You are on page 1of 14

International Journal of Fatigue 68 (2014) 10–23

Contents lists available at ScienceDirect

International Journal of Fatigue


journal homepage: www.elsevier.com/locate/ijfatigue

Fatigue life prediction for some metallic materials under constant


amplitude multiaxial loading
Jing Li a,b,⇑, Chun-wang Li b, Yan-jiang Qiao c, Zhong-ping Zhang b
a
Harbin Air Force Flight Academy, Nan Tong Street, Harbin 150001, China
b
The Science Institute, Air Force Engineering University, East Chang Le Road, Xi’an 710051, China
c
Air Field Service Technology Research Center of Air Force, Min Hang Road, Beijing 100195, China

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

Article history: Based on the critical plane approach, the Sun-Shang-Bao (SSB) model is analyzed and verified. It is
Received 26 March 2014 discovered that SSB model cannot take the non-proportional cyclic hardening into account and gives
Received in revised form 14 June 2014 non-conservative fatigue life predictions under the non-proportional loading. To solve this problem, a
Accepted 17 June 2014
stress-correlated factor is introduced to describe the degree of the non-proportional cyclic hardening
Available online 26 June 2014
as well as the effect of the non-zero mean stress. The accuracy of the proposed method is systematically
checked against the experimental data found in literature for 16 different materials under constant
Keywords:
amplitude multiaxial loading paths.
Multiaxial fatigue
Critical plane
Ó 2014 Elsevier Ltd. All rights reserved.
Life prediction
Non-proportional cyclic hardening
Effective Poisson’s ratio

1. Introduction (shear stress/strain amplitudes equal on all plane orientations)


gives the worst fatigue performance. Kanazawa et al. [8] found that
Engineering components and structures in service, such as out-of-phase loading results in a shorter fatigue life as compared
axles, crankshafts, and turbine disks and blades, are generally sub- with in-phase loading after conducting a number of axial–torsional
jected to multiaxial fatigue loading, some proportional and others low cycle fatigue (LCF) tests with various phase angles in 1%Cr–
non-proportional [1–3]. During recent decades, fatigue life predic- Mo–V steel. Especially, the 90° out-of-phase loading gives the
tions for such components and structures under non-proportional shortest fatigue life [8]. Jordan et al. [9] found that the relative
loading have become a growing research interest. However, the phase angle between normal and shear strains on the maximum
conventional multiaxial fatigue criteria may lead to non-conserva- shear plane did not affect the fatigue life. Hence, he suggested a life
tive results under non-proportional loading since they are usually correlating parameter using an integral root mean square of posi-
based on proportional fatigue tests. Therefore, accurate fatigue life tive normal strain on the maximum shear plane to replace the nor-
prediction for these components under non-proportional loading mal strain amplitude [9]. However, Wang and Brown [10] found
should be further studied. that the above-mentioned phase angle did have influence on fati-
For non-proportional loading, fatigue behavior under 90° axial– gue life and suggested to take it into account in a life correlating
torsional loading is often investigated since it is considered to be a parameter. Fatemi and Stephen [11] and Itoh et al. [12] also found
critical example. Nishihara et al. [4] showed that the fatigue limits that out-of-phase loading is detrimental for LCF life in terms of
of ductile materials are apparently higher for out-of-phase bending maximum shear strain amplitude. Socie [13] claimed that the
and torsion loading than that for in-phase loading. However, Guo shorter fatigue life observed under non-proportional loading may
[5] showed that for a given applied stress, the maximum shear be ascribed to the additional hardening associated with complex
stress amplitude under out-of-phase loading is lower than that loading paths. In the proposed Fatemi–Socie damage parameter
under in-phase loading. Among all out-of-phase loading paths, [14], the normal stress was introduced to replace the normal strain
Garud [6] and McDiarmid [7] reported that the special loading case amplitude. For LCF life prediction, some improved models based on
the equivalent strain parameter were also proposed by considering
⇑ Corresponding author at: The Science Institute, Air Force Engineering Univer-
the additional hardening and correcting the strain parameter for
sity, East Chang Le Road, Xi’an 710051, China. Tel.: +86 02984786693; fax: +86 non-proportional loading path [12,15–19]. Jahed et al. [20–21] pro-
02984786547. posed a fatigue damage parameter by the energy-based critical
E-mail address: lijing02010303@163.com (J. Li). plane fatigue damage analysis, which successfully correlated the

http://dx.doi.org/10.1016/j.ijfatigue.2014.06.009
0142-1123/Ó 2014 Elsevier Ltd. All rights reserved.
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 11

Nomenclature

Amax circle area with radius of maximum shear strain dur- k empirical constant contained in FS model
ing one cycle K0 cyclic strength coefficient
Aa,max swept area of ca–a polar coordinate space of each L non-proportional hardening coefficient
cycle n0 cyclic strain hardening exponent
ac orientation angle of the maximum shear strain range meff effective Poisson’s ratio
plane me elastic Poisson’s ratio
b fatigue strength exponent u phase shift
c fatigue ductility exponent k strain ratio, Dcapp/Deapp
E modulus of elasticity U factor of non-proportionality
Deapp applied axial strain range Dreq von-Mises equivalent stress range
em mean axial strain rnpro
eq;a non-proportional equivalent stress amplitude
en normal strain excursion rpro proportional equivalent stress amplitude
eq;a
Deeq von-Mises equivalent strain range rn,m normal mean stress
D en normal strain range rn,max maximum normal stress
e0f fatigue ductility coefficient
epro in-phase normal strain amplitude rpro
n;a in-phase normal stress amplitude
n;a
Decr equivalent strain range
r0f fatigue strength coefficient
eq
Dcapp applied shear strain range ry yield strength
cm mean shear strain n, g phase angle
Dcmax maximum shear strain range Dc, De, rn shear strain range, normal strain range and normal
stress on the maximum damage plane, respectively
H hardening factor

fatigue lives of metallic materials under various non-proportional or life as the multiaxial cyclic stresses. However, the equivalent
loading paths. The cyclic energy, for a given component knowing stress criteria are usually limited to high cycle fatigue (HCF) regime
material, geometry and cyclic loading, was calculated by perform- where stresses can be easily estimated by the elastic stress–strain
ing elastic–plastic analysis using cyclic behavior of material [20– relations. The advantage of these models lies in their relative sim-
21]. Shahrooi et al. [22] verified the Jahed’s model for a series of plicity of implementation. It is necessary to point out here that the
non-proportional loading conditions on 1%Cr–Mo–V steel based critical plane-based maximum shear stress and/or maximum nor-
on the nonlinear kinematic hardening model of Chaboche [23] mal stress criteria should not be classified as the equivalent stress
and the multi-surface model of Garud [6]. It is found that a weight- criterion, which can be used as the critical plane parameters. The
ing factor on shear plastic work should be introduced. So, a factor energy-based models, in general, utilize the scalar parameter as a
of 0.5 was used by Shahrooi et al. [22] to reduce the life scatter measure of fatigue damage. The plastic strain energy parameters
band. In fact, from the point of microscopic view the additional have been preferred because of their inherent capability to reflect
cyclic hardening can be attributed to different dislocation struc- the stress–strain path dependence of the fatigue process. A short-
tures between in-phase loading and out-of-phase loading. Doong coming of the energy-based models is in their inability to portray
et al. [24] found that in planar slip materials, single slip occurs the physics of the damage process. Experimental evidence has
under proportional loading, while multi-slip occurs under non- shown that cracks nucleate and grow along shear planes or other
proportional loading. As a result, ladder and planar structures are crystallographic orientations in many polycrystalline metals [14].
found for proportional cycling, while structures such as cell and However, a scalar parameter cannot distinguish among planes on
labyrinths structures are observed for non-proportional cycling which cracks may form. Critical plane approaches are based on
[24]. Different materials show different amounts of non- the physical observations that cracks initiate and grow on specific
proportional cyclic hardening [24]. Materials such as the 300 series planes and that the crack growth is assisted by the stress and/or
stainless steels show a large amount of non-proportional cyclic strain normal to those planes [14,30–35]. According to this
hardening, whereas aluminum alloys usually do not exhibit any approach, the fatigue evaluation is performed on one plane across
additional cyclic hardening. The different additional hardening a critical location in the component. This plane is called the critical
behaviors in different materials may also be related to the different plane, which is usually different for different fatigue models. In
dislocation structures, which are the results of different slip using these parameters, damage is calculated in terms of cyclic
characters in different materials [5,24]. Therefore, fatigue life pre- stresses or strains on each plane within the material to identify
diction should be connected with both loading history and the plane containing the greatest amount of damage or alterna-
material. tively on planes experiencing the maximum level of a predeter-
A significant amount of researches have been devoted over the mined damage parameter, such as cyclic shear strain. In recent
past few decades to gain a better understanding of the mechanisms years, criteria based on the critical plane approach for multiaxial
by which fatigue damage accumulates under multiaxial loading fatigue evaluation are becoming more popular because they gener-
and to develop damage parameters to model the observed behav- ally give more accurate predictions of the fatigue damage, espe-
ior [25–31]. The majority of these models can be broadly classified cially under non-proportional loading [26].
into equivalent stress-based models, energy-based models, and In the present study, the shortcoming of the Sun–Shang–Bao
critical plane approaches [32]. The early development of the equiv- (SSB) model is firstly analyzed, and then a modified multiaxial fati-
alent-stress models are usually based on extensions of static yield gue life prediction model for some metallic materials is proposed
theories to fatigue under combined stresses. Constant amplitude by taking into account the degree of the non-proportional cyclic
multiaxial stresses are transformed into equivalent uniaxial stress hardening as well as the effect of the non-zero mean stress. In
amplitude by the von-Mises or Tresca yield criteria. This equivalent order to verify the fatigue life prediction capability of the SSB
quantity of stress is assumed to produce the same fatigue damage and the modified model, a comparison using the test results of
12 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

16 metallic materials (i.e. 1050N steel, 1050QT steel, 1050IH steel, (a) 1
1045HR steel, 16MnR steel, S45C steel, 45 steel, 6061-T6 alumi- Maximum shear strain on the critical plane
0.75 Normal strain on the critical plane
num alloy, Inconel 718, S460N steel, AISI 304 stainless steel,
A533B steel, 1Cr–18Ni–9Ti stainless steel, Pure Titanium, BT9 tita-
0.5
nium alloy, and AL6XN stainless steel) is carried out. The results
show that the predicted lives by SSB model are satisfactory under 0.25

Strain (%)
proportional loading, while it gives non-conservative fatigue life
predictions under non-proportional loading. In contrast to the 0
SSB model, fatigue life predictions obtained by the modified model
-0.25
are in good agreement with the considered test results under both
proportional and non-proportional loadings. -0.5

-0.75
2. Proposal of fatigue life prediction model and analysis
-1
0 2 4 6 8 10
Kanazawa et al. [8] and Ohkawa et al. [36] reported that, during
ωt (rad)
in-phase fatigue loading, crack initiation and early propagation
occur on a maximum shear plane, and the crack growth plane 1
becomes normal to the maximum principal stress or strain. Under (b) Maximum shear strain on the critical plane
out-of-phase loading, there are two sets of maximum shear planes, 0.75 Normal strain on the critical plane
one is submitted to a larger normal stress/strain than the other
(the same result can also be obtained from Fig. A4) [8,36]. As a con- 0.5
sequence, the cracking may occur on one set of maximum shear
0.25

Strain (%)
planes, on which the normal stress/strain is maximum. It has been
widely accepted that fatigue crack initiation involves localized 0
plastic deformation in persistent slip bands even in the HCF region
[14]. Meanwhile, it has been experimentally observed that the -0.25
directions of these persistent slip bands are very closely aligned
-0.5
with that of maximum shear strain direction [8,37] and fatigue
cracks have always been found to initiate on the maximum shear -0.75
strain plane under different loading conditions. Therefore, for shear
mode materials, it is reasonable to choose the maximum shear -1
0 2 4 6 8 10
strain range plane having the larger normal strain as the critical
plane. ωt (rad)
Using 1045HR steel as an example, Fig. 1a–c illustrate the vary-
ing relationships of cmax and en with respect to xt under in-phase (c) 1
Maximum shear strain on the critical plane
and out-of-phase loadings. It can be seen that the phase angle dif- Normal strain on the critical plane
0.75
ference between the normal strain, en, and the maximum shear
strain, cmax, equals to zero (Fig. 1a) under in-phase loading, while 0.5
during out-of-phase loading this phase angle difference is non-zero
(Fig. 1b and c). Fig. 2a–c shows that the maximum shear strain 0.25
Strain (%)

range, Dcmax, decreases with increasing phase shift, u, (Fig. 2a)


0
whereas the normal strain range, Den, and maximum normal
stress, rn,max, increase with increasing phase shift (Fig. 2b and c). -0.25
The gradual increasing tendency of the normal strain range and
the maximum normal stress are consistent with the phenomenon -0.5
of fatigue life reduction with increasing phase shift, u. How the
ratio of the normal strain range, Den, to the maximum shear strain -0.75
range, Dcmax, varies with phase shift, u, and strain ratio, k were
-1
shown in Fig. 3. This figure indicates that the effect of the normal 0 2 4 6 8 10
strain range on fatigue damage becomes more prominent for ωt (rad)
non-proportional loading, even the ratio between Den and Dcmax
can reach 0.666 for the most severe non-proportional loading, Fig. 1. Variation of maximum shear strain and normal strain on the critical plane
while in the case of proportional loading, this ratio is always less with Deeq = 0.8%, k = 2: (a) in-phase loading; (b) 45° out-of-phase loading; and (c)
than 0.167. The variations in en,a and ca for arbitrary directions 90° out-of-phase loading.

during one cycle are analyzed for a special case (the most severe
non-proportional loading condition): sinusoidal wave form, 90° one plane (a = 0°) corresponding to the greatest normal strain
out-of-phase loading with k = 1 + meff, and their maximum values amplitude. For this special case, Verreman and Guo [38] reported
in each direction are plotted in r–a polar coordinate in Fig. 4. Here, that the cracks are scattered in a large range of orientations since
meff is the effective Poisson’s ratio, a is the angle inclination to the all planes have the same shear strain amplitude. However, a higher
specimen axis, en,a and ca are normal strain amplitude and shear peak is found at 0°, where the normal strain amplitude is maxi-
strain amplitude on the a-plane, respectively. From this figure, it mum. Therefore, for shear mode materials, it is reasonable to use
can be seen that the maximum shear strain range is identical on the maximum shear strain, the normal strain and stress acting on
all planes of the specimens. Fig. 5 shows how the maximum shear the Dcmax plane as the fatigue damage parameters.
strain amplitude and normal strain amplitude vary on different In Ref. [39], Sun et al. combined the maximum shear strain
planes for this special loading case. It can be seen that there is only range, Dcmax, with the normal strain excursion, en , between
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 13

(a) 0.7
1
ϕ = 0°
0.75 0.6 ϕ = 30°
ϕ = 45°
Maximum shear strain (%)

0.5
0.5 ϕ = 60°
0.25 ϕ = 90°

Δ εn/Δ γ max
0.4
0
0.3
-0.25

-0.5 0.2
ϕ = 0°
-0.75 ϕ = 45° 0.1
ϕ = 90°
-1
0 2 4 6 8 10 0
0 2 4 6 8 10
ωt (rad)
Strain ratio, λ

(b) 0.5 Fig. 3. Relation between the ratio Den/Dcmax and the strain ratio, k.

0.25
Normal strain (%)

-0.25
ϕ = 0°
ϕ = 45°
ϕ = 90°
-0.5
0 2 4 6 8 10
ωt (rad) Fig. 4. Normal and shear strain states in 90° out-of-phase loading with k = 1 + meff.

(c) 800 0.5


ϕ = 0°
maximum shear strain amplitude
700 ϕ = 45°
normal strain amplitude
600
ϕ = 90° 0.4
Normal stress (Mpa)

Strain range (%)

500
0.3
400

300
0.2
200

100 0.1

0 Critical Plane
0 2 4 6 8 10
0
ωt (rad) -90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90

Fig. 2. Variation of damage parameters on the critical plane with Deeq = 0.8%, k = 2:
Angle, α
(a) cmax; (b) en; and (c) rn,max.
Fig. 5. Variations of strain amplitude in 90° out-of-phase loading with k = 1 + meff.

adjacent turning points of the maximum shear strain on the critical 1


plane as the basic damage parameters leading to the following en ¼ Den ½1 þ cosðn þ gÞ ð2Þ
2
equivalent damage parameter for the SSB model.
In Eq. (2), the phase angles n and g can be determined by Eqs.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u (A11) and (A12), respectively.
De uv eff ð2  v eff Þ Dc 2
cr
¼t
eq max Sun et al. [39] and Shang et al. [35] considered that the normal
þ en2 ð1Þ
2 ð1 þ v eff Þ2 2 strain excursion en can be used to reflect the effect of non-
proportional cyclic hardening. However, it is not particular useful.
where the normal strain excursion en can be expressed by [39] Fig. 6 clearly shows the dependence of the normal strain excursion
14 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0.5 Decr
eq Deapp ð3  v eff Þ  ð1 þ v eff Þ rn;max ð1  v eff Þ2
¼ þ pro  ð7Þ
Δε = 0.52% 2 2 4 rn;a 4
normal strain excursion (%)

0.4 Besides, for uniaxial loading without mean stress, the maxi-
λ=2 3 mum normal stress, rn,max, reduces to the in-phase normal stress
amplitude, rpro
n;a . Then Eq. (7) can be rewritten as
0.3 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Decr
eq Deapp ð3  v eff Þ  ð1 þ v eff Þ ð1  v eff Þ2 Deapp
¼ þ ¼ ð8Þ
λ= 3 2 2 4 4 2
0.2
Combining Eq. (8) with the Manson-Coffin equation, the multi-
λ= 3 2 axial fatigue damage model may be given by
0.1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
λ =3 3 2 Decr
eq 3  v eff Dcmax 2 rn;max Den 2
¼ þ pro
2 4ð1 þ v eff Þ 2 rn;a 2
0 r0f
0 10 20 30 40 50 60 70 80 90
¼ ð2Nf Þb þ e0f ð2Nf Þc ð9Þ
phase shift (deg) E
where r0f and b are the fatigue strength coefficient and fatigue
Fig. 6. Relation between the phase shift and the normal strain excursion en . strength exponent, respectively. e0f and c are the fatigue ductility
en on phase shift u for axial-torsion loading with constant equiva- coefficient and fatigue ductility exponent, respectively. The four
lent strain. It is seen that en varies little increasing, first, and then uniaxial fatigue constants (i.e. r0f , e0f , b and c) can be determined
decreasing slightly with increasing phase shift between axial p and by the uniaxial fatigue tests. In the absence of the uniaxial fatigue
ffiffiffi tests, r0f , e0f , b, c can be theoretically estimated by the formulas as
torsional loading. Furthermore, under high strain ratios (k > 3),
en is almost unchanging for any given value of phase shift, u, while
 follows
en is changes For steel [43]
pffiffiffilittle with increasing phase shift, u, under low strain
ratios (k > 3). Hence, the normal strain excursion en is not suit- r0f ¼ 4:25HB þ 225 ð10Þ
able as a damage parameter to account for non-proportional cyclic
hardening. It is widely accepted that the maximum normal stress b ¼ 0:09 ð11Þ
can reflect the phenomenon of non-proportional cyclic hardening
[1,14,31,34,40–42]. Therefore, the maximum normal stress is
0:32HB2  487HB þ 191000
introduced to take the non-proportional cyclic hardening into e0f ¼ ð12Þ
E
account in the SSB model, and the following damage parameter
is proposed c ¼ 0:56 ð13Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
Decr
eq 3  v eff Dcmax 2 rn;max Den 2 For aluminum alloy [44]
¼ þ pro ð3Þ
2 4ð1 þ v eff Þ 2 rn;a 2 0
r ¼ 3:66HB þ 370:8 ð14Þ
f

where the maximum shear strain range plane with the larger nor-
1
mal strain is taken as the critical plane. rn,max is the maximum nor- b¼ log½ð3:66HB þ 370:8Þ=ð1:632HB þ 7:047Þ ð15Þ
mal stress and rpro 6
n;a is the in-phase normal stress amplitude at the
same strain amplitude level from a plot of normal stress amplitude
e0f ¼ 0:281 ð16Þ
versus normal strain amplitude. Since in-phase multiaxial fatigue
behavior is usually similar to uniaxial fatigue behavior, the in-phase
c ¼ 0:664 ð17Þ
normal stress amplitude response, rpro n;a , can be related to the in-
phase normal strain amplitude, epro n;a , by the following Ramberg– For titanium alloy [45]
Osgood form equation
r0f ¼ rb þ 355 ð18Þ
pro  pro n0
1
r n;a r n;a  
epro
n;a ¼ þ 0 ð4Þ 1 rb þ 355
E K b¼ log ð19Þ
6 0:446rb
0 0
where K is the cyclic strength coefficient, n is the cyclic strain hard-
ening exponent, E is the modulus of elasticity. epro n;a can be deter- e0f ¼ ef ð20Þ
mined by Eq. (A15) under axial–torsional loading. For bending-
torsion loading, one of the plasticity models such as the Garud c ¼ 0:664 ð21Þ
model [6] should be applied to obtain the in-phase normal stress
amplitude, rpro
n;a .
Under uniaxial loading (k = 0), it can be derived from Eqs. (A17) 3. Validation of fatigue life prediction model
and (A18) that
Sixteen sets of experimental fatigue life data are selected from
Den 1  meff the technical literature [14,42,46–57] for comparison with the
¼ Deapp ð5Þ
2 4 modified SSB model in this section. All the experiments are con-
ducted on smooth specimens at room temperature under strain
Dcmax 1 þ meff
¼ Deapp ð6Þ controlled mode. The monotonic and fatigue properties of these
2 2 investigated materials are summarized in Table 1. It is noted that
Substituting Eqs. (5) and (6) into Eq. (3), then the following the purpose of these comparisons is to validate the model’s gener-
equation is obtained ality to different materials and conditions. The collected data cover
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 15

Table 1
Summery of the investigated materials and their monotonic and fatigue properties.

Materials Ref E (GPa) ry (MPa) rb (MPa) me r0 f (MPa) e0 f b c K0 (MPa) n0 Failure mode


1050N [42] 206 421 709 0.3 1109 0.292 0.10 0.456 1480 0.223 Shear
1050QT [42] 203 1009 1164 0.3 1346 2.01 0.062 0.725 1558 0.123 Shear
1050IH [42] 198 2156 – 0.3 4974 0.529 0.152 0.91 3328 0.083 Shear
1045HR [14,43] 202 382 621 0.29 1027 0.322 0.107 0.487 1258 0.208 Shear
16MnR [44] 212.5 337.9 544.5 0.31 966.4 0.842 0.101 0.618 1106 0.186 Mixed
S45C [45] 186 496 770 0.28 923 0.359 0.099 0.519 1215 0.217 Shear
45 steel [46] 190 370 610 0.3 843 0.327 0.105 0.546 864.3 0.192 Shear
6061T6 [47] 71.5 300 330 0.33 373 0.104 0.0326 0.473 436 0.069 Shear
Inconel 718 [48] 209 1160 1445 0.3 1640 2.67 0.06 0.82 1530 0.07 Shear
S460 N [49] 208.5 500 643 0.3 969.6 0.281 0.086 0.493 1115 0.161 Tensile
AISI 304 [50] 183 325 650 0.3 1043 0.157 0.0876 0.419 1660 0.287 Tensile
A533B [51] 193 432 576 0.3 872.1 1.46 0.084 0.665 827 0.13 Shear
1Cr–18Ni–9Ti [52] 193 310 605 0.3 1124 0.807 0.0905 0.665 1115 0.13 Shear
Pure Ti [53] 112 475 558 0.4 647 0.548 0.033 0.646 668.8 0.0515 Shear
BT9 [53] 118 910 1080 0.37 1180 0.278 0.025 0.665 1232 0.0361 Shear
AL6XN [54] 195 390 – 0.3 760.8 0.565 0.0997 0.482 1104 0.2 Shear

a range of materials used in industries, for example, the automo- 4. Discussions


tive and aerospace. The comparison of the predicted and experi-
mental fatigue lives is summarized in Fig. 7. The comparisons shown in Fig. 7 indicate that the proposed
All the fatigue tests were conducted under combined tension model assesses fatigue lifetime well. Most of the predicted fatigue
and torsion loading except the A533B steel case. For the A533B lives for various materials are within a factor of ±3 when compared
steel, the low-to-intermediate cycle fatigue behavior is investi- with experimental fatigue lifetime data. These results illustrate
gated using solid round bar specimens tested in combined bending that the proposed method is promising.
and torsion [54]. Both in-phase and 90° out-of-phase loading con- Principal directions remain fixed for proportional multiaxial
ditions are applied to produce cases of proportional and non-pro- loading, whereas non-proportional multiaxial loading results in
portional biaxial fatigue [54]. It is necessary to point out here the rotation of principal axes in time [1–3]. The rotation of the
that, for A533B steel, the damage parameters (Den, Dcmax, and principal stress and strain axes during non-proportional loading
rn,max) are taken from Ref. [58]. The experimental data of Inconel often causes additional cyclic hardening. This non-proportional
718 [51], S460N steel [52], and AISI 304 stainless steel [53] include cyclic hardening usually reduces the fatigue lifetime of materials.
in-phase/out-of-phase axial–torsional data with non-zero mean Socie [13] performed in-phase and 90° out-of-phase fatigue tests
stress. For 1050 steel, constant amplitude in-phase and 90° out- with the same shear strain range on AISI 304 stainless steel. It
of-phase axial–torsional fatigue tests are conducted on tubular was found that cyclic stabilized stress–strain hysteresis loops in
specimens with three hardness levels obtained from normalizing the 90° out-of-phase tests had stress ranges twice as large as those
(1050N steel), quenching and tempering (1050QT steel) and induc- of the in-phase tests even though both loading histories had the
tion hardening (1050IH steel). Since uniaxial strain-life fatigue same shear strain range. Socie et al. [1] concluded that the higher
properties (r0 f, e0 f, b and c) are not available, equivalent fatigue magnitude of stress ranges in the out-of-phase test is due to the
properties obtained from the in-phase tests are used to predict effect of an additional strain hardening in the material. In order
the fatigue lives. Additionally, for 1050 steel [42], pure titanium to describe this phenomenon, the hardening factor was defined as
[56], and titanium alloy BT9 [56], the maximum normal stress,
rn,max, is taken from the original paper. For the other considered rnpro
eq;a
H¼ 1 ð22Þ
materials, incremental plasticity theory with Garud’s hardening rpro
eq;a
rule [6,59–60] was used to compute the maximum normal stress,
rn,max. The reliability of the incremental plasticity model with Gar- where rnpro
eq;a is non-proportional equivalent stress amplitude and
ud’s hardening rule was evaluated by using the experimental rpro
eq;a is proportional equivalent stress amplitude at the same strain
results of 1045HR steel [46] during proportional and non- amplitude level.
proportional loadings. The loading paths were illustrated in In Ref. [12], Itoh et al. performed various strain paths using type
Fig. 8. The detailed data were listed in Table 2. In this table, Dr0x 304 stainless steel hollow cylinder specimens at ambient temper-
and Ds0xy are the stress ranges calculated by the incremental plas- ature to show the path dependence of the non-proportional cyclic
ticity model. Drx and Dsxy are the experimental values of stress hardening. Experimentally measured stress–strain histories for
range [46]. The relative deviations of the calculated principal Path I through Path IV (see Fig. 9) have been tabulated in Ref. [1].
and shear stress ranges with the test ones are defined as Each of the strain paths has the same maximum applied shear
dDr0x ¼ ðDr0x  Drx Þ=Drx , dDs0xy ¼ ðDs0xy  Dsxy Þ=Dsxy , respectively. and normal strain ranges, which equal to 1.39% and 0.8%, respec-
It can be seen from Table 2 that the maximum error is 7.7% and tively [1,12]. Path I is proportional, and Path II through Path IV
12.3% for the principal stress range and the shear stress range, have increasing degrees of non-proportionality [1,12]. In more
respectively. Most errors are within 10%. Experimental verification details, the non-proportional factors of Path I through Path IV are
shows that the incremental plasticity model with Garud’s hardening 0, 0.1, 0.2, and 0.77, respectively [12]. The variation of maximum
rule is reliable to calculate the stress of the thin tubular specimen normal stress with respect to the hardening factor, H, is shown
under axial–torsional loading. It can be observed from Fig. 7a that in Fig. 10. It can be seen from Fig. 10 that during non-proportional
the predicted lifetime by the proposed model under uniaxial and straining the magnitude of the normal stress increases with
proportional loading is satisfactory. The data points fall mainly increasing the hardening factor, H. This tendency means that the
within an error factor of 2. Under the non-proportional loading proposed parameter, via its normal stress term to account for the
(see Fig. 7b), the predicted lifetime is somewhat more scattered. effect of the additional hardening caused by out-of-phase tests, is
However, most of the data points fall within an error factor of 3. able to account for this effect.
16 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

0.4 0.5
A B
0.2 0.25

0 0

-0.2 -0.25

-0.4 -0.5
-0.4 -0.2 0 0.2 0.4 -0.5 -0.25 0 0.25 0.5

0.44 0.6
C D
0.22 0.3

0 0

-0.22 -0.3

-0.44 -0.6
-0.44 -0.22 0 0.22 0.44 -0.6 -0.3 0 0.3 0.6

Fig. 8. The loading paths for validation of Garud’s hardening rule.

Table 2
Comparison between the incremental plasticity model with Garud’s hardening rule
and the test results of stress for thin tubular specimen.

Paths Deapp (%) Dcapp (%) Dr0x ðMPaÞ jdDr0x j (%) Ds0xy ðMPaÞ jdDs0xy j (%)
Drx ðMPaÞ Dsxy ðMPaÞ

A 0.29 0.62 356/358 0.6 273.6/262 5.5


B 0.83 0.444 675.4/680 0.7 127.6/120 6.3
C 0.82 0.426 783.8/728 7.7 334.6/298 12.3
D 0.528 1.13 722.4/690 4.7 442.8/408 8.5

Fig. 7. The accuracy of proposed model in assessing the fatigue lives of the
considered materials [14,42,46–57]: (a) uniaxial and proportional loading; (b) non-
proportional loading; and (c) with non-zero mean stress.

As mentioned previously, Fatemi–Socie [14] proposed to utilize Fig. 9. Proportional and non-proportional loading histories [12].

the maximum normal stress, rn,max, to account for the effects of between the experimental data under fully reversed tension–
additional cyclic hardening instead of the normal strain term in compression and those under cyclic torsion if k sets as a constant
Brown-Miller [33,61] parameter. The Fatemi–Socie parameter
[62]. Subsequently, life predictions using FS model will be scattered
(FS) can be expressed as [14] since predictions are sensitive to the value of k [42]. In Ref. [42], a
  simplified expression for k was obtained using the von-Mises
Dcmax rn;max
1þk ¼ constant ð23Þ criterion, G = E/2(1 + me), and me = 0.3, i.e.
2 ry
where ry is the yield strength, and k is a constant found by fitting 0:31ry
k¼ ð24Þ
the uniaxial data against the pure torsion data. In more details, r0f ð2Nf Þb
the constant k is determined iteratively until the FS–Nf curves under
fully reversed tension–compression and cyclic torsion are con- From Eq. (24), Shamsaei et al. [42] obtained that k increases as
verged to a single curve [62]. However, for some materials such fatigue life increases, especially in the long-life regime and for
as extruded AZ61A magnesium alloy, divergence can be found higher-hardness materials. This is may be one of the reasons why
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 17

600 Therefore, the proposed uniform material law is mainly applicable


Path I to ductility materials such as the steel, aluminum, titanium, super
Path II alloy, and so on.
Maximum normal stress (MPa)

550
Path III Under multiaxial fatigue loading, mean stress has a substantial
Path IV effect on fatigue life. Mean stress effects are included in fatigue
500 damage parameters in different ways [1]. One approach, applied
earlier by Fatemi and Socie [14], incorporates mean stress using
450
(a) 8
λ= 0
400 7 λ = 1.0
λ = 2.0
6
λ = 3.0
350
5

σn,m/σn,apro
300
0 0.1 0.2 0.3 0.4 0.5 4

Hardening factor, H 3

Fig. 10. Relation between the maximum normal stress, rn,max, and the hardening
2
factor, H.

1
the FS model gives non-conservative life predictions (see Ref. [42])
0
in the long-life regime even by up to a factor of 10. Additionally, if
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
test data are not available, Shamsaei et al. [42] suggested that one
may consider k = 1 when using the axial form of the FS equation. R=εx,min/εx,max=γ xy,min/γ xy,max
In Ref. [33], Brown and Miller argue that the normal strain influ-
ences fatigue ductility which in turn is related to the fatigue (b) 9
strength and then conclude that the normal strain across the max- λ = 1.0
8
imum shear strain plane assists in crack propagation. That is to say, λ = 2.0
the normal strain term is still an important damage parameter in 7 λ = 3.0
multiaxial fatigue life prediction. In the FS model, only two damage
parameters (i.e. rn,max, and Dcmax) may be insufficient for certain 6
σn,m/σn,apro

materials and loading conditions. Take 30CrNiMo8HH steel for 5


example, Noban et al. [63] observed that the additional hardening
in 90° out-of-phase loading is not significant, while fatigue lives 4
were considerable shorter under this loading path, as compared
3
to in-phase loading. Therefore, the drop in fatigue life of 30CrNiM-
o8HH cannot be attributed to additional hardening, but rather 2
related to a path-dependent non-proportional sensitivity factor
[63]. Fortunately, the normal strain term can be used to describe 1
the non-proportionality of loading path [64]. In fact, from the point
0
of microscopic view, fatigue crack growth is a de-cohesion process -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
along crack tip shear bands, both normal strain and stress on the
crack plane makes this de-cohesion behavior accelerate. Hence, R=εx,min/εx,max=γ xy,min/γ xy,max
the normal strain term should also be considered in the fatigue
(c) 8
damage parameter. As can be seen from Eq. (23), only the normal
λ = 1.0
stress term is considered as the secondary factor. This may be
the other reason why the FS model gives scattered life predictions
7 λ = 2.0
[42,62,65]. In contrast to the FS model, the basis experiments (i.e. λ = 3.0
6
the uniaxial and torsional tests) are unnecessary to determine
the material constant in the proposed model. Besides, both the nor- 5
σn,m/σn,apro

mal stress and strain terms are considered to describe the fatigue
damage. Hence, the proposed model may be more suitable to pre- 4
dict the fatigue lives of metallic materials.
The failure mode of all the materials considered in the present 3
study is listed in Table 1. It can be seen from this table, all the con-
sidered materials are failed with shear failure mode except AISI 2
304 stainless steel, S460N steel and 16MnR steel. AISI 304 stainless
1
steel and S460N steel display tensile failure mode, while 16MnR
steel displays mixed crack behavior. Based on the failure mode of
0
the considered materials, it can be obtained that the proposed -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
model is intended for cracks in shear failure mode materials.
R=εx,min/εx,max=γ xy,min/γ xy,max
Besides, Sonsino [66] pointed out that, for low-ductility materials,
damage is activated by normal stresses or strains and, for Fig. 11. Sensitivity of rn;m =rpro
n;a to mean strain: (a) u = 0°; (b) u = 45°; and (c)
ductile materials, damage is activated by shear stresses or strains. u = 90°.
18 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

Fig. 12. Some specific load paths designed to challenge FS model [69].

0.45 7
(a) 10
Δ εeq /2= 0.85%, λ = 1.732 Uniaxial
6 Proportional
10
Effectvie Poisson’s Ratio

0.44 3 scatter band


Predicted life (cycles)

5
10

4
0.43 10

3
10
0.42
2
10

1
0.41 10
0 10 20 30 40 50 60 70 80 90 1 2 3 4 5 6 7
10 10 10 10 10 10 10
Phase Shift (deg) Experimental life (cycles)
Fig. 13. Correlation between the ratio effective Poisson’s ratio, meff, and phase 7
shift, u. (b) 10
the maximum normal stress to modify the damage parameter. 6
Similar to the Fatemi–Socie damage parameter, the proposed mul- 10
tiaxial fatigue damage parameter can easily account for mean
Predicted life (cycles)

5
stress effects through the normal stress term. This is justified on 10
the basis that only the mean tensile stresses normal to the crack
will contribute to damage and the mean stresses parallel to the 4
10
crack will have minimal influence on the fatigue crack growth rate
or closure [14,67]. Thus, in the presence of mean stresses, the nor-
3
mal stress term contained in Eq. (3) will be composed of the alter- 10
nating normal stress and the mean or static normal stress. So Eq. Non-proportional
(3) can be rewritten as 2
10 Non-zero mean stress
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   
De cr
eq 3  v eff Dcmax 2 rn;a þ rn;m Den 2 3 scatter band
¼ þ ð25Þ 1
2 4ð1 þ v eff Þ 2 rpro
n;a 2 10 1 2 3 4 5 6 7
10 10 10 10 10 10 10
where rn,m is the normal mean stress on the critical plane. Experimental life (cycles)
In order to take into account the contribution of the mean stress
and strain to the overall fatigue damage, it is initially useful to Fig. 14. Fatigue life predictions of studied materials by SSB model: (a) uniaxial and
proportional loading; and (b) non-proportional loading and non-zero mean stress.
rewrite rn;max =rpro
n;a as
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 19

rn;max rn;a rn;m


¼ pro þ pro ð26Þ
rpro
n;a rn;a rn;a
Using 1045HR steel as an example, Fig. 11 shows rn;m =rpro n;a vary-
ing with the strain ratio, R (= ex,min/ex,max = cxy,min/cxy,max), when in-
phase and out-of-phase axial–torsional loadings with mean strains
are considered. It can be seen from Fig. 11a–c that for either in-
phase or out-of-phase loading rn;m =rpro n;a increases as the strain ratio
R increases. This simply means that the proposed model is in
agreement with the experimental evidence that fatigue damage
increases by increasing the magnitude of the superimposed tensile
static stresses [68].
It is necessary to point out here that, for the loading case, i.e.,
the cyclic shear with a superimposed mean tensile normal stress
(See Path E in Fig. 12), Eq. (3) is incapable since the second term
in Eq. (3) which corrects for mean stress will be zero for this case
Fig. 15. Fatigue life predictions of studied materials by the proposed model
as the normal strain amplitude term is zero. Under such cases, [14,42,46–57].
the critical plane was redefined as the maximum damage plane
on which the largest fatigue damage parameter (Eq. (3)) was
obtained. Then Eq. (3) can be rewritten as
Poisson’s ratio, meff, and phase shift, u, is drawn in Fig. 13. This dia-
2sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 gram is in conformity with the above analyses. In other words, the
 2  
Decr
eq 3  v eff Dc rn De 2 5 proposed damage parameter can consider the common elastic–
¼4 þ pro ð27Þ
2 4ð1 þ v eff Þ 2 rn;a 2 plastic states of materials through employing the effective Pois-
max
son’s ratio, meff.
where Dc, De, and rn are the shear strain range, the normal strain In addition, the SSB model was checked by using the fatigue
range and the normal stress on the maximum damage plane, data of the aforementioned materials, for comparison purposes
respectively. Using Inconel 718 as an example [1], validation of (see Fig. 14). It can be seen from Fig. 14a, under uniaxial and pro-
Eq. (27) using the limited data shows that the predicted life is portional loadings, the predicted lifetime by the SSB model are sat-
16,769, which is closed to the experimental one (equals to isfactory. However, under non-proportional loading and/or loading
10,300). Although this predicted result is encouraging, for this spe- conditions with non-zero mean stress the predicted lifetime by SSB
cial loading condition, more test data are still needed to verify the model is non-conservative (see Fig. 14b). The predicted results of
reliability of the proposed model further. Fig. 14b clearly show that the use of the normal strain excursion
In fact, the interaction of different stress and strain parameters en is incapable of correctly accounting for the presence of the
(including Path E in Fig. 12) on the critical plane and their effect on non-zero mean stress as well as the additional cyclic hardening
fatigue damage are not yet well understood. Take the load paths caused by the non-zero out-of-phase angles. In order to clearly
presented in Fig. 12 for example, Shamsaei [69] reported that all evaluate the capability of the proposed life prediction model, all
these load paths have the same shear strain range and maximum of the predicted fatigue lives are plotted vs. the experimental lives
normal stress on the maximum shear strain range plane. However, in Fig. 15. Considering all the data of the investigated materials in
shear strain and normal stress components for each load path Fig. 15, 84.9% and 96.6% of the data fall within scatter of 2 and 3,
interact differently [69]. In Fig. 12, Path E which tensile normal respectively. For the SSB model, 72.9% and 82.4% of the data fall
stress is constantly applied may have larger effect on fatigue dam- within scatter of 2 and 3, respectively. However, during non-pro-
age than Path F which normal stress is pulsating only once at the portional loading, over 85% of data points predicted by the SSB
zero shear strain [69]. However, the FS model cannot differentiate model fall within the non-conservative side, even by up to a factor
these load paths. Using 1050QT steel as an example, Shamsaei [69] of 11. Therefore, the proposed model for multiaxial fatigue life pre-
reported that the conservative life prediction was obtained by the diction of metallic materials improves prediction accuracy to the
FS model for Path E, whereas the non-conservative life prediction SSB model.
was obtained by the FS model for Path F. To give better fatigue life
prediction, the interaction of different stress and strain compo- 5. Conclusions
nents on the critical plane should be studied further in multiaxial
fatigue analysis.  A critical plane-based fatigue life prediction model is proposed,
As mentioned earlier, the increased hardening under non-pro- i.e.
portional loading results from the change of dislocation substruc- Decr
eq r0f
¼ ð2Nf Þb þ e0f ð2Nf Þc
tures from single slip to multi-slip structures. Doong et al. [24] 2 E
reported that formation of the multi-slip structures results from
with
the change of slip direction during the cyclic loading. Ideally, the sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
stress history under elastic deformation can be viewed as a record    ffi
Decr
eq 3  v eff Dcmax 2 rn;max Den 2
for the change of atomic spacing [24]. On the other hand, the plas- ¼ þ pro
2 4ð1 þ v eff Þ 2 rn;a 2
tic strain of metals results from the dislocation slip. Each macro-
scopic plastic increment corresponds to certain microscopic  The maximum normal stress can correctly account for the pres-
dislocation movements [24]. In comparison with proportional ence of non-zero mean stress as well as the non-proportional
loading, the microscopic dislocation movements decrease under cyclic hardening.
non-proportional loading, which is in agreement with the experi-  Under the same equivalent strain loading, the value of effective
mental observation [24,70]. Therefore, under the same equivalent Poisson’s ratio, meff, decreases with increasing the non-propor-
strain loading, the value of effective Poisson’s ratio, meff, should tional cyclic hardening. The proposed damage parameter can
decrease with increasing the non-proportional cyclic hardening. consider the common elastic–plastic states of materials through
Take 1045HR steel for example, correlation between effective employing the effective Poisson’s ratio, meff.
20 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

2 1
3
 The SSB model can give satisfactory fatigue life predictions ex c
2 xy
0
under uniaxial and proportional loadings, while it is incapable 61 7
eij ¼ 4 2 cxy meff ex 0 5 ðA1Þ
of accounting for the presence of the non-zero mean stress as
0 0 meff ex
well as the non-proportional cyclic hardening. The predicted
lives by the SSB model are non-conservative under non-propor- If the applied axial and shear strains of the thin-walled tubular
tional loading. Experimental validations using 16 different specimen are given with the sine wave, i.e.
materials show that the proposed model is reliable for assessing
the fatigue life of materials under both proportional and non-
Deapp
ex ¼ sin xt þ em ðA2Þ
proportional loading. 2

Dcapp
cxy ¼ sinðxt  uÞ þ cm ðA3Þ
2
Acknowledgement
where u is the phase angle between the axial strain and shear
The authors gratefully acknowledge the financial support of strain. Deapp and Dcapp are the applied axial and shear strain ranges,
Innovation Funds of Air Force University of Engineering respectively. em and cm are the mean axial and shear strains, respec-
(Dx2010403). tively. meff represents the effective Poisson’s ratio, which can be
determined by [71]
Appendix A. Strain varying behavior under axial–torsional ð0:5  me ÞDreq
loading meff ¼ 0:5  ðA4Þ
EDeeq

If the X-axis is conveniently chosen to be parallel to the loading where me is the elastic Poisson’s ratio. E is the elastic modulus. Dreq
and the Z-axis coincides with the outside normal of the surface, the and Deeq are the von-Mises equivalent stress and strain ranges,
strain tensor for thin-walled tubular specimen under strain-con- respectively. Fig. A1 shows the step-by-step procedure required to
trolled loading can be given by calculate the effective Poisson’s ratio. In this figure, K0 and n0 are

Fig. A1. Procedure for determining the effective Poisson’s ratio meff.
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 21

the cyclic strength coefficient and the cyclic strain hardening expo- 1
nent, respectively. L is the non-proportional hardening coefficient, Δεapp = 1.2%, λ = 0.5, ϕ = 90°
which is defined as the ratio of the equivalent stress under 90°
out-of-phases loading to the equivalent stress under in-phase load-
ing [40–41]. This coefficient may be defined at any strain level as 0.5
γα

Strain Amplitude (%)


follows [40–41]


rOP
1 ðA5Þ ε n ,α
rIP 0

where rOP is the 90° out-of-phase equivalent stress and rIP is the in- ξ+η
phase equivalent stress at the same strain level. The non-propor-
tionality factor, U, is determined by the following formulation [72] -0.5
A
U¼2 a 1 ðA6Þ
Amax
where Amax is the area of the circle with radius equal to the maxi- -1
0 90 180 270 360
mum shear strain during one cycle. Aa is the swept area of ca –a
polar coordinate (shadow area in Fig. A2) space. ωt (deg)
The normal and shear strains on the plane which make an angle
Fig. A3. Schematic showing of the phase angle n + g.
a with the thin-walled tubular specimen axis are expressed as
1  meff 1 þ meff 1
en;a ¼ ex þ ex cos 2a þ cxy sin 2a ðA7Þ
2 2 2 k cos 2a sin u
tan g ¼ ðA12Þ
k cos 2a cos u  ð1 þ meff Þ sin 2a
ca ¼ ð1 þ meff Þex sin 2a þ cxy cos 2a ðA8Þ

It can be derived from Eqs. (A2)–(A8) that Dcapp


k¼ ðA13Þ
Deapp
1
ca ðtÞ ¼ Deapp
2 The phase angle between en,a and ca is (n + g) (schematic show-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 ffi
2 ing in Fig. A3), ranging between –90° and 90°. From Eqs. (A9) and
 k cos 2a cos u  ð1 þ meff Þ sin 2a þ ½k cos 2a sin u
(A10), the Dca and Den,a can be obtained as
1  meff 1 þ meff
 sinðxt þ gÞ þ em þ em cos 2a Dca ¼ Deapp
2 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1  2 2
þ cm sin 2a  k cos 2a cos u  ð1 þ meff Þ sin 2a þ ½k cos 2a sin u
2
ðA14Þ
ðA9Þ
1
1 Den;a ¼ Deapp
en;a ðtÞ ¼ Deapp 2
4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2 2
 2 2  2ð1 þ meff Þ cos2 a  2meff þ k sin 2a cos u þ ½k sin 2a sin u
 2ð1 þ meff Þ cos2 a  2meff þ k sin 2a cos u þ ½k sin 2a sin u
ðA15Þ
 sinðxt  nÞ  ð1 þ meff Þem sin 2a þ cm cos 2a
ðA10Þ Eq. (A14) is differentiated with respected to a which gives the
maximum or minimum value of angle ac for the shear strain range.
where
2kð1 þ meff Þ cos u
k sin 2a sin u tan 4ac ¼ ðA16Þ
tan n ¼ ðA11Þ ð1 þ meff Þ2  k2
ð1 þ meff Þ cos 2a þ ð1  meff Þ þ k sin 2a cos u
The angles determined by Eq. (A16) are substituted into Eq.
(A15), the one giving the maximum value for the normal strain is
taken as the critical plane angle. Then the maximum shear strain
range plane is determined. Using 1045HR steel as an example,
the variation of maximum shear strain range and normal strain
range with respect to the angle, a, are illustrated in Fig. A4. It
can be seen from this figure that the above method can determine
a fixed critical plane. The maximum shear strain range and the nor-
mal strain range on the critical plane are given by
Dcmax ¼ Deapp
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 ffi
2
 k cos 2ac cos u  ð1 þ meff Þ sin 2ac þ ½k cos 2ac sin u
ðA17Þ

1
Den ¼ Deapp
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 2
 2ð1 þ meff Þ cos2 ac  2meff þ k sin 2ac cos u þ ½k sin 2ac sin u
ðA18Þ
Fig. A2. Illustration of Aa in ca–a polar coordinate space.
22 J. Li et al. / International Journal of Fatigue 68 (2014) 10–23

1.5 [24] Doong SH, Socie DF, Robertson IM. Dislocation substructure and non-
proportional hardening. ASEM J Eng Mater Technol 1990;112:456–64.
[25] Garud YS. Multiaxial fatigue: a survey of the state of the art. J Test Eval
1981;9:165–78.
1.2 [26] You BR, Lee SB. A critical review on multiaxial fatigue assessments of metals.
Int J Fatigue 1996;18:235–44.
Strain Amplitude (%)

[27] Papadopoulos IV, Davoli P, Gorla C, Filippini M, Bernasconi A. A comparative


study of multiaxial high-cycle fatigue criteria for metals. Int J Fatigue
0.9 1997;19:219–35.
[28] Macha E, Sonsino CM. Energy criteria of multiaxial fatigue failure. Fatigue Fract
Eng Mater Struct 1999;22:1053–70.
[29] Wang YY, Yao WX. Evaluation and comparison of several multiaxial fatigue
0.6 criteria. Int J Fatigue 2004;26:17–25.
[30] Karolczuk A, Macha E. A review of critical plane orientations in multiaxial
fatigue failure criteria of metallic materials. Int J Fracture 2005;134:
267–304.
0.3 [31] Socie DF. Critical plane approaches for multiaxial fatigue damage assessment.
Advances in multiaxial fatigue, ASTM STP, 1191, ASTM, Philadelphia; 1993, p.
Critical Plane 7–36.
[32] Erickson M, Kallmeyer AR, Van Stone RJ, Kurath P. Development of a multiaxial
0 fatigue damage model for high strength alloys using a critical plane
-90 -60 -30 0 30 60 90
methodology. ASME J Eng Mater Technol 2008;130:041008.
Angle, α (deg) [33] Brown MW, Miller KJ. A theory for fatigue failure under multiaxial stress-
strain conditions. Proc Inst Mech Eng 1973;187(65):745–55.
Fig. A4. Variations of strain range with the angle a. [34] Li J, Zhang ZP, Sun Q, Li CW, Qiao YJ. A new multiaxial fatigue damage model
for various metallic materials under the combination of tension and torsion
loadings. Int J Fatigue 2009;31:776–81.
[35] Shang DG, Wang DJ. A new multiaxial fatigue damage model based on the
critical plane approach. Int J Fatigue 1998;20:241–5.
References [36] Ohkawa I, Takahashi H, Moriwaki M, Misumi M. A study on fatigue crack
growth under out-of-phase combined loadings. Fatigue Fract Eng Mater Struct
[1] Socie DF, Marquis GB. Multiaxial fatigue. Warrendale, PA: Society of 1997;20:929–40.
Automotive Engineers Inc.; 2000. [37] Tairs S, Inoue T, Yoshida T. Low cycle fatigue under multiaxial stresses (in the
[2] Stephens RI, Fatemi A, Stephens RR, Fuch HO. Mental fatigue in engineering. case of combined cyclic tension-compression and cyclic torsion) at room
2nd ed. John Wiley & Sons Inc.; 2000. temperature. In: Proc 12th Japan cong test mater, vol. 2; 1969. p. 50–5.
[3] Schijve J. Fatigue of structures and materials. 2nd ed. Berlin: Springer; 2009. [38] Verreman Y, Guo H. High-cycle fatigue mechanisms in 1045 steel under non-
[4] Nishihara T, Kawamoto M. The strength of metals under combined alternating proportional axial-torsional loading. Fatigue Fract Eng Mater Struct
bending and torsion with phase difference. Memoirs, College of Engineering, 2007;30:932–46.
11, Kyoto Imperial University; 1945. p. 85–112. [39] Sun GQ, Shang DG, Bao M. Multiaxial fatigue damage parameter and life
[5] Guo H. Biaxial fatigue mechanisms and life predictions for 1045 steel and 7075 prediction under low cycle loading for GH4169 alloy and other structural
aluminum alloy including anisotropy and notch effects. Ph.D. Dissertation, materials. Int J Fatigue 2010;32:1108–15.
Ecole Polytechnique de Montreal, Quebec, Canada; 2003. [40] Shamsaei N, Fatemi A. Effect of microstructure and hardness on non-
[6] Garud YS. A new approach to evaluation of fatigue under multiaxial loadings. proportional cyclic hardening coefficient and predictions. Mater Sci Eng A
ASME J Eng Mater Technol 1981;103:118–25. 2010;A 527:3015–24.
[7] McDiarmid DL. Fatigue under out-of-phase bending and torsion. Fatigue Fract [41] Shamsaei N, Fatemi A, Socie DF. Multiaxial cyclic deformation and non-
Eng Mater Struct 1987;9:457–75. proportional hardening employing discriminating load paths. Int J Plast
[8] Kanazawa K, Miller KJ, Brown MW. Low cycle fatigue under out-of-phase 2010;26:1680–701.
loading conditions. ASME J Eng Mater Technol 1977;99:222–8. [42] Shamsael N, Fatemi A. Effect of hardness on multiaxial fatigue behaviour and
[9] Jordan EH, Brown MW, Miller KJ. Fatigue under severe non-proportional some simple approximations for steels. Fatigue Fract Eng Mater Struct
loading. Multiaxial Fatigue, ASTM STP 853, Miller KJ, Brown MW, eds., ASEM, 2009;32:631–46.
Philadelphia; 1985. p. 569–85. [43] Roessle ML, Fatemi A. A strain-controlled fatigue properties of steels and some
[10] Wang CH, Brown MW. A path-independent parameter for fatigue under simple approximations. Int J Fatigue 2000;22:495–511.
proportional and non-proportional loading. Fatigue Fract Eng Mater Struct [44] Li J, Zhang ZP, Sun Q, Li CW, Li RS. A modified method to estimate fatigue
1993;16:1285–97. parameters of wrought aluminum alloys. J Mater Eng Perform
[11] Fatemi A, Stephens RI. Biaxial fatigue of 1045 steel under in-phase and out-of- 2011;20:1323–9.
phase loading. Multiaxial fatigue: analysis and experimental, AE-14, SAE; [45] Lee KS, Song JH. Estimation methods for strain-life fatigue properties from
1987. p. 121–37. hardness. Int J Fatigue 2006;28:386–400.
[12] Itoh T, Sakane M, Ohanami M, Socie DF. Non-proportional low cycle fatigue [46] Fatemi A. Fatigue and deformation under proportional and non-proportional
criterion for type 304 stainless steel. ASME J Eng Mater Technol biaxial loading. Ph.D. Thesis, University of Iowa; 1985.
1995;117:285–92. [47] Gao Z, Zhao T, Wang X, Jiang Y. Multiaxial fatigue of 16MnR Steel. ASME J Eng
[13] Socie DF. Multiaxial fatigue damage models. ASME J Eng Mater Technol Mater Technol 2009;131:021403.
1987;109:293–8. [48] Kim KS, Park JC, Lee JW. Multiaxial fatigue under variable amplitude loads.
[14] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage ASEM J Eng Mater Technol 1999;121:286–93.
including out of phase loading. Fatigue Fract Eng Mater Struct [49] Shang DG, Sun GQ, Deng J, Yan CL. Multiaxial fatigue damage parameter and
1988;11:149–65. life prediction for medium-carbon steel based on the critical plane approach.
[15] Itoh T, Sakane M, Hata T, Hamada N. A design procedure for assessing low Int J Fatigue 2007;29:2200–7.
cycle fatigue life under proportional and non-proportional loading. Int J [50] Lin H, Nayed-Hashemi H, Pelloux RM. Constitutive relations and fatigue life
Fatigue 2006;28:459–66. prediction for anisotropic Al-6061-T6 rods under biaxial proportional loadings.
[16] Itoh T, Yang T. Material dependence of multiaxial low cycle fatigue lives under Int J Fatigue 1992;14(4):249–59.
non-proportional loading. Int J Fatigue 2011;33:1025–31. [51] Socie DF, Waill LA, Dittmer DF. Biaxial fatigue of Inconel 718 including mean
[17] Li B, Reis L, de Freitas M. Simulation of cyclic stress/strain evolutions for stress effects. In: Miller KJ, Brown MW, editors. Multiaxial fatigue. ASTM STP
multiaxial fatigue life prediction. Int J Fatigue 2006;28:451–8. 853. Philadelphia, PA: American Society for Testing and Materials; 1985. p.
[18] Gladskyi M, Sergiy S. A new model for low cycle fatigue of metal alloys under 463–81.
non-proportional loading. Int J Fatigue 2010;32:1568–72. [52] Jiang YY, Hertel O, Vormwald M. An experimental evaluation of three critical
[19] Borodii MV, Adamchuk MP. Life assessment for metallic materials with the plane multiaxial fatigue criteria. Int J Fatigue 2007;29:1490–502.
used of the strain criterion for low cycle fatigue. Int J Fatigue [53] Chiou YC, Yip MC. An energy-based damage parameter for the life prediction of
2009;31:1579–87. AISI 304 stainless steel subjected to mean strain. J Chinese Inst Eng
[20] Jahed H, Varvani-Farahani A. Upper and lower fatigue life limits model using 2006;29(3):507–17.
energy-based fatigue properties. Int J Fatigue 2006;28:467–73. [54] Nelson DV, Rostami A. Biaxial fatigue of A533B pressure vessel steel. J Pressure
[21] Jahed H, Varvani-Farahani A, Noban M, Khalaji I. An energy-based fatigue life Vessel Technol 1997;119:325–31.
assessment model for various metallic materials under proportional and non- [55] Chen X, An K, Kim KS. Low-cycle fatigue of 1Cr–18Ni–9Ti stainless Steel and
proportional loading conditions. Int J Fatigue 2007;29:647–55. related weld metal under axial, torsional and 90° out-of-phase loading. Fatigue
[22] Shahrooi S, Metselaar IH, Huda Z. Evaluating a strain energy fatigue method Fract Eng Mater Struct 2004;27:439–48.
using cyclic plasticity models. Fatigue Fract Eng Mater Struct 2010;33:530–7. [56] Shamsaei N, Gladskyi M, Panasovskyi K, Shukaev S, Fatemi A. Multiaxial
[23] Chaboche JL. On some modifications of kinematic hardening to improve the fatigue of titanium including step loading and load path alteration and
description of ratcheting effects. Int J Plast 1991;7:661–78. sequence effects. Int J Fatigue 2010;32:1862–74.
J. Li et al. / International Journal of Fatigue 68 (2014) 10–23 23

[57] Kalnaus S, Jiang YY. Fatigue of AL6XN stainless steel. ASME J Eng Mater [65] Li J, Zhang ZP, Sun Q, Li CW. Multiaxial fatigue life prediction for various
Technol 2008;130:031013. metallic materials based on the critical plane approach. Int J Fatigue
[58] Park J. Life prediction for mechanical components experiencing multiaxial 2011;33:90–101.
fatigue, including effects of stress concentrations. Ph.D, Dissertation, Stanford [66] Sonsino CM. Influence of material’s ductility and local deformation mode on
University; 1998. multiaxial fatigue response. Int J Fatigue 2011;33:930–47.
[59] Garud YS. Multiaxial fatigue of metals. Ph.D. Dissertation, Stanford University; [67] Socie DF, Shield TW. Mean stress effects in biaxial fatigue of Inconel 718. ASME
1981. J Eng Mater Technol 1984;106:227–32.
[60] Garud YS. Prediction of stress-strain response under general multiaxial [68] Susmel L. Multiaxial fatigue limits and material sensitivity to non-zero mean
loading. In: Rohde RW, Swearengen JC, editors. Mechanical testing for stresses normal to the critical planes. Fatigue Fract Eng Mater Struct
deformation model development, ASTM STP 765, American Society for 2008;31:295–309.
Testing and Materials; 1982. p. 223–38. [69] Shamsaei N. Multiaxial fatigue and deformation including non-proportional
[61] Kandil FA, Brown MW, Miller KJ. Biaxial low cycle fatigue fracture of 316 hardening and variable amplitude loading effects. Ph.D. Dissertation, The
stainless steel at elevated temperatures, vol. 280. London: The Metal Society; University of Toledo; 2010.
1982. p. 203–10. [70] Doong SH, Socie DF. Deformation mechanisms of metals under complex non-
[62] Yu Q, Zhang JX, Jiang YY, Li QZ. Multiaxial fatigue of extruded AZ61A proportional cyclic loading. In: Proceeding, third international conference on
magnesium alloy. Int J Fatigue 2011;33:437–47. biaxial/multiaxial fatigue, April 3–6, Stuggart, FRG; 1989.
[63] Noban M, Jahed H, Ibrahim E, Ince A. Load path sensitivity and fatigue life [71] Wang CH, Brown MW. Life prediction techniques for variable amplitude
estimation of 30CrNiMo8HH. Int J Fatigue 2012;37:123–33. multiaxial fatigue – part 1: theories. ASME J Eng Mater Technol
[64] Li BC, Jiang C, Han X, Li Y. A new path-dependent multiaxial fatigue model 1996;118:367–70.
for metals under different paths. Fatigue Fract Eng Mater Struct 2014;37: [72] Chen X, Gao Q, Sun XF. Low cycle fatigue under non-proportional loading.
206–18. Fatigue Fract Eng Mater Struct 1996;19(7):839–54.

You might also like