Beruflich Dokumente
Kultur Dokumente
Mississippi State University, Department of Mechanical Engineering, 210 Carpenter Hall, Mail Box 9552, Mississippi State, MS 39762, United States
Chrysler Group LLC, 800 Chrysler Dr., Auburn Hills, MI 48326, United States
a r t i c l e
i n f o
Article history:
Received 25 September 2013
Received in revised form 10 February 2014
Accepted 23 February 2014
Available online 6 March 2014
Keywords:
Multiaxial loading
Fatigue life predictions
Classical theories
Critical plane approaches
Non-proportional loading
a b s t r a c t
The aim of this study is to estimate fatigue life of steels and super alloys under multiaxial loading based
on commonly available tensile properties. The state of loading for most components and structures is
multiaxial resulting from multidirectional loading or stress concentrations. Multiaxial fatigue models
have been developed to predict fatigue behavior under multiaxial loading. These models relate multiaxial
stress/strain components to uniaxial fatigue properties in order to predict fatigue life. In this study, MuralidharanManson, BumelSeeger, and RoessleFatemi prediction methods are employed to predict uniaxial fatigue properties based on simple tensile properties in the absence of any fatigue data. Appropriate
multiaxial fatigue models representing the damage mechanism are then used along with the estimated
uniaxial fatigue properties to predict fatigue lives under in-phase and out-of-phase multiaxial loading.
Predictions are compared with experimental multiaxial data for sixteen different steels and super alloys
from literature. Some approximation techniques to predict stress response for in-phase and out-of-phase
loading based on simple tensile properties are also reviewed. Stress estimated based on these approximation techniques are then used in multiaxial fatigue life predictions and results are compared with experimental observations. It is concluded that fatigue life of steels and super alloys under multiaxial loading
may be predicted reasonably well using appropriate damage models only requiring monotonic
properties.
2014 Elsevier Ltd. All rights reserved.
1. Introduction
Multiaxial states of loading are very typical in many industrial
applications. The multiaxial stresses/strains in critical elements
of components and structures can result from multidirectional
loading, stress concentrations due to geometrical complexity, and
residual stresses generated from manufacturing processes. Multiaxial loading can be categorized as in-phase (IP) and out-of-phase
(OP) loading. For in-phase loading, the ratio of torsion to axial loading and principal directions remain xed. However, under out-ofphase loading, principal directions and consequently maximum
shear directions rotate in time.
Fatigue lives under out-of-phase loading are usually shorter
than in-phase loading at the same equivalent strain level. Kanazawa
et al. [1] related the shorter fatigue lives under out-of-phase
(non-proportional) loading to the non-proportional cyclic
hardening phenomenon. They [1] explained this additional nonproportional cyclic hardening phenomenon with the change in slip
Corresponding author. Tel.: +1 662 325 2364; fax: +1 662 325 7223.
E-mail addresses: shamsaei@me.msstate.edu (N. Shamsaei), SM1453@Chrysler.
com (S.A. McKelvey).
1
Tel.: +1 248 944 5083.
http://dx.doi.org/10.1016/j.ijfatigue.2014.02.020
0142-1123/ 2014 Elsevier Ltd. All rights reserved.
plane from one crystallographic slip system to another one resulting from the rotation of maximum shear plane under nonproportional loading. The interaction of active slip systems then
may cause an additional hardening under non-proportional cyclic
loading.
Multiaxial fatigue models can be used to relate multiaxial state
of loading to uniaxial fatigue properties. Classical models, such as
Maximum Principal Strain and von Mises, were rst proposed in
the early twentieth century as failure theories under static or
monotonic loading. These hypotheses were later extended to cyclic
loading and fatigue strength. For tensile failure mode materials, the
Maximum Principal Strain model has been commonly used to predict fatigue life. The Maximum Principal Strain is related to fatigue
properties and life as shown below:
e1;max
r0f
E
63
Nomenclature
OP
Dr
Dr/2
Ds/2
e
ef
e1,max
ea
ee
ep
b
c
E
FS
HB
IP
k
K
K0
n
n0
2Nf
OP
SWT
1
ea p
21 m
s
2
2
De
3 Dc
2
1 m
2
2 2
2
e0f
w
k
me
mp
m
r
rmax
1
rmax
n
ru
ry
r IP
r OP
r0f
s
where De/2 is the axial strain amplitude, Dc/2 is the shear strain
is the equivalent Poissons ratio and can be calcuamplitude, and m
lated from Eq. (3):
m
me Dee mp Dep
De
where ee , ep , and e are equivalent elastic, plastic, and total strains
and me and mp are elastic and plastic Poissons ratios. The von Mises
equivalent strain and fatigue life are related through the following
CofnManson equation (i.e. Eq. (4)), and therefore, this equation
can be used to calculate fatigue life based on the von Mises criterion, when equivalent strain is calculated from Eq. (2):
ea
r0f
E
However, these classical models may only work for proportional or in-phase loading. For the case of non-proportional or
out-of-phase loadings, using classical models often leads to significant errors as these models do not consider the effects of load nonproportionality. Critical plane models which reect the damage
mechanism and predict the failure on the specic critical plane(s)
within the material have been developed over the last few decades
[2]. These models may be used for fatigue life estimations under
both IP and OP loading and also for predicting the direction of crack
initiation. Among all types of critical plane approaches, strain
stress-based models have the advantage of reecting the constitutive behavior of material such as non-proportional cyclic hardening. These models include both a strain component as the driving
parameter and a secondary stress component taking into account
the cyclic hardening due to non-proportionality of loading as well
as mean and residual stresses. SmithWatsonTopper (SWT) [3]
rmax
1
De1;max r02
f
max
Dcmax
r
1k n
2
ry
r0f
r0f
2Nf b
1 me 2Nf b 1 mp e0f 2Nf c 1 k
E
2r y
where ry is the material monotonic yield strength, and k is a material constant found by tting fatigue data from uniaxial tests to fatigue data from torsion tests.
Fatigue data are not always available and generating fatigue
properties is an expensive process. Furthermore, a slight change
in material chemical composition or any surface enhancements
such as shot peening or hardening may greatly affect the fatigue
behavior. Therefore, developing predictive techniques for fatigue
64
2. Experimental data
In this study, multiaxial fatigue data from eleven shear failure
mode materials and ve tensile failure mode materials are
Fig. 1. Cyclic stressstrain curves for (a) shear failure mode materials and (b) close
up of lower strength shear failure mode materials considered in this study.
65
Fig. 2. Cyclic stressstrain curves for tensile failure mode materials considered in
this study.
max h
ih
i
Dcmax
r
A2N f 0:09 B2Nf 0:56 1 kC2Nf 0:09
1k n
2
ry
A
steel. In addition aluminum materials have low hardness levels
which are outside the range of applicability of the hardness method. For these reasons, only steels and super alloys are included in
the following analysis.
r0:832
u
32; 000
B 19e0:155
r0:53
f
u
C
2:42r0:832
u
1:17ru 426
0:075ru 32
In order to estimate strain-life fatigue properties based on tensile properties, Muralidharan and Manson [6] modied the universal slopes method, previously proposed by Manson [25] as follows:
r0f 0:623E
r 0:832
u
E
b 0:09; c 0:56
e0f 0:0196e0:155
f
r 0:53
u
;
7
r0:832
2Nf 0:09
u
or k 1
0:31ry
r0f 2Nf b
Table 1
Material properties for shear failure mode materials.
Monotonic properties
Modulus of Elasticity, E (GPa)
Shear modulus, G (GPa)
Yield strength, ry (MPa)
Ultimate strength, ru (MPa)
Brinell hardness, HB
Strength coefcient, K (MPa)
Strain hardening exponent, n
1050 N
[9,11]
1050 QT
[9,11]
1050 IH S45C
[9]
[12]
1045 N
[13,14]
SNMC630 304L SS
[15]
[11]
206
81
421
709
198
1455
0.253
203
81
1009
1164
360
1461
0.060
198
79
2200
2248
565
NA
NA
186
NA
496
770
182
NA
NA
204
80
382
620
189
1185
0.230
196
NA
951
1103
348
NA
NA
195
NA
390
760
185
NA
NA
1027
195
77
208
585
130
680
0.214
193
NA
310
605
160
NA
NA
193
NA
379
767
172
NA
NA
209
78
1160
1850
370
1910
0.080
Cyclic properties
Fatigue strength coefcient, r0f (MPa) 1109
1346
4974
932
1008
1287
1157
1124
838
3950
0.100
0.292
0.062
2.010
0.152
0.529
0.099 0.107
0.359 0.322
0.042
1.168
0.145
0.122
0.115
1.165
0.091
0.807
0.058
0.178
0.151
1.500
0.456
1480
0.725
1558
0.910
3328
0.519 0.487
1246
1258
0.792
NA
0.394
2939
0.568
1036
0.665
1115
0.446
1047
0.761
1564
0.223
0.123
0.083
0.199
0.208
NA
0.374
0.189
0.130
0.129
0.068
1249
1420
7980
728
1358
NA
5056
6122
1680
NA
NA
0.177
0.113
0.221
0.093
0.208
NA
0.373
0.354
0.158
NA
NA
66
Table 2
Material properties for tensile failure mode materials.
Monotonic properties
Modulus of elasticity, E (GPa)
Shear modulus, G (GPa)
Yield strength, ry (MPa)
Ultimate strength, ru (MPa)
Brinell hardness, HB
Strength coefcient (MPa)
Strain hardening exponent
Cyclic properties
Fatigue strength coefcient,
r0f (MPa)
SA333Gr6 [20]
304 SS [21,22]
310 SS [23]
203
NA
302
450
137
NA
NA
183
83
325
650
160
1210
0.193
NA
NA
NA
641
145
NA
NA
170
NA
268
490
130
512
0.093
156
61
244
581
130
NA
NA
921
1000
NA
823
899
0.124
0.392
0.114
0.171
NA
NA
0.082
0.489
0.089
0.084
0.532
1015
0.211
683
0.078
0.402
1660
0.287
4668
0.344
NA
NA
NA
4450
0.338
0.730
891
0.113
851
0.088
0.446
1474
0.199
2562
0.237
Fig. 3. Predicted fatigue lives versus observed fatigue lives with factor of 3 and 5 scatter bands for predictions made using modied universal slopes with (a) von Mises
equivalent strain theory and (b) FatemiSocie critical plane approach, using uniform material law with (c) von Mises equivalent strain theory and (d) FatemiSocie critical
plane approach, and using the hardness method with (e) von Mises equivalent strain theory and (f) FatemiSocie critical plane approach for shear failure mode materials.
results in 89% of total data and 57% of OP data within scatter bands
of 5, as presented in Fig 4(a). In addition, 70% of all data and 30% of
OP data are within scatter bands of 3.
The SWT critical plane multiaxial model modied based on
MuralidharanManson method as presented by Eq. (10) is also
used here to predict fatigue lives for tensile failure mode materials
and results are presented in Fig. 4(b). Approximately 79% of total
data and 43% of OP data are within scatter bands of 5 and the only
material properties required are ultimate strength and true fracture ductility. About 57% of all data and 26% of OP data are within
scatter bands of 3. Comparing these results with predictions from
Maximum Principal Strain theory, the SWT critical plane model
did not improve the predictions.
De1;max
P2Nf 0:18 Q 2Nf 0:65 ;
2
P 0:00012ru1:664 ; Q 61e0:155
r0:302
f
u
rmax
1
10
11
Fatigue lives were predicted using von Mises criterion, Eqs. (2)
and (4), and estimated fatigue properties using BumelSeeger
method for the shear failure mode materials, as presented in
Fig. 3(c). It should be again mentioned that the modulus of elasticity
is approximated as 200 GPa for steels and super alloys and this value
was used in all life predictions in this study. As can be seen from this
gure, approximately, 73% of total data and 58% of OP data are within scatter bands of 5. About 46% of total data and 25% of OP data are
within scatter bands of 3. To take into consideration the non-proportional cyclic hardening and damage mechanism effects to improve
the predictions, FS critical plane model, Eq. (6), is used and fatigue
properties are estimated from BumelSeeger uniform material
law method (i.e. Eq. (11)), as presented below:
max h
ih
i
Dcmax
r
A2Nf 0:087 B2Nf 0:58 1kC2Nf 0:087
1k n
2
ry
ru
ru
; B 0:885w; C
;
103; 000
1:56ru 570
0:248ru 88
or k 1
k
ru 2Nf 0:087
12
The parameter k here is calculated from Eq. (9). In Eq. (12), yield
strength can be estimated from ry = 1.17ru 426; therefore, ultimate strength is the only required material property in this model.
Using this equation, predicted and experimentally observed fatigue
lives for eleven different shear failure mode materials are compared
in Fig. 3(d). Approximately, 85% of total data and 83% of OP data are
within scatter bands of 5. About 79% of all data and 77% of OP data
are within scatter bands of 3; hence, life predictions for OP loading
are signicantly improved using FS critical plane approach.
For tensile failure mode materials, Maximum Principal Strain criterion, Eq. (1), with fatigue properties estimated from BumelSeeger method (i.e. Eq. (11)) are employed in this study and 89% of total
data and 57% of OP data fall within scatter bands of 5, as can be seen
from Fig. 4(c). Approximately 69% of total data and 35% of OP data
are within the factor of 3 scatter bands. To account for non-proportional cyclic hardening effects, SWT critical plane model is also used
and fatigue properties are estimated from Eq. (11) as follows:
rmax
1
P
De1;max
P2Nf 0:174 Q 2Nf 0:667 ;
2
r2u
89; 000
; Q 0:885ru w
6 0:003;
c 0:58; w 1 for
E
r
ru
u
> 0:003
for
w 1:375 125
E
E
13
67
i
1h
0:32HB2 487HB 191; 000 ;
E
14
Using von Mises criterion, Eqs. (2) and (4), and estimated fatigue properties based on Brinell hardness (i.e. Eq. (14)), 82% of data
for shear failure mode materials are within scatter bands of 5, as
presented in Fig. 3(e). However, only 62% of out-of-phase data fall
within scatter bands of 5. Furthermore, about 62% of all data and
27% of OP data are within scatter bands of 3. Shamsaei and Fatemi
[9] combined FS critical plane multiaxial model and Roessle
Fatemi hardness method as presented with Eq. (15) to predict multiaxial fatigue lives for several steels and reported satisfactory results [9,27].
max h
ih
i
Dcmax
r
A2Nf 0:09 B2Nf 0:56 1 kC2Nf 0:09
1k n
2
ry
2:53HB 293
0:48HB2 731HB 286; 500
; B
;
200; 000
200; 000
1
C
0:0022HB 0:38
15
In
Eq.
(15),
ry = 0.0044(HB)2 + 1.33(HB) [26] and
k = [0.0003(HB) + 0.0585](2Nf)0.09 [9], or k = 1 can simply be used;
therefore, hardness is the only required material property. More
details regarding these equations can be found in [9,27]. In this
study, the modulus of elasticity is approximated as 200 GPa for
steels. Eq. (15) is also employed here to predict fatigue lives for
shear failure mode materials and results are presented in
Fig. 3(f). In this study, k = 1 was used since the effect of k is not signicant for the uniaxial form of the FS equation. However, a more
accurate k value may be necessary when the shear form of the FS
equation is used. As can be seen in Fig. 3(f), 95% of total data and
92% of OP data are within scatter bands of 5. In addition, 88% of total data and 82% of OP data are within scatter bands of 3. Better OP
fatigue life predictions using the FS critical plane approach can be
explained by the fact that this model represents the failure mechanism and takes into consideration the constitutive behavior of
materials including non-proportional cycle hardening effects.
For tensile failure mode materials, Maximum Principal Strain
criterion is used to predict fatigue lives under multiaxial loading.
Comparing the experimental fatigue lives with predicted ones
using Maximum Principal Strain criterion and fatigue properties
estimated from RoessleFatemi hardness method (Eq. (14)), 82%
of total data and 52% of OP data are within scatter bands of 5, as
presented in Fig. 4(e). Fig. 4(e) also shows that 60% of total data
and only 13% of OP data are within scatter bands of 3.
The SWT critical plane model (i.e. Eq. (5)) for tensile failure
mode materials takes into account material constitutive behavior
including mean and residual stresses as well as non-proportional
68
Fig. 4. Predicted fatigue lives versus observed fatigue lives with factor of 3 and 5 scatter bands for predictions made using modied universal slopes with (a) Maximum
Principal Strain theory and (b) SmithWatsonTopper critical plane approach, using uniform material law with (c) Maximum Principal Strain theory and (d) SmithWatson
Topper critical plane approach, and using the hardness method with (e) Maximum Principal Strain theory and (f) SmithWatsonTopper critical plane approach for tensile
failure mode materials.
De1;max
P2N f 0:18 Q2Nf 0:65 ;
2
P 0:0854HB 12:36; Q 0:0013HB2 0:22HB 567 16
rmax
1
69
of 137 HB. This suggests a need for developing another t for materials with hardness below 150 HB.
The life predictions for 1050 (IH) steel with a hardness of 565
HB, the highest strength material used in this study, had the worst
correlation with experimental data as seen in Fig. 3(f). However,
this material showed signicant scatter in experimental fatigue
test results [9]. This is due to the increased sensitivity of high hardness materials to impurities. As a result of the scatter in experimental data, the predictions were not expected to have great
correlation. Although the strainstress-based multiaxial fatigue
models such as FS and SWT may take into consideration the constitutive behaviors of material and load interaction effects [28], they
do not fully reect microstructural properties, and therefore, cannot explain the data scatters observed for different materials. To
account for such scatters in fatigue data, there is a need for developing microstructural sensitive multiaxial fatigue models. However, these microstructural sensitive fatigue models may require
signicantly more material constants [31,32] as compared to fatigue models which are only based on stress and/or strain.
4. Stress response predictions from tensile properties
One of the main advantages of using strainstress-based critical
plane approaches is their capabilities of taking into consideration
the effects of constitutive behavior of materials by the stress term.
However, one of the challenges of using strainstress-based critical
plane approaches such as FS and SWT is then requiring the stress
response of the material under multiaxial loading. Lopez and
Fatemi [33] recently developed a model to predict uniaxial cyclic
deformation of steels (i.e. RambergOsgood type equation)
employing common tensile properties and hardness. Since inphase multiaxial fatigue behavior is usually similar to uniaxial fatigue behavior, the LopezFatemi model can also be used to predict
cyclic deformation under in-phase multiaxial loading. Lopez and
Fatemi developed several equations for predicting the cyclic deformation behavior, but the following, based on ultimate and yield
strengths, provided the best correlation with the experimental
data, and therefore, was used in this study:
0:75ry 82
1:16ru 593
for
n 0:37log
ru =ry 6 1:2
17
0
1
IP
IP n0
DeIP Dr
Dr
0
2
2E
2K
18
It should be noted that a von Mises relationship is used to determine the in-phase and out-of-phase stress and strain amplitudes.
Details on deriving the in-phase strain amplitudes and stress
amplitudes used in Eq. (18) can be found in [9]. Derivation of the
out-of-phase stress and strain amplitudes are also discussed in [9].
In order to predict stress response under nonproportional multiaxial loading, Shamsaei and Fatemi [29] proposed a predictive
70
2 2nn0
nn0
K
De
K
De
3:8
2:2
0
0
2
2
K
K
a 1:6
19
Dr
a OP 1
DrIP
20
Fig. 6. Predicted fatigue lives using SWT method versus experimentally observed
fatigue lives for tensile failure mode materials based on (a) experimental fatigue
properties and (b) fatigue properties estimated using hardness method.
Ds Dr k
2
2 3
21
71
Fig. 7. Predicted fatigue lives using FatemiSocie and hardness methods versus
experimentally observed fatigue lives for shear failure mode materials based on (a)
estimated stress response and (b) measured stress response.
and Fig. 8(b) shows the life predictions based on the measured
stress response. It can be seen in these gures that the predictions
made using the estimated stress response were as good as the predictions made with the measured stress response. These observations can be explained by the fact that maximum normal stress
in both FS and SWT critical plane multiaxial fatigue models is a secondary term. Strain terms in both models are the primary parameters driving the crack and the maximum normal stress is the
secondary parameter opening the crack and accelerating the damage if tensile. Therefore, the effects of normal stress in these multiaxial fatigue parameters are not as signicant as the effects of
strain terms.
5. Conclusions
The following conclusions can be made from the analyses performed in this study:
(1) Acceptable multiaxial fatigue life predictions were obtained
employing FatemiSocie strainstress-based critical plane
approach and fatigue properties estimated from Roessle
Fatemi, MuralidharanManson, or BumelSeeger method.
Therefore, using an appropriate multiaxial fatigue model
seems to be more essential than the selected approximation
technique for uniaxial properties.
72
(2) FatemiSocie critical plane model always provided signicantly better multiaxial fatigue life predictions for shear failure mode materials compared to von Mises criterion no
matter which method was used to predict uniaxial fatigue
properties.
(3) SmithWatsonTopper critical plane approach also
improved fatigue life predictions for tensile failure mode
materials compared to Maximum Principal Strain criterion
when it was combined with RoessleFatemi and Bumel
Seeger methods. However, when fatigue properties were
estimated from MuralidharanManson modied universal
slopes method, Maximum Principal Strain criterion provided
better fatigue life predictions than SmithWatsonTopper
critical plane model.
(4) Fatigue properties estimated from RoessleFatemi hardness
method, when used in the appropriate critical plane multiaxial model (i.e. FatemiSocie and SmithWatsonTopper)
yield better multiaxial fatigue life predictions compared to
other estimation methods for fatigue properties. In addition,
hardness can be measured nondestructively even for in-service components, whereas measuring ultimate strength is a
destructive test requiring a sample specimen.
(5) Stress response under in-phase and 90 out-of-phase loading
may also be approximated only knowing tensile properties.
Life predictions made using approximated stress response
from tensile properties were found to be as good as predictions made with measured stress response.
Acknowledgments
Chrysler Group LLC provided the support for this research
study. Authors would like to express their gratitude to Professor
Ali Fatemi from The University of Toledo, Professor Darrell Socie
from University of Illinois at Urbana-Champaign, and Dr. Yung-Li
Lee from Chrysler Group LLC for very useful discussions on this
subject.
References
[1] Kanazawa K, Miller KJ, Brown MW. Cyclic deformation of 1% Cr-Mo-V steel
under out-of-phase loads. Fatigue Fract Eng Mater Struct 1979;2:21728.
[2] Socie DF, Marquis GB. Multiaxial fatigue. SAE Inc.; 2000.
[3] Smith RN, Watson PP, Topper TH. A stress-strain parameter for fatigue of
metals. J Mater 1970;5:76778.
[4] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage
including out-of-phase loading. Fatigue Fract Eng Mater Struct 1988;11:
14965.
[5] Kim KS, Chen X, Han C, Lee HW. Estimation methods for fatigue properties of
steel under axial and torsional loading. Int J Fatigue 2002;24:78393.
[6] Muralidharan U, Manson SS. Modied universal slopes equation for estimation
of fatigue characteristics. ASME Eng. Mater. Tech. 1988;110:558.