International Journal of Fatigue 129 (2019) 105209

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Development of new critical plane model for assessment of fatigue life under T
multi-axial loading conditions
Punit Arora , Suneel K. Gupta, M.K. Samal, J. Chattopadhyay
⁎

Bhabha Atomic Research Centre, Mumbai 400085, India

ARTICLE INFO ABSTRACT

Keywords: The present work is aimed at development of new critical plane model based on envelope of strain energy
Fatigue crack initiation life density as fatigue damage parameter. This model eliminates the use of hypotheses for evaluation of resultant
Multi-axial shear. The fatigue life has been predicted and compared with test life for a large set of multi-axial test data
Non-proportional available on SA 333 Gr. 6, 16MnR, S460N and 7075-T651 Al materials. Predicted fatigue life is in good
Critical plane
agreement with test life. The predicted crack initiation orientations are also comparable with the measured
Fatigue damage parameter
Crack orientation
angles.
The shortcomings of existing models have been brought out.

1. Introduction such small length scale, however, for engineering structures, reasonable
assessment can be made by analyzing continuum resolved stresses. With
The mechanical components such as pipes, elbows/bends, T-junc- this perspective, many models based on planes with maximum damage
tions, vessels etc. are subjected to periodic cyclic loading during normal (termed as ‘critical plane’) have been proposed [3–21,28–35]. These
operating conditions as well as during the design basis events and may models have used different definitions of the critical plane orientation
fail due to the fatigue cycling. Generally, the fatigue damage is eval- and the Fatigue Damage Parameter (FDP). The FDP correlates fatigue
uated according to the design codes, which use fatigue design curves. life and is based on resolved stresses, strains, stress/strain invariants or
These fatigue design curves are derived from uniaxial cyclic tests on strain energy density. The concepts of critical plane based methodology
standard Low Cycle Fatigue (LCF) specimens under completely re- were first studied for High Cycle Fatigue (HCF) [28–35]. Mc-Diarmid
versible strain-controlled condition. However, the real piping compo- [29] proposed the linear multiaxial fatigue model based on maximum
nents and vessels experience multi-axial state of stresses/strains due to normal and shear stress amplitudes on a critical plane. The regime of
complex geometry and loadings. Nozzle-shell junction in pressure high cycle fatigue was extensively investigated by Papadopoulos
vessel, T-junction in piping network and other geometric discontinuities [32–35] to devise an equivalent parameter for fatigue life evaluation.
are the locations where combined pressure and thermal transients re- Recently, Matsubara [30] proposed unified equivalent shear stress
sults in multi-axial stress/strain cycling. amplitude for life predictions and Maktouf [31] performed the root-
It is widely reported [1–22] that fatigue life of a material depends cause analyses of gas turbine blade of Inconel 718 grade in HCF regime.
on strain cycling with stationary principal directions (proportional, in The commonly used fatigue life estimation techniques are based on
general) or varying principal directions (non-proportional, in general). strain components or the combination of stress and strain such as
These types of tests are generally carried out on standard tube specimen proposed by Smith et al. [3], Fatemi and Socie [9], Brown et al. [10],
subjected to axial and torsion strain cycling [23]. Glinka [11], Chu [8], Jiang [13] and Jing Li et al. [14], Ince and Glinka
The phenomenological fatigue damage has been studied by many [18], Lu et al. [15] etc. For non-proportional loading conditions, both
researchers for decades and has been reported to be caused by micro- magnitude and direction of resultant shear stress vary with time. In
structural discontinuity arising due to formation of intrusions & ex- order to evaluate the mean and amplitude of this resultant shear stress,
trusions. The formations of intrusions and extrusions take place on the different methods such as Longest Cord (LC), Minimum Circumscribed
planes aligned closely to maximum shear planes [24–27]. The micro- Circle (MCC) [33] and Minimum Circumscribed Ellipse (MCE) were
structurally localized stresses cause strain peaking and lead to fatigue proposed [36].
crack initiation on material surface. Although fatigue damage occur at Jiang [13], Zhu et al. [16], Ince and Glinka [18] and Gupta et al.

⁎
Corresponding author.
E-mail address: [email protected] (P. Arora).

https://doi.org/10.1016/j.ijfatigue.2019.105209
Received 6 March 2019; Received in revised form 3 July 2019; Accepted 23 July 2019
Available online 01 August 2019
0142-1123/ © 2019 Elsevier Ltd. All rights reserved.
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Nomenclature n , nl, nr (or n ,

a a a a a
nl,
a
amplitude of normal and shear stress (or
nr )
strain) components on lr plane
t
ij (or
t| )
xyz stress time history in xyz coordinate system n , nl , nr (or n , nl , nr ) mean value of normal and shear stress (or
m m m m m m
_
strain) components on lr plane
t
pq (or
t| )
nlr stress time history in nlr coordinate system
_ n
max
, nlmax , nr
max
(or nmax , nlmax , nrmax ) maximum of normal and shear
ij , pq strain time history in xyz, nlr coordinate systems
t t
stress (or strain) components on lr plane
n outward normal unit vector ns (or ns , ns , ns , ns , ns ) time history, ampli-
t a m max t a m max
ns , ns , ns , ns ,
r , l^ in plane unit vectors of oblique plane tude, mean, maximum and range of resultant shear stress
azimuth angle (or strain) on lr plane
, c (or ct , c , cm, cmax , c time history, amplitude,
a
angle between n and z axes t a m
c, c , c , c
max

api direction cosines of transformation matrix mean, maximum and range of resultant shear stress (or
k material parameter strain) on critical plane
, c (or ct , c , cm, cmax , c time history, amplitude,
a a
E Young’s modulus
t m max
c, c , c , c
phase shift angle mean, maximum and range of normal stress (or strain) on
Poisson’s ratio critical plane
y yield strength
ultimate tensile strength Acronyms
u
t
ns angle history between ns t
and nlt
T time period of one cycle LCF Low Cycle Fatigue
( c, c) angular coordinates of critical plane PHWR Pressurized Heavy Water Reactor
PSB Persistent Slip Bands
n elastic and plastic normal strain ranges on oblique plane
el pl
n,
MCC Minimum Circumscribed Circle
ns elastic and plastic resultant shear strain ranges on oblique
el pl
ns ,
plane FDP Fatigue Damage Parameter
el pl
elastic and plastic normal strain ranges on critical plane HCF High Cycle Fatigue
c , c
PHT Primary Heat Transport
c elastic and plastic resultant shear strain ranges on critical
el pl
c ,
plane LC Longest Cord
MCE Minimum Circumscribed Ellipse
n, nl, nr (or n, nl, nr ) time histories of normal and shear stress (or
t t t t t t

strain) components on lr plane SWT Smith-Watson-Topper

[20] have made attempts to combine strain energy density and critical 2. Scope of present work
plane approaches to develop a parameter which is representative of
strain energy on the critical plane. All these models used LC/MCC In the present work, a new model is proposed which is based on a
methods for evaluation of resultant shear stress/strain amplitude/ hybrid scheme of plastic strain energy density envelope on critical
mean. Recent literature [37–40] proposes corrections in existing plane. The proposed model does not need evaluation of resultant shear
models vis-à-vis new developments to account the effect of mean stress amplitude/mean and hence it eliminates the scope of subjectivity in
on fatigue life under complex multi-axial loading conditions. assessment of fatigue life.
Therefore, a large number of critical plane models have been de- The proposed model includes a material parameter (load in-
veloped till date, many of which produce good results for a particular dependent) to account for relative dissipation of shear and normal
grade of material under a given range of multiaxial non proportional strain energies on critical plane till fatigue crack initiation (typical
fatigue cycling, but very limited models have been used for wide variety crack size ~1 mm). This parameter is based on phenomenon of fatigue
of loading conditions and materials used for engineering applications. damage at continuum scale. Such parameter is missing in most of strain
The attempt of this paper is to develop a model which eliminates the energy based critical plane models.
need of resultant shear evaluation and can cater to large spectrum of Further, limited literature is available which predict both fatigue
complex multi-axial loading conditions for steel structures generally life and cracking orientation reasonably well [41,42]. The proposed
used for engineering applications. To cover the aspect of loading model not only predicts comparable fatigue life w.r.t. test outcome for
spectrum, various loading paths have been selected and analyzed with three different materials, but it estimates plane associated with max-
more rational way of evaluation of resultant shear without use of hy- imum damage also. The measured crack orientations are comparable to
pothetical approximations. To cover the materials’ variability aspect, predicted crack planes angles.
three materials from three different applications have been chosen.
These analyses systematically follow extensive experimental in-
vestigations on C-Mn piping material [1]. Some of the promising 3. Review of existing fatigue models
models from literature were selected for fatigue life assessment fol-
lowed by development of new model and its validation on other ma- It is widely studied that fatigue crack initiation is caused by cyclic
terials. Although, a few models produce reasonable assessment for the plasticity. The Persistent Slip Bands (PSB) are found to form in the
materials considered here, however they involve subjectivity in eva- grains which are closely aligned with the plane of maximum shear
luation of resultant shear mean/amplitude components and hence stress/strain. The critical plane models are based on micro-structurally
predicted fatigue life might be different using different approximations observed physical phenomenon of PSB. Further, the fatigue strength of
of LC/MCC or MCE for a complex loading path. In present work, much a material is also related to its true fracture ductility. The true fracture
effort has been devoted towards developing appropriate simplification ductility depends on the hydrostatic strain components around the
which overcame these approximations and resulted in minimization of micro-structurally initiated crack. Therefore, shear strain and the
computational time of critical plane based fatigue calculations. normal strain components on critical plane can be used as the key
governing parameters till the crack become mechanical size (~1 mm)
vis-à-vis criterion for fatigue crack initiation life [23].

2
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

More than 50 critical plane based models are available in literature The typical variations of all six stress components in xyz system with
[3–21,28–35]. The critical plane based models define fatigue damage sine family of waveform under proportional (in-phase) and non-pro-
parameter as a function of normal and shear stresses/strains on certain portional (out-of-phase) conditions are shown in Fig. 1(b). Upon rota-
material planes. There are various definitions of the critical plane as tion to nlr system, the corresponding variations of nt, nlt, nr
t
in nlr space
explained by different models. Because of large number of the critical would be an oblique line and oblique ellipse for proportional and non-
plane fatigue models reviewed, it is realized to explain calculation proportional conditions respectively as depicted in Fig. 1(e).
methodology of the stress and strain quantities arising in their FDP The normal stress component, nt , along n direction varies only in
definitions. magnitude, therefore the mean and amplitude of normal stress
( nm ( , ) and na ( , ) ) can be evaluated simply by Eqs. (2) and (3)
respectively.
3.1. Evaluation of resolved stress/strain on oblique material plane
m 1 t t
n ( , )= [ max ( n( , )) + min ( n( , ))]
At the maximum stress location of an engineering component, the 2 0 t T 0 t T (2)
stress/strain tensor history is required for fatigue life assessment. Let
1
ij (or ij ) be the stress (or strain) components at time ‘t’. It is assumed
t t a t t
n ( , )= [ max ( n( , )) min ( n( , ))]
2 0 t T (3)
that a material point is large enough to be depicted by sphere of radius
0 t T

r, where r → 0 (see Fig. 1(d)) with its center located at highly stresses The projections of the oblique line (for proportional loading con-
location. An oblique plane is defined on the surface of this sphere by a dition) and oblique ellipse (for non-proportional loading condition) can
unit radial outward vector n . The vector n makes a meridian angle be taken on lr plane. These projections in Fig. 1(e) show the resultant
with respect to + z-axis. The projection of this vector on xy plane, shear stress ( ns
t
) , as given by Eq. (4), changes its magnitude (only) for
vector nsin makes azimuth angle with + x-axis such that, −90° ≤ proportional loading. Therefore, it is easier to find the amplitude and
ϕ ≤ 90° and −180° ≤ ψ ≤ 180° to form complete sphere. mean of resultant shear for proportional loading condition. However,
Any material plane on the surface of this small sphere can be de- for non-proportional load scenario, both magnitude and direction of
signated using local orthogonal co-ordinate system nlr at the material resultant shear stress/strain change. This makes the evaluation of mean
point such that ^l × r = n . The stress tensor specified along any arbi- and amplitude of resultant shear stress difficult.
trary co-ordinate system (say xyz) can be rotated to local nlr co-ordi- t t 2 t 2
nates using the rotation tensor as given by (1). ns = ( nl ) +( nr ) (4)

t t
pq = api ij aqj (1)
3.2. Evaluation of resultant shear stress/strain: Mean and amplitude
where pq t
and ijt are the histories of stress tensors in nlr and xyz co-
ordinate systems respectively and api, aqj are the coefficients of rotation For an arbitrarily variation of stress components (non-sine family)
matrix, such that, api = cos pi ; where, pi is angle between p and i axes, with non-proportional condition prevailing and having non-zero mean
where, p , q (nlr ) and i , j (xyz ) . stress components, a typical variation of shear stress components

Fig. 1. (a) Typical shell-nozzle junction subjected to (b) varying six stress components (proportional/non-proportional), (c) tubular specimen subjected to axial-
torsion cycling, (d) infinitesimal material sphere at point ‘O’ and (e) transformed state of stress in nlr coordinate system.

3
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

a 1 tj ti 2 tj ti 2
ns ( , )= max ( nr nr ) +( nl nl )
2 0 ti T
0 tj T (6)

m c 2 c 2
ns ( , )= ( nr ) +( nl ) ; where, centerof LC is ( nlc, c
nr ) (7)

3.2.2. Minimum Circumscribed Circle (MCC) method

Papadopoulos [33] also proposed Minimum Circumscribing Circle
(MCC) approach that defined ns a
as the radius of minimum circum-
scribing circle to the shear stress loading path on lr plane. The resultant
mean shear ( nsm
) is evaluated by connecting the center of MCC to zero
shear stress point. The equations to evaluate nsa
and nsm
on a ( , ) plane
using MCC method are given by Eqs. (8) and (9) respectively.
a
ns ( , ) = radius of MCC (8)

m c 2 c 2
ns ( , )= ( nr ) +( nl ) ; where, center of MCC is ( nlc, c
nr )

(9)

3.2.3. Minimum Circumscribed ellipse (MCE) method

Anes et al. [36] proposed Minimum Circumscribing Ellipse (MCE)
methodology which works on circumscription of minimum ellipse
covering entire loading path. The mean and amplitude are evaluated as
indicated in Fig. 2(b). The amplitude and mean of resultant shear stress,
that are, ns
a
and nsm
, on a ( , ) plane using MCE method are given by
Eqs. (10) and (11) respectively.

(10)
a
ns ( , )= Ra2 + Rb2
Fig. 2. The mean and amplitude of resultant shear for arbitrarily loading path where Ra : half of major axis length, Rb : half of minor axis length
using (a) LC and MCC methods and (b) MCE method.
m c 2 c 2
ns ( , )= ( nr ) +( nl ) ; where, center of MCE is ( nlc, c
nr )

( nl, nr ) on
t t
oblique plane ( , ) is shown in Fig. 2. The direction of re- (11)
sultant shear, ns
t
w.r.t. nlt is given by Eq. (5). These methodologies for evaluation of amplitude and mean com-
t ponent of shear stress/strain are computationally challenging for rea-
t 1 nr
ns = tan t listic loading history encountered during service transients of reactor
nl (5)
components. For estimation of fatigue life, present study used LC
The amplitude ( and mean (
a
ns ) ns ) of
resultant shear have been
m
method for analyses on few popular models (Section 3.3).
proposed by Papadopoulos [33] and Anes et al. [36] as explained in
subsequent sections. 3.3. Fatigue life assessment using existing popular models

3.2.1. Longest Cord (LC) method Table 1 summarizes some of the promising critical plane models
Papadopoulos [33] has proposed the Longest Cord (LC) approach which use strain energy density as fatigue damage parameter on critical
(Fig. 2(a)) to evaluate amplitude and mean of resultant shear stress on plane. Table 1 indicates definition of critical plane ( c , c ) and FDP
any ( , ) plane. The LC method defines the resultant shear stress ( ns a
) evaluated on this plane as prescribed by different models. The fatigue
amplitude (or strain amplitude ( nsa )) as half of the longest cord length of life is generally assessed after evaluating FDP for complex multiaxial
the shear stress (or shear strain) locus as shown in Fig. 2(a). loading condition. This FDP is used to predict fatigue life from FDP-
The resultant mean shear ( nsm
) is evaluated by connecting the mid- versus-fatigue life curve (Ni ) under uniaxial case. The FDP is considered
point of LC to zero shear stress point. The equations to evaluate ns a
and as an equivalent parameter for all loading cases.
ns on a ( , ) plane using LC method are given by Eqs. (6) and (7) Initially, Smith, Watson and Topper (SWT) [3] brought out the
m

respectively. concept of dissipation of strain energy density due to resolved normal

Table 1
The summary of a few classical and recent strain energy density based critical plane models.
Model Critical Plane: ( c , c) FDP on( c , c)

Smith, Watson & Topper [3] max( na ) a max

c c
( , )
Chu [8] max( a max
n n +2 a max
ns ns )
a max
c c +2 a max
c c
( , )
Shun-Peng Zhu et al. [16] n max ns max k c max
+ c max
max k + c c
( , ) 2 n 2 ns 2 2

Ince & Glinka [18] el pl el pl el pl el pl

n max n n ns max ns ns c max c c c max c c
max + + + + + +
( , ) 2 n 2 2 2 ns 2 2 2 c 2 2 2 c 2 2

Gupta et al. [20] max[ a max

n n + a a
ns ns ]
a max
c c + a a
c c
( , )

4
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

strain and normal stress components. Later, Chu [8] proposed bi-axial The predicted and test fatigue lives have been compared under all
SWT model in which the dissipation due to both normal and shear the loading paths (Table 2) for SA 333 Gr. 6 [1], S460N [13] and
components was considered. Subsequent to bi-axial SWT proposal by 16MnR [6] materials in Fig. 4. This figure shows that Chu model seems
Chu with reasonable fatigue life predictions w.r.t. test results, many to be more promising than SWT model as most of the data points for
authors ([16,18–20]) started working on his idea for further refinement pure axial, pure torsion and combined axial-torsion fall on conservative
of life predictions under various load paths and materials of their in- predictions side. This is due to the definition of damage parameter
terest. based on the energy dissipation in both shear and normal mode on
The tensile properties and type of loadings adopted on these three maximum damage plane. However, the reason for giving higher fixed
materials are furnished in Table 2. The numbers of tests analyzed in weightage to shear energy dissipation in FDP by a multiple of 2
each category of loading (or strain path) have also been listed in (Table 1) is not well explainable. The data scatter goes out of the per-
Table 2. missible scatter band for all the three materials. However, Chu model
results in conservative fatigue life predictions.
3.3.1. Smith, Watson and Topper (SWT) model Further, the critical plane and FDP definitions of Shun-Peng Zhu
Smith, Watson and Topper [3] proposed a fatigue damage para- model [16] are nearly similar to Chu model. Due to this reason, the
meter which is representative of strain energy density due to normal fatigue life predictions using Zhu model are expected to be similar to
components of stress ( nmax ) and strain ( na ) on maximum damage plane. that of Chu model.
The maximum damage plane ( c , c ) has been defined to be associated
with maximum value of the normal strain amplitude ( na ). This plane
3.3.3. Ince-Glinka model
inherently includes the effect of additional strain hardening in case of
Ince and Glinka [18] proposed more rational fatigue damage
non-proportional load condition. This model also includes the effect of
parameter that accounts dissipation of normal and shear strain energy
mean stress. However, the damage due to shear strain energy has not
density on critical plane. The plastic strain components are multiplied
been considered in fatigue parameter. This limits the scope of the model
with the energy conjugate stress amplitudes and mean stress effect has
to nearly brittle class of materials. The critical plane criterion and fa-
been considered on the elastic strain part. This could be due to non-
tigue damage parameter are as given in Table 1.
relaxation of mean stress component in high cycle fatigue regime. Ince
The assessed fatigue lives are compared with test lives for wide
and Glinka have given equal weightage to normal and shear strain
variety of loading conditions and all three materials in Fig. 3.
energies on critical plane. For the present analyses on SA 333 Gr. 6,
Fig. 3 indicates that the pure torsion and most of non-proportional
S460N and 16MnR materials, LC method has been used for evaluation
tests do not fall in the acceptable band for data scatter of 2. The pre-
of resultant shear stress range, ns (or strain range, ns ).
dicted fatigue life using SWT model is higher for tests conducted under
Fig. 5 shows that fatigue life predictions for most of the loading
pure torsion and axial-torsion strain cycling with 90° phase shift.
paths and nearly all the materials are comparable with the test out-
come. However, the model used does not explain the basis of equal
3.3.2. Chu model
weightage given to normal and shear components on an oblique plane
Chu [8] has analyzed the fatigue damage function on sufficiently
in the definition of fatigue damage parameter. In addition, this model
large numbers of oblique planes ( , ) and defined the plane as critical
requires LC, MCC or MCE hypotheses to arrive at FDP for fatigue life
plane which attains its maximum value (or FDP) in one complete cycle.
assessment.
Chu investigated multiaxial test data on SAE 1045 steel using this fa-
The fatigue damage parameter of Gupta et al. [20] also considered
tigue damage parameter which is a representative of strain energy
equal weightage to normal and shear strain energies on oblique mate-
density on critical plane in an approximate way.
rial plane. In presence of zero mean stress, Ince & Glinka model [18]
The damage parameter of this model is the measure of energy dis-
and Gupta et al. [20] would result in identical fatigue life assessments.
sipation in shear and normal mode per loading cycle for a unit volume
of material. The maximum shear stress and strain amplitude are eval-
uated using LC or MCC hypothesis. In this model, higher weightage has 3.3.4. Discussions on existing models
been given to the shear mode of energy dissipation. Further, all these models require computation of amplitude, mean

Table 2
Tensile properties, number of tests and type of tests carried out on SA 333 Gr. 6 [1], S460N [13] and 16MnR steel [6] materials.
Strain path
(→)

Designation A B C D E F G H I
(→)

Pure axial/pure torsion/axial- Pure axial Pure torsion Axial-torsion

torsion (→)
λ (°) – – 0 90 90 90 45 180 45
Waveform (→) – – – Triangular Sine Trapezoidal Triangular – Sine
Material with tensile properties (↓) Number of tests conducted in each strain path

SA 333 Gr. 6: 24 6 15 15 5 5 2 1 –
E = 203 GPa, ν = 0.3, σy
= 307 MPa, σu = 463 MPa
S460N: 5 4 3 3 8 – – – 3
E = 209 GPa, ν = 0.3,
σy = 500 MPa, σu = 970 MPa
16MnR: 12 10 7 8 – – – – –
E = 213 GPa, ν = 0.31,
σy = 324 MPa, σu = 544 MPa

5
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Fig. 3. The plot of predicted fatigue life with test fatigue life using SWT model for SA 333 Gr. 6, S460N and 16MnR materials.

Fig. 4. The plot of predicted fatigue life with test fatigue life using Chu model for SA 333 Gr. 6, S460N and 16MnR materials.

or maximum of resultant shear stress/strain using LC, MCC or MCE the fatigue crack initiation life predictions. The strain energy density
schemes. The MCC and MCE schemes are computationally expensive for due to excursion of normal component of stress and strain on an oblique
highly distorted loading paths which are generally encountered during plane is represented by the product of their maximum/amplitude, as it
service life of a nuclear power plant. In addition, LC, MCC and MCE was proposed in SWT model [3]. In addition to this, the strain energy
hypotheses produce significantly different mean and amplitude of re- density due to excursion of two orthogonal shear components of stress
sultant shear for highly distorted loading path as depicted in Fig. 2(b). and strain on the oblique plane is represented by sum of products of the
This results in difference in assessment in fatigue life. corresponding amplitude (or maximum) of shear stresses ( nlmax , nr max
)
To address these issues, a new critical plane model based on strain and strains ( nla, nra ) . The proposed model considers the energy density
energy density has been worked out which eliminated the intensive due to shear in l and r directions separately (Fig. 1) in spite of evalu-
computations involved in LC, MCC or MCE hypotheses. ating the resultant shear strain energy. This modification has been
suggested on the basis of energy being a scalar and invariant quantity.
4. A new critical plane-cum-strain energy density envelope model The critical plane and proposed fatigue damage parameter are given
by Eqs. (13) and (14) respectively. Eq. (14) shows that FDP requires
In view of above observations, a new model has been proposed for calculation of individual shear components (( nlmax , nr
max
)and( nla, nra ) ) on

6
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Fig. 5. The plot of predicted fatigue life with test fatigue life using Ince-Glinka model for SA 333 Gr. 6, S460N and 16MnR materials.

critical plane. These parameters change only in magnitude and not in conducted on SA 333 Gr. 6 material. The fatigue damage parameter has
direction even for non-proportional loading scenario. Hence, the pro- been plotted w.r.t. all possible planes at a material point lying on outer
posed model eliminates the use of LC, MCC [33] or MCE [36] hy- surface in gauge region of tube specimen in Fig. 7. The peaks of FDP
potheses. represent the probable crack orientations in terms of ( , ) .
Fatigue Damage Function (fFD ) on any ( , ) plane = k a max
n n + (1 k )( a max
nl nl + a max
nr nr ) (12)

The predicted fatigue life has been compared with test life (Ni)
Critical Plane( c , c ): max (fFD )
( , ) (13) under the loading scenario studied by Arora et al. [1] in Fig. 8. This
figure indicates that the large spectrum of loading considered by Arora
a max a max a max
FDP on ( c , c) =k c c + (1 k )( nl nl + nr nr )|( c, c ) (14) et al. [1] fall in the acceptable band of data scatter for this model.
Further, Fig. 9 shows that the predicted fatigue life (using uniaxial
Fatigue life prediction criterion: Uniaxial FDP Ni curve FDP-versus-life curve) is in good agreement with actual test life within a
Secondly, the proposed model considers an additional material scatter band of 2 for nearly all the load paths for S460N and 16MnR
parameter k which is determined from the pure axial and pure torsion materials. However, some of the data points fall on the conservative life
fatigue tests. This parameter as explained earlier signifies the dominant assessment side. Additionally, the proposed model has been verified
mode of failure in a material. The basic idea of using such a parameter with respect to loading paths having static axial mean stress component
is that the fundamental mode of material failure does not depend upon with torsion cycling on 7075-T651 Al material [39]. Fig. 9 indicates
the quasi-statically applied loading. that the proposed model also works satisfactory for mean stress loading
Further, the present analyses consider fatigue crack initiation life paths reported by Kujawski D. [39].
corresponding to mechanical size of crack (~1 mm). The shear strain Hence, the proposed model predicts the fatigue life which is in good
energy plays an important role till micro-crack initiation. Once the
micro-crack initiates, both normal and shear stress/strain components
cause further damage till the crack becomes mechanical size. Hence, the
proposed model includes k parameter in FDP to quantify relative extent
of shear and normal strain energies in causing damage. This parameter
would be different for different engineering materials as plastic flow
behavior of materials is different due internal slip resistance.

4.1. Fatigue life predictions using proposed model

The value of k is calibrated from pure axial and pure torsion cyclic
tests using an iterative procedure. The value of k was varied from 0 to 1
with a typical step size of 0.1 and the plots between FDP and test fatigue
lives were analyzed. The k which resulted in least data scatter for all
pure axial and pure torsion cyclic fatigue tests [1], has been chosen for
predicting fatigue life of multiaxial tests cases. Typical evaluation of k
parameter for SA 333 Gr. 6 material [1] is shown in Fig. 6. The value of
k is found to be 0.6 for this material. Further, the analyses have been Fig. 6. Evaluation of k for proposed model using best fit curve between FDP and
performed for in-phase axial-torsion and out-of-phase axial-torsion tests experimental life for pure axial and pure torsion tests data on SA 333 Gr. 6.

7
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Fig. 7. The variation of FDP of proposed model on different (ϕ (°), ψ (°)) planes for determining the critical planes under (a) pure axial (AMTL-2/R = −1), (b) pure
torsion (CA0S87), (c) Axial-torsion with 0° phase shift (CA20S79p0) and (d) Axial-torsion with 90° phase shift (CA20S79p90) tests (Table 4 and Table 5 of [1]).

agreement with test life for loading paths and materials considered (SA nearly comparable to that of Ince & Glinka model (Fig. 5). Despite this
333 Gr 6, S460N, 16MnR and 7075-T651 Al) in this study. The extent of fact, the fatigue damage parameter in proposed model has been for-
data scatter in predicted and test fatigue life using proposed model is mulated more rationally by giving unequal weightage to normal and

Fig. 8. Predicted fatigue life (with k = 0.6) versus experimental life for SA 333 Gr 6 material using proposed model.

8
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Fig. 9. Predicted fatigue life versus experimental life for S460N, 16MnR and 7075-T651 Al [39] using proposed model.

shear energies. This aspect covers the fundamental failure mode of a initiation have also been predicted using the proposed model and the
material under any loading condition (quasi-statically applied) and may predicted cracking orientations are compared with the actual measured
cater to large variety of engineering materials/structures. However, this cracking orientations for SA 333 Gr. 6 [1] and S460N [13] materials. In
attribute needs detailed analyses on widely different materials in terms case of axial-torsion biaxial cyclic loading on tubular specimen, the
of their ductility. Further using the principle of “energy being invariant”, plane normal associated with maximum FDP lies in circumferential-
the proposed model offers better quantification of damage without longitudinal plane of tube (Fig. 1 (c)) viz. plane of maximum FDP are
hypothetical approximations. associated with angle as 90° (or −90°) ([8], Fig. 7). For this reason,
Fig. 10 shows the variation of normalized FDP on critical plane under a special case of bi-axial loading condition, maximum FDP is a
(FDPC) with phase shift angle (λ). The normalization has been sole function of azimuth angle . The typical predicted maximum da-
done w.r.t. corresponding proportional load case (that is λ = 0°). The mage planes are shown with actual cracked tube specimens for pure
normalized FDPC has been evaluated using SWT, Chu, Ince & torsion, in-phase axial torsion and out-of-phase axial-torsion cases in
Glinka and proposed models for tests on SA 333 Gr. 6 material Fig. 11.
[1]. The tests considered are, ( x = ±0.4%, xy = ± 0.52%, = 0°) ; For a given loading condition, it was found that the several material
( x = ±0.4%, xy = ± 0.52%, = 45°) ; ( x = ±0.4%, xy = ± 0.52%, = 90°) ; planes or a set of material planes experience identical or nearly the
( x = ±0.4%, xy = ± 0.52%, = 180°) . same fatigue damage as depicted in Fig. 7. Due to the inherent data
The criterion as proposed by Smith-Watson-Topper (SWT) predicts scatter in fatigue test studies, it is preferable to identify the material
lesser damage for axial-torsion test with 90° phase shift case, that is, planes with similar fatigue damage. In the current investigations, a
( x = ±0.4%, xy = ± 0.52%, = 90°) in comparison with corresponding range of 10% from the maximum fatigue damage is used for the pre-
in-phase shift test. The criteria based on Chu, Ince & Glinka and pro- dictions of possible cracking material planes. It is to be noted that
posed models support the test investigations [1] by predicting the lesser
damage for 0° phase shift test and higher damage for corresponding test
with 90° phase shift. However, Chu model does not converge to an
acceptable range of fatigue life prediction due to large scatter in data.
It is inferred from the above studies on fatigue crack initiation life
predictions that the proposed model agrees with the test observations
by showing higher damage for non-proportional loading condition
leading to lesser fatigue life. The predicted-versus-test fatigue life data
points for pure axial, pure torsion, in-phase axial-torsion and out-of-
phase axial-torsion tests mostly fall within the acceptable data scatter
bounds.

4.2. Crack plane predictions and comparisons

Very limited literature exists on the predictions of fatigue crack

initiation plane using critical plane based models and its comparison
with measured orientations. Wang et al. [41] and Masao et al. [42]
have brought out reasonable comparison between predicted and mea-
sured crack angles for different grades of engineering materials. Fig. 10. The variation of normalized FDPC with phase shift angle for
The maximum damage planes viz. the planes with probable crack CA40S52p0, CA40S52p45, CA40S52p90 and CA40S52p180 (Table 5 of [1]).

9
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

Fig. 11. Typical predicted and measured [1] crack plane orientation for (a) pure torsion, (b) in-phase axial-torsion case and (c) out-of-phase axial-torsion case.

fatigue crack initiation is a localized phenomenon and it occurs mainly 90% of the maximum FDP value.
due to the formation of persistent slip bands in most of the metallic This clearly shows that the proposed model not only predicts the
materials. However, in stress analyses, on the macroscopic level, the fatigue life for materials SA 333 Gr. 6, 16MnR steel, S460N and 7075-
material is assumed to be homogeneous and isotropic which means T651 Al reasonably well, but the predicted and measured cracking
identical resistance to slip on all the planes and in all directions. But, in planes are also found in good agreement for representative test cases of
reality, on the microscopic length scale, different resistances to slip SA 333 Gr. 6 and S460N. The proposed modification in bi-axial SWT
phenomenon are there for different planes and directions. Therefore, model is well supported by fundamental mechanics and test in-
the fatigue crack initiation plane may not be exactly located with vestigations.
maximum FDP plane. To account for this aspect, a band of 10% with
respect to maximum FDP was arbitrarily taken.
Fig. 12 summarizes the comparison of the measured crack angles 5. Conclusions
and the predicted cracking orientations using the proposed model for
SA 333 Gr. 6 [1] and S460N [13] materials. The dots represent the The salient summary of fatigue analyses performed on SA 333 Gr. 6,
measured crack orientations and the range bars in this figure are the S460N and 16MnR materials under proportional and non-proportional
predicted band of crack plane orientation having damage more than loading condition is as follows:

Fig. 12. Typical comparison between observed and predicted (proposed model) crack angles for SA 333 Gr. 6 and S460N materials.

10
P. Arora, et al. International Journal of Fatigue 129 (2019) 105209

(1) The SWT model results in over-prediction of fatigue life for most of including out-of-phase loading. Fatigue Fract Eng Mater Struct 1988.
non-proportional loading conditions. This could be due to absence of [10] Kandil FA, Brown MW, Miller KJ. Biaxial low-cycle fatigue fracture of 316 stainless
steel at elevated temperatures, vol. 280. London: The Metals Society; 1982. p.
shear stress/strain components in its FDP. However, bi-axial SWT 203–10.
class of models such as Chu, Gupta et al. and Ince & Glinka produce [11] Glinka G, Wang G, Plumtree A. Mean stress effects in multiaxial fatigue. Fatigue
comparatively better predictions for fatigue life. Despite this fact, Chu Fract Eng Mater Struct 1995;18:755–64.
[12] Socie DF. Multiaxial fatigue damage models. J Eng Mater Technol
model results in large scatter in FDP exceeding the permissible limits. 1987;109(4):293–8.
(2) To overcome these shortcomings, a new model is proposed that falls [13] Yanyao Jiang, Hertel O, Wormwald M. An experimental evaluation of three critical
in bi-axial SWT class of strain energy density family and includes plane multiaxial fatigue criteria. Int J Fatigue 2007;29:1490–502.
[14] Li Jing. A new multiaxial fatigue damage model for various metallic materials under
the followings: the combination of tension and torsion loadings. Int J Fatigue 2009;31:776–81.
(a) A material parameter k, has been included to account for re- [15] Lu C, Melendez J, Martínez-Esnaola JM. Multiaxial fatigue criterion considering the
lative dissipation of shear and normal strain energies leading to influence of out-of-phase failure and loading condition. Int J Fatigue
2018;114:323–30.
fatigue crack initiation. This parameter is anticipated to address
[16] Zhu Shun-Peng, Yu Zheng-Yong, Correia José, De Jesus Abílio, Berto Filippo.
a wide variety of engineering materials having small to high Evaluation and comparison of critical plane criteria for multiaxial fatigue analysis
ductility. This material parameter attributes to fundamental of ductile and brittle materials. Int J Fatigue 2018;112:279–88.
failure mode which is intrinsic characteristic of a material. [17] de Freitas Manuel. Multiaxial fatigue: From materials testing to life prediction.
Theor Appl Fract Mech 2017;92:360–72.
(b) In the existing critical plane based fatigue models, the calcu- [18] Ince Ayhan, Glinka Grzegorz. A generalized fatigue damage parameter for multi-
lation of shear stress/strain amplitudes using LC, MCC or MCE axial fatigue life prediction under proportional and non-proportional loadings. Int J
hypotheses may lead to different results for some cases of non- Fatigue 2014;62:34–41.
[19] Chen X, Xu S, Huang D. A critical plane-strain energy density criterion for multi-
proportional loading. This subjectivity has been eliminated in axial low-cycle fatigue life under non-proportional loading. Fatigue Fract Eng Mater
proposed model by adopting shear energy dissipation in both Struct 1999;22(8):679–86.
shear directions separately, hence ruling our use of LC/MCC/ [20] Gupta Suneel K, Fesich TM, Schuler X, Roos Evie. A Critical plane based model for
fatigue assessment under fixed and rotating principal direction loading. Structural
MCE methodologies. The evaluation of fatigue damage para- mechanics in reactor technology conference, SMiRT-21, New Delhi, India, Volume:
meter using proposed model is simple and less time consuming. Div-II, Paper ID: 624.
(3) The fatigue life assessment analyses performed on SA 333 Gr. 6, [21] Braccesi C, Morettini G, Cianetti F, Palmier M. Development of a new simple energy
method for life prediction in multiaxial fatigue. Int J Fatigue 2018;112:1–8.
S460 N, 16MnR and 7075-T651 Al materials using proposed model [22] Sivaprasad S, Bar HN, Gupta Suneel Kumar, Arora Punit, Bhasin V, Tarafder S. A
result in reasonably acceptable predictions for nearly all cyclic comparative assessment of cyclic deformation behaviour in SA333 Gr.6 steel using
loading paths considered. The predicted life using this model is solid, hollow specimens under axial and shear strain paths. Int J Fatigue
2014;61:76–86.
comparable to test life within a data scatter band of 2. Further, the
[23] Standard Practice for Strain-Controlled Axial-Torsional Fatigue Testing with Thin
scatter band of predicted fatigue life using proposed model is Walled Tubular Specimens (Designation: E2207 – 08), American Society of Testing
comparable to that of Ince & Glinka model. However, Ince & Glinka Materials (ASTM).
model is based on hypothetical approximation of resultant shear on [24] Xiaoshan Liu, Guoqui He, Xiangqun Ding, Defeng Mo, Weihua Zhang. “Fatigue
behavior and dislocation substructures for 6063 aluminum alloy under non-pro-
critical plane which might produce different assessments for dif- portional loadings. Int J Fatigue 2009;31:1190–5.
ferent methodologies of LC, MCC and MCE. [25] Masing G. Eigenspannungen und verfestigung beim messing (Self stretching and
(4) The critical planes, that is, the planes with maximum damage have hardening for brass). In: Proceedings of the second international congress for ap-
plied mechanics, Zurich, Switzerland; 1926. p. 332–35.
been identified using proposed model. The planes having FDP more [26] Ellyin F. Fatigue damage, crack growth and life prediction, edition; 1997. Chapman
than 90% of maximum FDP have been considered to arrive at prob- & Hall. Pohl K, Mayr P, Macherauch E. Persistent slip bands in the interior of a
able cracking orientation band. The cracking orientation band cor- fatigued low-carbon steel. Scrip Metall 1980;14:1167–9.
[27] Polak J, Kruml T. Topography of the crack nuclei at the emerging persistent slip
responding to FDP more than 90% of maximum FDP was arbitrarily band in austenitic 316L steel. In: Rie KT, Portella PD, editors. Low cycle fatigue and
chosen to account for anisotropy and inhomogeneity at micro- elasto-plastic behavior of materials. Elsevier Amsterdam; 1998. p. 559–64.
structural level. The measured crack plane angles ([1,13]) compare [28] Ninic D. A stress-based multiaxial high cycle fatigue damage criterion. Int J Fatigue
2006.
well with predicted crack orientation band using proposed model.
[29] McDiarmid DL. A general criterion for high cycle multiaxial fatigue failure. Fatigue
Fract Eng Mater Struct 1991;14(4):429–53.
Acknowledgement [30] Matsubara Go, Hayashida Atsushi, Kano Daichi. Predicting the multiaxial fatigue
limit and the multiaxial high-cycle fatigue life based on the unified equivalent shear
stress from axial strength characteristics with various waveforms. Int J Fatigue
Authors gratefully and sincerely acknowledge the technical support 2018;112:52–62.
and guidance of Mr. Vivek Bhasin, Associate Director, Nuclear Fuel [31] Maktouf W, Ammar K, Ben Naceur I, Saï K. Multiaxial high-cycle fatigue criteria and
Group, Bhabha Atomic Research Centre, Mumbai, India. life prediction: application to gas turbine blade. Int J Fatigue 2016;92:25–35.
[32] Papadopoulos Ioannis V. A new criterion of fatigue strength for out-of-phase
bending and torsion of hard metals. Int J Fatigue 1994.
References [33] Papadopoulos Ioannis V. A comparative study of multiaxial high-cycle fatigue cri-
teria for metals. Int J Fatigue 1997;19(3):219–35.
[34] Papadopoulos Ioannis V. Critical plane approaches in high-cycle fatigue: on the
[1] Arora Punit, Gupta Suneel K, Bhasin V, Singh RK, Sivaprasad S, Tarafder S. Testing definition of the amplitude and mean value of the shear stress acting on the critical
and assessment of fatigue life prediction models for Indian PHWRs piping material plane. Fatigue Fract Eng Mater Struct 1998;7:195–211.
under multi-axial load cycling. Int J Fatigue 2016;85:98–113. [35] Papadopoulos IV. Long life fatigue under multiaxial loading. Int J Fatigue 2001.
[2] Chattopadhyay Somnath. AC 2012-3221: Investigation of proportional and non- [36] Anes V, Reis L, Li B, Freitas M, Sonsino CM. Minimum Circumscribed Ellipse (MCE)
proportional loadings using Mohr’s circle. American Society for Engineering and Stress Scale Factor (SSF) criteria for multiaxial fatigue life assessment. Theor
Education; 2012. Appl Fract Mech 2014;73:109–19.
[3] Smith RN, Watson P, Topper TH. A stress strain function for the fatigue of metals. J [37] Ince A. A mean stress correction model for tensile and compressive mean stress
Mater 1970;5(4):767–78. fatigue loadings. Fatigue Fract Engng Mater Struct 2017;40:939–48.
[4] Skibicki Dariusz. Phenomena and computational models of non-proportional fatigue [38] Ince Ayhan. A generalized mean stress correction model based on distortional strain
of materials. Springer-publisher; 2014. energy. Int J Fatigue 2017;104:273–82.
[5] Wang CH, Brown MW. A path-independent parameter for fatigue under propor- [39] Kujawski Daniel. A deviatoric version of the SWT parameter. Int J Fatigue
tional and non-proportional loading. Fatigue Fract Engng Mater Struct 2014;67:95–102.
1993;16(12):1285–98. [40] Zhu Shun-Peng, Yu Zheng-Yong, Liu Qiang, Ince Ayhan. Strain energy-based mul-
[6] Gao Z, Zhao T, Wang X, Jiang Y. Multiaxial fatigue of 16MnR steel. J Pressure Vessel tiaxial fatigue life prediction under normal/shear stress interaction. Int J Damage
Technol 2009;131:021403-1-9. Mech 2018:1–32.
[7] Fatemi Ali, Shamsaei Nima. Multiaxial fatigue: an overview and some approxima- [41] Wang Y, Susmel L. Estimation of orientation of crack initiation planes under con-
tion models for life estimation. Int J Fatigue 2011;33:948–58. stant and variable amplitude fatigue loading. Proc Struct Integrity 2016;2:3233–9.
[8] Chu CC. Fatigue damage calculations using the critical plane approach. J Eng Mater [42] Sakane Masao, Itoh Takamoto. A synthesis of cracking directions in tension-torsion
Technol 1995. multiaxial low cycle fatigue at high and room temperatures. Theor Appl Fract Mech
[9] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage 2018;98:13–22.

11