metals

Article
Fatigue Life Evaluation Considering Fatigue
Reliability and Fatigue Crack for FV520B-I in VHCF
Regime Based on Fracture Mechanics
Jinlong Wang 1, *, Yuxing Yang 1 , Jing Yu 1 , Jingsi Wang 1 , Fengming Du 1 and Yuanliang Zhang 2
1 Marine engineering college, Dalian Maritime University, Dalian 116000, China;
yangyuxing@dlmu.edu.cn (Y.Y.); yj_lunji@dlmu.edu.cn (J.Y.); wjs@dlmu.edu.cn (J.W.);
dfm@dlmu.edu.cn (F.D.)
2 School of mechanical engineering, Dalian University of Technology, Dalian 116000, China;
zylgzh@dlut.edu.cn
* Correspondence: wjl19890806@dlmu.edu.cn; Tel.: +86-150-4111-8408




Received: 12 February 2020; Accepted: 11 March 2020; Published: 13 March 2020


Abstract: This paper focuses on the fatigue reliability analysis and the development of a new life
model of reliability and crack growth mechanisms in FV520B-I (high strength martensitic-type
stainless steels) in the very-high cycle fatigue (VHCF) regime, which haven’t been studied well. First,
the fatigue test was carried out to clarify the fatigue failure mechanism in the very-high cycle regime.
Based on the test results and fatigue reliability theory, the fatigue life distribution and P-S-N curves
were modeled. A new fatigue life evaluation model for FV520B-I is proposed according to the fracture
mechanics and classic life evaluation method. With the comprehensive application of P-S-N curves
and a new proposed fatigue life evaluation model, a new assumption of a P-Sc-N curve is developed
and verified, to quantitatively express the relationship between fatigue life, reliability and fatigue
cracking. This is novel and valuable for further fatigue study of FV520B-I.

Keywords: FV520B-I; fatigue reliability; P-Sc-N curve

1. Introduction
FV520B-I, as an important high-strength engineering metal, has been widely adopted in the
manufacturing of centrifugal compressor vanes for its positive mechanical properties, including high
strength, high corrosion resistance, high abrasive resistance and good welding characteristics [1,2]. In
practical engineering, the fatigue life of the FV520B-I component is required to be more than 107 cycles,
a very-high cycle fatigue (VHCF) level. Quantitatively expressing the fatigue reliability of FV520B-I is
a key research topic, especially for those who can use the information to improve the manufacturing of
mechanical equipment and avoid fatigue failure. Fatigue reliability [3–8] refers to the probability that
theoretical fatigue life will meet the actual required life time of the material, which can be quantitatively
expressed by a probability of the fatigue life without the fatigue failure in solving real engineering
fatigue problems, especially for the core components of large mechanical equipment. With FV520B-I
being more and more widely used, there is increasing interest in detailed study modelling its fatigue
reliability in the VHCF regime, and developing a new precise fatigue life evaluation model (which
is beneficial to clarify fatigue failure characteristics, and also very essential to avoid fatigue failure
efficiently in real engineering practice).
A few investigations have been carried out to explore the fatigue reliability of metallic materials.
Wirsching [9–12] studied probabilistic and statistical methods for assessing fatigue in design and
analysis. New methods were proposed to analyze the reliability of both individual joints and the
overall system, with consideration of crack initiation. Wang [13] conducted a study of the fatigue

Metals 2020, 10, 371; doi:10.3390/met10030371 www.mdpi.com/journal/metals

Metals 2020, 10, 371 2 of 16

mechanisms of conveyor slide rails and wheels under constant amplitude cyclic loading, and a normal
distribution model was employed to make a contrastive analysis between the fatigue reliabilities of
rails and wheels. Wang [14] conducted a bending fatigue experiment on metal 8822H to establish
the P-S-N curve and analyze fatigue reliability of metal 8822H. Wurzel [15] focused on the fatigue
of aerospace components, and a new fatigue evaluation model was proposed based on the classical
theory and experimental data. Paolino [16] put forward an investigation about the effect of fatigue
cracks on the P-S-N curve for metallic material in the VHCF regime.
For predicting the fatigue life of FV520B-I, some previous studies have investigated the relationship
Metals 2020, 10, x FOR PEER REVIEW 2 of 16
between fatigue cracks and fatigue life. Liu [17,18] performed a finite element analysis for the open
impeller to clarify
reliabilities the and
of rails stress distribution
wheels. Wang [14] of FV520B-I
conducted impeller
a bending fatigueinexperiment
actual operational
on metal 8822H conditions, and
performed a fatigue analysis based on classic fatigue theory. Zhang [19–22] focusedonon a study of
to establish the P-S-N curve and analyze fatigue reliability of metal 8822H. Wurzel [15] focused
the fatigue of aerospace components, and a new fatigue evaluation model was proposed based on
the effect of fatigue influence parameters (including internal inclusion, surface roughness and loading
the classical theory and experimental data. Paolino [16] put forward an investigation about the effect
frequency) on the
of fatigue fatigue
cracks on thefailure
P-S-Nmechanism
curve for metallic of FV520B-I
material ininthe VCHF
VHCFand regime.HCF regime. Also, the quantitative
expression model between
For predicting thethe influence
fatigue life of parameters
FV520B-I, some andprevious
fatigue studies
life were proposed
have andthe
investigated verified.
relationship
Some research between
has fatigue cracks andout
been carried fatigue
on life. Liu [17,18]
fatigue performed
reliability a finite element
analysis and fatigueanalysisproperties of
for the open impeller to clarify the stress distribution of FV520B-I impeller in actual operational
FV520B-I, and these studies provide a theoretical foundation for the development of fatigue reliability
conditions, and performed a fatigue analysis based on classic fatigue theory. Zhang [19–22] focused
analysis onmethods
a study ofand the establishment
effect of fatigue of FV520B-I
influence fatigue
parameters life equation.
(including However,
internal inclusion, up to now, there
surface
is very roughness
little scientific
and loading understanding
frequency) on the of the fatigue
fatigue failurereliability
mechanism of ofFV520B-I
FV520B-I in theand
in VCHF VHCF
HCF regime. No
previous specific study of FV520B-I in the VHCF regime has been performed to addresslife
regime. Also, the quantitative expression model between the influence parameters and fatigue the question of
were proposed and verified.
P-S-N curve establishment of FV520B-I, and neither has a fatigue
Some research has been carried out on fatigue reliability analysis and fatigue properties of
life evaluation model considering
reliability and fatigue
FV520B-I, and these cracks in FV520B-I
studies provide been
a theoretical well developed.
foundation for the developmentMore ofthanfatiguethat, fatigue reliability
reliability
analysis and life
analysis methodsevaluation for FV520B-I
and establishment of FV520B-I in engineering
fatigue life equation.practiceHowever,is very
up tocase-oriented—different
now, there is
theoriesveryand little scientific
models areunderstanding
specific to desiredof the fatigue reliability ofand
applications FV520B-I in the VHCF
conditions, and regime. No
the parameters of the
previous specific study of FV520B-I in the VHCF regime has been performed to address the question
model are not the same when applying to dissimilar material modeling.
of P-S-N curve establishment of FV520B-I, and neither has a fatigue life evaluation model considering
The study presented
reliability here isinone
and fatigue cracks of thebeen
FV520B-I firstwell
specific investigations
developed. More than that, to explore the fatigue reliability
fatigue reliability
analysis
of FV520B-I inand
thelifeVHCF evaluation
regime for toFV520B-I
address in the
engineering
questions practice
above.is veryThe case-oriented—different
primary aim of this paper is to
theories and models are specific to desired applications and conditions, and the parameters of the
contribute to the quantitative expression of fatigue reliability for FV520B-I in the VHCF regime, and
model are not the same when applying to dissimilar material modeling.
attempt to The develop a new fatigue life evaluation model combining reliability and fatigue cracking
study presented here is one of the first specific investigations to explore the fatigue reliability
together. As shown
of FV520B-I in thein VHCFFigure
regime 1. toFirst,
address antheultrasonic
questions above. fatigue test foraim
The primary FV520B-I
of this paperwas carried out to
is to
contribute fatigue
collect effective to the quantitative
test dataexpression
for FV520B-I of fatigueandreliability
provide forscientific
FV520B-I in the VHCF
support forregime, and
the following fatigue
attempt to develop a new fatigue life evaluation model combining reliability and fatigue cracking
reliability estimation. The characteristics of the fracture surface were observed with SEM to make clear
together. As shown in Figure 1. First, an ultrasonic fatigue test for FV520B-I was carried out to collect
the leading cause
effective of fatigue
fatigue test data failure. Next, and
for FV520B-I the provide
fatigue scientific
life distributions
support forofthe FV520B-I
followinginfatigue
the VHCF regime
and P-S-N curve
reliability modelsThe
estimation. with different reliability
characteristics of the fracture values
surface were
wereestablished,
observed with with SEM to the comprehensive
make
clear the
application of aleading cause of fatiguemodel,
three-parameter failure. Next,
test datathe fatigue
and alife distributionsnormal
logarithmic of FV520B-I in the VHCFfunction. The
distribution
regime and P-S-N curve models with different reliability values were established, with the
classic fatigue life formula was employed to develop a new fatigue life evaluation model for FV520B-I
comprehensive application of a three-parameter model, test data and a logarithmic normal
with the consideration
distribution function. ofThe
fatigue
classicreliability, and thewas
fatigue life formula new proposed
employed modela was
to develop verified
new fatigue lifeto be reliable
through additional
evaluation model fatigue tests.with
for FV520B-I Thisthe demonstrates
consideration of fatigue that the new model
reliability, and the newcan proposed
provide satisfactory
model
reliability was verified
analysis results to be
and reliable through additional
life predictions fatigue tests.
for FV520B-I. This demonstrates
Lastly, an assumption that the
of anewP-Sc-N curve is
model can provide satisfactory reliability analysis results and life predictions for FV520B-I. Lastly, an
proposed which combines fatigue reliability, fatigue cracking, and fatigue life together, and the P-Sc-N
assumption of a P-Sc-N curve is proposed which combines fatigue reliability, fatigue cracking, and
curve isfatigue
modeled with three
life together, and thedifferent
P-Sc-N curvereliability
is modeledvalues.
with Further study
three different is still values.
reliability necessaryFurtherto demonstrate
its rationality
study is and correctness.
still necessary A framework
to demonstrate has been
its rationality created inAestablishing
and correctness. framework hasthe been created P-S-N and
analytical
in establishing
life evaluation models the analytical P-S-N and
for FV520B-I inlife
theevaluation
VHCF regime,models forand FV520B-I
this in the enhance
will VHCF regime, theandaccuracy of the
this will enhance the accuracy of the fatigue reliability analysis in real engineering practice. This
fatigue reliability analysis in real engineering practice. This study is valuable to engineering practice
study is valuable to engineering practice and very supportive to the theoretical study of fatigue
and very supportive to the theoretical study of fatigue reliability.
reliability.

Figure 1. The structure of the research on proposed fatigue reliability.

Figure 1. The structure of the research on proposed fatigue reliability.
Metals 2020, 10, 371 3 of 16
Metals 2020, 10, x FOR PEER REVIEW 3 of 16

2. Materials and Methods Fatigue Test

2. Materials and Methods Fatigue Test
2.1. Test Method
2.1. Test Method
As mentioned above, lack of fatigue data has existed as a primary problem in research on the
As mentioned above, lack of fatigue data has existed as a primary problem in research on the
fatigue properties of FV520B-I. Fatigue testing is the most common and effective method to obtain
fatigue properties of FV520B-I. Fatigue testing is the most common and effective method to obtain
useful data and provide practical support for theoretical fatigue analysis. In order to investigate the
useful data and provide practical support for theoretical fatigue analysis. In order to investigate the
fatigue reliability of FV520B-I in VHCF regime, a fatigue test was carried out in a USF-2000 ultrasonic
fatigue reliability of FV520B-I in VHCF regime, a fatigue test was carried out in a USF-2000 ultrasonic
fatigue test system. This fatigue test system consists of a control and data acquisition part, a monitoring
fatigue test system. This fatigue test system consists of a control and data acquisition part, a monitoring
system, and a test implementation part, as shown in Figure 2a. The operating frequency was set to
system, and a test implementation part, as shown in Figure 2a. The operating frequency was set to 20
20 KHz. The maximum stress amplitude used in the test was 600 MPa and the minimum was 500 MPa.
KHz. The maximum stress amplitude used in the test was 600 MPa and the minimum was 500 MPa.
The interval was 25 MPa. The mean stress in the test was set to 0, that is, with stress ratio r = −1
The interval was 25 MPa. The mean stress in the test was set to 0, that is, with stress ratio r = −1
(symmetric cyclic load) and the stress amplitude means the alternating stress from 0. Air cooling and a
(symmetric cyclic load) and the stress amplitude means the alternating stress from 0. Air cooling and a
pulse-pause loading model are employed to avoid the effect of temperature rise on the test results.
pulse-pause loading model are employed to avoid the effect of temperature rise on the test results.
2.2. Material and Specimen
2.2. Material and Specimen
The test object was high-strength metal FV520B-I; the chemical composition of FV520B-I is shown
The
in Table 1. test object was properties
The mechanical high-strength metal FV520B-I;
of FV520B-I are shownthe
in chemical
Table 2. composition of FV520B-I is
shown in Table 1. The mechanical properties of FV520B-I are shown in Table 2.
Table 1. Chemical composition of FV520B-I (wt%).
Table 1. Chemical composition of FV520B-I（wt%）.
Chemical Composition C Si Ni Cu S Cr Mo Nb Fe
Chemical Composition C Si Ni Cu S Cr Mo Nb Fe
Content 0.02–0.07 0.15–0.7 5–6 1.3–1.8 <0.025 13–14.5 1.3–1.8 0.25–0.45 Bal
Content 0.02–0.07 0.15–0.7 5–6 1.3–1.8 <0.025 13–14.5 1.3–1.8 0.25–0.45 Bal

Table 2. Mechanical property of FV520B-I.

Table 2. Mechanical property of FV520B-I.
Elastic Tensile Vickers
Mechanical Elastic Tensile Yield Strength
Yield Vickers Elongation
Mechanical Modulus Strength Hardness Elongation
Parameters Modulus Strength Rp0.2 (MPa)
Strength Hardness A (%)
parameters E (GP) Rm (MPa) HV A (%)
E (GP) Rm (MPa) Rp0.2 (MPa) HV
FV520B-I
FV520B-I 194 194 1180 1180 1029 1029 380 380 16.07
16.07

AAstandard
standard specimen
specimen was
was employed
employedininthe
thetest. TheThe
test. geometry of the
geometry of specimen is shown
the specimen in Figure
is shown in
2b, and the units in Figure 2b are mm.
Figure 2b, and the units in Figure 2b are mm.

Figure2.2.The
Figure Thefatigue
fatiguetest
test system
system and
and specimen. (a) The
specimen. (a) The fatigue
fatiguetest
testsystem.
system.(b)
(b)The
Thespecimen
specimenand
andits
its dimensions.
dimensions.

2.3. The Results and Observations

The test results include the stress amplitude and fatigue life, as shown in Figure 3.
Metals 2020, 10, 371 4 of 16

2.3. The Results and Observations

Metals 2020, 10, x FOR PEER REVIEW 4 of 16
MetalsThe
2020,test
10, xresults include
FOR PEER the
REVIEW stress amplitude and fatigue life, as shown in Figure 3. 4 of 16

Figure 3. The test results for fatigue.

Figure 3. The test results for fatigue.
Several fatigue test phenomena of FV520B-I have been observed with the application of scanning
Several
electron fatigue test
microscopy
microscopy phenomena
(SEM),
(SEM), suchasas
such ofGranular
FV520B-I
Granular have
Bright
Bright been
Facet observed
Facet
(GBF)(GBF)
(JEOL with
(JEOL theTokyo,
Ltd, application
Ltd, Japan)of
Tokyo, scanning
Japan)
and and
“fish-
electron microscopy
“fish-eye”. Shiozawa (SEM),
[23–25] such
and as Granular
Murakami Bright
[26–28] Facet
focused (GBF)
on the(JEOL
study Ltd,
of Tokyo,
VHCF
eye”. Shiozawa [23–25] and Murakami [26–28] focused on the study of VHCF failure mechanisms. It Japan)
failure and “fish-
mechanisms.
eye”.
It
waswas Shiozawa
proposed
proposed [23–25]
that and crack
thatprimary
primary Murakami [26–28]
crackinitiation
initiation focused
sites
sites on the
arearound
are around study
the
the of VHCF
subsurface
subsurface failure
crack:
crack: mechanisms.
during
during the fatigue
the It
was proposed
process, that primary
some cracks form incrack
the GBFinitiation sites arepropagating
and continue around the to subsurface
“fish-eye”crack:
until during
failure. the fatigue
These two
process, some cracks form
fracture features are prominent in the GBF and
prominent characteristics continue
characteristics of propagating
of VHCF
VHCF failures to
failures of “fish-eye”
of metallic until
metallic materials. failure.
materials. These two
fracture features
The GBF region are prominent
region isiscaptured characteristics
capturedclearly
clearlythrough of
throughSEM VHCF failures
SEMobservation of
observationofofmetallic
thethe materials.
specimen
specimen in in
thethe
experiment
experiment as
shown
as The
in GBF
shown Figureregion
in Figure is
4. captured
4. This is the is
This mostclearly
the mostthrough
significant SEMcharacteristic
observation
characteristic
significant ofofthe
of the fatigue specimen
thefailure
fatigue in the
caused
failurebyexperiment
subsurface
caused by
as shown
crack in Figure
(or internal
subsurface crack 4. This inclusion).
inclusion).
(or internal is the most significant characteristic of the fatigue failure caused by
subsurface crack (or internal inclusion).

Figure
Figure 4.
4. The
The GBF
GBF (Granular
(Granular Bright
Bright Facet)
Facet) region.
region. (a)
(a) the
the GBF
GBF region,
region, (b)
(b) the
the GBF
GBF region.
region.
Figure 4. The GBF (Granular Bright Facet) region. (a) the GBF region, (b) the GBF region.
Except in the GBF region, from the fracture surface of the specimen, the “fish-eye” boundary can
Except
be clearly in the GBF
observed region,infrom
as shown the5.fracture
Figure surface
“Fish-eye”
“Fish-eye” of most
is the
is the the
mostspecimen, the
significant
significant “fish-eye” boundary
characteristic
characteristic can
of very-high
be clearly observed as shown in Figure 5. “Fish-eye” is the most significant characteristic
cycle fatigue fracture, and the change of the crack propagation rate during the crack propagation of very-high
cycle fatigue
process leads fracture, and theofchange
to the formation of the crack propagation rate during the crack propagation
“fish-eye.”
process leads to the formation of “fish-eye.”
Metals 2020, 10, 371 5 of 16

cycle fatigue
Metals 2020, 10, xfracture,
FOR PEER and the change of the crack propagation rate during the crack propagation
REVIEW 5 of 16
process leads to the formation of “fish-eye.”

(a) the “fish-eye” (b) the “fish-eye”

(c) the “fish-eye” (d) the “fish-eye”

Figure 5. The “fish-eye” region.

3. Fatigue Reliability
3. Fatigue Reliability of
of FV520B-I
FV520B-I in
in the
the VHCF
VHCF Regime
Regime

3.1. Very-High Cycle Fatigue Life Distribution of FV520B-I in the VHCF Regime
3.1. Very-High Cycle Fatigue Life Distribution of FV520B-I in the VHCF Regime
Based on the test results, as shown in Figure 2, the fatigue life distribution of FV520B-I in the
Based on the test results, as shown in Figure 2, the fatigue life distribution of FV520B-I in the
VHCF regime with different stress amplitudes can be estimated. The logarithmic normal distribution
VHCF regime with different stress amplitudes can be estimated. The logarithmic normal distribution
model has been widely used to quantitatively analyze the reliability of various mechanical parts for its
model has been widely used to quantitatively analyze the reliability of various mechanical parts for
numerous advantages: a simple calculation process, high accuracy, and wide applicability. Logarithmic
its numerous advantages: a simple calculation process, high accuracy, and wide applicability.
normal distribution is employed as the base model in this paper to obtain the fatigue life distribution
Logarithmic normal distribution is employed as the base model in this paper to obtain the fatigue life
of FV520B-I in the VHCF regime. The fatigue life probability density function [29] of the logarithmic
distribution of FV520B-I in the VHCF regime. The fatigue life probability density function [29] of the
normal distribution is written as:
logarithmic normal distribution is written as:
 1 ln x − µ2 2 
 ! 
11 μ
f (xf )( x=) = √ exp 1  ln x - 
exp−−2  σ    (1)
 
(1)
xσ 2π
xσ 2π 2
  σ  

where xx is
where is the
the actual
actual fatigue
fatigue life
life in
in the
the engineering condition, N
engineering condition, µ is
Nff;; μ is location
location parameter,
parameter, and σ is
and σ is
proportional parameter. µ
proportional parameter. and σσ are
μ and are expressed
expressed as:
as:

i=1lnx
Pn n
lnxi
µμ== i =1 i
(2)
(2)
n
n

 ( lnx − μ )
n 2
i
σ= i =1 (3)
n −1

In order to guarantee the accuracy of the life distribution model, appropriate data must be
selected from the test to estimate μ and σ. xi is the ith variable; here, it is the fatigue life of FV520B-I
with a certain stress amplitude as shown in Figure 3. As displayed in Equation (2), the model
Metals 2020, 10, 371 6 of 16

s
− µ)2
Pn
i=1 (lnxi
σ= (3)
n−1
In order to guarantee the accuracy of the life distribution model, appropriate data must be selected
from the test to estimate µ and σ. xi is the ith variable; here, it is the fatigue life of FV520B-I with a
certain stress amplitude as shown in Figure 3. As displayed in Equation (2), the model parameter µ
means the average of logarithm of fatigue life, and the model parameter σ is the standard deviation of
the logarithm of fatigue life.
The P-S-N (P is reliability, S is stress and N is fatigue life) curve is an important and practical
method in engineering conditions, especially for those who can use S-N curves to quickly determine
the fatigue reliability and life, to improve the reliability of the FV520B-I component design and avoid
fatigue failure for FV520B-I. Thus, it is a key issue to address the problem of lacking P-S-N curves
for FV520B-I, and establish a P-S-N curve model to quantitatively analyze the fatigue reliability of
FV520B-I in the VHCF regime. According to the definition of the P-S-N curve, the P-S-N curve actually
means the S-N curve with a certain reliability value R. Thus, determining the S-N curve model is the
primary work in P-S-N curve modeling. The commonly used S-N basics formula [30] is written as
Equation (4):
Three − parameter model : (S − S0 )α Nf = C (4)

where S is the stress amplitude, which is equal to the σa in the fatigue test; Nf is fatigue life; S0 is
the stress parameter, and α and C are constant parameters. These three parameters are unknown for
FV520B-I. A three-parameter model [29–31] was chosen as the basic model to deduce the P-S-N curve
of FV520B-I in this paper, for its two advantages compared to other models: flexible form and better
ability of data fitting.
Furthermore, at least three groups of test data with different stress amplitudes are required, from
500 MPa to 600 MPa, as employed in the fatigue test. Adequate amounts of test data are required to
ensure the accuracy of estimation results. The fatigue results in Figure 3 include the fatigue life with
five different stress amplitude values; most of the test data are in the very-high cycle fatigue regime,
which means the test data are sufficient for the estimation of the unknown model parameters.
In order to realize the simple calculation process and obtain accurate calculation results, the
fatigue test data with three stress amplitude 500 MPa, 525 MPa and 600 MPa were employed with a
comprehensive consideration of the distribution range of test data and the amount of test data. By
employing the test data in the very-high cycle fatigue regime in Figure 3, the three groups of the
unknown parameter µ and σ are estimated; the results are shown in Table 3.

Table 3. The parameters µ and σ.

Stress Amplitude 500 MPa 525 MPa 600 MPa

location parameter µ 2.1673 1.6429 1.0239
proportional parameter σ 0.4372 0.4213 0.8989

By substituting the parameters in Table 3 into Equation (1), the probability density functions of the
logarithmic normal distribution under different stress amplitudes are obtained, and the corresponding
probability density curves are shown in Figure 6.
 location parameter μ 2.1673 1.6429 1.0239
proportional parameter σ 0.4372 0.4213 0.8989

By substituting the parameters in Table 3 into Equation (1), the probability density functions of
the logarithmic normal distribution under different stress amplitudes are obtained, and the
Metals 2020, 10, 371 7 of 16
corresponding probability density curves are shown in Figure 6.

Metals 2020, 10, x FOR PEER REVIEW 7 of 16

Figure 6. The
Figure6. The fatigue
fatigue life
life probability
probability density
densityfunction
functioncurves
curvesof
ofFV520B-I.
FV520B-I.
The reliability of the logarithmic normal distribution function can be expressed as:
The reliability of the logarithmic normal distribution function can be expressed as:
 ln x − μ 
R ( X ) = P ( X > x) = 1− Φ   , ! ( x > 0) (5)
 lnσx −µ
R(X ) = P(X > x) = 1 − Φ , (x > 0) (5)
where x is Nf’, which presents the actual fatigue life in theσengineering condition (it conforms to the
fatiguex life
is Ndistribution); X is NfT, the required fatigue life without failure; and R is the fatigue
where f ’, which presents the actual fatigue life in the engineering condition (it conforms to the
fatigue life distribution); X isthe
reliability, which presents NfTprobability
, the required with NfT as
fatigue lifethe variable
without when
failure; andtheR isactual fatigue
the fatigue life has
reliability,
reached or exceeded the required life without fatigue failure (that is the probability
which presents the probability with NfT as the variable when the actual fatigue life has reached or of N fT > Nf’). In a

word, thethe
exceeded reliability
requiredis life
always defined
without as the
fatigue probability
failure (that isthat, for a given of
the probability stress
NfT > amplitude
Nf ’). In aand
word, defect
the
size, the random variable fatigue life, N fT, is larger than the required fatigue life, Nf’: R = P (NfT > Nf’).
reliability is always defined as the probability that, for a given stress amplitude and defect size, the
randomFatigue
variablereliability
fatiguecurves
life, NfTbased on than
, is larger the logarithmic
the requirednormal
fatigue distribution
life, Nf ’: R = model
P (NfT >are obtained
Nf ’).
according
Fatigueto reliability
Equation (5), as shown
curves based in onFigure 7. A negative
the logarithmic correlation
normal between
distribution fatigue
model are reliability
obtained
and fatigue life can be clearly detected: with the increased fatigue life, reliability
according to Equation (5), as shown in Figure 7. A negative correlation between fatigue reliability without failure
and
decreases.
fatigue life can be clearly detected: with the increased fatigue life, reliability without failure decreases.

Figure 7. The fatigue reliability curves.

Figure 7. The fatigue reliability curves.
3.2. The P-S-N Curve Estimation
3.2. The P-S-N Curve Estimation
As discussed above, a three-parameter model was employed to estimate the P-S-N curve of
As discussed
FV520B-I, which has above, a three-parameter
not been published before. model
The was employed tomodel
three-parameter estimate the(Equation
[32,33] P-S-N curve (6)) of
is
FV520B-I,
the which
extension andhas not been published
improvement before.
of the Basquin The three-parameter
model, and is proposed model [32,33] (Equation
for estimating P-S-N curves(6)) to
is
the extension
describe andpart
the slope improvement of the
and horizontal Basquin
part. With anmodel, and isofproposed
expansion forscope,
application estimating P-S-N curves
the three-parameter
to describe
model the slope
was also part to
employed and horizontal
estimate P-S-Npart. With
curves in an expansion
VHCF of P-S-N
[34]. The application
curve scope, thethe
refers to three-
S-N
parameter
curve model wastoalso
corresponding employed
different to estimate
survival = 1 −curves
rates P (PP-S-N in VHCF the
R), considering [34].dispersion
The P-S-Nofcurve
fatiguerefers
life.
to the S-N curve corresponding to different survival rates P (P = 1 − R), considering the dispersion of
fatigue life. Each certain fatigue reliability R corresponds to a certain P-S-N curve, so the three-
parameter model with a certain fatigue reliability R can be re-expressed as:

(S − S )
αR
0, R N f = CR (6)
Metals 2020, 10, 371 8 of 16

Each certain fatigue reliability R corresponds to a certain P-S-N curve, so the three-parameter model
with a certain fatigue reliability R can be re-expressed as:

(S − S0,R )αR Nf = CR (6)

where S0,R is the stress parameter when the reliability is R, and αR and CR are the unknown material
parameters with fatigue reliability R. At least three S-N curves with different fatigue reliability values
should be modeled to constitute the P-S-N curves of FV520B-I.
In most situations, the generic S-N curve is the P-S-N curve with reliability R = 0.5, and is also the
most widely used P-S-N curve in practical applications to analyze the fatigue reliability and fatigue life
of FV520B-I in VHCF regime. Thus, the P-S-N curve with R = 0.5 is chosen as an example to show the
deduction procedure of P-S-N curves with different fatigue reliability values. As the three-parameter
model has been widely accepted as a way to obtain the P-S-N curve, so the derivation procedure of the
P-S-N curve for FV520B-I is straightforward. The basic three-parameter model with R = 0.5 under
different stress amplitude values can be written as:

(S1 − S0,0.5 )α0.5 Nf0.5S1 = C0.5



(S2 − S0,0.5 )α0.5 Nf0.5S2 = C0.5


(7)


α0.5


 (S − S
3 0,0.5 ) Nf0.5S2 = C0.5

where S0,0.5 , α0.5 and C0.5 are unknown model parameters for R = 0.5; S1 , S2, and S3 are three stress
amplitude values: S1 = 500 MPa, S2 = 525 MPa and S3 = 600 MPa; and N0.5,S1 , N0.5,S2 and N0.5,S3 are
fatigue life with R = 0.5 under different stress amplitudes, which can be calculated with the combined
application of the standard normal distribution table and Equation (5).
Next, the fatigue life values with different stress amplitude values are calculated as expressed in
Equation (8):
 N0.5S1 = exp (µ1 ) = exp (1.0239) = 2.78(×10 )

 7

 N0.5S2 = exp (µ2 ) = exp (1.6429) = 5.17(×107 )


(8)


0.5S3 = exp (µ3 ) = exp (2.1673) = 8.73(×10 )

 N 7

Moreover, the model data that are employed to obtain the P-S-N curve with R = 0.5 are shown in
Table 4.

Table 4. Stress amplitude, fatigue reliability, and fatigue life.

Stress Amplitude σ a /MPa S1 S2 S3

The fatigue reliability R 0.5 0.5 0.5
Fatigue life Nf 2.78 × 107 5.17 × 107 8.73 × 107

This basic model (Equation (7)) is established based on classical fracture mechanics and
mathematical derivations, so the suitability of the model for the high-cycle fatigue life prediction
of FV520B-I is assured in nature. Each individual material will have its own model with particular
parameters, but there are three unknown model parameters for FV520B-I, including S0,0.5 , α0.5 and
C0.5 . These parameters are dependent on metal FV520B-I and no shared values can be adopted for
the model identification directly. These three parameters are essential in the establishment of the life
model. A small change in them will result in a significant alteration in the fatigue life prediction results.
This implies that there is no published model that can cover a range of different materials. Thus, it
is key to identify the unknown parameters to develop a specific model for FV520B-I with R = 0.5 in
VHCF regime.
As shown in Table 4, the data are diverse, not following certain rules. A nonlinear fitting algorithm
should be adopted to identify the unknown parameters. Such a fitting algorithm uses continuous
curves or analytical expressions to represent discrete data and the functional relationship between the
Metals 2020, 10, 371 9 of 16

variables. The underdeveloped base model is established in Equation (9), and iterative computation is
employed to determine the value of S0,0.5 , α0.5 and C0.5 for the function f (Si ,Nfi ) using the minimizing
operation of Equation (9), i.e.,

3
X
(S0,0.5 , α0.5 , C0.5 ) = argmin ( f (Sci , Nfi ) − Ni )2 (9)
S0,0.5 ,α0.5 ,C0.5 i=1

With employing the model data in Table 4 into Equation (6), the unknown model parameters are
estimated as shown in Table 5.

Table 5. Model parameters.

S0,0.5 α0.5 C0.5

481.1823 0.6205 53.9676
Metals 2020, 10, x FOR PEER REVIEW 9 of 16

By substituting the model parameters into Equation (7), the S-N curve model of FV520B-I with
( S − 481.1823 )
0.6205
R = 0.5 is developed as: N = 53.9676 (10)
(S − 481.1823)0.6205 N = 53.9676 (10)
A P-S-N curve with R = 0.5 has been determined in Equation (10), and more P-S-N curves with
different fatigue
A P-S-N curve with R =
reliability values
0.5 hasshould
been be developedinto
determined enrich the
Equation P-S-N
(10), and curves
more P-S-Nof FV520B-I. Two
curves with
more P-S-N
different curves
fatigue in critical
reliability conditions
values should with R = 0.9 and to
be developed R =enrich
0.1 arethe
employed as additional
P-S-N curves examples.
of FV520B-I. Two
WhenP-S-N
more the reliability
curves inRcritical
= 0.9, conditions
it means that with R = is
there 0.9a and = 0.1 are employed
highRprobability that the fatigue life of examples.
as additional FV520B-I
can satisfy
When the reliability R = 0.9, itConversely,
the requirement. means that there whenisthe reliability
a high R = 0.1,
probability thethe
that fatigue lifelife
fatigue of of
FV520B-I
FV520B-I likely
can
can’t satisfy
satisfy the requirement.
the requirement. According
Conversely, whento the
the reliability R = 0.1, the
same deduction process
fatiguedisplayed above, likely
life of FV520B-I S-N curves
can’t
with Rthe
satisfy = 0.9 and R = 0.1
requirement. are modeled,
According to the sameand three-parameter
deduction process models with above,
displayed three different
S-N curves fatigue
with
R = 0.9 andvalues
reliability R = 0.1
(0.1,
are0.5 and 0.9)and
modeled, are three-parameter
expressed below,models respectively:
with three different fatigue reliability
values (0.1, 0.5 and 0.9) are expressed below, respectively:
 R = 0.1: ( S − 499.2021) 0.0715
N f = 11.5501
 
0.5: :((SS−− 499.2021
481.1823))
0.0715
0.6205
 = =0.1 NNf f==53.9676
 RR 11.5501

 (11)
 
 0.6205
R = 0.5 : ( S − 481.1823 ) 1.1550 Nf = 53.9676 (11)
 R= 0.9 : (S( − 478.0581))1.1550 Nff = 226.8041
R = 0.9 : S − 478.0581 N = 226.8041





Based on Equation (11), three P-S-N curves for FV520B-I in VHCF regime can be drawn as shown
Based on Equation (11), three P-S-N curves for FV520B-I in VHCF regime can be drawn as shown
in Figure 8.
in Figure 8.

Figure 8. P-S-N curves with the R = 0.1, 0.5, 0.9.

Figure 8. P-S-N curves with the R = 0.1, 0.5, 0.9.

3.3. Discussion
Based on Figure 8 and Equation (11), the combination (R,Nf,σar) can be obtained directly, and this
has important practical application value for actual engineering. Two characteristics of the FV520B-I
fatigue reliability can be figured out according to the P-S-N curves. First, in the case of constant stress
amplitude, there is obvious negative correlation between the reliability and fatigue life of FV520B-I:
Metals 2020, 10, 371 10 of 16

3.3. Discussion
Based on Figure 8 and Equation (11), the combination (R,Nf ,σar ) can be obtained directly, and this
has important practical application value for actual engineering. Two characteristics of the FV520B-I
fatigue reliability can be figured out according to the P-S-N curves. First, in the case of constant stress
amplitude, there is obvious negative correlation between the reliability and fatigue life of FV520B-I:
with increased fatigue life, fatigue reliability decreases. For example, with σar = 575 MPa, when the
fatigue life is around 107 cycles, the fatigue reliability is about 0.9. This means the fatigue life of
FV520B-I can achieve the VHCF level with a high probability. However, when the fatigue life increases
to about 108 cycles, fatigue reliability decreases from 0.9 to 0.5, even as low as 0.1, which means that
the fatigue failure can easily occur in the regime at 108 cycles.
Second, Figure 8 indicates that the decreasing the stress amplitude is an effective way to increase
fatigue life and enhance fatigue reliability at the same time. It should be noticed that the stress
amplitude used in the fatigue test is the maximum stress; this is calculated using the applied fatigue
load set by the test system and the interface size of the specimen. However, in practical engineering
conditions, the influence of stress amplitude depends on the mean stress and, particularly, on fatigue
crack size. Under the same applied loading conditions, the specimen with smaller fatigue cracks will
suffer smaller stress amplitudes, which will result in long fatigue life and high reliability. Meanwhile,
large fatigue cracks will lead to serious stress concentration and big stress amplitude on the FV520B-I
component, so fatigue life will be reduced and failure will easily occur.
So, an assumption is proposed based on the above theoretical analysis: that the influence of stress
amplitude on fatigue reliability can be replaced by the influence of fatigue cracks on fatigue reliability;
there should be a quantitative relationship between reliability, fatigue cracking, and fatigue life.

4. The Fatigue Life Evaluation for FV520B-I

4.1. The Fatigue Life Evaluation Model Considering Reliability

It has been reported and demonstrated that the fatigue cracking is the leading cause of VHCF
failure of FV520B-I; the fatigue test results in this paper are consistent with the reported conclusions.
Up to now, the reported fatigue life prediction model for FV520B-I can only quantitatively describe the
single relationship between fatigue cracking and fatigue life, and the relevant fatigue reliability in this
situation is not clear. In order to provide more comprehensive theoretical support for the maintenance
of FV520B-I specimens in practical engineering, a new fatigue life evaluation model for FV520B-I in
the VHCF regime is required to quantitatively express fatigue reliability and life. This new model is
creatively developed with the combined application of the reported life model and a P-S-N curve,
which can be employed to evaluate reliability and life at the same time, to improve the applicability of
FV520B-I fatigue analysis results in practical engineering.
The fatigue life of FV520B-I in VHCF regime is employed as the transition variable of to integrate
the reliability and crack into one united model. Thus, the first step in the development of this new
model is to determine the fatigue life of FV520B-I. Zhang [20] proposed an specific very-high cycle
fatigue life model for FV520B-I, as displayed in Equation (12)

2 G
9∆Kth Sc(1−n/2)
Nf = + (12)
2E(σa − σr )2 Sc
 
Cσa n βn1 πn/2 n2 − 1

where Sc is the fatigue cracking in µm (here the inclusion size for FV520B-I); HV is Vickers hardness,
HV = 380, kgf/mm2 ; σa is the stress amplitude, MPa; elasticity modulus E = 2.1 × 105 , MPa; shear
modulus G = 8 × 104 , MPa; β1 is the geometric constant β1 = 0.5 × π0.5 ; ∆Kth , C, n and α are the
material parameters: ∆Kth = 6.3 MPa·m1/2 , and C and n are dimensionless constants: C = 3.14 × 10−14 ,
n = 2.95 [20].
Metals 2020, 10, 371 11 of 16

In order to keep the same stress amplitude conditions in the derivation process of the P-S-N
Metals 2020, 10, x FOR PEER REVIEW 11 of 16
curve model of FV520B-I, stress amplitudes of 600 MPa, 525 MPa and 500 MPa were employed. The
quantitative relationship curves between fatigue crack and fatigue life with 600 MPa, 525 MPa and
500 MPaMetals 2020, drawn
were 10, x FOR in
PEER REVIEW
Figure 9, based on Equation (12). 11 of 16

Figure 9. Fatigue life curves with

Figure 9. with 500
500 MPa,
MPa, 525
525 MPa,
MPa, and
and 600
600 MPa.
MPa.
Figure 9. Fatigue life curves with 500 MPa, 525 MPa, and 600 MPa.

The
The second
second step is
is to
stepstep to combine the fatigue reliability, fatigue life and cracking into one unified
The second is combine
to combine thethefatigue
fatigue reliability, fatiguelife
reliability, fatigue lifeand
and cracking
cracking intointo
one one unified
unified
model.
model. As mentioned above, in most situations, the P-S-N curve with reliability
model. As mentioned above, in most situations, the P-S-N curve with reliability R = 0.5 has beenbeen
As mentioned above, in most situations, the P-S-N curve with reliability RR =
= 0.5
0.5 has
has been
widely
widely accepted
accepted
widely as
as the
accepted the generic
generic
as the S-N
genericS-N
S-Ncurve,
curve,for
curve, forinvestigating
for investigating
investigating or or quantitatively
orquantitatively
quantitatively expressing
expressing
expressing the the
the fatigue
fatigue
fatigue
properties
properties of metallic
of metallic
properties material.
material.
of metallic Thus,
Thus,
material. R = 0.5
Thus,RR= =0.5 is also
0.5isis alsochosen
also chosen as
chosen as the
asthebasic
thebasic
basicsituation
situation
situation in the new
in the
in the newnew fatigue life
fatigue
fatigue
evaluation model
life evaluation
life evaluation ofmodel
model FV520B-I in the in
of FV520B-I
of FV520B-I VHCF
inthe regime.
theVHCF
VHCF regime.The fatigue
regime. The life with
Thefatigue
fatigue life
life =with
Rwith 0.5
R under
=R0.5 three
under
= 0.5 different
three
under three
stress different
amplitude stress amplitude
values is values
estimated is
as estimated
shown in as shown
Table 4. in
By Table 4. By
substituting substituting
different stress amplitude values is estimated as shown in Table 4. By substituting the fatigue life the fatigue the fatigue
life with life
R = 0.5
into with R = 0.5
Equation into
(12), theEquation (12), the corresponding
corresponding fatigue crackingfatigue
can cracking
be can beas
obtained obtained
displayedas displayed
in Table in
6 and
with R = 0.5 into Equation (12), the corresponding fatigue cracking can be obtained as displayed in
Figure Table 6 and Figure 10.
Table 610. and Figure 10.
Table 6. Stress amplitude, fatigue reliability, fatigue life and fatigue cracking.
Table 6. Stress
Table 6. Stress amplitude,
amplitude, fatigue
fatigue reliability,
reliability,fatigue
fatiguelife
lifeand
andfatigue
fatiguecracking.
cracking.
Stress Amplitude σa/MPa 600 525 500
Stress Amplitude σ a /MPa
Stress Amplitude
Fatigue σa/MPa
reliability R 600 600
0.5 525 525 0.5500
0.5 500
Fatigue
Fatigue
Fatigue reliability R life N 2.78
reliability R 0.5 0.5
f × 10 7 5.17 × 107 0.58.73 ×0.5
0.5 107 0.5
Fatigue crack Sc/μm
Fatigue life Nf Fatigue life Nf 2.78 × 2.78 7322
10 × 10
7 139.4
5.17 5.17
× 10 × 10
7 90.6
8.73 × 107
7 8.73 × 107
Fatigue crack Sc/μm 322 322
Fatigue crack Sc/µm 139.4139.4 90.6 90.6

Figure 10. The fatigue life with R = 0.5.

It is clear that there is a non-linear relationship between fatigue cracking (Sc) and fatigue life (Nf)
Figure 10. The fatigue life with R = 0.5.
with R = 0.5. However, no previous Figure model
10. Thehas beenlife
fatigue reported
with R =to0.5.
fit this relationship based on the
distribution characteristics of fatigue life with R = 0.5 that is a nonlinear exponential distribution.
It is clear that there is a non-linear relationship between fatigue cracking (Sc) and fatigue life (Nf)
with R = 0.5. However, no previous model has been reported to fit this relationship based on the
distribution characteristics of fatigue life with R = 0.5 that is a nonlinear exponential distribution.
Metals 2020, 10, 371 12 of 16

It is clear that there is a non-linear relationship between fatigue cracking (Sc) and fatigue life
Metals 2020, 10, x FOR PEER REVIEW 12 of 16
(N f ) with R = 0.5. However, no previous model has been reported to fit this relationship based on
the distribution characteristics of fatigue life with R = 0.5 that is a nonlinear exponential distribution.
Thus, the form of the two-parameter exponential model is employed as a reference to develop the
Thus, the form of the two-parameter exponential model is employed as a reference to develop the new
new fatigue life evaluation model. This model is employed to develop the quantitative expression
fatigue life evaluation model. This model is employed to develop the quantitative expression between
between reliability, fatigue cracking, and fatigue life, as displayed:
reliability, fatigue cracking, and fatigue life, as displayed:
ScaaRR N f = bR (13)
Sc Nf = bR (13)
where Sc is fatigue cracking, μm; Nf is fatigue life; and aR and bR are model parameters related to the
fatigueScreliability
where is fatigueofcracking,
FV520B-I. Nf isthe
µm;With fatigue
samelife; and aR and
derivation bR areasmodel
process parameters
Equation (10) andrelated to the
the data in
fatigue
Tab.6, areliability
nonlinear of FV520B-I. With theshould
fitting algorithm same derivation
be adopted process as Equation
to identify (10) and theparameters
the unknown data in Tab.6,
in
aEquation
nonlinear fitting
(13). Thealgorithm
estimatedshould
resultsbe
foradopted
the modelto identify the unknown
parameters are: a0.5 = parameters in Equation
0.91, bR = 484.5. (13).
So, the new
The estimated results for the model parameters
fatigue life evaluation model for FV520B-I in VHCF regime are: a 0.5 = 0.91, b = 484.5. So, the
with RR= 0.5 is written as: new fatigue life
evaluation model for FV520B-I in VHCF regime with R = 0.5 is written as:
Sc 0.91 N f = 484.5 (14)
0.91
Sc Nf = 484.5 (14)
4.2. Further
4.2. Further Assumption:
Assumption: The
The P-Sc-N
P-Sc-N Curve
Curve
As discussed
As discussed above,
above, the the new
new fatigue
fatigue life
life evaluation
evaluation model model can can reasonably
reasonably and and quantitatively
quantitatively
describe the relationship between fatigue cracking and fatigue life in
describe the relationship between fatigue cracking and fatigue life in a general condition (that a general condition (that is,
is, the
the
condition of reliability R = 0.5). The fatigue life model with R = 0.5 can
condition of reliability R = 0.5). The fatigue life model with R = 0.5 can satisfy most of application satisfy most of application
requirementsof
requirements ofgeneral
generalpractical
practicalengineering,
engineering,just justasasthethe commonly
commonly used
used S-NS-N curve
curve represents
represents a P-
a P-S-N
S-N curve with a reliability of 0.5, which can be applied in most cases. But P-S-N
curve with a reliability of 0.5, which can be applied in most cases. But P-S-N curves with other reliability curves with other
reliability
values arevalues are still to
still required required
produce to aproduce a comprehensive
comprehensive reliability reliability
analysisanalysis for FV520B-I
for FV520B-I in the
in the VHCF
VHCF regime. Thus, it is also necessary to improve the application range
regime. Thus, it is also necessary to improve the application range of the new fatigue life evaluation of the new fatigue life
evaluation model by establishing fatigue life evaluation models with
model by establishing fatigue life evaluation models with different reliability values, especially for different reliability values,
especially
some forsituations
critical some criticallike Rsituations
= 0.1 and like R = Based
R = 0.9. 0.1 and on Rthe=above
0.9. Based
pointson and the above pointsa new
requirements, and
requirements, a new comprehensive assumption is proposed to describe
comprehensive assumption is proposed to describe the relationship between fatigue reliability, fatigue the relationship between
fatigue reliability,
cracking and fatigue fatigue cracking
life (that is, theand fatiguecurve).
P-Sc-N life (thatThe is,proposed
the P-Sc-N curve).curve
P-Sc-N The proposed
can improve P-Sc-N
the
curve can improve the accuracy of the fatigue reliability analysis by considering
accuracy of the fatigue reliability analysis by considering the influence of fatigue cracking on fatigue the influence of
fatigue cracking on fatigue failure, and it is also beneficial to the design
failure, and it is also beneficial to the design of procedures in re-manufacturing engineering under the of procedures in re-
manufacturing
material fatigue engineering
failure control. under the material
The logical relationshipfatigue failurefatigue
between control. The logical
reliability, relationship
fatigue crack and
betweenlife
fatigue fatigue
can bereliability,
expressedfatigue
as Figurecrack
11:and fatigue life can be expressed as Figure 11:

Figure 11. The logical relationship between fatigue reliability, fatigue crack and fatigue life.
Figure 11. The logical relationship between fatigue reliability, fatigue crack and fatigue life.
As displayed in Equation (14), the new fatigue life evaluation model for FV520B-I is the P-Sc-N
curveAs displayed
with R = 0.5,inand
Equation
P-Sc-N(14), the in
curves new fatigue
critical life evaluation
conditions with Rmodel for FV520B-I
= 0.9 and is the
R = 0.1 will P-Sc-N
be further
curve with to
developed R enrich
= 0.5, and P-Sc-N curves
the P-Sc-N curves ofinFV520B-I
critical conditions
in the VHCFwithregime.
R = 0.9 and R = 0.1 will be further
developed to enrich the P-Sc-N curves of FV520B-I in the VHCF regime.
Referring to the derivation process of the fatigue life evaluation model with R = 0.5(Equation
(14)),Referring to thefitting
the nonlinear derivation process
algorithm wasofalso
theadopted
fatigue life evaluation
to identify the model
unknown with R = 0.5(Equation
model parameters.
(14)), the
Then, the nonlinear
new P-Sc-N fitting
modelalgorithm wasinalso
of FV520B-I VHCFadopted
regime to with
identify the unknown
the reliability of R model parameters.
= 0.9 and 0.1 could
Then, the new P-Sc-N model of FV520B-I in VHCF regime with the reliability of R = 0.9 and 0.1 could
be specifically developed. For situations with R = 0.9, 0.5 and 0.1, the quantitative expressions
between fatigue reliability, fatigue life and fatigue cracking for FV520B-I in the VHCF regime are
written as:
Metals 2020, 10, 371 13 of 16

Metals 2020, 10, x FOR

be specifically For situations with R = 0.9, 0.5 and 0.1, the quantitative expressions between
PEER REVIEW
developed. 13 of 16
fatigue reliability, fatigue life and fatigue cracking for FV520B-I in the VHCF regime are written as:
 R =0.9 : Sc 0.68 N = 232.1
0.68 f
R = 0.9 : Sc0.91 Nf = 232.1


RR==0.5 Sc0.91NNf f==484.
0.5 ::Sc 5

(15)

484.5 (15)




1.28
0.1 : Sc NNf f ==107
R R==0.1: 1.28
4.1

 1074.1

The new fatigue life evaluation models, with the consideration of fatigue reliability and fatigue
cracking
cracking as expressed in Equation (15), also present P-Sc-N curve models of FV520B-I with three
different fatigue reliabilities. Based on Equation (15), the P-Sc-N curves of FV520B-I with R =
= 0.1, 0.5
and 0.9 can be drawn as shown in Figure 12.

Figure
Figure 12.
12. The P-Sc-N curve
The P-Sc-N curve of
of FV520B-I
FV520B-I with R == 0.1,
with R 0.1, 0.5
0.5 and
and 0.9.
0.9.

4.3. Discussions
4.3. Discussions
Figure 12 indicates some observations. First, both the fatigue life (Nf ) and the corresponding
Figure 12 indicates some observations. First, both the fatigue life (Nf) and the corresponding
fatigue reliability (R) can be determined based on the P-Sc-N curves with a certain fatigue crack value.
fatigue reliability (R) can be determined based on the P-Sc-N curves with a certain fatigue crack value.
The fatigue reliability analysis of FV520B-I in the VHCF regime is simplified to some extent.
The fatigue reliability analysis of FV520B-I in the VHCF regime is simplified to some extent.
Second, the P-Sc-N curve becomes more gradual with the increase of R from 0.1 to 0.5, and this
Second, the P-Sc-N curve becomes more gradual with the increase of R from 0.1 to 0.5, and this
continues to 0.9. A small change in fatigue cracking will result in a large variation in fatigue life. This
continues to 0.9. A small change in fatigue cracking will result in a large variation in fatigue life. This
implies that fatigue life is more sensitive to the change in fatigue cracking when fatigue reliability
implies that fatigue life is more sensitive to the change in fatigue cracking when fatigue reliability is
is high.
high.
Third, with an increase in fatigue life, the fatigue reliability appears to be more sensitive to the
Third, with an increase in fatigue life, the fatigue reliability appears to be more sensitive to the
change in fatigue cracking. A small decrease in fatigue cracking will realize an improvement in
change in fatigue cracking. A small decrease7 in fatigue cracking will realize an improvement in
reliability. For example, in the case of a 5 × 10 -cycle fatigue life, fatigue reliability based on fatigue
reliability. For example, in the case of a 5 × 107-cycle fatigue life, fatigue reliability based on fatigue
cracking is improved from 0.5 to 0.9 by a big decrease in fatigue cracking from about 150 µm to about
cracking is improved from 0.5 to 0.9 by a big decrease in fatigue cracking from about 150 μm to about
70 µm (the difference value is around 80 µm). However, in the case of a 1 × 108 -cycle fatigue life, the
70 μm (the difference value is around 80 μm). However, in the case of a 1 × 108-cycle fatigue life, the
fatigue reliability based on fatigue cracking is improved from 0.5 to 0.9 by a relatively small decrease
fatigue reliability based on fatigue cracking is improved from 0.5 to 0.9 by a relatively small decrease
in fatigue cracking, about 30 µm.
in fatigue cracking, about 30 μm.
Fourth, in the case of the certain fatigue life, decreasing the fatigue crack is an effective and direct
Fourth, in the case of the certain fatigue life, decreasing the fatigue crack is an effective and direct
method to improve the fatigue reliability of FV520B-I in VHCF regime.
method to improve the fatigue reliability of FV520B-I in VHCF regime.
More than that, the proposition of the P-Sc-N curve is also useful to the decision-making process
More than that, the proposition of the P-Sc-N curve is also useful to the decision-making process
in remanufacturing based on the fatigue crack message. Remanufacturing is a kind of high-tech repair
in remanufacturing based on the fatigue crack message. Remanufacturing is a kind of high-tech repair
and transformation industry for the mechanical component with defect or to be scrapped parts. Based
and transformation industry for the mechanical component with defect or to be scrapped parts. Based
on the failure analysis, life assessment and other analysis, remanufacturing engineering design is
on the failure analysis, life assessment and other analysis, remanufacturing engineering design is
carried out, and a series of related advanced manufacturing technologies are adopted to make the
carried out, and a series of related advanced manufacturing technologies are adopted to make the
quality of remanufactured products reach or exceed new products. In remanufacturing applications, if
quality of remanufactured products reach or exceed new products. In remanufacturing applications,
the fatigue crack can satisfy the specification requirement, that is the existing fatigue crack Sc is smaller
if the fatigue crack can satisfy the specification requirement, that is the existing fatigue crack Sc is
smaller than the fatigue crack critical threshold of remanufacturing Ac, the object will be re-
manufactured. And the critical situation for remanufacturing is that Sc = Ac [1]. Then, different
combinations of reliability and fatigue life (R,Nf) in the critical situation for remanufacturing are also
Metals 2020, 10, 371 14 of 16

than the fatigue crack critical threshold of remanufacturing Ac, the object will be re-manufactured. And
the critical situation for remanufacturing is that Sc = Ac [1]. Then, different combinations of reliability
and fatigue life (R,Nf ) in the critical situation for remanufacturing are also determined by the P-Sc-N
curves. At this time, the fatigue reliability also presents the reliability that the fatigue crack and fatigue
life satisfy the requirement of remanufacturing. The P-Sc-N curve of FV520B-I can also provide some
useful information to the decision-making of remanufacturing judgment. The result can be optimized
based on the P-Sc-N curve, by analyzing the changing relationship between the fatigue reliability and
fatigue life under the certain fatigue crack, which provide a strong theoretical connection between
material science and re-manufacturing practice.

5. Conclusions
A few conclusions can be drawn from this study:

(1) The ultrasonic fatigue test for FV520B-I was performed. Obvious evidence of very-high cycle
fatigue failure in FV520B-I, including “fish-eye” and GBF regions, were detected on the fracture
surface; these are constant with the reported results of the very-high cycle fatigue study.
(2) A three-parameter model for FV520B-I was achieved. The unknown parameters related to
FV520B-I were obtained, and the fatigue reliability of FV520B-I in the VHCF regime was modeled.
The corresponding P-S-N curves with different fatigue reliability values were drawn, respectively.
The determination of the three-parameter model and the P-S-N curve for FV520B-I make up for
the lack of a fatigue reliability study of FV520B-I in the VHCF regime.
(3) A new fatigue evaluation model for FV520B-I, considering fatigue reliability and fatigue crack,
was developed based on the P-S-N curves. This new model was verified using the test data and
the errors were in an acceptable range.
(4) A new relationship curve between fatigue reliability, fatigue crack and fatigue life N for FV520B-I
in the VHCF regime was proposed: the P-Sc-N curve. The P-Sc-N curve model with different
reliability values was developed with the combined application of the classic formula and test data.
The corresponding P-S-N curves with different fatigue reliability values were drawn, respectively.
(5) Using the P-Sc-N curve for FV520B-I, it was observed that there is a negative correlation between
fatigue cracking and the reliability for FV520B-I. The new P-Sc-N curve is also useful for the
decision-making process in re-manufacturing, based on the fatigue cracking information.

However, this research on fatigue reliability based on fatigue crack, is still a framework for potential
research, and needs further investigation and deeper understanding. Specifically: (i) the rationality of
the P-Sc-N curve has to be tested by a comprehensive application of fatigue experimentation and in
actual engineering cases; (ii) more fatigue testing of FV520B-I is still required to modify the new model
and the unknown model parameters, to improve the accuracy of the fatigue reliability analysis. The
concerns discussed in this paper are obviously open to future research investigation and study.

Author Contributions: Conceptualization, J.W. (Jinlong Wang); methodology, J.W. (Jinlong Wang) and Y.Y.;
software, Y.Y.; investigation, J.Y. and F.D.; resources, J.W. (Jingshi Wang); writing—original draft preparation, J.W.
(Jinlong Wang); writing—review and editing, J.W. (Jinlong Wang), J.Y. and Y.Z. All authors have read and agreed
to the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China, grant number No.
51875082; National Defense Basic Scientific Research program of China, No.2019-JCJQ-JJ-547.
Acknowledgments: Thanks to Zhang Yuanliang, Ding Mingchao in Dalian University of Technology for their
technical supports in this research work. And the authors have no conflict of interest to declare.
Conflicts of Interest: The authors declare no conflict of interest.
Metals 2020, 10, 371 15 of 16

References
1. Wang, J.L.; Zhang, Y.L.; Zhao, Q.C.; Zhang, H.C. Study of fatigue crack critical threshold of remanufacturing
cores and judgment of re-manufacturability based on fatigue crack. Chin. J. Mech. Eng. CHS 2017, 53, 41–49.
[CrossRef]
2. Zhang, M.; Wang, W.Q.; Wang, P.F.; Liu, Y.; Li, J.F. Fatigue behavior and mechanism of FV520B-I welding
seams in a very high cycle regime. Int. J. Fatigue. 2016, 87, 22–37. [CrossRef]
3. Fujii, T.; Maekawa, Z. Fatigue Life Analysis of Particle Dispersed Composite Materials by Reliability Concept.
J. Soc. Mater. Sci. Jpn. 1978, 27, 984–990. [CrossRef]
4. Weibull, W. Fatigue Testing and Analysis of Results; Pergamon Press: London, UK, 1961.
5. Weibull, W. A Statistical function of the wide applicability. J. Appl. Mech. 1951, 18, 293–300.
6. Smaga, M.; Boemke, A.; Daniel, T.; Skorupski, R.; Sorich, A.; Beck, T. Fatigue behavior of metastable austenitic
stainless steels in LCF, HCF and VHCF regimes at ambient and elevated temperatures. Metals 2019, 9, 704.
[CrossRef]
7. Rozumek, D.; Marciniak, Z.; Lesiuk, G.; Correia, J.A.; de Jesus, A.M. Experimental and numerical investigation
of mixed mode I+ II and I+ III fatigue crack growth in S355J0 steel. Int. J. Fatigue 2018, 113, 160–170.
[CrossRef]
8. Hong, Y.S.; Liu, X.L.; Lei, Z.Q.; Sun, C.Q. The formation mechanism of characteristic region at crack initiation
for very-high-cycle fatigue of high-strength steels. Int. J. Fatigue 2016, 89, 108–118. [CrossRef]
9. Wirsching, P.H.; Wu, Y.T. Probabilistic and Statistical Methods of Fatigue Analysis and Design. Press. Vessel
Pip. Technol. A Decade Prog. 1985, 793–819.
10. Wu, Y.T.; Wirsching, P.H. Advanced Reliability Method for Fatigue Analysis. J. Eng. Mech. 1984, 110, 536–553.
[CrossRef]
11. Martin, W.S.; Wirsching, P.H. Fatigue crack initiation-propagation reliability model. J. Mater. Civil. Eng. 1991,
3, 1–18. [CrossRef]
12. Martindale, S.G.; Wirsching, P.H. Reliability-based Progressive Fatigue Collapse. J. Struct. Eng. 1983, 109,
1792–1811. [CrossRef]
13. Wang, H.; Guo, L.Z.; Lin, H.Q.; Yang, Y.L. Research on Life of Sliding Rail and Idler Wheel and Reliability
Under Constant Amplitude Load. Bearing 2016, 10, 36–39.
14. Wang, W.S.; Zhao, L.; Huang, Y.; Zhu, H.C.; Su, Y.N. The bending fatigue strength test based on the reliability
of material 8822H. Locomot. Roll. Stock Technol. 2015, 3, 35–37.
15. Wurzel, G.; Alderliesten, R. Reliability modeling for rotorcraft component fatigue life prediction with
assumed usage. Aeronaut. J. 2016, 1, 1–35.
16. Paolino, D.S.; Tridello, A.; Chiandussi, G.; Rossetto, M. Effect of crack size on P-S-N curves in Very-High-Cycle
Fatigue. Procedia Struct. Integr. 2017, 7, 335–342. [CrossRef]
17. Liu, S.J.; Liu, C.; Hu, Y.W.; Gao, S.B.; Wang, Y.F.; Zhang, H.C. Fatigue life assessment of centrifugal compressor
impeller based on FEA. Eng. Fail. Anal. 2016, 60, 383–390. [CrossRef]
18. Liu, C.; Liu, S.J.; Gao, S.B.; Hu, Y.W.; Zhang, S.X.; Zhang, H.C. Fatigue life assessment of the centrifugal
compressor impeller with cracks based on the properties of FV520B. Eng. Fail. Anal. 2016, 66, 177–186.
[CrossRef]
19. Wang, J.L.; Zhang, Y.L.; Zhao, Q.C.; Zhang, M.; Guan, Z.M.; Lu, H.T. The fatigue failure analysis and fatigue
life prediction model of FV520B-I as a function of surface roughness in HCF regime. J. Mater. Res. 2017, 32,
634–643. [CrossRef]
20. Zhang, Y.L.; Wang, J.L.; Sun, Q.C.; Zhang, H.; Jiang, P.S. Fatigue Life Prediction of FV520B with Internal
Inclusions. Mater. Des. 2015, 69, 241–246. [CrossRef]
21. Wang, J.L.; Zhang, Y.L.; Sun, Q.C.; Liu, S.J.; Shi, B.W.; Lu, H.T. Giga-fatigue life prediction of FV520B-I with
surface roughness. Mater. Des. 2016, 89, 1028–1034. [CrossRef]
22. Wang, J.L.; Zhang, Y.L.; Shi, B.W.; Zhao, Q.C.; Sun, Q.C.; Zhang, Z.; Lu, H.Y. Fatigue behavior and mechanism
of FV520B-I owing to the effect of loading frequency on the fatigue property in HCF and VHCF regime. Arch.
Metall. Mater. 2018, 63, 347–357.
23. Shiozawa, K.; Lu, L.; Ishihara, S. S-N curve characteristics and subsurface crack initiation behaviour in
ultra-long life fatigue of a high carbon-chromium bearing steel. Fatigue Fract. Eng. Mater. 2010, 24, 781–790.
[CrossRef]
Metals 2020, 10, 371 16 of 16

24. Shiozawa, K.; Lu, L. Very high-cycle fatigue behaviour of shot-peened high-carbon-chromium bearing steel.
Fatigue Fract. Eng. Mater. 2010, 25, 813–822. [CrossRef]
25. Shiozawa, K.; Morii, Y.; Nishino, S.; Lu, L. Subsurface crack initiation and propagation mechanism in
high-strength steel in a very high cycle fatigue regime. Int. J. Fatigue 2006, 28, 1521–1532. [CrossRef]
26. Murakami, Y. Mechanism of fatigue failure in ultralong life regime and application to fatigue design. Fatigue
Fract. Eng. Mater. 2002, 15, 2927–2938.
27. Murakami, Y.; Takada, M.; Toriyama, T. Super-longlife tension-compression fatigue properties of quenehed a
tempered 0.46% Carbon steel. Int. J. Fatigue 1998, 16, 661–667. [CrossRef]
28. Murakami, Y. Metal Fatigue-Effects of Small Cracks and Nonmetallic Inclusions; Elsevier: Amsterdam, The
Netherlands; Boston, MA, USA, 2002; Volume 6–7, pp. 314–316.
29. Chen, J.Q. Models for fatigue life distribution. J. Mech. Strength 1991, 13, 42–45.
30. Fu, H.M.; Gao, Z.T. A method for fitting P-S-N curve. Acta Aeronaut. ET Astronaut. Sin. 1988, 9, 338–341.
31. Basquin, O.H. The Exponential Law of Endurance Tests. Am. Soc. Test. Mater. Proc. 1910, 10, 625–630.
32. Ling, J.; Pan, J. A maximum likelihood method for estimating P-S-N curves. Int. J. Fatigue 1997, 19, 415–419.
[CrossRef]
33. Zhao, Y.X.; Wang, J.N.; Gao, Q. A unified classical maximum likelihood approach for estimating P-S-N
curves of three commonly used fatigue stress-life relations. Chin. J. Appl. Mech. 2001, 18, 83–90.
34. Sun, Z.D.; Li, W.; Zhang, Z.Y.; Deng, H.L.; Zhang, X.H. A method to building probabilistic S-N curves for
multiple failure modes in very high cycle fatigue. Eng. Mech. 2016, 33, 218–224.

© 2020 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/).