International Journal of Fatigue 150 (2021) 106321

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: [Link]/locate/ijfatigue

On the application of fracture fatigue entropy to multiaxial loading

M. Mehdizadeh, M.M. Khonsari *
Louisiana State University, Department of Mechanical and Industrial Engineering, 3283 Patrick Taylor Hall, Baton Rouge, LA 70803, USA

A R T I C L E I N F O A B S T R A C T

Keywords: An experimental and theoretical analysis of low carbon steel 1018 subjected to multiaxial loading is presented.
Multiaxial fatigue Different loading conditions, including tension–compression, torsion, in-phase, and out-of-phase, are applied to
In-phase loading investigate the effect of loading type on fatigue life. The fracture fatigue entropy (FFE) framework is then used to
Out-of-phase loading
predict life using both hysteresis loops and thermography. The results show that the out-of-phase loading leads to
Fracture fatigue entropy
the minimum fatigue life, while the torsion exhibit the maximum life at a similar equivalent strain. The FFE
predictions using thermography successfully captured the specimens’ fatigue life subjected to the multiaxial
fatigue loading independent of the loading condition. Furthermore, the results show that the strain energy ob­
tained from the hysteresis loop is a good measure of dissipation energy in the cases of tension–compression,
torsion, and in-phase loading conditions. Accordingly, the predicted life using the hysteresis loop agrees well
with the thermographic and experimental results. However, in the case of out-of-phase loadings, the plastic strain
energy obtained from the hysteresis loop deviates from the dissipation energy because the stored energy origi­
nated from the internal state variable variation. Accordingly, the predicted life using the thermography is more
accurate than the hysteresis loop, which agrees well with the experimental results.

1. Introduction proportional and non-proportional loading. This method also considers

the effect of cyclic hardening during fatigue loading.
Most components and structures in service experience multiaxial More recently, Shamsaei et al. [3] proposed a methodology based on
stress/strain rather than uniaxial loading during their lifetime. There­ the critical plane approach for the axial-torsion strain paths. They
fore, it is vital to establish a methodology capable of reliable prediction employed the von Mises, Fatemi-Socie [6], and Wang-Brown [4] cycle-
of multiaxial fatigue life. One approach to convert the multiaxial fatigue counting models and used Palmgran-Miner linear damage model to
calculation to that of the uniaxial loading is based on the equivalent evaluate cumulative fatigue damage. They showed that some materials
stress/strain prescribed by the von Mises and Teresa criteria [1,2]. These are more sensitive to the non-proportionality of loadings during the
models are not sensitive to the hydrostatic stress/strain and do not fatigue phenomenon [3].
consider the effect of mean stress/strain. Also, they are not path- Another approach to capture the multiaxial fatigue life is based on
dependent and, consequently, do not work well in the case of non- assessing energy dissipation. It has been shown that the energy dissi­
proportional fatigue loading [3]. Wang and Brown [4] modified the pation during a fatigue experiment can be employed as fatigue damage
von Mises equivalent strain to address the non-proportional fatigue [7–10]. Feng et al. [11] proposed a multiaxial fatigue model based on
loading using the rainflow modeling approach. However, the applica­ energy dissipation that works under different strain paths, including in-
tion of this method requires introducing several material parameters and phase and out-of-phase loadings. Specifically, they performed a set of
leads to higher sensitivity and complexity. multiaxial experiments on stainless steel 316L under five different
Another approach to predict the multiaxial fatigue behavior was loading paths and successfully evaluated the efficacy of their approach.
proposed by Brown and Miller [5]. They assumed that cracks initiate and In another investigation, Skibicki et al. [12,13] used the plastic strain
propagate on the critical plane that experiences the maximum stress/ energy obtained from thermography, as well as hysteresis loops, to
strain and, consequently, become susceptible to maximum damage [5]. predict the multiaxial fatigue life. They conducted a set of multiaxial
Fatemi and Socie [6] modified Brown and Miller’s critical plane experiments with various paths and waveforms using CuZn37 specimens
approach and developed a model to predict fatigue life under and used the models proposed by Ellyin [14–16] and Gruad [17] to

* Corresponding author.
E-mail address: khonsari@[Link] (M.M. Khonsari).

[Link]
Received 5 March 2021; Received in revised form 7 May 2021; Accepted 10 May 2021
Available online 13 May 2021
0142-1123/© 2021 Elsevier Ltd. All rights reserved.
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

predict fatigue life using the plastic strain energy. Skibicki et al. reported Fig. 2 shows a typical thermographic image of a specimen under cyclic
that their approach can successfully predict the experimental results loading captured by an infra-red (IR) camera. The surface specimens
under in-phase and under various strain paths for out-of-phase loading. were sprayed with a thin layer of black color to increase thermal emis­
We now turn our attention to the recent development on the ther­ sivity and reduce IR reflection.
modynamic framework of fracture fatigue entropy (FFE), which is a Fig. 3 illustrates the strain pattern and sinusoidal waveforms of a
unique material parameter. FFE is the cumulative thermodynamic en­ series of experiments performed in this study. Specimens were subjected
tropy generated in a material under cyclic loading, beyond which frac­ to four different strain paths, including pure tension–compression (T-C),
ture occurs. Research shows that FFE is independent of the specimen’s pure torsion, in-phase (proportional), and 90◦ out-of-phase(non-
geometry, applied load, and frequency [18–20]. While the application of proportional) under different equivalent strain amplitudes.
the FFE framework to uniaxial tension–compression, torsion, and
bending has been successfully reported [18,20–23], to date, its efficacy 3. Theory
for treating general multiaxial fatigue has not been demonstrated. To
fulfill this gap, in this paper, we report the results of series of multiaxial 3.1. Plastic strain energy by the hysteresis loop
fatigue experiments with low carbon steel (LCS) 1018 and test the val­
idity of the thermodynamic framework for their prediction of fatigue In order to compare the results of different loading conditions in
life. The effect of loading type on the fatigue life is studied by incorpo­ fatigue experiment, an equivalent stress/strain, σeq /εeq , should be
rating different loading paths composed of tension–compression, tor­ introduced. Here, different combinations of uniaxial and torsion are
sion, in-phase, and out-of-phase loadings. The plastic strain energy is used as multiaxial fatigue loading conditions. A scheme introduced by
calculated using the thermography approach and the hysteresis is Shamsaei et al. [27] is incorporated to define σeq /εeq is as follows:
measured via a biaxial extensometer. The FFE values of the specimens √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
are evaluated from the plastic strain energy derived from both hysteresis σ eq = σx 2 + 3τxy 2 (1)
loops and thermography, and results are reported. The theoretical pre­
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
dictions are then compared versus the experiments, and the underlying 1 3
mechanisms of the observed deviations are described. εeq = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 2εx 2 (1 + ϑeq )2 + γ xy 2 (2)
2(1 + ϑeq ) 2

2. Experimental procedure where σx and τxy are the applied normal stress and shear stress
induced by uniaxial and torsion loadings, respectively. εx and γ xy are the
Fig. 1 shows the dimensions and schematic of hollow tubular spec­ applied normal and shear strains, respectively, and ϑeq is the equivalent
imens made of LCS 1018 designed according to the ASTM E2207-15 for Poisson’s ratio, which can be obtained as follow:
strain-controlled axial–torsional fatigue. The inner surface of specimens
ϑe εe + ϑe εp
are honed, and the outer surface of their gage section is polished up to ϑeq = (3)
Ra = 0.2μm using the silicon carbide papers with grit Numbers 400, 600, ε
800,1000, 1500, and 2000, respectively. where εe , εp and ε represent the elastic, plastic and the total axial
The physical, mechanical, and thermal properties of LCS 1018 used strains, and ϑe and ϑp are elastic and plastic Poisson’s ratios, respec­
in this investigation are presented in Table 1. Also, the chemical com­ tively. In the case of LCS 1018, the elastic Poisson’s ratio is ≈ 0.29, and
positions of the material are given in Table 2. Based on the published the plastic Poisson’s ratio is commonly considered as 0.5, i.e., perfectly
result, the average value of FFE for this material is 22.2 mMJ
3 K [20,24–26]. plastic behavior [28]. The axial and shear strain values are directly
Constant-amplitude, strain-controlled cyclic tests at the frequency of measured using a biaxial extensometer.
1 Hz are performed using a servo-hydraulic fatigue tester with the In the case of pure axial or pure torsion experiments, the plastic
capability of a maximum of ±50 kN axial force, ±2 kN.m torque, and a strain energy is equal to the area confined by the hysteresis loop of
maximum frequency of 75 Hz (Fig. 2). An axial–torsional extensometer uniaxial ΔWaxial
P or pure torsion ΔWTorsional
P .
with the axial gauge length of 25 mm and accuracy of ±10% axial strain In the case of combined uniaxial and torsion loading, both hysteresis
and ±2 torsional shear strain angle is used to record the applied strain loops of axial and shear are recorded. Accordingly, the plastic strain
◦

to the specimen under cyclic loading. The specimen surface temperature energy of a hysteresis loop, ΔWHY P , is calculated as follows:
during the fatigue tests was monitored and recorded using a Mikron
7500 IR camera with the resolution of 320 × 240 pixels, and the accu­ ΔW HY axial
P = ΔW P + ΔW Torsional
P (4)
racy of ±2% of temperature reading in the range of − 40 ◦ C to 500 ◦ C. The last step is to define the in-phase and out-of-phase loading used
here to assess the multiaxial fatigue behavior. In the case of in-phase
loading, the specimen experiences the maximum and minimum values
of axial strain and shear strain at the same time. In the case of out-of-

Table 2
The chemical composition of low carbon steel 1018.
Fig. 1. Dimensions and schematic illustrations of hollow tubular specimens C Cr Mn Mo Si S Ni N P Fe
made of low carbon steel 1018 designed according to the ASTM E2207-15 (all
0.15–0.2 – 0.6–0.9 – – 0.5 – – 0.04 Balance
the units are millimeter).

Table 1
The physical, mechanical, and thermal properties of low carbon steel 1018.
Physical properties Mechanical properties Thermal properties

Density Elastic modulus Shear modulus Ultimate tensile strength Yield tensile strength Specific heat capacity Thermal conductivity

kg/m3 MPa GPa MPa MPa J/kg K W/m K

7912 205 ± 5 80 493 ± 5 370 ± 5 486 51.9

2
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

Fig. 2. Hydraulic Machine and thermographic image.

Fig. 3. Strain patterns and sinusoidal waveform used in this investigation including (a) pure tension–compression, (b) pure torsion, (c) in-phase, and (d)90 out-
◦

of-phase.

steady-state and measuring the rate at which it begins to cool upon a

sudden interruption of the fatigue test. Fig. 5 shows the evolution of a
typical temperature graph of a specimen under fatigue loading as a
function of time. Based on Meneghetti’s method, the cooling gradient of
temperature, dT *
dt , at t = t (the time that the experiment is suddenly
stopped) can be used to calculate the plastic strain energy per cycle as
follow:
ρcp dT
ΔW TG
P = − |t = t* (5)
f dt
where f is the applied frequency, and ρ and cp represent the density
and specific heat capacity, respectively. This procedure requires stop­
ping the experiment for each specimen tested at different loading con­
ditions. Mehdizadeh et al. [21] showed that the cooling curve of
Fig. 4. A typical schematic of the temperature evolution of a component under
materials under fatigue loading is a unique curve for a specific geometry
cyclic loading. and can be obtained via a single test. They performed an extensive set of
bending and tension–compression fatigue tests and showed the cooling
phase loading, however, due to the phase difference between axial and
shear strains, the maximum and minimum values of axial and shear
strains do not coincide.

3.2. Plastic strain energy by thermography

Generally, the temperature evolution of a component under cyclic

loading consists of three different phases (Fig. 4). In the first phase, the
specimen experiences a rapid temperature rise due to the formation of
dislocations. Then, in the second phase, the temperature becomes nearly
constant. Note that the specimen spends most of its life in this phase. In
the third phase, microcracks coalesce into a rapidly propagating mac­
rocrack(s) that causes a sharp rise in temperature and specimen fractures
shortly thereafter.
To obtain the plastic strain energy value, Meneghetti [29,30] sug­ Fig. 5. Schematic of temperature evolution of specimen under fatigue loading
gested a thermographic technique by allowing the specimen to attain including a sudden stop at.t = t *

3
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

and plastic deformation—which is damaging and causes fatigue degra­

dation—as well as heat dissipated due to the internal friction mecha­
nism, which is non-damaging and must be excluded from total heat
dissipation. This stems from the fact that internal friction includes
reversible deformations such as the reversible motion of atoms and
dislocations. Research shows that contribution of internal friction can be
significant, particularly in HCF regime [25]. Niblett and Wilks [36]
showed that dislocations could move easier in the slip plane. Thereby,
the greater portion of dislocation moves in the case of torsional loading
compared to uniaxial and consequently the internal friction increases,
leading to higher temperature rise and heat dissipation [36]. Reported
experimental results show that this behavior varies based on the mate­
rials tested [12,37–42].
Based on the concept of fracture fatigue entropy introduced by
Fig. 6. Schematic of temperature evolution for different tests with the steady-
Naderi et al. [19], the accumulated entropy generation of a pristine
state temperature of Ts1 , Ts2 and.Ts3 component under fatigue loading from the onset of fatigue process at t =
0 up to the failure at t = tf is a material parameter, and can be used as a
curve for specific geometry is independent of loading conditions. In criterion to predict fatigue life [19,25,43–45]. The FFE can be obtained
other words, no matter what the steady-state temperature (Ts ) is, if the from Eq. (9).
test is stopped, the cooling curve follows a unique curve independent of ∫ tf
Ẇ p
an applied loading condition (Fig. 6). Therefore, the cooling gradient of FFE = ( )dt (9)
t=0 T
temperature at any steady-state can be obtained from this unique curve
by performing a single experiment. Accordingly, the plastic strain en­ Eq. (9) can be rewritten in terms of the number of cycles, N, instead
ergy can be obtained in-situ for any maximum steady-state temperature of time as follow:
without any need to interrupt the experiment. Note that the maximum ∫ Nf ( HY
)
ΔẆ P
steady-state temperate occurs in the middle of the gauge section. FFE = dt (10)
N=0 T

where ΔWHY P denotes for the hysteresis plastic energy per cycle.
3.3. Fatigue life prediction
When the specimen surface temperature becomes steady, (T = Ts ), the
plastic strain energy obtained by each hysteresis loop becomes a con­
In this section, a model is developed to predict multiaxial fatigue life
stant value. Therefore, Eq. (10) can be rewritten as follow:
according to the entropy generation during cyclic loading. Based on the
second law of thermodynamics, the rate of entropy generation, Ṡi , for Nf ΔWPHY Nf ΔW TH
FFE = = P
(11)
dissipative processes is always equal to or greater than zero. The rate of Ts Ts
entropy generation in the case of plastic deformation of a specimen
where ΔWTG represents the plastic strain energy obtained via
under cyclic loading can be defined as follows [31,32]: P
thermography.
σ : ε̇p → ̅̅→
q .gradT Ai V̇ i In very low-cycle fatigue experiments, the temperature in the second
Ṡi = − − ≥0 (6) phase may increase due to the higher applied stress compared to the
T T2 T
high-cycle fatigue (Fig. 7). Therefore, the plastic strain energy per cycle
where σ and ε̇p are the stress tensor and the corresponding rate of the
is increasing as well. Note that the specimen experiences most of its
plastic strain tensor, T represents the absolute temperature, q stands for
fatigue life in the second phase. So, to accurately obtain the entropy
heat flux, V̇ i denotes the internal state variables such as hardening and generation, this phase is divided into n segments, and the FFE will be
Ai is the conjugate thermodynamic force related to V̇ i . Eq. (6) can be obtained as an approximation in the sense of quasi-steady as follow.
rewritten in the form of heat dissipation as follows:
Nf
∑ W TH
→ ̅̅→
q .gradT FFE = Pi
(12)
T Ṡi = σ : ε̇p − − Ai V̇ i ≥ 0 (7) i=1
Tsi
T
This equation shows the relation between entropy generation of a
thermodynamic system and heat dissipation. All the terms on the right-
hand side of Eq. (7) are in the form of heat dissipation consisting of three
different mechanisms: dissipations due to the plastic deformation, Ẇp =
→ ̅̅→
q .gradT
σ : ε̇p ; heat conduction, T
; and the internal variable. In the case of
uniaxial fatigue experiments, it has been shown that the generated heat
due to the last term of this equation, Ai V̇ i , is only about 5–10% of the
entropy generation due to the plastic strain energy term and can be
neglected [33,34]. Furthermore, the heat dissipation by the conduction
mechanism is commonly neglected compared to heat dissipation due to
the plastic deformation in the case of a low-cycle fatigue regime [20,35].
Therefore, Eq. (6) can be simplified as follows:

Ẇ p
Ṡi = (8)
T
According to Mehdizadeh et al. [22,25], heat dissipation of a spec­
imen subjected to cyclic loading is due to the formation of microcracks Fig. 7. The temperature evolution of a component in very low cycle fatigue.

4
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

where Ts i and WPTH i are the specimen temperature and plastic energy
obtained using thermography for ith cycle, respectively. WPTH i can be
obtained as follows:
ρcp dT i
TH
WPi = − |N = Ni (13)
f dti

The average values of Ts i and WPTH , i.e., Ts and ΔWTH

P , respectively,
can be used to further simplify Eq. (12) as follows:

Nf ΔWPHY Nf ΔW TH
FFE = = P
(14)
Ts Ts

4. Results and discussions

The multiaxial fatigue behavior of LCS 1018 is investigated here by

applying different loading conditions summarized in Table 3. The
equivalent strain and stress amplitudes at the mid-life cycle are calcu­
lated for each loading condition using Eqs. (1, 2) and shown in Fig. 8.
Fig. 8. Comparison of cyclic deformation behavior of LCS under different
The results show that the cyclic hardening behavior of in-phase, T-C, and
equivalent strain.
torsion loadings are similar. The fatigue response of the specimen during
torsion is the lower bound of the in-phase loading, while the T-C
dT
response is the upper bounds of the equivalent stress-equivalent strain of gradient, dt , and consequently using Eq. (5), the plastic strain ener­
in-phase loading. Any in-phase loading path as a combination of T-C and gy,ΔWTH P , was calculated for different loading paths, including torsion,
torsion will be located between these two curves. Fig. 8 shows that the T-C, in-phase, and out-of-phase loading paths. Next, the plastic strain
out-of-phase loadings exhibit higher equivalent stress at the same energy was numerically calculated by measuring the area contained
equivalent strain. In other words, the material shows higher hardening within the hysteresis loop.
in the case of the out-of-phase loading. The stabilized hysteresis loops were recorded in the mid-life for the
The non-proportional hardening coefficient of each loading condi­ same equivalent strain of εeq = 0.008 under various strain paths, and the
tion, α, is then calculated. α is defined as the ratio of the equivalent stress results are presented in Fig. 10. The ΔWPHY values for T-C, torsion, in-
amplitude for a 90 out-of-phase experiment, σeq OP , to the equivalent phase, and out-of-phase are 0.38 MJ/m3 cycle ,0.19 MJ/m3 cycle, 0.22
◦

stress amplitude of in-phase experiment, σeq IP , at the same strain. Eq. MJ/m3 cycle, and 1.61 MJ/m3 cycle, respectively. It can be observed that
(15) gives the formulation [46]. the 90 out-of-phase loading exhibit the highest plastic strain energy at a
◦

σ eq OP similar equivalent strain. A high plastic strain energy leads to high en­
α= − 1 (15) ergy dissipation and temperature rise, and consequently, the fatigue life
σeq IP
is reduced. A high ΔWPHY can be attributed to high equivalent stress
In Eq. (15), σeq OP and σeq IP are obtained at mid-life, where the stress (Fig. 8) and the rounded shape of the hysteresis loop in the out-of-plane
values become relatively stable. The values of α for the equivalent loading. These can be compared to the sharper shape of in-phase, T-C,
strains of 0.007, 0.008, and 0.009 are 0.1, 0.16, and 0.2, respectively. and torsion loading types (Fig. 10). The different shapes of the out-of-
The results show that as the equivalent strain increases, so does the non- phase loading compared to in-phase loading types can be attributed to
proportional hardening coefficient. the difference in the nature of their loading patterns. Both the in-phase
As discussed earlier, the plastic strain energy for every steady-state and out-of-phase may have a similar equivalent strain amplitude. In the
temperature can be obtained from the unique cooling curve [21]. case of in-phase loading, the maximum T-C and torsional strains occur
Accordingly, the specimen was allowed to reach a steady-state temper­ simultaneously at the maximum equivalent strain, while in the case of
ature and then the experiment was paused to capture the cooling trend. out-of-phase loading, the maximum of T-C and torsion strains occur at
Fig. 9 shows the variation of temperature as a function of time under 90 different strain values, which may be different from the maximum
◦

out-of-phase experiment with εeq = 0.009. Using this graph, the cooling equivalent strain amplitude.
Table 4 summarizes the equivalent stress and strain amplitudes,

Table 3
T-C, torsion, in-phase and 90 out-of-phase fatigue results.
◦

Test Number Loading Path εeq f (Hz)

I T-C 0.007 1
II T-C 0.008 1
III T-C 0.009 1
IV T-C 0.010 1
V Torsion 0.007 1
VI Torsion 0.008 1
VII Torsion 0.010 1
VIII Torsion 0.012 1
IX IP 0.007 1
X IP 0.008 1
XI IP 0.009 1
XII IP 0.010 1
XIII OP 0.006 1
IXV OP 0.007 1
XV OP 0.008 1
Fig. 9. The temperature variation of the specimen’s surface under 90 out-of-
◦

XVI OP 0.009 1
phase experiment with.εeq = 0.009

5
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

Fig. 10. Mid-life stress–strain hysteresis loops under (a) pure axial, (b) pure torsional, (c) in-phase and (d) 90 out-of-phase loading.
◦

Table 4
T-C, torsion, in-phase, and 90 out-of-phase fatigue results.
◦

Test Number Load Path εeq σeq ΔT ΔWHY ΔWTH 2Nf

P P
- MPa ◦
C 3
MJ/m cycle 3
MJ/m cycle cycle

I T-C 0.007 405 1.0 0.095 0.104 124,256

II T-C 0.008 432 3.8 0.38 0.36 37,424
III T-C 0.009 449 6.2 0.93 0.84 13,520
IV T-C 0.010 461 14.7 1.87 1.80 6994
V Torsion 0.007 388 0.7 0.054 0.06 207,082
VI Torsion 0.008 411 1.3 0.19 0.18 63,846
VII Torsion 0.010 446 6.1 0.81 0.80 13,504
VIII Torsion 0.012 452 8.5 1.35 1.28 8234
IX IP 0.007 398 0.9 0.07 0.08 159,884
X IP 0.008 426 2.3 0.22 0.24 51,338
XI IP 0.009 444 4.5 0.67 0.64 19,448
XII IP 0.010 453 7.3 1.15 1.12 10,214
XIII OP 0.006 398 2.4 0.28 0.24 49,592
IXV OP 0.007 438 3.7 0.42 0.34 39,614
XV OP 0.008 496 8.6 1.61 1.24 8938
XVI OP 0.009 532 28.9 3.70 3.48 3238

6
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

applied frequency, specimen’s surface temperature rise, fatigue lives,

and plastic strain energy obtained by thermographic model and hys­
teresis loop for various loading paths. The number of cycles to failure for
different loading can be generally summarized as:
Nftorsion > Nfin− phase > NfT− C > Nfout− of − phase

The results show that the out-of-phase leads to minimum fatigue life,
much shorter than the other loading conditions. The torsion exhibits the
maximum life at a similar equivalent strain.
To investigate the effect of internal friction on LCS 1018, results of
the dissipated energies obtained by the thermographic method, ΔWTH
P ,
and hysteresis loop, ΔWHY P , are plotted as a function of the number of
cycles and shown in Fig. 11. It can be seen that both the uniaxial and
torsional dissipated energies have the same slope. This implies that the
internal friction on torsional and T-C loading has an equal effect on the
material loading fatigue damage, at least in LCF. Therefore, the effect of
internal friction is negligible in the current work.
The results also show that the plastic strain energy calculated from
the hysteresis loop, ΔWHY
P , is close to those obtained via thermography
ΔWTH HY TH Fig. 12. Plastic strain energy obtained by thermographic analysis as a function
P . The variation of ΔW P versus ΔW P is further investigated, as
shown in Fig. 12. Referring to this figure, in the case of T-C, torsion and of those obtained by the hysteresis loop. The blue dotted line presents the
ΔW TH
P ,=ΔW P
HY
and the black dotted lines illustrate 20% error. (For interpre­
in-phase loading conditions, ΔWHY TH
P ≈ ΔW P . In other words, T Ṡi =
TH p HY tation of the references to color in this figure legend, the reader is referred to
ΔWP ≈ σ : ε̇ = ΔWP , according to Eq. (7). This suggests that the the web version of this article.)
contribution of internal state variables stored in the material, such as
hardening, is negligible. In the case of out-of-phase loading, however,
versus the experimental fatigue life. Fig. 14 (a) shows that the maximum
the results show that T Ṡi = ΔWTH HY
P < ΔW P in the mid-life cycle. In other errors of the life predicted using the plastic strain energy of the hys­
words, the effect of the internal state variable, which is non-damaging,
teresis loop ΔWPHY occurs in the cases of 90 out-of-phase loading at the
◦

still exists in the mid-life and cannot be neglected in the case of out-of-
equivalent strain of 0.009 and torsional loading at the equivalent strain
phase loading for dissipated energy obtained via the hysteresis loop.
of 0.01, which are 21.2% and 23.5%, respectively. One can note that T-C
Therefore, the non-damaging portion of ΔWHYP , due to the internal state errors and in-phase loadings are less than 12.6% and 18.6%, respec­
variables, should be excluded from total plastic strain energy obtained tively. Fig. 14 (b) shows the comparison of the predicted fatigue life
by hysteresis loop. However, the effect of internal state variables is based on the proposed thermographic model versus the experimental
already considered by the thermographic method. The results show that results. The maximum errors occur in out-of-phase loading cases at the
the difference between the ΔWTH HY
P and ΔW P (on average 19% for the equivalent strain of 0.009 and torsion loading at the equivalent strain of
tested specimens in the current work) could be related to the effect of 0.012 that are 19.0% and 18.7%, respectively. The errors for T-C and in-
internal variables in entropy generation (Fig. 12). phase experiments are less than 10%.
Fig. 13 illustrates the experimental FFE of LCS 1018 as a function of As Fig. 14 shows that in the cases of in-phase, T-C, and torsion
equivalent strain and number of cycles to failure using the plastic strain loadings, the errors in fatigue life predictions obtained from both ΔWPTH
energy obtained via the thermographic method. As the results reveal,
and ΔWPHY are close. However, in the case of out-of-phase loadings, the
the obtained FFE in the current work is in the range of FFE reported in
thermographic plastic strain energy can clearly predict fatigue life more
the literature [20,24–26] for this material and also is limited in a small
accurately than the hysteresis loop method. To unravel the underlying
band independent of loading conditions such as strain amplitude,
mechanisms of this observation, one can focus on Eq. (7). The dissipated
loading type, loading path, etc.
energy, which is an accurate measure of damage in the FFE method, can
In the final step, the fatigue life of specimens subjected to different
be precisely obtained by the dissipated heat measured by thermography.
loading conditions is investigated using the FFE framework. Accord­
In the case of the hysteresis loop method, however, the ΔWPHY becomes
ingly, two different plastic energy dissipations, i.e., ΔWPTH and ΔWPHY ,
close to the dissipated energy only if the variation in internal state
are used to calculate the fatigue life. Fig. 14 shows the comparison of
variables are negligible, which is the case for any general in-phase
two different fatigue life predictions obtained from ΔWPTH and ΔWPHY loading, including T-C and torsion. In the case of out-of-phase

Fig. 11. Experimental results for (a) dissipated energy obtained by the thermographic method as a function of the number of cycles to failure (b) dissipated energy
obtained by hysteresis loop as a function of the number of cycles to failure.

7
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

Fig. 13. The experimental fracture fatigue entropy of LCS 1018 as a function of (a) equivalent strain amplitude (b) number of cycles to failure for different types of
loading paths.

Fig. 14. Experimental fatigue lives as a function of (a) fatigue life based on the thermographic model (b)fatigue lives based on the hysteresis loop (The black dotted
lines present 20% error and blue dotted lines are related to the error of 0.0%). (For interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)

loadings, however, as shown in Fig. 12, the plastic strain energy ob­ • The results showed that the out-of-phase leads to the minimum fa­
tained by the hysteresis loop is not entirely dissipated, i.e., TṠi = tigue life, while the torsion exhibited the maximum life at a similar
ΔWPTH < ΔWPHY . In other words, part of ΔWPHY is stored in the form of equivalent strain.
internal state variable change, including cyclic hardening, which is, on • The FFE is a constant independent of loading conditions such as
average 19% in the current work for the tested samples. strain amplitude, loading type, loading path, etc.
• The FFE prediction using thermography was successfully captured
5. Conclusions the fatigue life of the samples subjected to the multiaxial fatigue
loading independent of the loading condition.
In the present work, the multiaxial cyclic response of the LCS 1018 • The strain energy obtained from the hysteresis loop is shown to be a
was investigated using both theory and experiment. Tension- good measure of dissipation energy in the cases of ten­
Compression, torsion, in-phase, and out-of-phase loadings were sion–compression, torsion, and in-phase loading conditions. In these
applied to capture the multiaxial aspects of fatigue behavior. The frac­ cases, the average error is less than 6.8%.
ture fatigue entropy framework which was validated for simple loading • The predicted life of the sample subjected to tension–compression,
types, was tested here for complex multiaxial loading patterns. Two torsion, and in-phase loading conditions using the hysteresis loop
different methodologies—by thermography technique and by mea­ agreed well with the thermographic and experimental results.
surement of the area within the hysteresis loop—were employed to • In the case of out-of-phase loadings, the plastic strain energy ob­
obtain the plastic strain energy. The effect of the loading pattern was tained from the hysteresis loop deviated from the energy dissipation
investigated on these calculations. Finally, the fatigue life of LCS 1018 because the stored energy originated from the internal state variable
was predicted using the FFE framework along with both methodologies, variation.
and the obtained results were compared with the experimental results. • The predicted life of the specimen subjected to out-of-phase loading
The main results and conclusions of the work can be summarized as obtained via thermography was found to be more accurate than the
follows: hysteresis loop. The corresponding average errors are 8.2% for
thermography and 14.3% for hysteresis loop. This can be attributed
to the error introduced into the dissipated energy calculation due to
the stored energy originated from the internal state variables.

8
M. Mehdizadeh and M.M. Khonsari International Journal of Fatigue 150 (2021) 106321

Declaration of Competing Interest [20] Liakat M, Khonsari MM. Rapid estimation of fatigue entropy and toughness in
metals. Mater Des 2014;62:149–57.
[21] Mehdizadeh M, Haghshenas A, Khonsari M. In-situ Technique for Fatigue Life
The authors declare that they have no known competing financial Prediction of Metals Based on Temperature Evolution. Int J Mech Sci 2020;192:
interests or personal relationships that could have appeared to influence 106113.
the work reported in this paper. [22] Mehdizadeh M, Haghshenas A, Khonsari MM. On the effect of internal friction on
torsional and axial cyclic loading. Int J Fatigue 2021;145:106113. [Link]
10.1016/[Link].2020.106113.
References [23] Haghshenas, A., J. Jang, and M. Khonsari, On the intrinsic dissipation and fracture
fatigue entropy of metals. Mechanics of Materials: p. 103734.
[1] Lohr RD, Ellison EG. A simple theory for low cycle multiaxial fatigue. Fatigue Fract [24] Mehdizadeh M, Khonsari MM. On the application of fracture fatigue entropy to
Eng Mater Struct 1980;3(1):1–17. variable frequency and loading amplitude. Theor Appl Fract Mech 2018;98:30–7.
[2] Ellison EG, Andrews JMH. Biaxial cyclic high-strain fatigue of aluminum alloy [25] Mehdizadeh M, Khonsari MM. On the role of internal friction in low-and high-cycle
RR58. Journal of Strain Analysis 1973;8(3):209–19. fatigue. Int J Fatigue 2018;114:159–66.
[3] Shamsaei N, Fatemi A, Socie DF. Multiaxial fatigue evaluation using discriminating [26] Jirandehi AP, Mehdizadeh M, Khonsari MM. Temperature-induced buckling of
strain paths. Int J Fatigue 2011;33(4):597–609. ductile metals during cyclic loading and the subsequent early fracture. Int J Mech
[4] Wang, C. and M. Brown, Life prediction techniques for variable amplitude Sci 2020;176:105525. [Link]
multiaxial fatigue—part 1: theories. 1996. [27] Shamsaei N, Fatemi A. Effect of hardness on multiaxial fatigue behaviour and some
[5] Brown MW, Miller KJ. A theory for fatigue failure under multiaxial stress-strain simple approximations for steels. Fatigue Fract Eng Mater Struct 2009;32(8):
conditions. Proceedings of the Institution of Mechanical engineers 1973;187(1): 631–46.
745–55. [28] Shamsaei N, Fatemi A. Effect of microstructure and hardness on non-proportional
[6] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage cyclic hardening coefficient and predictions. Mater Sci Eng, A 2010;527(12):
including out-of-phase loading. Fatigue Fract Eng Mater Struct 1988;11(3):149–65. 3015–24.
[7] Meneghetti G, Ricotta M, Atzori B. A synthesis of the push-pull fatigue behaviour of [29] MENEGHETTI G. Analysis of the fatigue strength of a stainless steel based on the
plain and notched stainless steel specimens by using the specific heat loss. Fatigue energy dissipation. Int J Fatigue 2007;29(1):81–94.
Fract Eng Mater Struct 2013;36(12):1306–22. [30] Meneghetti G, Quaresimin M. Fatigue strength assessment of a short fiber
[8] Wang XG, Crupi V, Guo XL, Zhao YG. Quantitative Thermographic Methodology composite based on the specific heat dissipation. Compos B Eng 2011;42(2):
for fatigue assessment and stress measurement. Int J Fatigue 2010;32(12):1970–6. 217–25.
[9] Wang XG, Crupi V, Jiang C, Guglielmino E. Quantitative Thermographic [31] Lemaitre J, Chaboche J-L, editors. Mechanics of Solid Materials. Cambridge
Methodology for fatigue life assessment in a multiscale energy dissipation University Press; 1990.
framework. Int J Fatigue 2015;81:249–56. [32] Marigo JJ. Modelling of brittle and fatigue damage for elastic material by growth
[10] Hajshirmohammadi B, Khonsari M. A simple approach for predicting fatigue crack of microvoids. Eng Fract Mech 1985;21(4):861–74.
propagation rate based on thermography. Theor Appl Fract Mech 2020;107: [33] Halford G. The energy required for fatigue(Plastic strain hystersis energy required
102534. for fatigue in ferrous and nonferrous metals). Journal of materials 1966;1:3–18.
[11] Feng ES, Wang XG, Jiang C. A new multiaxial fatigue model for life prediction [34] Lemaitre, J. and J.-L. Chaboche, Mechanics of solid materials. 1994: Cambridge
based on energy dissipation evaluation. Int J Fatigue 2019;122:1–8. university press.
[12] Skibicki D, Pejkowski Ł. Low-cycle multiaxial fatigue behaviour and fatigue life [35] Hajshirmohammadi B, Khonsari M. On the entropy of fatigue crack propagation.
prediction for CuZn37 brass using the stress-strain models. Int J Fatigue 2017;102: Int J Fatigue 2020;133:105413.
18–36. [36] Niblett DH, Wilks J. Dislocation damping in metals. Adv Phys 1960;9(33):1–88.
[13] Skibicki D, Lipski A, Pejkowski Ł. Evaluation of plastic strain work and multiaxial [37] Lee B, Kim K, Nam K. Fatigue analysis under variable amplitude loading using an
fatigue life in CuZn37 alloy by means of thermography method and energy-based energy parameter. Int J Fatigue 2003;25(7):621–31.
approaches of Ellyin and Garud. Fatigue Fract Eng Mater Struct 2018;41(12): [38] Sugiura K, Chang KC, Lee GC. Evaluation of low-cycle fatigue strength of structural
2541–56. metals. J Eng Mech 1991;117(10):2373–83.
[14] Ellyin, F., Fatigue damage, crack growth and life prediction. 2012: Springer [39] SHAHROOI S, METSELAAR IH, HUDA Z. Evaluating a strain energy fatigue method
Science & Business Media. using cyclic plasticity models. Fatigue Fract Eng Mater Struct 2010;33(8):530–7.
[15] Ellyin F, Golos K, Xia Z. In-Phase and Out-of-Phase Multiaxial Fatigue. ASME. J. [40] HAN C, CHEN X, KIM K. Evaluation of multiaxial fatigue criteria under irregular
Eng. Mater. Technol. January 1991;113(1):112–8. [Link] loading. Int J Fatigue 2002;24(9):913–22.
1.2903365. [41] Glinka G, Shen G, Plumtree A. A multiaxial fatigue strain energy density parameter
[16] Ellyin F, Kujawski D. Plastic Strain Energy in Fatigue Failure. ASME. J. Pressure related to the critical fracture plane. Fatigue Fract Eng Mater Struct 1995;18(1):
Vessel Technol. November 1984;106(4):342–7. [Link] 37–46.
1.3264362. [42] Liu, K., A method based on virtual strain-energy parameters for multiaxial fatigue
[17] Garud YS. A New Approach to the Evaluation of Fatigue Under Multiaxial Loadings. life prediction, in Advances in multiaxial fatigue. 1993, ASTM International.
ASME. J. Eng. Mater. Technol. April 1981;103(2):118–25. [43] Liakat M, Khonsari MM. On the anelasticity and fatigue fracture entropy in high-
[18] Jirandehi AP, Khonsari M. Microstructure-sensitive estimation of fatigue life using cycle metal fatigue. Mater Des 2015;82:18–27.
cyclic thermodynamic entropy as an index for metals. Theor Appl Fract Mech 2020; [44] Haghshenas A, Khonsari MM. Non-destructive testing and fatigue life prediction at
112:102854. different environmental temperatures. Infrared Phys Technol 2019;96:291–7.
[19] Naderi M, Amiri M, Khonsari MM. On the thermodynamic entropy of fatigue [45] Jang JY, Khonsari MM. On the evaluation of fracture fatigue entropy. Theor Appl
fracture. Proceedings of the Royal Society A: Mathematical, Physical and Fract Mech 2018;96:351–61.
Engineering Sciences 2010;466(2114):423–38. [46] Socie, D.F. and G.B. Marquis, Multiaxial fatigue. 2000: Society of Automotive
Engineers Warrendale, PA.