Professional Documents
Culture Documents
KTHR Ktho
KTHR Ktho
PII: S0020-7403(17)32128-8
DOI: 10.1016/j.ijmecsci.2017.12.014
Reference: MS 4078
Please cite this article as: Xiaojing Cai , Ri Xia , Mingchen Huo , Jinquan Xu , A threshold formula for
fatigue crack growth with mean stress intensity factors, International Journal of Mechanical Sciences
(2017), doi: 10.1016/j.ijmecsci.2017.12.014
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Highlights
The variation of threshold with mean SIF is similar to that of fatigue limit with mean stress.
The plastic damage at crack front would affect threshold greatly.
A threshold formula with respect to mean SIF has been derived.
A formula to estimate the critical mean stress intensity factor has been derived.
T
IP
CR
US
AN
M
ED
PT
CE
AC
* Corresponding author.
E-mail address: jqxu@sjtu.edu.cn (J.-Q. Xu).
1
ACCEPTED MANUSCRIPT
Graphical Abstract
T
IP
CR
US
AN
M
ED
PT
CE
AC
* Corresponding author.
E-mail address: jqxu@sjtu.edu.cn (J.-Q. Xu).
2
ACCEPTED MANUSCRIPT
Abstract: From the fatigue limit condition of crack front, a threshold formula has been derived
with respect to mean stress intensity factors. Examinations with experimental results show that
T
this formula can express threshold quantitatively and uniformly. It is found that threshold varies
with mean stress intensity factor just as that fatigue limit varies with mean stress if the mean stress
IP
intensity factor is smaller than a critical value, while if the mean stress intensity factor is larger
than the critical value, plastic deformation at crack front would affect threshold greatly. It has also
CR
been investigated that why there would be such a critical mean stress intensity factor behavior.
The reason lies on that whether damage induced by plastic deformation at crack front could be
neglected or not. A formula to estimate the critical mean stress intensity factor has also been
derived.
US
Keywords: Threshold, Mean stress intensity factor, Fatigue limit, Mean stress, Crack front
1. Introduction
AN
Fatigue crack growth behavior is a very important material’s property in application, thereby,
the growth rate curve [1-6] of metals has been widely investigated in the last decades. The
threshold is especially of interest in engineering. However, since it is strongly dependent on stress
M
ratio or mean stress intensity factor (SIF), determining thresholds systemically by experiments
would take too much time and too many costs. Besides, many other factors, such as environment
ED
and temperature, local behavior of crack tip, etc., would influence the value of threshold. Even the
measurement method would influence its value too. Of course, micro-mechanisms [7-12] also play
a very important role in detailed threshold behavior. On the other hand, from the view of
application, a simple relationship between thresholds and mean stress ratios or mean stress
PT
intensity factors (SIFs) is strongly expected. So various empirical relationships [13-19] have been
proposed already in literature. These empirical relationships have indicated that there are various
CE
patterns of threshold’s variation with stress ratio, and different empirical formulas must be adopted
for different materials. It must be noted that threshold can be regarded as a kind of material
constant (containing the effect of material’s micro-structure) only if the stress field at crack front is
AC
SIF-dominated, and must be under the plane strain state and the same non-corrosive environment,
moreover, material’s damage should be induced by cyclic loading only. Thereby, any relationship
between threshold and mean stress ratio or SIF should be applied only to cases that these
limitations have been satisfied. Assuming that threshold is a kind of material constant, i.e., for
cases that all mentioned limitations are satisfied, this study tried to derive a unified formula for
threshold with some material constants only.
Considering that fatigue limit of material has been widely applied in engineering application,
correlating the threshold formula with fatigue limit may be a better choice. Focusing on the
threshold of crack propagation only, this study has derived a threshold formula with respect to
mean SIF, from the view of fatigue limit condition at crack front. Examinations with experimental
* Corresponding author.
E-mail address: jqxu@sjtu.edu.cn (J.-Q. Xu).
3
ACCEPTED MANUSCRIPT
results show that the formula can well express threshold’s variation with mean SIF for various
metals.
Nomenclature
T
K Im Mean stress intensity factor
IP
K Im C Critical mean SIF at which decreasing style of threshold changes
Ka Amplitude of stress intensity factor
CR
K thR Threshold corresponding to stress ratio R
K thRC Threshold corresponding to critical stress ratio RC
K th ( K Im ) Threshold corresponding to mean SIF K Im
Lf
R
RC
Characteristic length
Stress ratio
Critical stress ratio
US
AN
SIF Stress intensity factor
1 Fatigue limit of smooth material under symmetric fatigue
b Ultimate tensile strength
m
M
Mean stress
eq Equivalent stress of multiaxial fatigue
f ( m1 ) Fatigue limit of smooth material under mean stress m
ED
Poisson’s ratio
Ratio of plastic region size to characteristic length
CE
where K max and K min are nominal maximum and minimum SIFs, respectively. If K K th
(threshold, a kind of material constant if necessary limitations are satisfied), no crack propagation
would occur, while if K K th , fatigue crack growth is unavoidable. Early crack growth laws
[1, 20] were usually just described by the nominal SIF range K . But lately, taking the effect of
crack tip closure into account [21-24], the effective SIF range which is defined by
K ef K max K op , where K op is the nominal SIF at which crack tip closure appears, has become
the most widely used parameter to describe the growth law. Obviously, there would also be crack
tip closure phenomenon at threshold condition. So it may be more reasonable to define threshold
* Corresponding author.
E-mail address: jqxu@sjtu.edu.cn (J.-Q. Xu).
4
ACCEPTED MANUSCRIPT
by the effective SIF range. It is true that K op varies with nominal K , so growth laws
described by nominal and effective SIF ranges may be quite different, and usually those described
by K ef can fit experimental results well indeed. However, focusing on the threshold condition
only, K is a specified critical SIF range only, so K op , and thereby K ef , should also be
specified values too. That is, limited to, and only to the threshold’s expression, it is equivalent to
adopt either K or K ef , though their critical values are of course different. For this reason,
T
taking the convenience of application into account, we use the critical nominal SIF range to
IP
express threshold in this study. However, it does not mean that K and K ef are equivalent
CR
in describing crack growth law, in which the latter is generally more effective.
Empirical formulas of threshold existing in literature are different according to testers and
materials. They can be classified into 6 types as shown in Fig. 1.
US
Kth
AN
I
II IV
M
VI V
ED
III
-1.0 -0.5 0.0 R 0.5 1.0
Type I: The threshold decreases linearly with stress ratio at first, and then keeps constant [13-15]
hereafter if the stress ratio reaches a critical value RC . Denoting the threshold as K thR (where
CE
suffix R denotes the corresponding stress ratio), the relationship of this type can be expressed
mathematically as:
K thR (1 R)K th0 for R R C
(1a)
K thR C (1 RC )K th0 for R R C
AC
or
K thR K th0 BR for R R C
(1b)
K thR C K th0 BRC for R R C
where B is the slope of threshold’s decrease with stress ratio, and K th0 is the threshold at
R 0 . Obviously, Eq. (1a) in fact is only a special case of Eq. (1b) with B K th0 . In this
empirical formula, a linear decreasing style is assumed, which can fit experimental results for
some materials. Though there is no theoretical evidence that threshold would decrease linearly
with slope of B K th0 , Eq. (1) is a frequently used formula to estimate threshold. The
well-known physical explanation [17] to Eq. (1a) is that the threshold is dominated by K max
5
ACCEPTED MANUSCRIPT
when stress ratio is smaller than RC , and is dominated by K when stress ratio is larger than
RC . The advantage of Eq. (1) is that no additional fitting parameter is necessary, but at the same
time, it is also the shortage of Eq. (1), since the variation of threshold is basically
material-dependent and material constants should be involved logically. On the other hand, an
obvious difference between experimental thresholds and estimated results would appear when
R 1 . Moreover, K th1 2K th0 can be estimated from Eq. (1a), but it is also not true for
many materials.
Type II: The threshold decreases non-linearly with R ratio at first, and then keeps constant [16,17]
hereafter if the stress ratio reaches a critical value RC . The non-linear decreasing style can be
expressed by various functions of R , so many function forms are possible to be used to express
T
the empirical relationship. The simplest expression (thereby, the most popular but quite rough
IP
expression) is
CR
(2)
K thR C (1 RC ) K th0 for R R C
where is a fitting parameter to express decreasing style. Obviously, Eq. (1a) is in fact involved
in Eq. (2) if one takes 1 . However, Eq. (1b) cannot be involved in Eq. (2). This fact would
US
mean that Eq. (2) might not always be the proper fitting function, even though a fitting parameter
has been introduced. Especially, the first formula in Eq. (2) is commonly known as Walker’s
AN
approach [16,17].
Type III: For R RC , this type is the same as Type II. The difference is that this type [15]
considered threshold’s non-linear decreasing for R RC too. Such a consideration is logically
more reasonable since that K th1 at R 1 must be zero. There were indeed experimental results
M
[13,18] which do not keep constant after RC . The simplest expression of this type is
KthR (1 R) Kth 0 for R R C
(3)
(1 RC )
ED
where is also a fitting parameter, and K thRC denotes the threshold corresponding to the
PT
summarized below.
Type IV: The threshold decreases with stress ratio non-linearly but smoothly [13,18]. The
AC
simplest expression is
In this type, there is no critical stress ratio RC . It should be noted that this empirical formula
usually cannot be extended to R 0 , since the fitting parameter is obtained by experimental
results for R 0 only.
Type V: This type is similar to Type III, but it is limited only to R 0 [19].
Type VI: This type is similar to Type I, but it is limited only to R 0 [25-28].
From above reviews, it can be concluded that threshold decreases with stress ratio or mean SIF
generally. There would be a critical stress ratio RC or mean SIF K Im C at which the decreasing
6
ACCEPTED MANUSCRIPT
style changes. Almost all empirical formulas on threshold were concluded with respect to stress
ratios. It is not difficult to transform them into the form with respect to mean SIFs by using
1 R
K Im K thR (5)
2(1 R)
In this study, we will develop a unified formula of threshold with respect to mean SIF theoretically.
To distinguish them clearly, we denote the threshold corresponding to mean SIF as K th ( K Im ) ,
while those corresponding to stress ratios as K thR .
T
Threshold behavior of crack growth can be well explained by dislocations’ pile-up ahead of
crack tip [7-10]. It is true that the threshold formula may be better to be derived from its
IP
micro-mechanisms. However, threshold itself is basically a macroscopic property, so correlating it
with macro fatigue properties such as fatigue limit may be a more applicable method. For the
CR
purpose of correlating, we consider a certain length L f at crack front (noting that it is introduced
only for correlating threshold condition to fatigue limit condition, it will vanish finally in the
x y
K Im K a sin t
0
US
threshold formula), as shown in Fig. 2. Stress distribution near the crack tip can be expressed as:
, z 2 ( K Im K a sin t )
Plane stress
at 0 (6)
AN
2r P l a ns et r a i n
2r
where is the Poisson’s ratio, K Im is the mean SIF induced by remote mean stress, K a is the
M
amplitude of SIF corresponding to remote stress amplitude, and r is the distance from crack tip
at crack front, as shown in Fig. 2. So a crack front is always under multiaxial stress state. To
correlate the threshold behavior with fatigue limit, introducing an equivalent fatigue stress to
ED
multiaxial stresses may be the simplest method. Various definitions [29-32] of equivalent fatigue
stress can be considered. However, whatever an equivalent stress definition is used, since the
stress field is determined by the SIF only, the equivalent fatigue stress at crack front can be always
PT
expressed as:
y
CE
Singular stress
r
AC
x
Lf
F ( K Im K a sin t )
eq meq aeq sin t (7a)
2r
FK a FK Im
aeq , meq (7b)
2r 2r
where F is a constant depended on the definition of equivalent fatigue stress. Here we remain it
7
ACCEPTED MANUSCRIPT
stress amplitude a1 and mean stress m1 within the length L f are considered.
FK
Lf
1 FK a 2
a1 dr FK a (8a)
Lf 0 2r L f 2L f
Lf
1 FK Im 2
m1 dr FK Im (8b)
Lf 0 2r L f
T
Focusing on the damage within length L f , it surely should be different from that in a smooth
IP
specimen due to the effect of crack. Denoting the initial damage within length L f as D front 0 , it
CR
can be considered to be composed by two parts, i.e., 1) the damage induced by plastic deformation
at crack front due to the mean SIF, 2) the fatigue damage induced by cyclic deformation due to
cyclic SIF. So the initial damage at crack front can be expressed as:
US
D front 0 Dac (K ) D p ( K Im ) (9)
AN
where Dac (K ) denotes the fatigue damage and D p ( K Im ) denotes the plastic damage. Focusing
on the threshold condition, Dac 0 (otherwise, threshold condition cannot be reached due to the
damage accumulation. It should be noted that Dac 0 at a propagating crack front), but plastic
M
deformation will always exist at crack front, so that D p 0 generally if there is mean SIF.
According to the concept of effective stress in continuum damage mechanics, the fatigue limit
ED
a1
f ( m1 ) (10)
PT
1 D front 0
where f ( m1 ) is the fatigue limit corresponding to mean stress m1 . This fatigue limit
CE
condition means that if the effective amplitude is lower than the fatigue limit, damage
accumulation would not happen, so the crack would not grow. That is, Eq. (10) is an alternative
expression for threshold condition of K K th ( K Im ) . From this threshold condition, deriving
AC
8
ACCEPTED MANUSCRIPT
consider the symmetrical fatigue at first, i.e., consider the case with K Im 0, m1 0 . Limited to
this case and its threshold condition, one has D front 0 0 , so one gets
1
K th (0) K th1 2L f (12)
F
Regarding K th1 , 1 as material constants, the length L f can be determined from Eq. (12) as:
2
1 FK th (0)
Lf (13)
2 1
T
IP
Obviously L f is a material constant too, so we call it fatigue characteristic length below, and can
CR
also be applied to non-symmetrical fatigue cases. From the fatigue limit condition at crack front
within this length, i.e., substituting Eq. (13) into Eq. (11), one gets the theoretical threshold
formula as:
K th ( K Im )
US
(1 D front 0 ) f ( m1 )
1
K th (0) (14)
AN
The parameter F disappears in form. This fact means that whatever a definition of equivalent
stress is used, the variation of threshold with mean SIF is just similar to that of fatigue limit with
mean stress. However, it should be noted that the mean stress here corresponds to the equivalent
mean stress at crack front, not to the remote mean stress. Since m1 is much larger than the
M
remote mean stress due to the singular behavior at crack front, thresholds would decrease more
ED
However, to apply the formula Eq. (14), following two difficulties have to be cleared yet.
PT
mean SIF, it is reasonable to assume that the damage induced by plastic deformation at crack front
can be neglected if the mean SIF is small enough (i.e., D front 0 D p ( K Im ) 0 ). But for cases with
AC
distinguish whether the plastic damage can be neglected or not, Eq. (14) can be rewritten as:
f ( m1 )
K th (0) for K Im K Im C
1
K th ( K Im ) (15)
f ( m1 p )
1 D p ( K Im ) 1
K th (0) for K Im K Im C
Although m1 can be calculated by Eq. (8b) for K Im K Im C , the mean stress m1 p at crack
9
ACCEPTED MANUSCRIPT
front for K Im K Im C cannot be calculated by singular elastic field since the stress distribution
within L f may be dominated by HRR field rather than the K-field if the mean SIF is large. It is
is also very difficult to be determined. These facts mean that the second equation in Eq. (15) could
only be an empirical one, but it can help us to give the function form of empirical formula by such
a consideration. That is, representing
1 D p ( K Im ) 1 cK Im K Im C
T
f o r KI m K I Cm
(16)
m1 p bK Im
d
IP
where a modifying factor 1 cK Im K Im C has been introduced to replace 1 D front 0 , for the
CR
convenience of expressing threshold formula by macro parameters only. But parameters c,
US
expression, but the function form has taken HRR distribution into account. Since c, are fitting
parameters, the plastic damage Dp ( KIm ) is not an exact expression either. Eq. (16) has introduced
AN
an assumption that initial plastic damage varies with K Im K Im C in a power law function style.
Fortunately, as it will be examined later, such a variation function form can describe threshold’s
variation for many materials well. However, exceptions cannot be excluded since it is only an
assumption, though exceptions have not been found within the range of examined examples.
ED
Considering that the mean stress at the critical mean SIF must be continuous with Eq. (8b),
and when the mean SIF reaches K IC (fracture toughness), the mean stress must reach b
PT
(ultimate tensile strength) simultaneously, from the second equation of Eq. (16), one has:
2 1 K Im C
m1 p ( K Im C ) bK Im
d
C at K Im K Im C
CE
2 K Im C 1
log
b K th (0)
d , b db (18)
log(K Im C / K IC ) K IC
It can be seen that b, d are constants if K Im C has been estimated and can be calculated if
material constants are known. Anyway, b, d are not independent parameters, they are notations
introduced just for the convenience of expression.
Quantitative relationship of f ( m )
10
ACCEPTED MANUSCRIPT
To obtain quantitative estimation of threshold from Eq. (14) or (15), quantitative relationship
between fatigue limit and mean stress is necessary. There are several well-known empirical
relationships such as Goodman’s and Gerber’s formulas. However, somewhat surprisingly, they
can provide only very rough estimation of fatigue limit, and may be quite different from
experimental results [33] as shown in Fig. 3. Actually, any quantitative relationship of fatigue limit
can be used to estimate the threshold from Eq. (15). As an example, we use the relationship [34,35]
shown below for following examinations.
T
where , are material constants, b is the ultimate tensile strength. This formula is in fact a
IP
generalized relationship. Taking 1, 1 , it becomes the well-known Goodman’s relationship.
In other words, Goodman’s relationship is accurate only for materials with constants 1, 1 .
CR
While for materials with 1, 1 , Goodman’s relationship cannot give accurate enough
estimations. Fig. 3 shows the comparison with Gao’s experimental results [33] for Aluminum
alloy LY12CZ. It can be seen that Eq. (19) can estimate fatigue limit exactly. It is noted that ,
US
are material’s fatigue properties, they would have different values for different materials, therefore
fatigue limit would have different decreasing styles with mean stress. Considering that the
AN
variation of fatigue limit is also a kind of material property, introducing two constants to represent
the effect of material is reasonable and thereby can give quantitative estimation for any material.
From Eq. (19), it can be seen that constant dominates the decrease of fatigue limit for small
mean stress since S m 1 , so one or two fatigue limits for small mean stresses is enough to
M
determine constant by Eq. (19). While constant dominates the decrease of fatigue limit
for large mean stress, so it can be determined by one or two fatigue tests with large mean stresses,
ED
100
LY12CZ
80 6E6 cycles
Fatigue Limit /MPa
Fatigue Limit
CE
60 Eq.(19) =7.5
=0.527
40
AC
Goodman
20
0
0 50 100 150 200 250 300 350 400 450 500
Mean stress /MPa
With above preparation, substituting Eqs. (8b) (12) and (16) (19) into (15), one gets the unified
formula of threshold as:
11
ACCEPTED MANUSCRIPT
for K Im K Im C
Kth ( K Im )
(20a)
(1 c( K Im K Im C ) ) p
for K Im K Im C
where
, S m 1
(20b)
2 b K th (0)
S mp K th (0) gK Im
d
S mp
K th (0) gK Im
d 2
4gS mp K th (0) K Im
d
p
2
(20c)
T
bK d bK th (0)
S mp 1 Im , g
b 1
IP
Since b is a constant connected with material’s properties as shown in Eq. (18), g can also be
determined according to material constants. Noting that , are the same material constants
CR
defined by Eq. (19) for fatigue limit. Only for the case that 1, 1 , the decreasing style of
threshold is a line for K Im K Im C . Various curved decreasing styles of threshold can be expressed
US
by Eq. (20) with taking different constants , , so it is a generalized form of threshold with
mean SIF. All types reviewed in Section 2 can be involved in Eq. (20). It should be noted that if
AN
the relationships between fatigue limit and mean stress are given in other forms (but should be a
quantitative one), thresholds can also be estimated by Eq. (15) directly. Formula Eq. (20) is
developed from Eq. (19) for the convenience of following examinations. If material constants
and have not been determined by fatigue limits already, they can also be determined directly
M
5. Examinations
Kloster [15] has reported experimental thresholds under various stress ratios for 34CrNiMo6.
Material’s yield and ultimate strength are Y 626MPa, b 819MPa , respectively, Poisson’s
PT
ratio is 0.3. From very few fatigue limit data [31,32] as shown in Fig. 4, material constants
contained in Eq. (19) can be determined as 0.05, 1.5 and 1 235MPa . The
CE
comparison of Eq. (20) with Kloster’s results is shown in Fig. 5. The critical value
K Im C 7.5MPa m is estimated from experimental data here. It must be noted that the curve for
AC
K Im K Im C shown in the figure is calculated by Eq. (20a) theoretically (noting that plane strain
condition is assumed) with the use of , determined in Fig. 4. Experimental results coincided
with theoretical curve well. While the curve for K Im K Im C is a fitted one by taking
c 0.095, 1.1 in the second equation of Eq. (20a). In other words, parameters c, can be
determined by fitting technology. On the other hand, since only two parameters are needed to be
determined, two or three experimental thresholds for large mean SIFs are enough to determine
12
ACCEPTED MANUSCRIPT
c, . Noting that K IC 80MPa m is assumed in fitting due to lack of data. Since the second
equation is intrinsically an empirical one which has to be fitted by experiments, here K IC can
also be explained as a pre-assumed parameter for the fitted region. However, if thresholds
corresponding to mean SIF near fracture toughness K IC are concerned, exact value of K IC
must be used. p has also been shown in Fig. 5, which corresponds to the threshold without the
effect of plastic damage, but with the effect of mean stress’s reduction due to the yielding. It can
T
be seen that the damage induced by plastic deformation at crack front plays a role of strengthening
IP
rather than weakening in this material. In other words, plastic deformation leads to a minus
CR
interesting and important on considering plastic damage.
20
34CrNiMo6 Y=626MPa,b=819MPa
300
US
1/2
1/2
Kth(0)=19MPa m Fixed R tests
Kth(KIm) /MPa m
15
=0.05
Stress amplitude /MPa
1/2
c=0.095, =1.1 KIC=80MPa m
100 5
M
p
0
0 0 5 10 15 20 25 30 35 40 45
0 200 400 600 800 1/2
ED
KIm /MPa m
Mean stress /MPa
Fig. 4. Fatigue limit for various mean stresses Fig. 5. Thresholds of 34CrNiMo6
PT
Fig. 6 compared Eq. (20) with Couper’s experimental thresholds [25] under various stress
ratios for Aluminum casting alloy Al-7Si-0.4Mg. The Poisson’s ratio is 0.28 , yield strength is
CE
K IC 18.5MPa m . , are fitted from experimental threshold results directly in this example.
AC
strengthen the material. It can be seen that threshold behavior for the mean SIF near to K IC can
also be well expressed by Eq. (20) if exact K IC has been used. This examination shows that Eq.
(20) can express experimental thresholds very well, even if the relationship between fatigue limit
13
ACCEPTED MANUSCRIPT
and mean stress is unknown. That is, at least, Eq. (20) can be used to express threshold curve in a
unified way for any material.
Fig. 7 compared Eq. (20) with Ohta’s results [38] for HT-80 steel. The material’s properties are:
due to lack of fatigue limit data. K Im C 6.0MPa m is estimated from experimental threshold’s
data. Noting that K IC 85MPa m is assumed in fitting for K Im K Im C due to lack of data. It
can be seen that p is larger than experimental results, so plastic deformation damage weakens
T
IP
the fatigue limit. That is, damage induced by plastic deformation does play the role of damage in
CR
whether strengthening or weakening effects of plastic deformation would appear. Here we try to
give an explanation. This material has a very large ratio of yield and ultimate tensile strength, i.e.,
US
Y / b 0.965 . This fact means that micro-cracking would accompany with plastic deformation
greatly. While if Y / b is small, micro-cracking would not happen or at least very few when the
AN
plastic strain is not very large. So focusing on damage, there are always two effects of plastic
deformation, i.e., hardening and micro defect effects. Representing them by damage conceptually,
it can be expressed as:
M
So minus damage would appear in total if Dmicrodefect Dhardening , while weakening effect would
Al-7Si-0.4Mg ( Y / b 0.7 ) belong to the former, while HT80 belongs to the latter. Due to these
CE
two effects of plastic deformation on damage, deriving threshold variation with mean SIF from
plastic strain only becomes nearly impossible, though many tries can be found in literature.
10
20
AC
1/2
Kth(0)=14.3MPa m
Kth(0)=8.4 MPa m
1/2
Kth(KIm) /MPa m
Fixed R tests 15
1/2
=0.16,=1.5 =0.12,=2.4
/MPa m
c=0.011,=2.5 c=-0.03,=1.1
1/2
KIC=18.5MPa m 1/2
2 5 KIC=85 MPa m
p
p
0
0
0 5 10 15 KIC 20
1/2 0 5 10 15 20
KIm /MPa m KIm /MPa m
1/2
14
ACCEPTED MANUSCRIPT
From above examinations, it can be seen that thresholds for K Im K Im C can be estimated
theoretically, while thresholds for K Im K Im C can be well characterized by the modifying factor
6. Discussions
T
It can be seen from Figs. 5-7 that determining the critical mean SIF from experiments would
IP
surely have some arbitrariness. The reason is simple, because it is basically a problem with
arbitrariness that how large yield region can be neglected. However, it may be very useful if a
CR
rough estimation formula of K Im C can be developed. Considering that the critical mean SIF
K Im C is related to the condition whether the effect of plastic damage can be neglected or not, it is
US
possible to derive it from the view of plastic deformation at crack front. The plastic deformation
can be described by von-Mises stress. According to the knowledge of fracture mechanics [39],
AN
stresses at a mode I crack front can be expressed as:
3
1 sin 2 sin 2
x
3
M
KI
y cos 1 sin sin (22)
2r 2 2 2
xy 3
sin 2 cos 2
ED
So von-Mises stress can be expressed as:
K Im K th ( K Im ) sin t / 2
PT
2 2
e max , f 2 2 2 (24)
2 r plane strain
3
So there would always be plastic deformation at the crack tip. However, if the yield region is very
small compared with the length L f as it has been considered for fatigue fracture of crack front,
the damage induced by such a very small yield region might be neglected.
The yielding region length LY under critical mean SIF can be determined by equivalent
15
ACCEPTED MANUSCRIPT
LY
dr f K Im C K th ( K Im C ) / 2
1 2
e max Y (25a)
LY
e max
LY 0
where e max is the maximum von-Mises stress at crack front that determined by Eq. (24), and
e max is the maximum equivalent von-Mises stress in one cycle. Using Eq. (13), denoting the
T
ratio of yielding region size to characteristic length as , one gets:
IP
2
L f ( K K th ( K Im C ) / 2)
Y 4 1 Im C
CR
(26)
Lf F Y K th (0)
To determine F , which is a constant related to the definition of equivalent stress for multiaxial
eq I12
US
stress state, following definition is adopted here to calculate LY / L f .
2(1 ) 2
AN
I2 (27)
1 2 2
where I1 , I 2 are the first and second stress invariant, respectively, and is the Poisson’s ratio.
M
Under this definition, F can be derived from Eq. (7) by substituting Eq. (6) into (27) as:
2(1 2 3 2 ) / (1 2 2 )
plane stress (28)
F
ED
Using K Im C , K th ( K imC ) estimated in Figs. 5-7, Fig. 8 shows LY / L f for above examined
PT
0.8
=Ly/Lf
0.6
AC
0.2 34CrNiMo6
0.0
Materials
It can be seen that locates within a very narrow range (0.29-0.34) for all examined materials.
That is, the critical mean SIF corresponds to the condition that the yielding region length reaches
16
ACCEPTED MANUSCRIPT
about 0.32 L f . From this assumption, the estimation formula of K Im C can be derived.
L f
e maxdr f K Im C K th ( K Im C ) / 2
1 2
e max Y (29)
L f 0 L f
where Y is the yield strength. Substituting Eq. (13) into Eq. (29), one gets
1 F Y
T
K Im C K th (0) K th ( K Im C ) (30)
2 f 1
IP
Substituting Eq. (20a) into (30) yields
CR
FK th (0) Y S m K th (0) 2 K Im C S m K th (0) 2K Im C 2 8S m K th (0) K Im C
K Im C
2 1 f 4
(31)
US
Eq. (31) could be solved for K Im C numerically. Taking 0.32 , it can be calculated that
AN
K Im C 9.1MPa m for 34CrNiMo6, K Im C 4.8MPa m for Al-7Si-0.4Mg, and
K Im C 5.0MPa m for HT80. It can be easily seen from Figs. 5-7 that these values are within the
M
arbitrariness on determining K Im C . So setting 0.32 in Eq. (31) can give the estimation of
2(1 RC )
PT
K th ( K Im C ) K Im C (32)
1 RC
7. Conclusions
A unified formula to estimate threshold of crack growth with respect to mean SIF has been
developed. Main results can be concluded as:
1) Below the critical mean SIF, the dependence of threshold on mean SIF is just similar to that of
fatigue limit on mean stress.
2) Beyond the critical mean SIF, plastic deformation would affect threshold. An empirical factor to
express the contribution of plastic deformation has been proposed.
3) Mean stress at crack front is different from that of remote applied loading, due to the multiaxial
effects.
17
ACCEPTED MANUSCRIPT
Acknowledgement
This work is supported by the National Natural Science Foundation of China (NSFC) under
grant No.10772116.
References
[1] Paris PC, Erdogan FA critical analysis of crack propagation laws. J Basic Eng Trans ASME
1963;85:528-534.
T
[2] Lin XB, Smith RA. Fatigue growth simulation for cracks in notched and unnotched round
bars. Int J Mech Sci 1998;40(5):405-419.
IP
[3] Huang XP, Torgeir M. Improved modeling of the effect of R-ratio on crack growth rate. Int J
Fatigue 2007;29(4):591-602.
CR
[4] Sun CQ, Lei ZQ, Hong YS. Effects of stress ratio on crack growth rate and fatigue strength
for high cycle and very-high-cycle fatigue of metallic materials. Mech Mater
2014;69(1):227-236.
US
[5] Bulloch JH, Callagy AG. A detailed study of the relationship between fatigue crack growth
rate and striation spacing in a range of low alloy ferritic steels. Eng Fail Anal
2010;17(1):168-178.
AN
[6] Zheng J, Powell BE. A method to reduce the scatter in fatigue crack growth rate data. Fatigue
Fract Eng Mater Struc 1997;20(9):1341-1350.
[7] Li XD. Dislocation pile-up model of fatigue thresholds for 2024-and 7075-alike aluminium
M
Fatigue 2017;94:131–139.
[9] Zhang P, Zhu Q, Hu C, Wang CJ, Chen G, Qin HY. Cyclic deformation behavior of a
nickel-base superalloy under fatigue loading. Mater Design 2015;69:12–21.
PT
[10] Sangid MD, Maier HJ, Sehitoglu H. A physically based fatigue model for prediction of crack
initiation from persistent slip bands in polycrystals. Acta Mater 2011;59(1):328–341.
[11] Tao YS, Fang QH, Zeng X, Liu YW. Influence of dislocation on interaction between a crack
CE
[13] Kujawski D, Ellyin F. A unified approach to mean stress effect on fatigue threshold
conditions. Int J Fatigue 1995;17(2):101-106.
[14] Lal DN, Namboodhiri TKG. A model for the effect of mean stress on the threshold condition
for fatigue crack propagation. Mater Sci Eng A 1990;130:37-49.
[15] Kloster V, Richard HA, Kullmer G. Experimental characterization of the threshold and
fatigue crack growth behaviour regarding negative stress ratios. Ann Nucl Energ
2013;40(1):14-24.
[16] Walker K. The effect of stress ratio during crack propagation and fatigue for 2024-T3 and
7075-T6 aluminum. In: Effect of environment and complex load history on fatigue Life.
ASTM International 1970:1-14.
18
ACCEPTED MANUSCRIPT
[17] Mann T. The influence of mean stress on fatigue crack propagation in aluminum alloys. Int J
Fatigue 2007;29:1393-1401.
[18] Radhakrishan VM. Endurance diagram. Int J Fatigue 1990;12(6):513-517.
[19] Doker H. Fatigue crack growth threshold: implications, determination and data evaluation.
Int J Fatigue 1997;19(93):145-149.
[20] Erdogan FA, Ratwani M. Fatigue and fracture of cylindrical shells containing a
circumference crack, Int J Fract Mech 1970;6(4):379-392.
[21] Correia JAFO, Jesus AMPD, Moreira PMGP, et al. Crack closure effects on fatigue crack
propagation rates: application of a proposed theoretical model. Adv Mater Sci Eng
2016;2016(3):1-11.
T
[22] Pippan R, Hohenwarter A. Fatigue crack closure: a review of the physical phenomena.
IP
Fatigue Fract Eng Mater Struc 2017;40(4):471-495.
[23] Antunes FV, Correia L, Ramalho AL. A parameter for quantitative analysis of plasticity
CR
induced crack closure. Int J Fatigue 2015;71:87-97.
[24] Gardin C, Fiordalisi S, Sarrazin-Baudoux C, et al. Numerical simulation of fatigue
plasticity-induced crack closure for through cracks with curved fronts. Eng Fract Mech
2016;160:213-225.
US
[25] Couper MJ, Griffiths JR. Effects of crack closure and mean stress on the threshold stress
intensity factor for fatigue of an Aluminum casting alloy. Fatigue Fract Eng Mater Struc
AN
1990;13(6):615-624.
[26] Priddle EK. The influence of grain size on threshold stress intensity for fatigue crack growth
in AISI 316 stainless steel. Scripta Metall 1978;12(1):49-56.
M
[27] Bignonnet A, Dias A, Lieurade HP. Influence of crack closure on fatigue crack propagation.
ICF6 1984;3:1861-1868.
[28] Ibrahim FK. The effects of stress ratio, compressive peak stress and maximum stress level on
ED
fatigue behavior of 2024-t3 Aluminum alloy. Fatigue Fract Eng Mater Struc 1989;12(1):1-8.
[29] Matsubara G, Nishio K. Multiaxial high-cycle fatigue criterion considering crack initiation
and non-propagation. Int J Fatigue 2013;47:222–231.
PT
[30] Cai XJ, Xu JQ. Effective fatigue stress and criterion for high-cycle multi-axial fatigue. J
Mater Eng Perform 2015;24(1):158-166.
[31] Mamiya EN, Castro FC, Algarte RD, Araujo JA. Multiaxial fatigue life estimation based on a
CE
[33] Gao ZT. A handbook on fatigue properties of aeronautical material. Beijing: Beijing Material
Research Institute; 1981 (in Chinese).
[34] Cai XJ, Xu JQ. A quantitative relationship of high cycle fatigue limit with mean stresses. Eng
Mech 2015;32(10):25-30.
[35] Hou SQ, Xu JQ. Relationship among S-N curves corresponding to different mean stresses or
stress ratios. J Zhejiang Univ Sci A 2015;16(11):885-893.
[36] Dzugan J, Novy Z, Maresova M. Fatigue-life enhancement of sheets made of 34CrNiMo6
steel. Mater Tehnologije 2014;48(6):959–963.
[37] Sander M, Müller T, Lebahn J. Influence of mean stress and variable amplitude loading on
the fatigue behaviour of a high-strength steel in VHCF regime. Int J Fatigue 2014;62:10-20.
19
ACCEPTED MANUSCRIPT
[38] Ohta A, Sasaki E. Influence of stress ratio on threshold level for fatigue crack propagation in
high strength steel. Eng Fract Mech 1977;9(2):307-315.
[39] Okamura H. Linear fracture mechanics. Tokyo: Baifukan; 1976.
T
IP
CR
US
AN
M
ED
PT
CE
AC
20