applied

sciences
Article
Fatigue Life Estimation of Medium-Carbon Steel with
Different Surface Roughness
Changyou Li *, Weibing Dai, Fei Duan, Yimin Zhang and David He
School of Mechanical Engineering and Automation, Northeastern University, Shenyang 110004, China;
yifanwb@163.com (W.D.); neufei@163.com (F.D.); ymzhang@mail.neu.edu.cn (Y.Z.); davidhe@uic.edu (D.H.)
* Correspondence: chyli@mail.neu.edu.cn

Academic Editor: Cem Selcuk

Received: 2 March 2017; Accepted: 24 March 2017; Published: 29 March 2017

Abstract: Medium-carbon steel is commonly used for the rail, wire ropes, tire cord, cold heading,
forging steels, cold finished steel bars, machinable steel and so on. Its fatigue behavior analysis and
fatigue life estimation play an important role in the machinery industry. In this paper, the estimation
of fatigue life of medium-carbon steel with different surface roughness using established S-N and
P-S-N curves is presented. To estimate the fatigue life, the effect of the average surface roughness on
the fatigue life of medium-carbon steel has been investigated using 75 fatigue tests in three groups
with average surface roughness (Ra ): 0.4 µm, 0.8 µm, and 1.6 µm, respectively. S-N curves and P-S-N
curves have been established based on the fatigue tests. The fatigue life of medium-carbon steel is
then estimated based on Tanaka-Mura crack initiation life model, the crack propagation life model
using Paris law, and material constants of the S-N curves. Six more fatigue tests have been conducted
to validate the presented fatigue life estimation formulation. The experimental results have shown
that the presented model could estimate well the mean fatigue life of medium-carbon steel with
different surface roughness.

Keywords: medium-carbon steel; fatigue life estimation; surface roughness

1. Introduction
Medium-carbon steel is usually referred to as the carbon steel that contains approximately
0.30–0.60% carbon content. It balances ductility and strength, and has good wear resistance [1]. It is
commonly used for rails, wire ropes, tire cord, cold heading, forging steels, cold finished steel bars,
machinable steel, and so on [2]. Its fatigue behavior analysis and fatigue life estimation play an
important role in the safety and reliability in the machinery industry. From a raw steel piece to a part
meeting the design requirements, a manufacturing process must be used. The surface roughness of a
part as a result of the manufacturing process plays a major role in determining the fatigue behavior
of the part [3]. To ensure the high safety and reliability of the part of medium-carbon steel in service,
it is important to investigate S-N curve, P-S-N curve, and fatigue life of medium-carbon steel with
different surface roughness.
A method has been reported where the effect of the surface roughness on fatigue behavior was
commonly considered by using surface condition modification factor, which was:
b
k a = b1 Sut1 (1)

where Sut was the minimum tensile strength, b1 and b2 are statistical values and determined by
the material and the machining method such as ground, machined or cold-drawn, hot-rolled, and

Appl. Sci. 2017, 7, 338; doi:10.3390/app7040338 www.mdpi.com/journal/applsci

Appl. Sci. 2017, 7, 338 2 of 11

as-forged [4]. Then, the endurance limit Se at the critical location of a machine part in the geometry
and condition of use is proportional to the product of k a and Se0 :

Se ∝ k a S`e (2)

where S`e is the rotary-beam test specimen endurance limit. However, it is very difficult for this method
to estimate the fatigue life of the given material with the different surface roughness. Other methods
have also been proposed. For example, the surface groove due to the manufacturing process was
considered to result in stress concentration similar to a notch. The fatigue stress concentration factor Kf
was used for describing the effect of the surface roughness on the fatigue limit. It was the function
of the stress concentration factor Kt which can be estimated by the finite element method [5,6] or the
empirical formula including the average geometric parameters of surface roughness [7,8]. A method
has been also reported to estimate the fatigue life by:

N f = Ni + Np (3)

where Nf , Ni , and Np were fatigue life (the number of cycles to failure), crack initiation life, and crack
propagation life, respectively. The fatigue life can be computed as:

Ni = β(Kt S)α (4)

where β and α are determined by plotting Ni as a function of the stress amplitude S. The crack
propagation life could be estimated by iterative numerical integration of Paris’ law where the surface
fatigue crack propagation threshold is multiplied by Kt and the fatigue crack propagation threshold
in depth is not affected [5]. For the large surface roughness, the stress concentration due to the
surface groove might be obvious and the methods of estimating the fatigue life based on the stress
concentration factor might have high accuracy. However, the stress concentration factor might be
very close to 1 for small surface roughness. Moreover, accurately computing the stress concentration
factor is a time-consuming and complex process. Therefore, the method has been reported where
the surface groove due to machining was considered to be an initial micro-defect. The analytic
formulation of the crack initiation life Ni can be estimated based on the dislocation dipole accumulation
model [9–11] and the crack propagation life Np can calculated using the analytic solution of Paris’ law
integral [12–14]. Since the reported method has been validated by six types of ultra-high strength
steel [11] and high-strength martensitic stainless steel [12], it will be employed to estimate the fatigue
life of medium-carbon steel with the different surface roughness in this paper.
The remainder of this paper is organized as follows: In Section 2, the material, the specimen
geometry, and the testing apparatus will be described, the scatterplot of the fatigue experiment
results shown, and the mean fatigue life of the specimens with different average surface roughness
contrasted. In Section 3, S-N and P-S-N curves will be established and their curve-fitting parameters
listed. In Section 4, the formulation for estimating the mean fatigue life of medium-carbon steel with
different average surface roughness will be explained along with its validation. The conclusions are,
finally, drawn in Section 5.

2. Material and Experimental Results

2.1. Material and Testing Apparatus

The material investigated in this work is the medium-carbon steel and its composition and
mechanical properties are presented in Tables 1 and 2, respectively. The raw material is a hot rolled
steel bar and is machined into the funnel specimen. The geometry of the specimen is shown in Figure 1.
The specimens were polished to the specified average surface roughness. To investigate the effect of the
different surface roughness of the specimens on the fatigue life, specimens were machined into three
Appl. Sci. 2017, 7, 338 3 of 11

groups with average surface roughness (Ra ): 0.4 µm, 0.8 µm, and 1.6 µm, respectively. The rotating
Appl. Sci. 2017, 7, x FOR PEER REVIEW 3 of 12
bending fatigue tests were performed. The testing apparatus is shown in Figure 2. The cycle frequency
Appl. Sci. 2017, 7, x FOR PEER REVIEW 3 of 12
was 25 Hz. Table 1. Chemical composition of the medium-carbon steel.
Table 1. Chemical composition of the medium-carbon steel.
Table 1. Chemical
Element C Crcomposition
Mo of theMn
Si medium-carbon
S steel.
P Ni
Element
Weight (%) 0.44C Cr 0.02
0.04 Mo 0.23 Si Mn 0.016
0.57 S P
0.024 Ni
0.002
Element C Cr Mo Si Mn S P Ni
Weight (%) 0.44 0.04 0.02 0.23 0.57 0.016 0.024 0.002
Weight (%)Table0.44 0.04 properties
2. Mechanical 0.02 0.23 0.57 carbon
of the medium 0.016 steel.
0.024 0.002
Table 2. Mechanical properties of the medium carbon steel.
Tensile Strength
Table 2.(MPa) Yield
Mechanical Strength
properties (MPa)
of the Elongation
medium at Break (%)
carbon steel.
Tensile Strength
710 (MPa) Yield Strength
490 (MPa) Elongation at Break (%)
19.17
710 (MPa)
Tensile Strength 490 (MPa)
Yield Strength 19.17
Elongation at Break (%)
710 490 19.17

Figure 1. Fatigue
Figure 1. Fatigue specimen
specimen geometry
geometry (dimensions
(dimensions in
in mm).
mm).
Figure 1. Fatigue specimen geometry (dimensions in mm).

Specimen
Specimen

Loading device
Loading device

Figure 2. The testing apparatus.

Figure 2. The testing apparatus.
Figure 2. The testing apparatus.
2.2. Experimental Results
2.2. Experimental Results
2.2. Experimental Results
In this work, 75 fatigue tests were conducted. They were divided into three groups with
In this work, 75 fatigue tests were conducted. They were divided into three groups with average
averageIn this work,
surface 75 fatigue
roughness (Ra) tests
of 0.4were conducted.
μm, 0.8 μm, and 1.6Theyμm,were divided The
respectively. into stress
three amplitudes
groups with
of
surface roughness (Ra ) of 0.4 µm, 0.8 µm, and 1.6 µm, respectively. The stress amplitudes of the
average surface(R
the first group roughness
a = 0.4 μm) (Rincluded
a) of 0.4 μm,
550, 0.8
500,μm,
450,and
400,1.6 μm,
and 380respectively. The stress
MPa, and those amplitudes
of other of
two groups
the first group
included (Ra =400,
500, 450, 0.4 μm)
380, included 550, 500,
and 360 MPa. 450, 400,
A total andfatigue
of five 380 MPa,
testsand those
were of other two
performed groups
under each
included 500, 450,The
stress amplitude. 400,scatterplot
380, and 360 MPa.
of all theAexperimental
total of five results
fatigue is
tests wereinperformed
shown under each
Figure 3. Under
stress amplitude.
amplitude, The
it canscatterplot
be seen thatof all
thethe experimental
fatigue results
life has the is showndispersion.
conspicuous in Figure 3.TheUnder each
means of
stress amplitude, it can be seen that the fatigue life has the conspicuous dispersion. The means of
Appl. Sci. 2017, 7, 338 4 of 11

first group (Ra = 0.4 µm) included 550, 500, 450, 400, and 380 MPa, and those of other two groups
included 500, 450, 400, 380, and 360 MPa. A total of five fatigue tests were performed under each
stress amplitude. The scatterplot of all the experimental results is shown in Figure 3. Under each stress
amplitude,
Appl. Sci. it can7, be
2017, seen
x FOR PEERthat the fatigue life has the conspicuous dispersion. The means of fatigue
REVIEW 4 of 12 life
with Appl. Sci. 2017,
different 7, x FOR PEER
average REVIEW
surface roughness and stress amplitude are compared in Figure 4. The 4 ofaverage
12
fatigue life with different average surface roughness and stress amplitude
surface roughness is clearly effective in reducing mean of fatigue life for medium-carbon steel. In are compared in Figure 4. high
fatigue
The life with
average different
surface average
roughness is surfaceeffective
clearly roughnessin and stress
reducing amplitude
mean of arelife
fatigue compared
for in Figure 4.
medium-carbon
stress amplitude (for example 500 MPa), there is about 15% decrease while there is about 30% decrease
The average
steel. In highsurface
stress roughness
amplitude is(for clearly effective
example 500 in reducing
MPa), theremean of fatigue
is about life for medium-carbon
15% decrease while there is
in low stress
steel. In amplitude
high (for example
stress amplitude 380 MPa)
(for amplitude
example when
500 MPa),the average
there surface decrease
is about roughness increased is from
about 30% decrease in low stress (for example 380 MPa)15%when the averagewhile there
surface
0.4 µm to 0.8 µm
about 30%increased or from
decrease from 0.8
in lowµm to 1.6 µm. This could confirm that the surface roughness plays a vital
roughness 0.4stress
μm toamplitude
0.8 μm or (for
fromexample
0.8 μm to380 1.6MPa) when
μm. This the confirm
could average that
surface
the
role in the
roughness
surface fatigue
roughnesslife
increased of medium-carbon
from
plays 0.4 μm
a vital roletoin0.8steel.
theμm or from
fatigue 0.8medium-carbon
life of μm to 1.6 μm. This
steel.could confirm that the
surface roughness plays a vital role in the fatigue life of medium-carbon steel.
Ra=0.4
550 =0.4
RRa=0.8
a
550 =0.8
Ra=1.6
Ra
Ra=1.6

500
(MPa)

500
(MPa)
amplitude
amplitude

450
450
Stress
Stress

400
400

350
350

0 100000 200000 300000 400000

0 100000 200000 300000 400000
Number of cycles to failure
Number of cycles to failure
Figure3.3.Scatterplot
Figure Scatterplot of
ofthe
the7575fatigue tests.
fatigue tests.
Figure 3. Scatterplot of the 75 fatigue tests.
380MPa
200000 380MPa
400MPa
200000
400MPa
450MPa
450MPa
500MPa
160000 500MPa
to failure

160000
to failure

120000
of cycles

120000
of cycles

80000
Number

80000
Number

40000
40000

0
0 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0.4 0.6 0.8 1.0 1.2 1.4 1.6
Average surface roughness (μm)
Average surface roughness (μm)
Figure 4. Contrast of mean fatigue life with Ra = 0.4 μm, Ra = 0.8 μm, and Ra = 1.6 μm.
Figure 4. Contrast of mean fatigue life with Ra = 0.4 μm, Ra = 0.8 μm, and Ra = 1.6 μm.
Figure 4. Contrast of mean fatigue life with Ra = 0.4 µm, Ra = 0.8 µm, and Ra = 1.6 µm.
Appl. Sci.
Sci. 2017,
2017, 7,
7, x338
FOR PEER REVIEW 5 of 12
11

3. S-N Curves and P-S-N Curves

3. S-N Curves and P-S-N Curves
The S-N curves with the different average surface roughness are shown in Figure 5. The input
The S-N
data points of curves with
the curve the different
fitting average
were obtained bysurface roughness
taking the logarithmareof
shown in Figure
the means of the5.fatigue
The input
life
data points of the curve fitting were obtained by taking the logarithm of the means of
data and stress amplitudes. The material constants were estimated by the least square method and the fatigue life
data and
are listedstress
in amplitudes.
Table 3. The The material
materialconstants weresatisfy
constants estimated
thebylogarithmic
the least square methodequation
Basquin and are
listed in Table 3. The material constants satisfy the logarithmic Basquin equation lgN = lgc − nlgS,
lg N f = lg c − n lg S , where S is the stress amplitude. As shown in Figure 5, (1) the mean of the fatigue
f
where S is the stress amplitude. As shown in Figure 5, (1) the mean of the fatigue life with Ra = 0.4 µm
life with Ra = 0.4 μm is very close to that with Ra = 0.8 μm or Ra = 1.6 μm when the stress amplitude is
is very close to that with Ra = 0.8 µm or Ra = 1.6 µm when the stress amplitude is very high; and (2)
very high; and (2) the mean of the fatigue life with Ra = 0.8 μm is very close to that with Ra = 1.6 μm,
the mean of the fatigue life with Ra = 0.8 µm is very close to that with Ra = 1.6 µm, while the mean of
while the mean of the fatigue life with Ra = 0.4 μm is significantly different from that with Ra = 0.8 μm
the fatigue life with R = 0.4 µm is significantly different from that with R = 0.8 µm or Ra = 1.6 µm
or Ra = 1.6 μm when athe stress amplitude is low. This phenomenon cana be explained as surface
when the stress amplitude is low. This phenomenon can be explained as surface damage due to
damage due to machining not being the key factor of fatigue crack initiation and development of
machining not being the key factor of fatigue crack initiation and development of medium-carbon
medium-carbon steel at higher-stress amplitudes than the endurance limit. It might be concluded
steel at higher-stress amplitudes than the endurance limit. It might be concluded that the mean of
that the mean of the fatigue life of the medium-carbon steel cannot be observably decreased when
the fatigue life of the medium-carbon steel cannot be observably decreased when the average surface
the average surface roughness is increased from Ra = 0.8 μm to Ra = 1.6 μm or even high at the
roughness is increased from Ra = 0.8 µm to Ra = 1.6 µm or even high at the low-stress amplitude.
low-stress amplitude.
500
Ra=1.6
Ra=0.8
450
Ra=0.4
Stress amplitude (MPa)

400

350

300

250
0 1000000 2000000 3000000 4000000 5000000 6000000
Number of cycles ot failure

Figure5.
Figure 5. The
The S-N
S-N curves
curves with
withRRaa =
= 0.4
0.4 μm, Raa == 0.8
µm, R 0.8 μm,
µm, and
and RRaa==1.6
1.6μm.
µm.

Table
Table 3.
3. Material
Material constants
constants with
with the
the different
different average surface roughness.
average surface roughness.

Average Surface Roughness (μm) n c

Average Surface Roughness (µm) n c
Ra = 0.4 9.84 4.56 × 1030
Ra = 0.4 9.84 4.56 × 1030
Ra = 0.8 9.44 3.16 × 102929
Ra = 0.8 9.44 3.16 × 10
Ra = 1.6
R = 1.6 9.80 9.80 2.32
2.32××10
103030
a

The P-S-N curves with Ra = 0.4 μm, Ra = 0.8 μm, and Ra = 1.6 μm are shown in Figures 6–8,
The P-S-N
respectively. R iscurves with Rprobability.
the survival a = 0.4 µm, R a = data
The 0.8 µm, and
point Ra =inverted
(pink 1.6 µm triangle,
are shown in triangle,
blue Figures 6–8,
red
respectively. R is the survival probability. The data point (pink inverted triangle,
dot and black square) is the confidence lower limit which is calculated by [13]: blue triangle, red dot
and black square) is the confidence lower limit which is calculated by [13]:
= ̅+ℎ (5)
x RL = x + hσx (5)
where x and σ x are the mean and the standard deviation of the experimental fatigue life, h is the
one-sided tolerance factor which is calculated by:
Appl. Sci. 2017, 7, 338 6 of 11

where x and σx are the mean and the standard deviation of the experimental fatigue life, h is the
one-sided tolerance
Appl. Sci. 2017, 7, x FOR factor which is calculated by:
PEER REVIEW 6 of 12
Appl. Sci. 2017, 7, x FOR PEER REVIEW 6 of 12
s
1
h = u R ξ − tγ 1 + u2 ( ξ 2 − 1) (6)
ℎ= − n1s + R ( − 1) (6)
ℎ= − + ( − 1) (6)
In Equation (6), u R = Φ−1 ( R), Φ−1 (>) is the inverse function of the standard normal distribution
In Equation (6), = Φ ( ) , Φ (⋇)−1is the inverse function of the standard normal
In Equation
function, ξ the correction = Φ tγ(=) T, −
(6), factor, Φ1 (γ(⋇)
) , T is(>)thetheinverse
inverse function
function ofofthethe standard function
distribution normal
distribution function, the correction factor, = T ( ) , T (⋇) the inverse function of the
distribution function, the correction factor, = T ( ) , T (⋇) the inverse
of Student’s t-distribution with ns − 1 degrees of freedom, and ns the sample size. When the survival function of the
distribution function of Student’s t-distribution with ns − 1 degrees of freedom, and ns the sample
distributionRfunction
probability and theofconfidence
Student’s level
t-distribution with the
γ are given, ns −confidence
1 degrees of freedom,
lower limit and ns the
can be sample
calculated
size. When the survival probability R and the confidence level are given, the confidence lower
size. When
byEquation the survival probability R and the confidence level are given,
(5). Then, the P-S-N curve is described by the similar logarithmic Basquin equation the confidence lower
limit can
 be calculated by Equation (5). Then, the P-S-N curve is described by the similar logarithmic
limit
lg N fcan=bea1calculated
− a2 lglgS, by Equation
where (5). Then, the P-S-N
a1 and a2 are curve is described by the similar logarithmic
Basquin equation = parameters
− lg S, where parameters estimated
a1 and a2byarethe least
estimated square method
by the with
least square
Basquin
Rmethod equation lg = − lg S, where parameters a1 and a2 are estimated by the least square
withRaR=
a = 0.4 µm, a =0.8
0.4µm, Ra =R0.8
μm,and a = μm,
1.6 µmandshown
Ra = 1.6inμm
Tables
shown4–6,in respectively.
Tables 4–6, respectively.
method with Ra = 0.4 μm, Ra = 0.8 μm, and Ra = 1.6 μm shown in Tables 4–6, respectively.

R=50% R=50%
R=50% R=50%
2.76 R=90% 2.76 R=90%
2.76 R=90% 2.76 R=90%
R=99% R=99%
R=99% R=99%
logarithmofofS S

R=99.9% R=99.9%
logarithmofofS S
2.72 R=99.9% 2.72 R=99.9%
2.72 2.72
base-10logarithm

base-10logarithm

2.68 2.68
2.68 2.68

2.64 2.64
2.64 2.64
base-10

base-10

2.60 2.60
2.60 2.60

2.56 2.56
2.56 -1 0 1 2 3 4 5 2.56 -1 0 1 2 3 4 5
-1 0base-10
1 logarithm
2 3 4 N 5
of -1 0 1 2 3 4 5
base-10 logarithm of Nf f base-10 logarithm of N
base-10 logarithm of Nf f
(a) (b)
(a) (b)
Figure 6.
6. The
The P-S-N
P-S-N curves
curves with
with Ra = 0.4 μm and confidence level (a)
(a) == 95%;
95%;and (b) =99%
and(b) 99%.
Figure
Figure 6. The P-S-N curves with RRaa==0.4
0.4μm andconfidence
µmand confidencelevel
level(a) γ = 95%; and (b) γ =
= 99%.

R=50% R=50%
R=50% R=50%
2.72 R=90% 2.72 R=90%
2.72 R=90% 2.72 R=90%
R=99% R=99%
R=99% R=99%
2.68 R=99.9% 2.68 R=99.9%
R=99.9% R=99.9%
logarithmofofS S

2.68 2.68
logarithmofofS S

base-10logarithm

2.64 2.64
base-10logarithm

2.64 2.64

2.60 2.60
2.60 2.60
base-10
base-10

2.56 2.56
2.56 2.56
0 1 2 3 4 5 0 1 2 3 4 5
0 1
base-10 2 3 of 4N
logarithm 5 0 1
base-10 2logarithm
3 of 4N 5
base-10 logarithm of N f base-10 logarithm of Nf f
f
(a) (b)
(a) (b)
Figure 7. The P-S-N curves with Ra = 0.8 μm and confidence level (a) = 95%; and (b) = 99%.
Figure
Figure 7.
7. The
The P-S-N
P-S-N curves
curves with
with RRaa==0.8
0.8μm
µmand
andconfidence
confidencelevel (a) γ =
level(a) = 95%;
95%; and
and (b) = 99%.
(b) γ = 99%
Appl. Sci. 2017, 7, 338 7 of 11
Appl. Sci. 2017, 7, x FOR PEER REVIEW 7 of 12

R=50% R=50%
2.72 R=90% 2.72 R=90%
R=99% R=99%
R=99.9% R=99.9%
2.68 2.68

base-10 logarithm of S
base-10 logarithm of S
2.64 2.64

2.60 2.60

2.56 2.56

0 1 2 3 4 5 -1 0 1 2 3 4 5
base-10 logarithm of Nf base-10 logarithm of Nf
(a) (b)
Figure 8. The P-S-N curves with Ra = 1.6 μm and confidence level (a) = 95%; and (b) = 99%.
Figure 8. The P-S-N curves with Ra = 1.6 µm and confidence level (a) γ = 95%; and (b) γ = 99%
Table 4. Parameters of P-S-N curves with Ra = 0.4 μm.
Table 4. Parameters of P-S-N curves with Ra = 0.4 µm.
R (%)
50 90 99 99.9 Parameters
(%)
R (%)
50 40.7137 9054.1796 66.9757
99 76.5668 99.9a1 Parameters
γ (%) 95
−13.9865 −19.3591 −24.4740 −28.3078 a2
40.7137 54.1796 66.9757 76.5668 a1
95 42.5752 57.2561 71.7960 82.7595 a1
−13.9865
99 −19.3591 −24.4740 −28.3078 a2
−14.7206 −20.5889 −26.4008 −30.7832 a2
42.5752 57.2561 71.7960 82.7595 a1
99
−14.7206
Table 5.− 20.5889 of P-S-N−with
Parameters Ra = 0.8 μm. −30.7832
26.4008 a2
R (%)
50 90 99 99.9 Parameters
(%) Table 5. Parameters of P-S-N with Ra = 0.8 µm.
28.6827 29.3268 29.9388 30.3976 a1
95
R (%) −9.4241 −10.1261 −10.7931 −11.2931 a2
50 28.7717 90 29.4739 30.1694
99 30.6938 99.9
a1
Parameters
γ (%)
99
28.6827 −9.521229.3268
−10.2864 −11.0444 −11.6159 30.3976
29.9388 a2 a1
95
−9.4241 −10.1261 −10.7931 −11.2931 a2
Table 6. Parameters of P-S-N curves with Ra = 1.6 μm.
28.7717 29.4739 30.1694 30.6938 a1
99 R9.5212
(%)
− −
50 10.2864
90 −
9911.0444 99.9 −Parameters
11.6159 a2
(%)
34.0511 39.5910 44.8554 48.8012 a1
Table
95 6. Parameters of P-S-N curves with Ra = 1.6 µm.
−11.4174 −13.8578 −16.1769 −17.9151 a2
34.8170 40.8567 46.8385 51.3489 a1
R (%) 99
50 −11.7547 90
−14.4154 −17.0504
99 −19.0374 99.9a2 Parameters
γ (%)
4. 34.0511 39.5910 44.8554 48.8012 a1
95Fatigue Life Estimation
−11.4174 −13.8578 −16.1769 −17.9151 a2
The fatigue life Nf is composed of the crack initiation life Ni and the crack propagation life Np
and 34.8170 40.8567 46.8385 51.3489 a1
99 can be calculated by Equation (3). The analytic formulation of the crack initiation life Ni was
− 11.7547 − 14.4154 − 17.0504 − 19.0374 a2
presented by Tanaka and Mura [9,10] and was based on the dislocation dipole accumulation model.
It was extended to estimate the crack initiation life Ni of the specimen with the different surface
4. Fatigue roughness by Wang et al. [12]. It is [9–12]:
Life Estimation
9∆ G
=
The fatigue life Nf is composed of the crack (7)
initiation
E(S − υ) Ni and the crack propagation life Np and
S ) π(1 −life
can be calculated by Equation (3). The analytic formulation of the crack initiation life Ni was presented
by Tanaka and Mura [9,10] and was based on the dislocation dipole accumulation model. It was
extended to estimate the crack initiation life Ni of the specimen with the different surface roughness by
Wang et al. [12]. It is [9–12]:
9∆Kth2 G
Ni = (7)
E(S − Se )2 π(1 − υ) a0
here ∆Kth is the threshold of stress intensity factor, G is the shear modulus, E is the elastic modulus,
Se is the endurance limit, υ is the Poisson’s ratio, and a0 is the equivalent initial micro-defect size. a0
could be estimated as [12,14]:
Appl. Sci. 2017, 7, 338 8 of 11

a0 = 2.97Ra (8)

The number of cycles necessary to grow the crack from the initial micro-defect size to failure can
be described by Paris’ law as:
da
= C (∆K )n (9)
dN 0
where ∆K is the amplitude of stress intensity factor, C the material constant, and a the crack size. The
crack propagation life Np can be calculated by the analytic solution of Paris’ law integral as [12,15,16]:

(1− n2 ) (1− n2 ) (1− n2 ) (1− n2 )

a0 − asc asc − af
Np = n n

+ n n
(10)
CSn βn1 π 2 CSn βn2 π 2


2 −1 2 −1

where: asc is the critical crack size for identifying long and short cracks, af the crack size of final failure,
√
β 1 = 0.5 π and β 12 = 1. Generally, the number of crack growth cycles from a0 to asc is much greater
than the cycles needed to grow from asc to af . Then Np can be taken as [17]:

(1− n2 )
a0
Np = n n
(11)
CSn βn1 π 2


2 −1

It can be written by:

S n Np = c (12)

where:
(1− n2 )
a0
c= n n
(13)
Cβn1 π 2


2 −1
By taking logarithms in Equation (12), then:

lgNp = lgc − nlgS (14)

The material constant n in Equation (12) is the same as n in the S-N curves. The material constant
C can be solved by Equation (13). n and c have been listed in Table 3 in Section 3. The material constants
C and n in Equation (11) are listed in Table 7.

Table 7. Material constants C and n in Equation (11) with different average surface roughness.

Average Surface Roughness (µm) n C

Ra = 0.4 9.84 4.11 × 10−34
Ra = 0.8 9.44 4.81 × 10−34
Ra = 1.6 9.80 3.02 × 10−36

Then, the fatigue life Nf of medium-carbon steel is computed as:

2 G (1− n2 )
9∆Kth a0
Nf = + n (15)
E ( S − Se ) 2 π ( 1 − υ ) a 0 CSn βn1 π 2 n


2 −1

where the shear modulus G is 79.40 GPa, the elastic modulus E is 206.00 GPa, and the threshold of stress
intensity factor ∆Kth is 6.26. Substituting the material constants C and n in Table 7 into Equation (12),
the mean fatigue life Nf of the medium-carbon steel with different average surface roughness is:
Appl. Sci. 2017, 7, 338 9 of 11

9×6.262 ×79.4

2
×π×(1−0.27)×1.188×10−3


 206 ×( S − 275 )
(1− 9.835 , ⇐⇒ Ra = 0.4 µm

2 )

 1.188


 + √ 9.835 9.835 9.835

 4.11×10 − 34 ×S 9.835 ×(0.5 π ) π 2 ( 2 −1 )

2 ×79.4


 9 × 6.26
206×(S−260)2 ×π×(1−0.27)×2.376×10−3


Nf = 2.376
(1− 9.436 )
2
, ⇐⇒ Ra = 0.8 µm (16)

 + √ 9.436 9.436 9.436
4.81×10−34 ×S9.436 ×(0.5 π ) π 2 ( 2 −1 )



9×6.262 ×79.4



206×(S−255)2 ×π×(1−0.27)×4.752×10−3



, ⇐⇒ Ra = 1.6 µm

(1− 9.802
2 )

4.752

 +

√

9.802 9.802 9.802
3.02×10−36 ×S9.802 ×(0.5 π ) π 2 ( 2 −1 )

When the average surface roughness is 0.4, 0.8, and 1.6, respectively, the mean fatigue life curves of
medium-carbon steel calculated by Equation (16) are shown in Figure 9. It can be seen that the fatigue
life estimation model agrees with the experiment value well. To validate the presented formulation for
estimating the mean fatigue life of medium-carbon steel, six more fatigue tests were conducted and
the experimental results are shown in Table 8. The stress amplitude was 420 MPa and was not equal to
any stress amplitude in the previous experiments. The maximum and minimum estimation errors are
15.83% and 1.99%, respectively. Therefore, it is obvious that the presented model for estimating the
mean fatigue life of medium-carbon steel gives good results for roughness Ra = 0.4 µm, Ra = 0.8 µm,
and R a =Sci.
Appl. 1.62017,
µm.7, x FOR PEER REVIEW 10 of 12

Ra=0.4 (estimated)
550 Ra=0.8 (estimated)
Ra=1.6 (estimated)
Ra=0.4 (experiment)
Ra=0.8 (experiment)
500
Stress amplitude (MPa)

Ra=1.6 (experiment)

450

400

350

0 50000 100000 150000 200000 250000

Number fo cycles to failure

Figure 9. Estimated mean fatigue life curves and the experiment values.
Figure 9. Estimated mean fatigue life curves and the experiment values.
5. Conclusions
Table 8. Validation experiment results.
To estimate the fatigue life of medium-carbon steel specimen, the effect of the average surface
roughness on the fatigue
Average Surface
life has been
Stress
investigated
Estimated
in this paper Mean
Experimental
usingof75 fatigue tests in three
Experimental
groups with(µm)
averageAmplitude
surface roughnesses Error
Roughness Fatigue of
Life(Ra): 0.4 μm,Life
Fatigue 0.8 μm, andFatigue
1.6 μm, Liferespectively. S-N
curves and P-S-N curves were established based on the 70,569 fatigue tests. It was shown that there is about
Ra = 0.4 of the mean fatigue life in72,420
15% decrease high stress amplitude 73,863 1.99%
77,156 (for example, 500 MPa), while there is
about 30% decrease in 420low
MPastress amplitude (for example,57,824 380 MPa) when the average surface
Ra = 0.8
roughness increased from 0.4 μm to 0.856,060
μm or from 0.844,416
μm to 1.6 μm. Then, 51,120
one new method 8.81%
was
presented to estimate the fatigue life of the medium-carbon 66,174 steel with different average surface
Ra = 1.6 45,190 52,345
roughness (Ra). The experiment results show that the maximum
38,516 estimation error of the mean 15.83%
fatigue
life by the presented model is less than 16%.

Acknowledgements: The work is supported by National Natural Science Foundation of China (grant no.
51575095, 51675089), and Major State Basic Research Development Program of China (973 Program) (grant no.
2014CB046303).

Author Contributions: Changyou Li, Weibing Dai and Fei Duan conceived and designed the experiments
and analyzed the experiment data; Yimin Zhang and David He contributed to the method for estimating the
fatigue life of the medium-carbon steel with different average surface roughness; David He refined the
Appl. Sci. 2017, 7, 338 10 of 11

5. Conclusions
To estimate the fatigue life of medium-carbon steel specimen, the effect of the average surface
roughness on the fatigue life has been investigated in this paper using 75 fatigue tests in three groups
with average surface roughnesses of (Ra ): 0.4 µm, 0.8 µm, and 1.6 µm, respectively. S-N curves and
P-S-N curves were established based on the fatigue tests. It was shown that there is about 15% decrease
of the mean fatigue life in high stress amplitude (for example, 500 MPa), while there is about 30%
decrease in low stress amplitude (for example, 380 MPa) when the average surface roughness increased
from 0.4 µm to 0.8 µm or from 0.8 µm to 1.6 µm. Then, one new method was presented to estimate the
fatigue life of the medium-carbon steel with different average surface roughness (Ra ). The experiment
results show that the maximum estimation error of the mean fatigue life by the presented model is less
than 16%.

Acknowledgments: The work is supported by National Natural Science Foundation of China (grant no.
51575095, 51675089), and Major State Basic Research Development Program of China (973 Program) (grant
no. 2014CB046303).
Author Contributions: Changyou Li, Weibing Dai and Fei Duan conceived and designed the experiments and
analyzed the experiment data; Yimin Zhang and David He contributed to the method for estimating the fatigue
life of the medium-carbon steel with different average surface roughness; David He refined the language of
this manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
a Crack size
a0 Equivalent initial micro-defect size
a1 Parameter of P-S-N curve
a2 Parameter of P-S-N curve
af Crack size of final failure
asc Critical crack size for identifying long and short cracks
b1 Parameter of surface condition modification factor
b2 Parameter of surface condition modification factor
C Material constant in Paris law
c Material constant in Basquin’s relation (NSn = c)
E Elastic modulus
G Shear modulus
h One-sided tolerance factor
ka Surface condition modification factor
Kf Fatigue stress concentration factor
Kt Stress concentration factor
N Number of cycle
Np Crack propagation life
Ni Crack initiation life
Nf Fatigue life (number of cycles to failure)
n Material constant in Basquin’s relation(NSn = c)
ns Sample size
P-S-N Probabilistic stress-life
Ra Average surface roughness
R Survival probability
S Stress amplitude
S-N Stress-life
Se Endurance limit of a machine part
S`e Endurance limit of the rotary-beam test specimen
Sut Minimum tensile strength
x RL Confidence lower limit
x Mean of the experimental fatigue life
α Parameter in Equation (4) Ni = β(Kt S)α
β Parameter in Equation (4) Ni = β(Kt S)α
γ Confidence level
Appl. Sci. 2017, 7, 338 11 of 11

ξ Correction factor in Equation (6)

υ Poisson’s ratio
σx Standard deviation of the experimental fatigue life
∆K Amplitude of stress intensity factor
∆Kth Threshold of stress intensity factor

References
1. The Wikipedia Page on Carbon Steel. Available online: https://en.wikipedia.org/wiki/Carbon_steel#cite_
note-10 (accessed on 25 October 2016).
2. Gladman, T. Medium/High Carbon Steels for Rails, Rods, Bars and Forgings. In Materials Science and
Technology; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2006; pp. 401–432.
3. Moussaoui, K.; Mousseigne, M.; Senatore, J.; Chieragatti, R. The effect of roughness and residual stresses on
fatigue life time of an alloy of titanium. Int. J. Adv. Manuf. Technol. 2015, 78, 557–563.
4. Budynas, R.G.; Nisbett, J.K. Shigley’s Mechanical Engineering Design, 10th ed.; McGraw-Hill Education:
New York, NY, USA, 2015; pp. 294–301.
5. Suraratchai, M.; Limido, J.; Mabru, C.; Chieragatti, R. Modelling the influence of machined surface roughness
on the fatigue life of aluminum alloy. Int. J. Fatigue 2008, 30, 2119–2126. [CrossRef]
6. As, S.K.; Skallerud, B.; Tveiten, B.W.; Holme, B. Fatigue life prediction of machined components using finite
element analysis of surface topography. Int. J. Fatigue 2005, 27, 1590–1596.
7. Arola, D.; Ramulu, M. An examination of the effects from surface texture on the strength of fiber-reinforced
plastics. J. Compos. Mater. 1999, 33, 102–123. [CrossRef]
8. Arola, D.; Williams, C.L. Estimating the fatigue stress concentration factor of machined surfaces. Int. J. Fatigue
2002, 24, 923–930. [CrossRef]
9. Tanaka, K.; Mura, T. A dislocation model for fatigue crack initiation. Int. J. Appl. Mech. 1981, 48, 97–103.
[CrossRef]
10. Tanaka, K.; Mura, T. A theory of fatigue crack initiation at inclusions. Metall. Mater. Trans. A Phys. Metall.
Mater. Sci. 1982, 13, 117–123.
11. Wang, Q.Y.; Bathias, C.; Kawagoishi, N.; Chen, Q. Effect of inclusion on subsurface crack initiation and
gigacycle fatigue strength. Int. J. Fatigue 2002, 24, 1269–1274. [CrossRef]
12. 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]
13. Fu, H.M.; Gao, Z.T.; Xu, R.P. A confidence lower limit of population percentile. J. Beijing Univ. Aeronaut.
Astronaut. 1990, 3, 1–8.
14. Murakami, Y. Metal Fatigue: Effects of Small Defects and Nonmetallic Inclusions; Elsevier Science Ltd.: Kidlington,
UK, 2002; pp. 11–24.
15. 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]
16. Wang, Q.Y.; Pidaparti, R.M.; Palakal, M.J. Comparative study of corrosion–fatigue in aircraft materials.
AIAA J. 2001, 39, 325–330. [CrossRef]
17. Wang, Q.Y.; Berard, J.Y.; Rathery, S.; Bathias, C. High-cycle fatigue crack initiation and propagation behaviour
of high-strength spring steel wires. Fatigue Fract. Eng. Mater. Struct. 1999, 22, 673–677. [CrossRef]

© 2017 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/).