General relation between tensile strength and fatigue strength

of metallic materials
J.C. Pang, S.X. Li, Z.G. Wang, Z.F. Zhang
n
Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, 72 Wenhua Road, Shenyang 110016, China
a r t i c l e i n f o
Article history:
Received 8 May 2012
Received in revised form
7 November 2012
Accepted 27 November 2012
Available online 1 December 2012
Keywords:
High-strength materials
Tensile strength
Fatigue strength
Fatigue crack initiation site
Fatigue damage mechanism
a b s t r a c t
With the development of high-strength materials, the existing fatigue strength formulae cannot
satisfactorily describe the relation between fatigue strength s
w
and tensile strength s
b
of metallic
materials with a wide range of strength. For a simple but more precise prediction, the tensile and
fatigue properties of SAE 4340 steel with the tensile strengths ranging from 1290 MPa to 2130 MPa
obtained in virtue of different tempering temperatures were studied in this paper. Based on the
experimental results of SAE 4340 steel and numerous other data available (conventional and newly
developed materials), through introducing a sensitive factor of defects P, a new universal fatigue
strength formula s
w
s
b
(CPds
b
) was established for the rst time. Combining the variation tendency
of fatigue crack initiation sites and the competition of defects, the fatigue damage mechanisms
associated with different tensile strengths and cracking sites are explained well. The decrease in the
fatigue strength at high-strength level can be explained by fracture mechanics and attributed to the
transition of fatigue cracking sites from surface to the inner inclusions, resulting in the maximum
fatigue strength s
max
w
at an appropriate tensile strength level. Therefore, the universal fatigue strength
formula cannot only explain why many metallic materials with excessively high strength do not display
high fatigue strength, but also provide a new clue for designing the materials or eliminating the
processing defects of the materials.
& 2012 Elsevier B.V. All rights reserved.
1. Introduction
Fatigue is referred to the degradation of mechanical properties
leading to failure of a material or a component under cyclic
loading. The fatigue strength of materials is often dened as the
maximum stress amplitude without failure after a given number
of cycles (e.g., 10
7
or 10
9
). It is estimated that 90% of service
failures of metallic components resulted from fatigue. However, it
is very time and money consuming to perform fatigue tests.
Therefore, many attempts have been made to determine the
fatigue strength in a cost-effective way relating fatigue strength
to other mechanical properties, such as yield strength [1], tensile
strength [24], hardness [57] and so on; accordingly, the rela-
tions between fatigue strength and other mechanical properties
have been of more interest. Engineers and scientists have pro-
posed many formulae to describe the relations between fatigue
strength and other mechanical properties [17]. In 1870s, W ohler,
one of the pioneers in the fatigue eld, found that the ratio of
fatigue strength s
w
to tensile strength s
b
for ferrous metals
followed a simple proportional relation as below [8],
s
w
0:40:5 s
b
1
Based on the numerous data of fatigue strength and tensile
strength available for steels, copper and aluminum alloys [24] in
the past century, a more general form can be summarized as
follows,
s
w
ms
b
2
However, it is found that the fatigue strength either maintains
constant or decreases with further increasing the tensile strength
[3,4]; in other words, the linear relation in Eq. (2) is no longer held
at high-strength level. The critical tensile strength s
bc
, above
which fatigue strength does not increase correspondingly, the
maximum fatigue strengths s
max
w
and the coefcient m in Eq. (2)
for steels, Cu and Al alloys are summarized from Refs. [24]. It is
apparent that the linear equation cannot adequately be applied to
estimate the fatigue strength of those high-strength materials.
On the other hand, in 1980s, another important nding by
Murakami [5] is that there is a quantitative relationship among
fatigue strength s
w
, hardness Hv and inclusion size

area
p
in high-
strength steels. Soon after, many related tests have been done by
ultrasonic fatigue testing machines [6,7,920] and some modied
fatigue strength formulae were proposed by Wang et al. [9], Liu
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/msea
Materials Science & Engineering A
0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.11.103
n
Corresponding author. Tel.: 86 24 23971043; fax: 86 24 23891320.
E-mail address: zhfzhang@imr.ac.cn (Z.F. Zhang).
Materials Science & Engineering A 564 (2013) 331341
et al. [10] and McGreevy et al. [21]; nevertheless, no report
indicated that those relations suit for other high-strength materi-
als. In addition, the fatigue strength is found to have linear
relation with hardness or the sum of tensile and yield strengths
only in lower strength range [25]. In a word, there is also no
suitable formula to satisfactorily describe the general relation
between tensile and fatigue strengths of both high- and low-
strength materials.
During recent decades, many new high-strength materials,
such as bulk metallic glasses [22], ultrane or nano-grained
materials [2326] and ultra-high strength steels [27] have been
successfully developed; however, their fatigue strengths are
found to be not as high as we expected, even become relatively
lower in comparison with their higher tensile strength [2732].
Therefore, this gives rise to two open questions for scientists and
engineers: (1) Why do the materials with excessively high tensile
strength not possess high fatigue strength? (2) Is there a more
universal equation to describe the general relation between
fatigue strength and tensile strength in a wide strength range?
Therefore, in this study, SAE 4340 steel with a very wide tensile
strength range, one of the excellent quenched and tempered low-
alloy steels [25,3336], was employed to study and establish a
simple but more precise relation between fatigue strength and
tensile strength of materials and provide a better clue for the
design of high-performance structural materials.
2. Experimental material and procedures
In the current study, SAE 4340 steel bars were received with a
diameter of 14 mm under annealing condition and its composition
is given in Table 1. To gain different strength levels, ve optimized
heat-treatment procedures as shown in Table 2 were employed
and the corresponding specimens are dened as AE, respectively.
The congurations of the tensile and fatigue specimens are shown
in Fig. 1. All fatigue samples were polished in the longitudinal
direction using an emery paper having a mesh of 2000#.
The tensile tests were conducted at a strain rate of 210
4
s
1
;
very high-cycle (VHC) fatigue tests were conducted at a frequency of
20 kHz up to 10
9
cycles using a ultrasonic fatigue testing system
(Shimadzu USF-2000). To avoid the sample heating, the middle
section of each ultrasonic fatigue specimen was cooled by com-
pressed air. All fatigue tests were performed with the sinusoidal
wave shape under applied stress ratio of R1. The fatigue
strength was determined by the staircase method in which at least
six pairs of specimens were tested according to ISO12107:2003.
The microstructures of the specimens with different strength levels
were examined by electron backscattered diffraction (EBSD) with
scanning electron microscope (LEO SUPRA35). The fatigue fracto-
graphies were observed by using a Quanta 600 scanning electron
microscope (SEM).
3. Results
3.1. Microstructures
With increasing the tempering temperature, the body-
centered tetragonal (BCT) martensite, which is a supersaturated
solution of carbon in a-Fe, transforms to different microstructures
as shown in Fig. 2. Referring to the textbook [36,37] and XRD
proles, the microstructure features of sample AE are illustrated
as follows: the sample A tempered at 180 1C contains many
needle- or plate-shaped tempered martensites (see Fig. 2(a))
and some retained austenites. The sample B tempered at 250 1C
consists of microstructure (Fig. 2(b)) similar to that of the sample
A, but the size of retained austenite decreases because of its
decomposition; however, the sample C tempered at 350 1C only
displays needle- or plate-shaped tempered martensite. The sam-
ples D and E, respectively tempered at 420 1C and 500 1C, exhibit
tempered troostite with plate-like appearance (Fig. 2(d) and (e)),
and the lath width of troostite in sample E increases in compar-
ison with sample D because a phase has obviously recovered after
tempering above 400 1C.
3.2. Tensile properties
After different tempering procedures, the specimens A to E
exhibit different tensile properties and their tensile stressstrain
curves are shown in Fig. 3(a). It can be seen that the specimens A to
E display different yield strength, work-hardening ability, ultimate
tensile strength and elongation. Fig. 3(b) and (c) show the relations
among the strength, elongation to fracture and reduction in area
versus tempering temperature: it can be seen that as tempering
temperature increases, tensile strength successively decreases,
however, elongation to fracture and reduction in area successively
increase, which are in agreement with other steels [34,35,38].
On the other hand, the yield strength rst increases slightly
and then decreases, which agrees with 300 M steel [39] and some
other ultrahigh-strength steels tempered below 200 1C [34,35].
Fig. 3(d) demonstrates the relationships of yield strength and
tensile strength versus elongation to failure of the SAE 4340 steel.
As tensile and yield strengths increase, elongation to failure
continuously decreases, which is consistent with the trade-offs
between strength and elongation of steels [34,35]. By comparison,
Table 1
Chemical composition of SAE 4340 steel (Wt%).
C Mn Si P S Cr Ni Mo Cu As
0.42 0.66 0.25 0.009 0.014 0.74 1.41 0.17 0.11 0.034
Table 2
Heat-treatment procedures of SAE 4340 steel.
No. Quenching Tempering
A preheating to 850 1C
for 10 min and quenching in oil
180 1C tempering for 120 min
B 250 1C tempering for 120 min
C 350 1C tempering for 120 min
D 420 1C tempering for 30 min
E 500 1C tempering for 30 min
Fig. 1. Conguration of specimens tested for: (a) tensile strength; (b) VHC fatigue
strength.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 332
the elongation to failure of specimen C appears smaller because of
the tempering brittleness at 350 1C [33].
3.3. Fatigue properties
Six groups of VHC fatigue tests (the highest strength level was
repeated once more) were conducted to obtain the fatigue
strengths by the staircase method. Fig. 4 shows the crack origin
of VHC fatigue. There are four types of crack origin modes:
(a) surface scratch; (b) surface inclusion; (c) subsurface inclusion;
(d) inner inclusion, which are in accordance with the crack
initiation modes of high-strength steels [5,6]. Besides, special
granular bright and rough area in comparison with other area,
which is named as Granular-bright-facet (GBF) by Shiozawa et al.
[40], appears around the inclusion, as indicated by the blue cycle
in Fig. 4(d), but is not shown in Fig. 4(c).
Fig. 2. EBSD microstructures of SAE 4340 steel processed at different tempering temperature: (a) 180 1C; (b) 250 1C; (c) 350 1C; (d) 420 1C and (e) 500 1C.
Fig. 3. Tensile properties of SAE 4340 steel: (a) tensile engineering stressstrain curves; (b) the relation between strength and tempering temperature; (c) relations of
elongation and reduction in area vs. the tempering temperature and (d) relation between strength and elongation to fracture.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 333
Broadly speaking, according to the fatigue cracking site, the
four crack initiation modes can be classied into two types: i.e.,
surface (surface scratch, surface/subsurface inclusion) and inner
(inner inclusion). Fig. 5 displays the statistic ratio of surface and
inner cracking sites for VHC fatigue test: the ratio of inner
cracking sites (RICS, the ratio of the number of samples originated
from the inner inclusion site to the total number of failure
samples) gradually increases with increasing tensile strength;
however, when tensile strength is lower than 1200 MPa, RICS
tends to zero, when the tensile strength is higher than
2000 MPa, RICS tends to 100%. The ratio of surface cracking
sites (RSCS, the ratio of the number of samples originated from
surface site to the total number of failure samples) gradually
decreases as tensile strength increases; when tensile strength is
higher than 2000 MPa, RSCS tends to zero. This indicates that
when tensile strength is lower than 1200 MPa, nearly all the
fatigue cracks originated from surface defects; when tensile
strength is higher than 2000 MPa, nearly all the fatigue cracks
initiated from inner inclusion; when tensile strength is in the
range of 12002000 MPa, there is a gradual transition for fatigue
crack initiation from surface defects to inner inclusions.
Fig. 6(a) shows the relationship between the fatigue strength
and tensile strength of the SAE 4340 steels heat-treated under
different conditions. It is found that the VHC fatigue strengths
increase rst and then decrease with increasing tensile strength,
which is consistent with that of the CuBe alloy after different
ageing technologies [41]. The highest fatigue strength can be as
high as 693 MPa, which occurs in the specimen B with a tensile
strength of 1830 MPa that is not the highest tensile strength,
implying that a higher tensile strength of materials cannot always
lead to a higher fatigue strength, which is also found in other
high-strength materials [2732,41]. In contrast, as the tensile
strength increases to a certain extent, a decrease in fatigue
strength can be observed, for example, the specimen A with the
highest tensile strength of 2124 MPa has a fatigue strength of
655 MPa, which is lower than the fatigue strength (693 MPa) of
the specimen B with a tensile strength of 1830 MPa (see Fig. 6a).
In order to conrm such results, a repeated VHC fatigue test
was conducted and a fatigue strength of 657 MPa was obtained,
which implies that the fatigue testing result for the steel with
the highest tensile strength but with lower fatigue strength is
reliable.
3.4. Fatigue strength formula
For better understanding on the varying trend of fatigue
strength for the materials at different tensile strength levels, it
is necessary to nd a general fatigue strength formula to describe
the intrinsic relation between fatigue strength and tensile
strength. Fig. 6(b) demonstrates that the fatigue ratio R (the ratio
of fatigue strength s
w
to tensile strength s
b
) gradually declines
with increasing tensile strength. From the data in Fig. 6(b), it is
interesting to nd that the fatigue ratio R declines approximately
in a linear relation with increasing tensile strength. To conrm
this tendency, a linear tting is performed and the tting
equation can be expressed as below,
R s
w
=s
b
0:701:85 10
4
s
b
3
As shown in Fig. 6(c), the scope of fatigue ratio is within the 5%
error band. The fatigue strength s
w
in Eq. (3) can be written in
Fig. 4. Crack origin of VHC fatigue: (a) surface scratch; (b) surface inclusion; (c) subsurface inclusion and (d) inner inclusion.
Fig. 5. Relation between the VHC fatigue crack sites ratio and tensile strength.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 334
a quadratic equation of the tensile strength s
b
as below,
s
w
0:701:85 10
4
s
b
s
b
4
Eq. (4) is drawn in Fig. 6(d), it can be seen that the fatigue
strengths are basically within the 5% error band. This illustrates
that Eq. (4) can well describe the general relation between fatigue
strength and tensile strength of the SAE 4340 steel in a wide
strength range.
4. Discussion
4.1. Verication of the general fatigue strength formula
In order to conrm the fatigue strength formula above, it is
necessary to nd out more fatigue strength data of metallic
materials available. First, the rotating bending fatigue strength
data of some high-strength steels with different strength levels in
literature [38] are used to verify the proposed equation. It is noted
that the tting relations between fatigue ratio and tensile
strength for SAE 4140, 4340, 2340 and 4063 steels displayed in
Fig. 7 are also linear and the scope of fatigue ratio is within the 5%
error band, which entirely agrees with the general relation
proposed above. The tting equations can be written as below,
respectively
R 0:872:65 10
4
s
b
For SAE 4140 5a
R 0:761:78 10
4
s
b
For SAE 4340 5b
R 0:741:89 10
4
s
b
For SAE 2340 5c
R 0:922:37 10
4
s
b
For SAE 4063 5d
therefore, their fatigue strength s
w
can be expressed in quadratic
equations of the tensile strength s
b
as below, respectively
s
w
0:872:65 10
4
s
b
s
b
For SAE 4140 6a
s
w
0:761:78 10
4
s
b
s
b
For SAE 4340 6b
s
w
0:741:89 10
4
s
b
s
b
For SAE 2340 6c
s
w
0:922:37 10
4
s
b
s
b
For SAE 4063 6d
Eqs. (6a)(d) are drawn in Fig. 8, it can be seen that the
fatigue strengths are basically within the 5% error band. This
indicates that the quadratic equations can well describe the
general relation between fatigue strength and tensile strength of
those steels.
Second, the fatigue strengths of steels with the minimum
sample diameter of 3 mm at a frequency of 20 kHz up to 10
9
cycles are collected from Refs. [920]. Those data are also tted
according to the previous method and displayed in Fig. 9 with the
following equations,
R 0:671:52 10
4
s
b
, 7a
s
w
0:67s
b
1:52 10
4
s
2
b
7b
The most data of fatigue ratios and fatigue strengths are also
within the 15% error band, which expresses that the fatigue
strength formula above is also applicable to other steels with a
very wide range of tensile strength.
From the analysis above, it is wondering whether the relation
above suits for other steels, even for those non-ferrous metals or
Fig. 6. (a) Relation between tensile and fatigue strengths of the 4340 steel; (b) relation between tensile strength and fatigue ratio; (c) tting relation between tensile
strength and fatigue ratio and (d) tting curves of tensile and fatigue strengths.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 335
Fig. 7. Fitting relation between tensile strength and rotating bending fatigue ratio of high strength steels: (a) SAE 4140; (b) SAE 4340; (c) SAE 2340 and (d) SAE 4063. (Data
are cited from Ref. [38]).
Fig. 8. Fitting relation between tensile strength and rotating bending fatigue strength of high strength steels: (a) SAE 4140; (b) SAE 4340; (c) SAE 2340 and (d) SAE 4063.
(Data are cited from Ref. [38]).
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 336
not. Thus, many fatigue data of engineering materials, such as
wrought steels, aluminum and copper alloys [3,4] were further
collected. Fig. 10(a), (c) and (e) shows the tting curves of the
fatigue ratios of wrought steels, copper alloys and aluminum
alloys available, and the tting equations can be expressed as
follows, respectively,
R 0:611:24 10
4
s
b
For wrought steels , 8a
R 0:543:72 10
4
s
b
For wrought Cu alloys 8b
R 0:535:66 10
4
s
b
For aluminum alloys 8c
It can be seen that most data of fatigue ratios are also within
the 20% error band and the relation between the fatigue ratio and
tensile strength of metallic materials still looks like linear. Based
on the results, the fatigue strength s
w
for the three kinds of
materials can be separately tted by the following quadratic
equations:
s
w
0:611:24 10
4
s
b
s
b
For wrought steels 9a
s
w
0:543:72 10
4
s
b
s
b
For wrought Cu alloys 9b
s
w
0:535:66 10
4
s
b
s
b
For aluminum alloys 9c
In Fig. 10(b), (d) and (f), most fatigue strength s
w
data are
within the 20% error band, which tells us that the fatigue strength
formula above is also appropriate for the conventional engineer-
ing materials.
Recently, many new high-strength materials have been suc-
cessfully developed as mentioned above. We curiously know
whether the quadratic relation applies to them or not. However,
there is almost no fatigue data of the identical materials with
varying tensile strength in a very wide range. Luckily, we have
carried out some fatigue tests on ultrane grained low carbon
steel [42], and collected some fatigue data of coarse grained
[43,44] and ultrane grained [42,44,45] low-carbon steels and
analyzed them as shown in Fig. 11. The tting relations between
tensile strength and fatigue ratio as well as fatigue strength can
be written as below:
R 0:602:13 10
4
s
b
For low carbon steel , 10a
s
w
0:602:13 10
4
s
b
s
b
For low carbon steel 10b
The data of fatigue ratios and fatigue strengths are also within
the 10% error band. This indicates that the fatigue strength
formula also ts the low-carbon steel.
After analyzing the fatigue data of four specic materials (SAE
4340, 4140, 2340, 4063) and three types of materials (steels, copper
alloys, aluminum alloys) the two general conclusions can be drawn:
(1) the relation between fatigue ratios and tensile strengths is linear;
(2) the relation between fatigue strengths and tensile strengths is
quadratic. Therefore, Eqs. (3), (7a), (10a) (5a)(d) and (8a)(c) can be
expressed in the general form as below,
R CP s
b
11
where C and P are two constants; moreover, each material has its
own constants as listed in Table 3. Furthermore, Eqs. (4), (7b), (10b)
(9a)(c) and (6a)(d), can be unied in a more universal formula as
below,
s
w
CP s
b
s
b
12
4.2. Fatigue damage mechanism of materials
To reveal the reasons caused the results above, it is necessary
to further study the fatigue damage mechanism of the high-
strength steel. Since it is inevitable that the real materials contain
different kinds of defects [2,5], they can be roughly categorized
into two types: one is the intrinsic defect (ID), such as vacancies,
interstitial atoms, dislocations, stacking faults (SF), grain bound-
aries (GB) and so on, as illustrated in Fig. 12. The IDs are related to
the materials itself and can not be completely avoided by the
technology innovation of processing or by controlling the forma-
tion of microstructure. The other is the processing defect (PD),
including nonmetallic inclusions, cavities, segregation and
scratches etc, as shown in Fig. 12. The PDs are associated with
the technology of production or processing and can be avoided in
practice to some extent.
First, for the low-strength ductile materials, fatigue cracks
normally do not nucleate at PDs (inclusion or scratch), but at slip
bands (SB) or GBs [46] on the sample surface (see Fig. 12). The
materials with ID and without PD can be regarded as ideal ones
because ID mainly controls the microstructure and determines
the main mechanical properties. Therefore, for the ideal materials,
the fatigue strength s
w
might only depend on its IDs and the
corresponding ideal fatigue strength s
I
w
should show a propor-
tional correlation with its tensile strength s
b
i.e.,
s
I
w
C s
b
13
where, s
I
w
: ideal fatigue strength; C: a constant, which agrees
with the proportional relation in intrinsic aw regime [21], where
intrinsic aw is ID mentioned above.
Fig. 9. (a) Fitting relations (a) between tensile strength and fatigue ratio and (b) relation between tensile strength and VHC fatigue strength of steels with the smallest
diameter of 3 mm. Data are from Ref. [920].
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 337
For the grain-renement strengthening, the smaller grain size
is, the larger will be the total boundary surface area per unit
volume. Therefore, this can increase the pileup of dislocation and
make the motion of dislocation difcult, resulting in the increase
of tensile strength, as explained by the well-known HallPetch
relation. Therefore, the relation between tensile strength and
grain size D
1/2
is linear, which can be expressed as
s
b
p
1

D
p 14
For fatigue strength, small sized grains can lead to the reduced
aw sizes and increase difculties for the imposed stress con-
centration at the aw to exceed the critical stress of the material,
thus suppressing early crack nucleation and propagation. Fatigue
strength also follows the HallPetch relation [24,25] as blew,
s
w
p
1

D
p 15
Therefore, the relation between tensile strength and fatigue
strength can be regarded as linear, which can be veried by
the experimental results of some ultrane-grained materials in
Refs. [25,26].
Second, for the high-strength materials, fatigue cracks do not
frequently initiate along SB or GBs and other IDs, but at PDs
(inclusion or scratch in Fig. 12). At this time, the fatigue strength
of materials can be considered as PD-controlled, and the effects of
IDs on fatigue strength could be minied to some extent. If the
material can be considered as an isotropic solid with a crack,
hence the fracture mechanics can be applied. By which, Murakami
Fig. 10. Fitting relations of the rotating bending fatigue ratio and fatigue strength vs. tensile strength for different engineering materials, (a) and (b) for wrought steels at
10
6
cycles; (c) and (d) for copper alloys at 10
8
cycles; (e) and (f) for aluminum alloys at 510
8
cycles. (Data are collected from Ref. [3,4]).
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 338
et al. [5] found a well-known fatigue strength formula related to
the parameters such as hardness Hv and the inclusion size

area
p
as detail below. If the area of a crack at inclusion is denoted by
area, then the maximum value, K
Imax
of the stress intensity factor
along its crack front is given approximately by [47]
K
Imax
0:5s
a

p

area
p
q
16
here, K
Imax
is the applied stress intensity factor given in MPa m
1/2
,
where s
a
is the applied stress (MPa) and

area
p
is crack size (m).
The threshold for crack growth can be written [5]
DK
th
3:3 10
3
HV 120

area
p
1=3
17
where DK
th
is in MPa m
1/2
and HVis in kgf mm
2
, and

area
p
is in
mm. When K
Imax
DK
th
, the fatigue strength can be determined.
Combining Eqs. (16) and (17), and noticing the difference of the
unit of the

area
p
between those equations, the fatigue strength
s
w
can be expressed as below [48],
s
w
C HV 120 =

area
p
1=6
18
where, C: material constant relative to crack initiation site.
Eq. (18) is widely used to predict the fatigue strength of high-
strength steel with inclusion. While, the fatigue strength mea-
sured by staircase method is a statistical value; If the specimens
were prepared with the same batch of steel bars but different heat
treatments, the average inclusion sizes contained in those speci-
mens should be close to each other. Therefore, the inclusion size

area
p
in Eq. (18) can be considered as a constant, thus, Eq. (18)
becomes
s
w
C
0
0
Hv120 19
where, C
0
0

C0

area
p

1=6
. In practice, the hardness Hv and tensile
strength s
b
show linear relation [2,39]; so that Eq. (19) is
approximately consistent with Eq. (13). It implies that the rela-
tion between tensile strength and fatigue strength of one mate-
rial, no matter whether in low-strength level or in high-strength
level, is still linear in theory.
However, a question is raised as below. For our fatigue test, the
specimens A to E were prepared with the same batch of steel bars
but different heat treatments, the inclusion size contained in
those specimens should be the same in the statistical meaning.
Therefore, the inuence of the inclusion size on fatigue strength
of those specimens could be neglected. According to Eqs. (13) or
(19), the fatigue strength should be proportional to the tensile
strength; however, it is inconsistent with the experiment results
as mentioned above. The reason may be as follows. By analyzing
of the relation between DK
th
and tensile strength s
b
of 300 M
steel in Ref. [49], as shown Fig. 13, it is found that DK
th
linearly
decreases with tensile strength s
b
increasing. The linear relation
can be written in the form like:
DK
th
pC
1
C
2
s
b
20
Fig. 11. (a) Relation between tensile strength and fatigue ratio and (b) relation between tensile strength and fatigue strength: CG [43,44], UFG [42,44,45].
Table 3
Values of the constants C, P, M and E and corresponding s
bc
and s
max
w
.
Material C P/10
4
MPa
1
M E/GPa s
bc
/MPa s
max
w
/MPa Error
band (%)
Our work 0.70 1.85 38.5 208 1891 662 75
SAE 4140 0.87 2.65 54.9 207 1642 714 75
SAE 4340 0.76 1.78 37.0 208 2135 811 75
SAE 2340 0.74 1.89 39.1 207 1958 724 75
SAE 4063 0.92 2.37 49.1 207 1941 893 75
Steel alloy
a
0.67 1.52 31.5 207 2204 738 715
Steel alloy
b
0.61 1.24 25.7 207 2460 750 720
LC Steel
c
0.60 2.13 45.3 208 1408 423 710
Cu alloy 0.54 3.72 44.3 119 726 196 720
Al alloy 0.53 5.66 40.6 71.7 468 124 720
a
VHC fatigue under tension and compression.
b
Rotating bending fatigue.
c
Low carbon steel.
Fig. 12. Schematic illustration of fatigue mechanisms and strength formula with
tensile strength increasing.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 339
where, C
1
and C
2
are material constants. The decrease of DK
th
can
be also attributed to the enhancement of defect sensitivity with
tensile strength increasing.
Now, the resistance DK
th
for fatigue crack growth near an
inclusion in steels with a wide strength range should be modied.
Here, as an assumption, the DK
th
in Eq. (17) is modied as
DK
th
C HV 120

area
p
1=3
C
0
1
C
0
2
s
b

21
Combining Eqs. (16) and (21), the fatigue strength formula can
be written as:
s
w

C
0
0
Hv120 C
0
1
C
0
2
s
b

area
p
1=6
22
Considering the inclusion size as a constant in average in all
fatigue tests, and the linear relation between tensile strength s
b
and hardness Hv as mentioned above, Eq. (22) can be further
simplied as Eq. (12), i.e., s
w
(CP s
b
) s
b
. This indicates that
to a rst approximation, the quadratic relation can be explained
by fracture mechanics. As for the fatigue damage mechanisms, it
may be illustrated in outline as follows, thus, Eq. (12) can be
rewritten as
s
w
CP s
b
s
b
Cs
b
Ps
2
b
s
I
w
s
P
w
23
From the analysis above, it is known that the rst term on the
right side of Eq. (23), i.e., Eq. (13) as mentioned above, is the ideal
fatigue strength, and it is valid no matter whether in low-strength
level or in high-strength level, and the parameter C is named as
intrinsic factor due to the inuence of the intrinsic defects. On the
other hand, the second term in Eq. (23) can be expressed by,
s
P
w
P s
2
b
24
This indicates that the reduction of fatigue strength due to the
competitive effects of ID and PD, and the parameter P is named as
the fatigue sensitivity coefcient to processing defects.
In fact, a real material contains both IDs and PDs. At low-
strength level, s
P
w
is so small that the inuence of PD on fatigue
strength can be completely negligible and fatigue cracks appear at
SBs or GBs (see Fig. 12); and Eq. (23) becomes equivalent to
Eq. (13), i.e., s
w
s
I
w
C s
b
, which can be called as ID regime in
Fig. 12. With increasing the tensile strength, s
P
w
becomes gradu-
ally large (violet line in Fig. 12) and can not be negligible.
However, the inuence of ID on the fatigue strength is still
dominative, fatigue cracks will emerge from SB, GB, scratches
and so on, the quadratic Eqs. (12) or (23) can be approximately
expressed as Eq. (2), i.e., s
w
ms
b
, it is similar to Eq. (13) that is
controlled completely by IDs. However, the coefcient m should
be slightly smaller than the constant C in Eq. (13) due to the
inuence of PDs. Eq. (2) has been widely applied to most
engineering materials [25]. This can be named as main ID regime
(see Fig. 12). With further increasing the tensile strength, in this
case, s
P
w
becomes so large (violet line in Fig. 12) that the inuence
of PD on the fatigue strength is signicant; therefore, in this
tensile strength level it is dened as main PD regime (see Fig. 12).
In this regime, the fatigue cracks often nucleate at scratches,
surface or subsurface inclusion (SI), internal inclusion (II), and
even that crack initiation at internal inclusion becomes the main
event in those high-strength steels [5,6] with tensile strength
increasing (main PD regime, light green zone in Fig. 12). The
fatigue strength is signicantly affected by the size of inclusions
[5,7]. If the inclusion size is very different, the fatigue data will be
very scatter and Eq. (18) can suit well.
When the tensile strength reaches very high level, nearly all
fatigue cracks originate from PDs (inner inclusion, segregation), s
P
w
becomes very large, which can be called as PD regime shown in
Fig. 12. As tensile strength increasing, the critical inclusion size [50]
of steels becomes smaller and smaller, more and more inclusions
can become fatigue crack initiation sites and the fatigue sensitivity
to defects is much more serious [5]. Therefore, fatigue strength
decreases as tensile strength further increases.
4.3. Prediction of the maximum fatigue strength
If introducing Youngs modulus E of materials into Eq. (24),
one can get
s
P
w
Ms
2
b
=E Ms
b
e
b
25
where MP E, e
b
can be dened as the maximum tensile elastic
strain. From Eq. (12), the values of the maximum fatigue strength
s
max
w
and critical tensile strength s
bc
can be obtained as follows:
s
max
w

C
2
4P

EC
2
4M
26
s
bc

C
2P

EC
2M
27
From Eqs. (26) and (27), it can be concluded that s
max
w
and s
bc
are dependent on C, and P or E, C and M. According to their values,
the corresponding s
max
w
and s
bc
can be calculated, as listed in
Table 3. Compared with the error bands in Table 3, the error
bands of one type of materials such as steels, copper or aluminum
alloys are within 20%, which is the maximum. The error bands of
many steels under the same experimental condition are within
15%; and within 10% for low-carbon steel, with 5% for a specic
materials (SAE 4140, 4340, 2340 and 4063), which is minimum.
This indicates that the less are the inuencing factors of fatigue
experiments, the more accurate the formula predicts. Besides, it is
well known that the newly developed high-strength materials
often possess very poor fatigue properties [2732,41], which can
be explained that their tensile strengths are already greater than
their s
bc
, as illustrated in Fig. 12.
For the quadratic equation, the fatigue strength and the
maximum fatigue strength can be designed or improved by
adjusting the constants P (M) and C. For example, with surface
roughness of specimens increasing, the fatigue strength continu-
ously decreases [4]; with the size of inclusion increasing, fatigue
strength continuously decreases [5,7], which will affect the
constant P (M). With renement of grain size, the fatigue strength
also increases [24], which should make a contribution to constant
C. Therefore, the universal fatigue formula may not only correctly
predict the fatigue strength, but also provide a possible clue to
improve the fatigue strength by choosing and designing micro-
structure or eliminating the processing defects of the materials in
industry.
Fig. 13. Relation between threshold stress intensity range DK
th
at R0.05 and
tensile strength of 300 M steel tempered by different temperatures, adapted from
Ref. [49].
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 340
Notice that the universal fatigue strength formula above is
proposed on the basis of smooth specimens during tension-
compression fatigue tests under the stress ratio of R1.
Actually, there are some other factors affecting the fatigue
strength, including surface condition, specimen size, loading
mode and so on. Small changes of those factors can also affect
the fatigue properties, making analytical prediction of fatigue life
or strength difcult. Therefore, it is necessary to quantify the
effects of those factors on the fatigue strength in the future.
5. Summary and conclusions
After a systemic study on the fatigue strength s
w
of SAE 4340
steel tempered at different tempering temperatures, a universal
fatigue strength formula s
w
s
b
(CP s
b
) is rstly proposed by
taking into account the constants C and P. Through inspection of
numerous data available of other materials such as steels, copper
and aluminum alloys, it is shown that this universal fatigue
strength formula can well describe the fatigue strength s
w
of
different materials with increasing their tensile strengths s
b
.
Combining the variation tendency of fatigue crack sites and the
competition of defects, the fatigue damage mechanisms with the
change of tensile strength are explained well. The decrease in the
fatigue strength at high-strength level can be explained by fracture
mechanics and attributed to the transition of fatigue cracking sites
from surface to the inner inclusions, resulting in the maximum
fatigue strength s
max
w
at an adequate tensile strength level.
In addition, the maximum fatigue strength s
max
w
and the
critical tensile strength s
bc
can be calculated and the universal
fatigue strength formula can explain why many metallic materials
with excessively high strength do not display high fatigue
strength. The current universal fatigue strength formula may
provide a new clue or principle to achieve the higher fatigue
strength s
w
of materials by adjusting the constants P and C in the
future.
Acknowledgements
The authors would like to thank Mrs. C. L. Dai and Dr. H. F. Zou
for their helps of the SEM and EBSD observations. This work is
supported by National Natural Science Foundation of China
(NSFC) under Grant Nos. 50625103, 50890173, 50931005 and
the National Basic Research Program of China under grant No.
2012CB631006.
References
[1] N.A. Fleck, K.J. Kang, M.F. Ashby, Acta Metall. Mater. 42 (1994) 365381.
[2] C.X. Shi, Q.P. Zhong, C.G. Li, China Materials Engineering Canon Fundamentals
of Materials Engineering, vol. 1, Chemical Industry Press, Beijing, 2005.
[3] P.G. Frorrest, Fatigue of Metals, Pergamon Press, Oxford, 1962.
[4] Y.L. Lee, J. Pan, R.B. Hathaway, M.E. Barkey, Fatigue Testing and Analysis
(Theory and Practice), Elsevier Butter-worth Heinemann, Amsterdam, Boston,
Heidelberg, 2005.
[5] Y. Murakami, Metal FatigueEffects of Small Defects and Nonmetallic
Inclusions, Elsevier Science Ltd, Oxford, 2000.
[6] C. Bathias, P.C. Paris, Gigacycle Fatigue in Mechanical Practice, Marcel Dekker,
New York, 2005.
[7] S.X. Li, Int. Mater. Rev. 57 (2012) 92114.
[8] L. To th, S.Y. Yarema, Mater. Sci. 42 (2006) 673680.
[9] Q.Y. Wang, J.Y. Berard, A. Dubarre, G. Baudry, S. Rathery, C. Bathias, Fatigue
Fract. Eng. Mater. Struct. 22 (1999) 667672.
[10] Y.B. Liu, Z.G. Yang, Y.D. Li, S.M. Chen, S.X. Li, W.J. Hui, Y.Q. Weng, Mater. Sci.
Eng. A 517 (2009) 180184.
[11] Y.J. Zhang, W.J. Hui, J.Z. Xiang, H. Dong, Y.Q. Weng, Acta Metall. Sin. 45 (2009)
880886.
[12] J.M. Zhang, S.X. Li, Z.G. Yang, G.Y. Li, W.J. Hui, Y.Q. Weng, Int. J. Fatigue 29
(2007) 765771.
[13] B. Zettl, H. Mayer, C. Ede, S. Stanzl-Tschegg, Int. J. Fatigue 28 (2006)
15831589.
[14] Z.G. Yang, S.X. Li, J.M. Zhang, J.F. Zhang, G.Y. Li, Z.B. Li, W.J. Hui, Y.Q. Weng,
Acta Mater. 52 (2004) 52355241.
[15] Y.H. Nie, W.T. Fu, W.J. Hui, H. Dong, Y.Q. Weng, Fatigue Fract. Eng. Mater.
Struct. 32 (2009) 189196.
[16] Y.B. Liu, Z.G. Yang, Y.D. Li, S.M. Chen, S.X. Li, W.J. Hui, Y.Q. Weng, Mater. Sci.
Eng. A 497 (2008) 408415.
[17] Y.D. Li, Z.G. Yang, S.X. Li, Y.B. Liu, S.M. Chen, Acta Metall. Sin. 44 (2008)
968972.
[18] S.M. Chen, Y.D. Li, Y.B. Liu, Z.G. Yang, S.X. Li, Z.F. Zhang, Acta Metall. Sin. 45
(2009) 428433.
[19] E. Bayraktar, I.M. Garcias, C. Bathias, Int. J. Fatigue 28 (2006) 15901602.
[20] W. Li, T. Sakai, Q. Li, L.T. Lu, P. Wang, Mater. Sci. Eng. A 528 (2011)
50445052.
[21] T.E. McGreevy, D.F. Socie, Fatigue Fract. Eng. Mater. Struct. 22 (1999)
495508.
[22] A. Inoue, B.L. Shen, H. Koshiba, H. Kato, A.R. Yavari, Nat. Mater. 2 (2003)
661663.
[23] L. Lu, Y.F. Shen, X.H. Chen, L.H. Qian, K. Lu, Science 304 (2004) 422426.
[24] R.Z. Valiev, T.G. Langdon, Prog. Mater. Sci. 51 (2006) 881981.
[25] A. Vinogradov, J. Mater.Sci. 42 (2007) 17971808.
[26] Y. Estrin, A. Vinogradov, Int. J. Fatigue 32 (2010) 898907.
[27] W. Wang, W. Yan, Q.Q. Duan, Y.Y. Shan, Z.F. Zhang, Y. Ke, Mater. Sci. Eng. A
527 (2010) 30573063.
[28] A. Vinogradov, S. Hashimoto, Mater. Trans. 42 (2001) 7484.
[29] L. Tang, L. Lu, Acta Metall. Sin. 45 (2009) 808814.
[30] K. Fujita, T. Hashimoto, W. Zhang, N. Nishiyama, C. Ma, H. Kimura, A. Inoue,
J. Jpn. Inst. Met. 70 (2006) 816823.
[31] G.Y. Wang, P.K. Liaw, A. Peker, B. Yang, M.L. Benson, W. Yuan, W.H. Peter,
L. Huang, A. Freels, R.A. Buchanan, C.T. Liu, C.R. Brooks, Intermetallics 13
(2005) 429435.
[32] H. Ishii, T. Yagasaki, H. Akagi, Fatigue Fract. Eng. Mater. Struct. 25 (2002)
831835.
[33] G.E. Totten, Steel Heat Treatment: Metallurgy and Technologies, CRC Press
Taylor & Francis Group, Boca Raton, London,New York, 2007.
[34] Y. Gan, Z.L. Tian, H. Dong, D. Feng, X.L. Xin, China Materials Engineering
Canon Steel materials Engineering, vol. 3, Chemical Industry Press, Beijing,
2005.
[35] G. Aggen, F.W. Akstens, C.M. Allen, H.S. Avery, P. Babu, A.M. Bayer, et al., ASM
Handbook Properties and Selection: Irons, Steels, and High-Performance
Alloys, vol. 1, ASM International, the United States of America, 1990.
[36] W.D. Callister, Materials Science and Engineering: An Introductin, John Wiley
& Sons, Inc., the United States of America, 2007.
[37] J.S. Pan, J.M. Tong, M.B. Tian, Materials Science and Engineering, Tsinghua
University Press, Beijing, 2011.
[38] M.F. Garwood, H.H. Zurburg, M.A. Erickson, Correlation of Laboratory Tests
and Service Performance, American Society for Metals, Philadelphia, PA,
1951, pp. 177.
[39] P. Zhang, S.X. Li, Z.F. Zhang, Mater. Sci. Eng. A 529 (2011) 6273.
[40] K. Shiozawa, L. Lu, S. Ishihara, Fatigue Fract. Eng. Mater. Struct. 24 (2001)
781790.
[41] J.C. Pang, Q.Q. Duan, S.D. Wu, S.X. Li, Z.F. Zhang, Scripta Mater. 63 (2010)
10851088.
[42] J.C. Pang, M.X. Yang, G. Yang, S.D. Wu, S.X. Li, Z.F. Zhang, Mater. Sci. Eng. A
553 (2012) 157163.
[43] Y. Gan, Z.L. Tian, H. Dong, D. Feng, X.L. Xin, China Materials Engineering
Canon Steel materials Engineering, vol. 2, Chemical Industry Press, Beijing,
2005.
[44] T. Sawai, S. Matsuoka, K. Tsuzaki, Tetsu To Hagane 89 (2003) 726733.
[45] M. Okayasu, K. Sato, M. Mizuno, D.Y. Hwang, D.H. Shin, Int. J. Fatigue 30
(2008) 13581365.
[46] Z.F. Zhang, Z.G. Wang, Prog. Mater. Sci. 53 (2008) 10271099.
[47] Y. Murakami, S. Kodama, S. Konuma, Trans. Jpn. Soc. Mech. Eng. 54 (1988)
688696.
[48] Y. Murakami, M. Endo, Eng. Fract. Mech. 17 (1983) 115.
[49] R.O. Ritchie, Met. Sci. 8 (1977) 368381.
[50] Z.G. Yang, J.M. Zhang, S.X. Li, G.Y. Li, Q.Y. Wang, W.J. Hui, Y.Q. Weng, Mater.
Sci. Eng. A 427 (2006) 167174.
J.C. Pang et al. / Materials Science & Engineering A 564 (2013) 331341 341