materials

Article
An Energy-Based Unified Approach to Predict the
Low-Cycle Fatigue Life of Type 316L Stainless Steel
under Various Temperatures and Strain-Rates
Nae Hyung Tak 1 , Jung-Seok Kim 2 and Jae-Yong Lim 3, *
1 Division of Industrial Metrology, Korea Research Institute of Standards and Science, Daejeon 34113, Korea;
nhtak@kriss.re.kr
2 Autonomous Vehicles Research Team, New Transportation Innovative Research Center, Korea Railroad
Research Institute, Uiwang-si 16105, Korea; jskim@krri.re.kr
3 Department of Safety Engineering, Seoul National University of Science and Technology, Seoul 01811, Korea
* Correspondence: jaeyong.lim@gmail.com; Tel.: +82-2-970-6388




Received: 10 March 2019; Accepted: 29 March 2019; Published: 2 April 2019


Abstract: An energy-based low-cycle fatigue model was proposed for applications at a range of
temperatures. An existing model was extended to the integrated approach, incorporating the
simultaneous effects of strain rate and temperature. A favored material at high temperature, type 316L
stainless steel, was selected in this study and its material characteristics were investigated. Tensile
tests and low-cycle fatigue tests were performed using several strain rates at a temperature ranging
from room temperature to 650 ◦ C. Material properties were obtained in terms of temperature using
the displacement-controlled tensile tests and further material response were investigated using
strain-controlled tensile tests. Consequently, no pronounced reduction in strengths occurred at
temperatures between 300 and 550 ◦ C, and a negative strain rate response was observed in the
temperature range. Based on the low-cycle fatigue tests by varying strain rates and temperature,
it was found that a normalized plastic strain energy density and a strain-rate modified cycle were
successfully correlated. The accuracy of the model was discussed by comparing between predicted
and experimental lives.

Keywords: type 316L stainless steel; low-cycle fatigue; fatigue life prediction; plastic strain energy
density; elevated temperature

1. Introduction
Reliable lifetime assessment is of primary concern in structures or mechanical components,
most importantly in those whose failure results in catastrophic disaster. Because inaccurate or
non-conservative estimations can cause significant losses in human lives and properties, the accuracy
of predicted lives must be assured in the design stage and in the remaining useful lifetime assessment
of the structures. Accuracy can be achieved on an integral basis of extensive explorations into material
response, of external loading analysis, and of guaranteed life prediction models [1–3].
Many mechanical components in power facilities or nuclear reactors must withstand repeated
loading at elevated temperatures and low-cycle fatigue (LCF) is considered as a predominant failure
mechanism. For example, turbine blades, disks, and boiler components in plants are subjected to LCF at
elevated temperatures. In those applications, the mechanical components can experience a wide range
of operating temperatures. Specifically, the components are subject to temperature gradient induced
cyclic thermal stresses owing to startups and shutdowns or the temperature change of a coolant that is
either a liquid, such as sodium, or a gas, such as helium, that flows in the pipe of a cooling system [4,5].
Moreover, when exposed to a temperature greater than the creep-activation temperature, materials

Materials 2019, 12, 1090; doi:10.3390/ma12071090 www.mdpi.com/journal/materials

Materials 2019, 12, 1090 2 of 13
Materials 2019, 12, x FOR PEER REVIEW 2 of 13

temperature,
of which the componentsmaterials are of madewhichcanthe components
experience complicatedare made can experience
microstructural changes, complicated
resulting in
microstructural
appearance changes,
of various failure resulting in appearance
mechanisms. Thus, fatigue of various failure mechanisms.
life prediction Thus, fatigue
at elevated temperature canlife
be
prediction at elevated
a challenging problem. temperature can be a challenging problem.
AA long
long history
history of of studies
studies existsexists toto discover
discover the the effects
effects associated
associated with with LCFLCF behavior
behavior at at elevated
elevated
temperatures: Temperature,
temperatures: Temperature, temperaturetemperature cycling,cycling, strain
strain rate,
rate, hold
hold time,
time, load
load sequence,
sequence, etc. etc. [6–10].
[6–10].
Each previous study concentrated on identifying and quantifying
Each previous study concentrated on identifying and quantifying the individual effects on fatigue life. the individual effects on fatigue
life. However,
However, a more a more
convincingconvincing
lifetime lifetime calculation
calculation at the stage
at the design designand stage
the and the remaining
remaining useful
useful lifetime
lifetime assessments
assessments requires requires a more inclusive
a more inclusive approach approach
that embracesthat embraces
a few ofathe fewabove-mentioned
of the above-mentioned effects
effects because those are involved
because those are involved simultaneously. simultaneously.
Two representative
Two representative parameters parameters describedescribethe theLCF LCFcharacteristics:
characteristics: Plastic strain
Plastic amplitude
strain amplitude and
plastic strain energy density. For example, the Coffin-Manson
and plastic strain energy density. For example, the Coffin-Manson model [11] assumes a linear model [11] assumes a linear
relationship between
relationship between the the plastic
plastic strain
strain amplitude
amplitude and and fatigue
fatigue life
life in
in the
the logarithmic
logarithmic coordinate
coordinate and and
the Morrow
the Morrow model model [12] [12] relates
relates aa plastic
plastic strain
strain energy
energy density
density and and fatigue
fatigue lifelife in
in the
the logarithmic
logarithmic
coordinate. Subsequently, the strain rate effect on fatigue life
coordinate. Subsequently, the strain rate effect on fatigue life was incorporated by developing was incorporated by developing the
frequency-modified parameters in the frequency-modified
the frequency-modified parameters in the frequency-modified Coffin-Manson model [13] and the Coffin-Manson model [13] and the
frequency-modified energy
frequency-modified energy model model [14],[14], respectively.
respectively. The The Coffin-Manson
Coffin-Manson model model and and thethe Morrow
Morrow
energy model can predict well under the isothermal conditions,
energy model can predict well under the isothermal conditions, but they cannot explain the effect but they cannot explain the effect of of
strain rate
strain rate onon fatigue
fatigue life.
life. To To consider
consider the the effect
effect ofof strain
strain rate
rate on
on fatigue
fatigue life,
life, the
the frequency-modified
frequency-modified
Coffin-Manson model or the frequency-modified
Coffin-Manson model or the frequency-modified energy model should be used. Even energy model should be used. Even though
though thesethese
two models can explain well the effects of strain rate and strain amplitude
two models can explain well the effects of strain rate and strain amplitude on fatigue life, they cannot on fatigue life, they cannot
account for
account for the
the effect
effect of of temperature.
temperature. They They onlyonly represent
represent how how thethe material
material constants
constants change
change with with
respect to the test temperatures. Therefore, it is better to improve
respect to the test temperatures. Therefore, it is better to improve and modify the existing models, and modify the existing models,
including the
including the Coffin-Manson
Coffin-Mansonmodel model and and thethe Morrow
Morrow model model mentioned
mentioned above above [11–14],
[11–14], to to account
account for for
the temperature
the temperaturedependency
dependencyof offatigue
fatiguelife.life.
In this study, type 316L stainless steel isisselected
In this study, type 316L stainless steel selectedto topropose
proposeaaphenomenological
phenomenological energy-basedenergy-based
life prediction
life predictionmodel modelincorporating
incorporatingstrain strainraterateandandtemperature
temperatureeffects. effects.TheThe material
material has has been
been favored
favored
for power facilities or nuclear fusion reactors for its resistance at
for power facilities or nuclear fusion reactors for its resistance at high temperatures. In this regard,high temperatures. In this regard,
tensile tests and strain-controlled LCF tests were conducted against
tensile tests and strain-controlled LCF tests were conducted against various temperatures and strain various temperatures and strain
rates. The
rates. Thetemperatures
temperaturesranges rangesfrom from roomroom temperature
temperature (RT) (RT) to 650
to 650 °C, and
◦ C, and the strain
the strain rate israte is varied
varied from
from 1
− ×
5 10 −5 to 1 ×−102 −2
− s
1 −1 to investigate the influence of strain rate and temperature on the material
1 × 10 to 1 × 10 s to investigate the influence of strain rate and temperature on the material
behaviorand
behavior and to to measure
measure the the mechanical
mechanical properties.
properties. Subsequently,
Subsequently,the thecapabilities
capabilitiesand andvalidity
validityof of the
the
developed model
developed model are arediscussed.
discussed.

2. Materials
2. Materials and
and Methods
Methods

2.1.
2.1. Material
Material and
and Specimen
Specimen
All
All specimens were
were made
madefrom froman anextruded
extruded316L 316L stainless
stainless steel
steel barbar of diameter
of diameter 16 mm
16 mm that that
was
was solution-treated at 1100 ◦
solution-treated at 1100 °C forC40formin
40 min
and and subsequently
subsequently cold-drawn
cold-drawn by 17%
by 17% coldcold
workwork (%CW).
(%CW). The
The chemical
chemical composition
composition of the
of the steel
steel is shown
is shown in in Table1 1and
Table andaverage
averagegrain
grain size
size determined
determined by linear
linear
intercept
interceptmethod
methodwas was53 μm. A
53µm. A type
type of
of specimen
specimen designed
designed by by ASTM
ASTM code
code E606-92
E606-92 [15]
[15] was
was employed
employed
for
for tensile
tensile tests
tests and
and LCF
LCF tests
tests as
as shown
shown inin Figure
Figure 1,1, and
and the
the specimen
specimen surface
surface was
was polished
polished along
along the
the
longitudinal
longitudinal direction
directionwith
withemery
emerypaper
paperup uptoto13 μmto
13µm toremove
removesurface
surfacedefects.
defects.

28.00 36.00 28.00

6.00 6.00

M12*1.75mm 8

Figure 1. Specimen for tensile and low-cycle fatigue (LCF) test (dimensions in mm).
Figure 1. Specimen for tensile and low-cycle fatigue (LCF) test (dimensions in mm).
Materials 2019, 12, 1090 3 of 13

Table 1. The chemical composition of 316L stainless steel (wt %).

C Si Mn P S Ni Cr Mo Cu N
0.025 0.41 1.41 0.025 0.025 10.22 16.16 2.09 – 0.043

2.2. Test Equipment

A closed-loop servo-hydraulic test system with a 5-ton capacity manufactured by MTS
(Eden Prairie, MN, USA) was used to perform the tensile tests and LCF tests. A resistance type
furnace that can control the ambient temperature with the variation of ±1 ◦ C at steady state was used
for temperature control. A specimen was fixed on the test machine using a hydraulic grip system.
A high temperature extensometer manufactured by MTS (Model No.: 632-13F-20, guage length: 25 mm)
was used to perform the strain control tests and the displacement signal, load signal, and strain signal
were recorded for each cycle during the test.

2.3. Tesile Test

Tensile tests were performed in two different control manners: Displacement-control and
strain-control. The tests were primarily performed using displacement-controlled tensile tests to
obtain the mechanical properties depending on temperature. Additionally, strain-controlled tests,
when further investigations were required, were subsequently performed to capture a particular
mechanical response in the considered range of strain rate (1 × 10−3 , 1 × 10−4 or 1 × 10−5 s−1 ).
In the displacement-control tensile tests, the actuator is displaced at 2 mm/min under isothermal
condition ranging from RT to 650 ◦ C. Prior to loading, the material to be tested was metallurgically
stabilized by maintaining the specimen at a test temperature for an hour. From the results, tensile
properties such as E, σy , σu , %EL (percent elongation) were extracted.
For the strain-controlled tensile tests, specimens were prepared similarly as in the
displacement-control test. Subsequently, they were stretched with a strain rate controlled by an
extensometer. Because the extensometer used in the tests has a limited travel length with −5% ~
+10%, tensile tests using strain control were stopped at 5% strain value and stress relaxation tests were
proceeded by maintaining this strain value.

2.4. LCF Test

In order to investigate the effects of temperature and strain amplitude on fatigue life, LCF tests
were performed using strain control. All tests were performed in air under a fully reversed total strain
control mode employing a triangular waveform. A specimen was maintained at the test temperature
for an hour before the fatigue test started while unloaded.
First, extensive fatigue tests were performed to investigate the fatigue characteristics over a
wide range of temperature, from RT to 650 ◦ C. A fully reversed strain cycle using a strain rate of
1×10−3 s−1 was imposed under a carefully controlled temperature environment, and the considered
strain amplitudes were 0.3%, 0.4%, 0.5%, 0.6%, 0.7%, 0.8% to construct a fatigue life curve for a given
temperature. Likewise, the LCF tests were repeated to obtain the fatigue characteristics at other
temperatures. Next, investigations to observe the effect of strain rate on fatigue life were performed.
Here, the strain rates were varied with a fixed strain amplitude of 0.5%.
In this study, the fatigue life was defined as the 70% load drop point of a stabilized load amplitude
at Nf /2 cycles because a few cycles are required from the 70% load drop point to reach the final failure.
The plastic strain amplitude, ∆εp /2 was determined from a stress-strain hysteresis loop as specified in
BS 7270 (or ISO/DIS) [16], which is defined as the distance between the intersections with the strain
axis at the stress-strain hysteresis loop. For each test condition, two to four specimens were used on
account of inherent scatter. The fatigue life for each specimen was recorded and the average value was
regarded as the fatigue life.
Materials 2019, 12, 1090 4 of 13

3. Results and Discussion

Materials 2019, 12, x FOR PEER REVIEW 4 of 13
3.1. Tensile Test Result
3.1. Tensile Test Result
3.1.1. Test by Displacement Control
3.1.1.
The Test by Displacement
mechanical propertiesControlof type 316L stainless steel were obtained from tensile tests using the
displacement The control
mechanical
of 2properties
mm/minofattype each316Lteststainless steel were
temperature. obtained
The elasticfrom tensilewas
modulus testsdetermined
using the by
displacement
measuring the slopecontrol of 2 mm/mincurve
of stress-strain at eachduring
test temperature.
unloadingThe in elastic modulus
the elastic was and
region determined
the yieldby stress
measuring
was done by thethe slope
0.2% of stress-strain
offset method. curve during unloading in the elastic region and the yield stress
was done by the 0.2% offset method.
The mechanical properties of type 316L stainless steel are summarized in Table 2. In Figure 2,
The mechanical properties of type 316L stainless steel are summarized in Table 2. In Figure 2,
the material strength (σy ,(σ
the material strength
σyu, )σagainst temperature is plotted, and the ductility change is superimposed
u) against temperature is plotted, and the ductility change is superimposed
to clarify the effect
to clarify of temperature.
the effect of temperature. AsAsexpected,
expected, the materialstrengths
the material strengths andand ductility
ductility are affected
are affected by by
temperature. The remarkable
temperature. The remarkable decrease in strength
decrease in strength is is
observed
observedininthethetemperature rangebetween
temperature range between RTRT and
300 ◦ Cand
(16% 300and 25% decrease
°C (16% in σy and
and 25% decrease in σσuy, and
respectively) and between
σu, respectively) and between550 and and ◦650
550 650 C (17% decrease
°C (17%
decrease in both σ and σ ). However, it appears that the temperature
in both σy and σu ). However, it appears that the temperature effect is not pronounced between 300
y u effect is not pronounced
and 550between 300 and the
◦ C. Instead, 550 °C. Instead,strength
material the material strength
at 400 ◦ C isatslightly
400 °C isgreater
slightly than
greater than
that atthat
300at◦ C.
300Likewise,
°C.
Likewise, the material ductility does not change significantly between 300 and 550 °C. However, it
the material ductility does not change significantly between 300 and 550 ◦ C. However, it decreased
decreased significantly between RT and 200 °C and exhibits a minimum value at approximately 400
significantly between RT and 200 ◦ C and exhibits a minimum value at approximately 400 ◦ C and
°C and increases when temperature exceeds 550 °C.
increases when temperature exceeds 550 ◦ C.
Table 2. The mechanical properties of 316L stainless steel.
Table 2. The mechanical properties of 316L stainless steel.
Temperature Strain Rate σy σu EL σmax
Temperature (oC) Strain Rate (s−1) σy(MPa) (MPa)
σu (%)EL (MPa) σmax
(◦ C) 20 (s−1 ) 9.34 × 10−4(MPa)489 (MPa)
684 (%) 985
50.90 (MPa)
20 −4 × 10−3 489 419
200 9.34 × 101.01 546
684 28.80
50.90 638 985
−3
200 300 1.01 × 101.01 × 10−3 419 409 546
510 28.80 583
24.80 638
−3
300 400 1.01 × 109.11 × 10−4 409 389 510
512 24.80 591
24.14 583
400 −4
9.11 × 109.42
550 × 10 −4 389 368 512
487 24.14
24.00 554 591
550 9.42 × 10−4 368 487 24.00 554
600 9.09 × 10−4 346 346 457 25.32 528
600 9.09 × 10−4 457 25.32 528
650 650 9.68 × 109.68
−4 × 10−4 315 315 403
403 33.65
33.65 472 472

Figure 2. Material properties depending on temperature.

Figure 2. Material properties depending on temperature.
3.1.2. Test by Strain Control
3.1.2. Test by Strain Control
Strain-controlled tensile
Strain-controlled testtest
tensile data were
data wereprovided
provided up up to
toaastrain
strainof of
5%5% because
because the tensile
the tensile test was
test was
deliberately halted, as mentioned in Section 2. The tensile test results at 550 ◦ C using strain rates of
deliberately halted, as mentioned in Section 2. The tensile test results at 550 °C using strain rates of 1
1 × 10×−10
3 , −3
1, ×
1 ×10
−4 and 1 ×−510−1−5 s−1 are shown in Figure 3. It is noteworthy that typical engineering
10−4 and 1 × 10 s are shown in Figure 3. It is noteworthy that typical engineering metallic
metallic materials
materials are generally
are generally strengthened
strengthened at rate
at a higher a higher rateand
of loading of isloading and is called
called “strain-rate “strain-rate
hardening”
hardening” [17]. However, interesting facts were observed in the strain-controlled tensile tests of type
316L stainless steel. In contrast with typical engineering metals, type 316L stainless steel hardens at lower
 Materials 2019, 12, x FOR PEER REVIEW 5 of 13

[17].
Materials However,
2019, 12, 1090 interesting facts were observed in the strain-controlled tensile tests of type 316L5 of 13
stainless steel. In contrast with typical engineering metals, type 316L stainless steel hardens at lower
applied strain rates at 550 °C and is known as “a negative strain rate response” [18]. In addition, at
small
applied strain
strain rates
rates 550 ◦as
at such C 10 −4 and 10−5 s−1, the tensile curve is serrated during plastic deformation and
and is known as “a negative strain rate response” [18]. In addition, at small
strainisrates
known
suchasas“the − 4
10 plastic
and 10 −5 s−1 , phenomena.”
serration Therefore,
the tensile curve the strain
is serrated duringrateplastic
affected the material
deformation and is
behavior at high temperature and it is speculated that complexities may occur at temperatures
known as “the plastic serration phenomena.” Therefore, the strain rate affected the material behavior at from
300 to 550 °C. ◦
high temperature and it is speculated that complexities may occur at temperatures from 300 to 550 C.

3. 3.Strain ◦ C.
Figure
Figure Strainrate
ratedependency
dependency atat550
550°C.

According to previous
According studies
to previous [18,19],
studies [18,19],such
suchabnormal characteristics
abnormal characteristics in in
thethe deformation
deformation behavior
behavior
and mechanical properties of type 316L stainless steel at 300–550 ◦ C arise from the interaction between
and mechanical properties of type 316L stainless steel at 300–550 °C arise from the interaction
mobile dislocations
between mobileand solute atoms
dislocations during
and solute plastic
atoms deformation,
during that is, dynamic
plastic deformation, that is, strain
dynamic aging (DSA).
strain
aging
DSA is (DSA).
known DSA is the
to impair known to impair
fatigue the fatigue
property by way property by way
of multiple of multiple
crack crack
initiation, initiation,
which comes which
from the
comes from
DSA-induced the DSA-induced
localization localization
of deformation, andof rapid
deformation, and rapid
crack growth duecrack growth
to the due to thehardening.
DSA-induced DSA-
induced hardening.
3.2. LCF Test Result
3.2. LCF Test Result
3.2.1. Effect of Temperature on Fatigue Life
3.2.1. Effect of Temperature on Fatigue Life
LCF tests using a strain control of 1 × 10− 3 s−1 were carried out for strain amplitudes of 0.3%
LCF tests using a strain control of 1 × 10 −3 s−1 were carried out for strain amplitudes of 0.3% to
to 0.8% at RT, 200, 400, 550, and 650 ◦ C. It has long been known that the plastic strain amplitude or
0.8% at RT, 200, 400, 550, and 650 °C. It has long been known that the plastic strain amplitude or
dissipated energy
dissipated during
energy a stabilized
during cycle
a stabilized cyclecan
canbebe a representativeparameter
a representative parameter to characterize
to characterize LCF LCF
life. life.
Figure 4 presents the plastic strain amplitude and plastic strain energy density
Figure 4 presents the plastic strain amplitude and plastic strain energy density at a stabilized at a stabilized
cycle.cycle.
Despite thethetotal
Despite total strain-controlled tests,
strain-controlled tests, the the
teststests
aboveabove are associated
are associated with temperature.
with temperature. Thus,
Thus,tests
testswere
wereperformed
performed to to determine
determine whether
whether those those parameters
parameters can be can be applied
applied correctly
correctly regardlessregardless
of
of thethe operatingtemperature.
operating temperature.
First,First, the relationship
the relationship betweenplastic
between plasticstrain
strain amplitude
amplitude and
andfatigue
fatiguelifelife
is shown
is shownin Figure 5a. 5a.
in Figure
The Coffin-Manson model, in which damage accumulates as the plastic
The Coffin-Manson model, in which damage accumulates as the plastic strain cycle is repeated, strain cycle is repeated,
predicts the fatigue life Nf in terms of the plastic strain amplitude Δεp/2, as shown in Equation (1):
predicts the fatigue life Nf in terms of the plastic strain amplitude ∆εp /2, as shown in Equation (1):
N mC  
m f P 2   CC (1)
N f C (∆εP /2) = CC (1)
where mC and CC are the fatigue ductility exponent and fatigue ductility coefficient, respectively. As
whereshown
mC andin Figure
CC are5a,the
fatigue life exhibits
fatigue a linear
ductility relationship
exponent with theductility
and fatigue plastic strain amplitude
coefficient, in the
respectively.
As shown in Figure 5a, fatigue life exhibits a linear relationship with the plastic strain amplitudeatin the
logarithmic coordinate at each temperature. The Coffin-Manson model is a good prediction model
each temperature
logarithmic coordinatecondition, presumably, despite
at each temperature. the occurrence model
The Coffin-Manson of microstructural changes with
is a good prediction model
increased temperature.
at each temperature condition, presumably, despite the occurrence of microstructural changes with
increased temperature.
Next, the fatigue lives were plotted using the plastic strain energy density in Figure 5b. In this
case, the Morrow model was examined, and the model predicts the fatigue life Nf in terms of the
plastic strain energy density ∆Wp , as shown in Equation (2):
m
N f E ∆WP = CE (2)
Materials 2019, 12, 1090 6 of 13

Here, mE and CE are material constants determined empirically. The plastic strain energy density
Materials 2019, 12, x FOR PEER REVIEW 6 of 13
was determined from the area enclosed by the stabilized stress-strain hysteresis loop at Nf /2. As shown
in Figure 5b, the plastic strain energy density and fatigue life exhibit a linear relationship in the
logarithmic coordinate for each temperature. Likewise, the Morrow model is as a good prediction
Materials 2019, 12, x FOR PEER REVIEW 6 of 13
model for isothermal conditions as is the Coffin-Manson model.

.
4. Temperature
FigureFigure dependency
4. Temperature dependency (a)(a)∆εΔε
of of p and (b)∆W
p and(b) at ε =
ΔWpp at  11 ×310s−−13. s−1 .
10
Figure 4. Temperature dependency of (a) Δεp and (b) ΔWp at   1103 s−1.

Figure 5. Cont.
Materials 2019, 12, 1090 7 of 13
Materials 2019, 12, x FOR PEER REVIEW 7 of 13

.
Figure 5. Life prediction by (a) the Coffin-Manson model and (b) the Morrow model at ε = 1 × 10−3 s−1 .
Figure 5. Life prediction by (a) the Coffin-Manson model and (b) the Morrow model at ε  1103 s−1.

For simple calculations, the plastic strain energy density of a Masing material is given by
Next, the fatigue lives were plotted using the plastic strain energy density in Figure 5b. In this
case, the Morrow model was examined, and the 1 −model
n predicts the fatigue life Nf in terms of the
∆WP = ∆σ · ∆εP , (3)
+n
plastic strain energy density ΔWp, as shown in 1Equation (2):
where n and ∆σ are the cyclic strain hardening E
ΔWP  C
N mfexponent in cyclic stress-strain equation and the stress(2)
E
range determined from the stabilized stress-strain hysteresis loop at Nf /2, respectively. On the other
Here, m
hand, E and
the CE are
plastic material
strain constants
energy density determined
of a non-Masingempirically.
materialThe
canplastic strain energy
be obtained from thedensity
cyclic was
and
determined
master curvefrom the area enclosed by the stabilized stress-strain hysteresis loop at Nf/2. As shown in
equations:
Figure 5b, the plastic strain energy density and fatigue life exhibit a linear relationship in the
logarithmic coordinate for each 1 − n∗ 2n∗
∆WPtemperature.
= Likewise,
∆σ × ∆εP + the Morrow
δσ ×∆εmodel
P,
is as a good prediction(4)
model for isothermal conditions as is1the
∗ + n∗ 0
+ nCoffin-Manson1 model.
For simple
Here, n* is calculations, the plasticexponent
the strain-hardening strain energy
of thedensity
masterofcurve
a Masing
and material
δσ0 is theis given by in the
increase
 n [20].
proportional stress limit in the master and cyclic1curves

Both life prediction models, Coffin-MansonW     P , are excellent life prediction models
andMorrow, (3)
1 n
P
for a given temperature condition; however, they cannot be applied irrespective of the temperature.
where
The n and 
material are the cyclic
constants, mC , CCstrain
, mE , hardening
and CE , areexponent
summarized in cyclic stress-strain
in Table 3. equation and the stress
range determined from the stabilized stress-strain hysteresis loop at Nf/2, respectively. On the other
hand, the plastic strain energy Materialofconstants
Table 3.density in each life
a non-Masing prediction
material models.
can be obtained from the cyclic and
master curve equations:
Coffin-Manson Model Morrow Model
Temperature
(◦ C) 1 n
(Equation*(1))
2n * (Equation (2))

mW  CCΔε P 
Δσ× δσ0 ×Δε P , (4)
C P 1 n*
1  n* mE CE
RT 0.395 8.95 0.580 552.33
Here, n* is200
the strain-hardening
0.418 exponent8.51 0.535 and 0 is221.38
of the master curve the increase in the
proportional stress
400 limit in the0.501
master and cyclic curves [20]. 0.610
12.20 301.30
550
Both life prediction 0.510
models, Coffin-Manson9.38 0.635
and Morrow, are 247.07
excellent life prediction models for
650 0.526 9.96 0.651 221.31
a given temperature condition; however, they cannot be applied irrespective of the temperature. The
material constants, mC, CC, mE, and CE, are summarized in Table 3.
3.2.2. Effect of Strain Rate on Fatigue Life
The effect of strainTable 3. Material constants in each life prediction models.
rate on material behavior and fatigue life is negligible at RT, but it is not
negligible as temperature increases. To investigate the
Coffin-Manson effect ofMorrow
Model strain rate on fatigue life, LCF tests
model
were conducted with Temperature 1 × 10−2 ,1)1 × 10−3 , 1(Equation
diverse strain rates of(Equation × 10−4 and −5 −1
2) 3 × 10 s . In those
( C) of 0.5% was applied and three test temperature conditions, i.e., 400,
o
tests, one total strain amplitude mC CC mE CE
◦
550, and 650 C, were considered.
RT 0.395 8.95 0.580 552.33
200 0.418 8.51 0.535 221.38
400 0.501 12.20 0.610 301.30
 Materials 2019, 12, x FOR PEER REVIEW 8 of 13

550 0.510 9.38 0.635 247.07

650 0.526 9.96 0.651 221.31

3.2.2. Effect of Strain Rate on Fatigue Life

Materials 2019, 12, 1090 8 of 13
The effect of strain rate on material behavior and fatigue life is negligible at RT, but it is not
negligible as temperature increases. To investigate the effect of strain rate on fatigue life, LCF tests
Thewere conducted
plastic strain with
energydiverse strainobtained
density rates of 1 ×from
10−2, 1a×stabilized
10−3, 1 × 10−4hysteresis
and 3 × 10−5 sloop
−1. In those tests, one
are summarized in
total strain amplitude of 0.5% was applied and three test temperature
Table 4. Depending on the imposed strain rate, the plastic strain energy, W P , exhibits conditions, i.e. 400, 550, variations.
slight and
650 °C, were considered.
The plastic strain energy density obtained from a stabilized hysteresis loop are summarized in
Table 4. The changes of plastic strain energy density, WP , with respect to strain rate at each temperature
Table 4. Depending on the imposed strain rate, the plastic strain energy, WP, exhibits slight variations.
(∆εt = ±0.5%).
Table 4. The changes of plastic strain energy density, WP, with respect to strain rate at each
Strain Rate
temperature (t   0.5%).
W P (MJ/m3 )
(s−1 ) 400 ◦ C 500 ◦ C 650 ◦ C
Strain Rate WP (MJ/m3)
1 × 10−2 2.768
(s−1) 400 °C 500 2.610
°C 650 °C 2.701
1 × 10−3 2.748
1 × 10−2 2.768 2.662 2.701
2.610 2.739
1 × 10−4 2.828
1 × 10−3 2.748 2.825 2.739
2.662 2.780
3.2 × 10−5 1–× 10−4 2.828 2.952 2.780
2.825 2.837
3.2 × 10−5 – 2.952 2.837

With increased strain rate, approximately 10% difference of W occurs between the strain rates of
With increased strain rate, approximately 10% difference of WPP occurs between the strain rates
− 2
1 × 10 ofand 10−14 ×s10
−1 .
1 ×110×
−2 and −4 s−1.

As shown As in Figure
shown 6, greater
in Figure strain
6, greater rates
strain result
rates resultin
in longer fatigue
longer fatigue lives.
lives. ThisThis canexplained
can be be explained
based on the DSA effect. DSA is known to degrade the fatigue resistance [18,19].
based on the DSA effect. DSA is known to degrade the fatigue resistance [18,19]. The DSA-induced The DSA-induced
hardening hardening accelerates
accelerates the crack
the crack growth
growth rate,
rate, consequently reducing
consequently reducing the
thecrack
crackpropagation life. Inlife.
propagation the In the
regime of DSA (300–550
◦ °C), the DSA effect becomes more pronounced with a decrease in strain rate
regime of DSA (300–550 C), the DSA effect becomes more pronounced with a decrease in strain rate
and this therefore results in a reduction of fatigue life with decreasing strain rate at a given
and this therefore results in a reduction of fatigue life with decreasing strain rate at a given temperature.
temperature.

Materials 2019, 12, x FOR PEER REVIEW 9 of 13

FigureFigure 6. Strain
6. Strain rate rate dependenceon
dependence onfatigue
fatigue lives
livesatat(a)(a)550 °C ◦and
550 (b) 650
C and 650 ◦ C.
(b)°C.

3.2.3. Proposed Energy-Based Fatigue Model

A new life prediction model is developed for a range of temperature and strain rate. It is
noteworthy that the material ductility coefficient, CE, depend significantly on temperature. Further,
the dependence of the fatigue exponent, mE, exists, but not as much as that of CE (Table 3). To
incorporate temperature dependency into a life prediction model, the temperature-dependent
parameters must be properly introduced into the Morrow model to eliminate the temperature
Materials 2019, 12, 1090 9 of 13

3.2.3. Proposed Energy-Based Fatigue Model

A new life prediction model is developed for a range of temperature and strain rate.
It is noteworthy that the material ductility coefficient, CE , depend significantly on temperature.
Further, the dependence of the fatigue exponent, mE , exists, but not as much as that of CE (Table 3).
To incorporate temperature dependency into a life prediction model, the temperature-dependent
parameters must be properly introduced into the Morrow model to eliminate the temperature
dependence of the constants, mE and CE . Thus, a possible model should be of the following form:

mE ∆WP
Nf = CE (5)
f (T )

where f (T) is a temperature-dependent normalizing function for ∆WP .

To account for the effect of strain rate on fatigue life, the frequency-modified fatigue life was
.
introduced into Equation (5) by replacing the loading frequency ν by the strain rate ε because the strain
rate was controlled in the LCF tests. Therefore, the strain-rate inclusive model becomes a multiplicative
form of the strain-rate modified fatigue life and normalized dissipate energy density.
" .  k −1 # m
∆WP

ε
Nf . =C (6)
ε0 f (T )

.
Here, ε0 is a reference strain rate (e.g., 1 × 10−3 s−1 ) and m, k, C are material constants. For a
proper selection of f(T), the function must be a material parameter that decreases with increasing
temperature such as yield stress, material toughness, tangent modulus, etc. Simultaneously, it is
reasonable to consider the function as an energy-related function to maintain consistency.
We begin with a material response to decide the function f(T). As shown in the stress-strain curve
presented in Figure 7, the material exhibits a strain-hardening in the plastic deformation region.
Materials 2019, 12, x FOR PEER REVIEW 10 of 13

Figure
Figure 7. 7.Rigid
Rigidlinear-strain
linear-strain hardening
hardeningassumption
assumptionof tensile response.
of tensile response.

Assuming
Assuming a rigid-linear
a rigid-linear strain
strain hardening
hardening material,
material, thethe energy
energy absorbed
absorbed prior
prior to unstable
to unstable necking
necking
is given by is given by
σmax − σY Y
   
   max
σmax   Y+
 ,σY
f (T ) : W
f (T=
): W   max 
·
   , (7) (7)
H H   2 2

where
where H and
H and σmaxσmax
arearethe
thetangent
tangent modulus
modulus and
andultimate
ultimatetensile stress
tensile in the
stress in true stress-strain
the true curve.curve.
stress-strain
However, the energy-related parameter above is not practical for the description of f(T), as
However, the energy-related parameter above is not practical for the description of f(T), as shown shown in in
Equation (7) because of the number of material constants. For simplification, using a single dominant
material constant,  max
2
, in place of f(T), Equation (6) becomes

k 1 m
    WP ,
N f    C (8)
 0   max (T ) 
2
 

When the fatigue life is presented by a function of the plastic strain energy density normalized
by  2 , the fatigue life data at the temperature range (RT to 650 °C) was reduced to a single curve
necking is given by
   Y    max   Y  ,
f (T ) : W   max    (7)
 H   2 

where
MaterialsH and12,σ1090
2019, max are the tangent modulus and ultimate tensile stress in the true stress-strain curve.
10 of 13
However, the energy-related parameter above is not practical for the description of f(T), as shown in
Equation (7) because of the number of material constants. For simplification, using a single dominant
Equation (7) because2 of the number of material constants. For simplification, using a single dominant
material constant,  max , in place of f(T), Equation (6) becomes
material constant, σ2max , in place of f(T), Equation (6) becomes
k 1 m
 .  k−1 #m WP ,
εNf   
"
∆WP 2  C (8)
N f .  0    max  2 = C
    (T ) (8)
ε0 (σmax ( T ))
When the fatigue life is presented by a function of the plastic strain energy density normalized
When the fatigue life is presented by a function of the plastic strain energy density normalized by
by2  max
2
, the fatigue life data at the temperature range (RT to 650 °C) was reduced to a single curve
σmax , the fatigue life data at the temperature range (RT to 650 ◦ C) was reduced to a single curve as
, canbe
2
as shown
shown in Figure
in Figure 8. is,
8. That That
the normalizing factor, σ2max
is, the normalizing factor, , can be employed to eliminate the
max employed to eliminate the temperature
dependence of
temperature the constant
dependence of C efficiently.
the constant C efficiently.

Figure 8.
Figure The verification
8. The verification of
of the
the new
new LCF life prediction model with respect to temperature.

The effects of strain rate on fatigue life are examined from Figure 9. Figure 9a does not account
for the strain rate dependency on fatigue life, which is based on Equation (6). Meanwhile, Figure 9b is
generated by considering the strain rate using Equation (8). When the strain rate modified fatigue life
is presented as a function of the temperature modified plastic strain energy density, the data points
obtained at the range of temperature (RT to 650 ◦ C) and strain rate (3 × 10−5 to 1 × 10−2 s−1 ) were
reduced into a single curve, as shown in Figure 9b. In this case, the average values of constants m, k,
and C in Equation (8) were found to be 0.576, 0.808, and 6.21 × 10−4 , respectively.
Comparisons between the predicted lives and test data are presented in Figure 10. The horizontal
and vertical axes represent the test data and predicted data, respectively. The more data points closer to
the line of 45◦ slope (solid line in the figure) implies that the test results agree well with the predicted
lives from the developed model. From the comparison, most of the data points are located between
the factor-of-two line (dotted lines in the figure). Therefore, the new life prediction model described in
Equation (8) can successfully describe the fatigue life with the inclusion of the effects of strain rate
and temperature.
for the strain rate dependency on fatigue life, which is based on Equation (6). Meanwhile, Figure 9b
is generated by considering the strain rate using Equation (8). When the strain rate modified fatigue
life is presented as a function of the temperature modified plastic strain energy density, the data
points obtained at the range of temperature (RT to 650 °C) and strain rate (3 × 10−5 to 1 × 10−2 s−1) were
reduced2019,
Materials into12,a1090
single curve, as shown in Figure 9b. In this case, the average values of constants11m, k,
of 13
and C in Equation (8) were found to be 0.576, 0.808, and 6.21 × 10 , respectively.
−4

Figure 9.
Figure 9. The verification of the new LCF
LCF life
life prediction
predictionmodel
modelwith
withrespect
respecttotostrain-rate:
strain-rate:(a) without
Materials 2019, 12, x FOR PEER REVIEW 12 of 13
using the strain-rate modified fatigue lives; (b) with using strain-rate modified fatigue lives in Equation (8).
Comparisons between the predicted lives and test data are presented in Figure 10. The
horizontal and vertical axes represent the test data and predicted data, respectively. The more data
points closer to the line of 45° slope (solid line in the figure) implies that the test results agree well
with the predicted lives from the developed model. From the comparison, most of the data points are
located between the factor-of-two line (dotted lines in the figure). Therefore, the new life prediction
model described in Equation (8) can successfully describe the fatigue life with the inclusion of the
effects of strain rate and temperature.

Figure
Figure 10. Comparison between
10. Comparison between predicted
predicted fatigue
fatigue life
life and
and test
test data.
data.

4. Conclusions
In this study, an energy-based fatigue life prediction model was developed that included the
effects of temperature and strain rate. Tensile behavior of type 316L stainless steel were investigated
by varied temperatures and strain rates. Low-cycle fatigue tests were performed over a wide range
of temperature (RT to 650 °C) and strain rate (1 × 10−5 to 1 × 10−2 s−1) with a constant total strain
amplitude from 0.3% to 0.8%. Fatigue lives obtained from the extensive experimental program were
plotted with respect to plastic strain amplitude, plastic strain energy density, and normalized plastic
Materials 2019, 12, 1090 12 of 13

4. Conclusions
In this study, an energy-based fatigue life prediction model was developed that included the
effects of temperature and strain rate. Tensile behavior of type 316L stainless steel were investigated
by varied temperatures and strain rates. Low-cycle fatigue tests were performed over a wide range
of temperature (RT to 650 ◦ C) and strain rate (1 × 10−5 to 1 × 10−2 s−1 ) with a constant total strain
amplitude from 0.3% to 0.8%. Fatigue lives obtained from the extensive experimental program were
plotted with respect to plastic strain amplitude, plastic strain energy density, and normalized plastic
strain energy density. The conclusions were as follows.
1. The developed fatigue model successfully described the temperature and strain-dependent
fatigue life. The developed model was a multiplicative form of the normalized plastic strain
energy density and frequency modified fatigue life, resulting from a good correlation between
the two parameters. Finally, the prediction capabilities were verified by comparing with the
predicted life and experimental fatigue life. Consequently, most of experimental data were found
to be between the factor-of-two lines.
2. The conventional Coffin-Manson model and Morrow model were recommended at one isothermal
temperature for a life prediction model. However, they were not applicable over the specified
temperature range, or, the fatigue ductility coefficient, C, and exponent, m, must be provided
with respect to temperature.
3. At the temperature range of 300—550 ◦ C, DSA occurred, causing abnormal features in the
deformation behavior (a serrated flow and negative strain-rate sensitivity) and mechanical
properties (a plateau in the variation of strength and ductility with temperature), and this led
to the deterioration of the fatigue resistance. Despite the DSA-induced complicated nature of
fatigue life behavior, the capability of the developed fatigue model could be proven over a range
of temperature and strain rate.

Author Contributions: Conceptualization, N.H.T. and J.-Y.L.; Methodology, N.H.T. and J.-Y.L.; Formal Analysis,
N.H.T., J.-S.K. and J.-Y.L.; Investigation, N.H.T., J.-S.K. and J.-Y.L.; Writing—Original Draft Preparation, N.H.T.;
Writing—Review and Editing, J.-Y.L.; Supervision, J.-Y.L.
Funding: This research was funded by the Korea Research Institute of Standards and Science through the
project of “Development of Reliability Technology of Standard Measurement for Hydrogen Convergence Station
(KRISS-2018-GP2018-0011).” This research was also partially supported by a grant from the R&D Program of the
Korea Railroad Research Institute, Republic of Korea and the National Research Foundation of Korea (NRF) grant
funded by the Korea Government (Grant No.: 2017R1C1B5017363), Republic of Korea.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Kamal, M.; Rahman, M.M. Advances in fatigue life modeling: A review. Renew. Sustain. Energy Rev.
2018, 82, 940–949. [CrossRef]
2. Santecchia, E.; Hamouda, A.M.S.; Musharavati, E.; Zalnezhad, E.; Cabibbo, M.; El Mehtedi, M.; Spigarelli, S.
A review on fatigue life prediction methods for metals. Adv. Mater. Sci. Eng. 2016, 2016, 9573524. [CrossRef]
3. Bezazi, A.; Pierce, S.G.; Worden, K.; Harkati, E. Fatigue life prediction of sandwich composite materials
under flexural tests using a Bayesian trained artificial neural network. Int. J. Fatigue 2007, 29, 738–747.
[CrossRef]
4. Zhang, Z.; Tan, J.; Wu, X.; Han, E.; Ke, W.; Rao, J. Effect of temperature on corrosion fatigue behavior of
316LN stainless steel in high temperature pressurized water. Corros. Sci. 2019, 146, 80–89. [CrossRef]
5. Yan, X.; Zhang, X.; Tu, S.; Mannan, S.; Xuan, F.; Lin, Y. Review of creep-fatigue endurance and life prediction
of 316 stainless steels. Int. J. Press Vessels Piping 2015, 126–127, 17–28. [CrossRef]
6. Suazay, M.; Mottot, M.; Allais, L.; Noblecourt, M.; Monnet, I.; Perinet, J. Creep fatigue behavior of an AISI
stainless steel at 550 ◦ C. Nucl. Eng. Des. 2004, 232, 219–236. [CrossRef]
7. Qiu, Y.; Pang, J.C.; Li, S.X.; Zhang, M.X.; Zhang, Z.F. Influence of temperature on the low-cycle fatigue
properties of compacted graphite iron. Int. J. Fatigue 2018, 117, 450–460. [CrossRef]
Materials 2019, 12, 1090 13 of 13

8. Hong, J.; Jang, C.; Kim, T. Effects of mixed strain rates on low cycle fatigue behaviors of austenitic stainless
steels in a simulated PWR environment. Int. J. Fatigue 2016, 82, 292–299. [CrossRef]
9. Srinivasan, V.S.; Valsan, M.; Rao, K.B.S.; Mannan, S.L.; Raj, B. Low cycle fatigue and creep-fatigue interaction
behavior of 316L(N) stainless steel and life prediction by artificial neural network approach. Int. J. Fatigue
2003, 25, 1327–1338. [CrossRef]
10. Sharkar, P.P.; De, P.S.; Dhua, S.K.; Chakraborti, P.C. Strain energy based low cycle fatigue damage analysis in
a plain C–Mn rail steel. Mater. Sci. Eng. A Struct. Mater. 2017, 707, 125–135. [CrossRef]
11. Coffin, L.F., Jr. A study of the effects of cyclic thermal stresses on a ductile metal. Trans. ASME 1954, 76, 931–950.
12. Feltner, C.E.; Morrow, J.D. Microplastic strain hysteresis energy as a criterion for fatigue fracture. J. Basic
Eng. 1961, 83, 15–22. [CrossRef]
13. Shi, X.Q.; Pang, H.L.J.; Zhou, W.; Wang, Z.P. Low cycle fatigue analysis of temperature and frequency effects
in eutectic solder alloy. Int. J. Fatigue 2000, 22, 217–228. [CrossRef]
14. Lee, J.H.; Jeong, H.Y. Fatigue life prediction of solder joints with consideration of frequency, temperature
and cracking energy density. Int. J. Fatigue 2004, 61, 264–270. [CrossRef]
15. ASTM E606-92 Standard Practice for Strain-Controlled Fatigue Testing; ASTM International: West Conshohocken,
PA, USA, 2004.
16. Kandil, F.A. Potential ambiguity in the determination of the plastic strain range component in LCF testing.
Int. J. Fatigue 1999, 21, 1013–1018. [CrossRef]
17. Xu, Z.; Huang, F. Plastic behavior and constitutive modeling of armor steel over wide temperature and strain
rate ranges. Acta Mech. Solida Sin. 2012, 25, 598–608. [CrossRef]
18. Hong, S.G.; Lee, S.B. Dynamic strain aging under tensile and LCF loading conditions and their comparison
in cold worked 316L stainless steel. J. Nucl. Mater. 2004, 328, 232–242. [CrossRef]
19. Hong, S.G.; Lee, S.B. Mechanism of dynamic strain aging and characterization of its effect on the low-cycle
fatigue behavior in type 316L stainless steel. J. Nucl. Mater. 2005, 340, 307–314. [CrossRef]
20. Ellyn, F. Fatigue Damage, Crack Growth and Life Prediction; Chapman & Hall: London, UK, 1997; pp. 87–90.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).