakamura M, Watanabe T. Fretting fatigue analysis by using fracture
mechanics. American Society of Mechanical Engineers Paper no. 84-WA/DE-10
1984.
[8] King RN Lindley TC. Fretting fatigue in a 3
1
/
2
NiCrMoV rotor steel. In:
Proceedings of ICF5, 1980. p. 631.
[9] Okamoto N, Nakazawa M. Finite element incremental contact analysis with
various frictional conditions. International Journal for Numerical Methods in
Engineering 1979 14:377.
[10] Hattori T, Sakata H, Watanabe T. A stress singularity parameter approach for
evaluating adhesive and fretting strength. American Society of Mechanical
Engineers, Book no. G00485, MD-vol.6, 1988. p. 43.
[11] Hattori T, Nakamura N. Fretting fatigue evaluation using stress singularity
parameters at contact edges, Fretting Fatigue, 18. ESIS Publication; 1994
p. 453.
[12] Hattori T, et al. Fretting fatigue analysis using fracture mechanics. JSME
International Journal Series l 1988;31:100.
[13] Hattori T, Nakamura M, Watanabe T. Simulation of fretting fatigue life by using
stress singularity parameters and fracture mechanics. Tribology International
2003;36:87.
[14] Suresh S. Fatigue of Materials. 2nd ed.. Cambridge University Press; 1998
p. 469.
[15] Goryacheva IG, Rajeev PT, Farris TN. Wear in partial slip contact. ASME Journal
of Tribology 2001;123(4):848.
[16] Hattori T, Yamashita M, Nishimura N. Fretting fatigue strength and life
estimation in high cycle region considering the fretting wear process. JSME
International Journal 2005;48(4):24650.
[17] Hattori T, Nakamura, Nishimura N, Yamashita M. Fretting fatigue strength
estimation considering the fretting wear process. Tribology International
2006;39:11005.
10
3
10
5
10
6
10
7
10
8
10
4
100
500
1000
Number of cycles to failure N
f
S
t
r
e
s
s

a
m
p
l
i
t
u
d
e

a

(
M
p
a
)




Plane specimen
Fretting (Low cycle)
Fretting
(Ultra high cycle)
Experimental
Smooth specimen
Fig. 12. Estimated and experimental fretting fatigue SN curves.
Fig. 11. Fretting fatigue test apparatus.
Fig. 10. Calculated result of stress distributions.
T. Hattori et al. / Tribology International 44 (2011) 13891393 1393