Accepted Manuscript

Fatigue crack tolerance design for stainless steel by crack growth analysis

Masayuki Kamaya

PII: S0013-7944(17)30133-9
DOI: http://dx.doi.org/10.1016/j.engfracmech.2017.03.038
Reference: EFM 5463

To appear in: Engineering Fracture Mechanics

Received Date: 31 January 2017

Revised Date: 20 March 2017
Accepted Date: 21 March 2017

Please cite this article as: Kamaya, M., Fatigue crack tolerance design for stainless steel by crack growth analysis,
Engineering Fracture Mechanics (2017), doi: http://dx.doi.org/10.1016/j.engfracmech.2017.03.038

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.
Fatigue crack tolerance design for stainless steel by crack growth analysis

Masayuki Kamaya

E-mail: kamaya@inss.co.jp

Institute of Nuclear Safety System, Inc.

64 Sata, Mihama-cho, Mikata-gun, Fukui 919-1205, Japan

TEL: +81-50-71050096
FAX: +81-770-372009

ABSTRACT

Fatigue damage of stainless steel was assessed by crack growth analysis. In order to characterize
small crack initiation and growth, fatigue cracking behavior was observed by periodical replica
investigations during a strain-controlled fatigue test in air at room temperature. It was shown that fatigue
cracks with depths of several tens of micrometers were initiated in the early stage of the fatigue test and
fatigue life could be estimated as a sum of the cycles for crack initiation of 100 m in depth and those for
growth to 3 mm. Use of the equivalent stress intensity factor allowed prediction of the crack growth for a
given strain range. The crack growth analysis was also made to estimate the fatigue life prescribed by the
design fatigue curve. Then, the growth prediction was made for thermal transient conditions to take
advantage of the stress gradient in the depth direction and the relationship between the cumulative usage
factor (CUF) and crack size was shown. It was concluded that fatigue damage accumulated in a stainless
steel component can be estimated from an identified crack size, or the maximum CUF can be determined
even if no crack was detected by the inspection.

HIGHLIGHTS

1. Fatigue life was estimated by predicting crack initiation and growth

2. The design fatigue curve was divided into cycles for initiation and growth
3. Crack growth analysis procedure for thermal transient was developed and validated
4. The cumulative usage factor was correlated to crack size

KEY WORDS

Crack growth; stainless steel; design fatigue curve; flaw tolerance; equivalent stress intensity factor;
fitness-for-service assessment; low-cycle fatigue

1. INTRODUCTION

In structural design of nuclear power plant components, magnitude of fatigue damage is assessed
using the cumulative usage factor (CUF), which is the ratio of the number of cyclic loadings expected
during plant operation to the number of allowable cycles determined from the design fatigue curve (DFC)
prescribed in design codes [1,2]. The CUF is used not only for designs but also for assessing structural
integrity of operating plants [3,4], for which the factual CUF is calculated from the loading history

1
actually experienced. Although the CUF is controlled at less than unity in the design, the factual CUF
may exceed the critical value due to long term plant operation. Particularly, for components in the
primary loop of nuclear power plants, the factual CUF is amplified by considering the detrimental effect
of the high-temperature coolant water environment [5-8].
Even if the factual CUF exceeds unity during plant operation, the component does not necessarily
fail due to the fatigue damage because design factors are considered in determining the DFC from test
results. Actually, to the authors knowledge, no crack has been found in nuclear plant components for
which fatigue damage was assessed in the design. In order to manage the fatigue damage of operating
plants reasonably, it is important to know the actual fatigue damage accumulated in the components.
Since the fatigue damage is brought about by crack initiation and growth [9,10] and failure of specimens
in fatigue tests is caused by a grown crack, the fatigue damage assessment can be replaced with crack
initiation and growth analysis. If the relationship between the crack size and the CUF is determined, the
actual CUF can be estimated from the crack size identified by inspections [11]. Even if no crack is
detected, possible damage can be estimated from the crack size detectable by the inspection technique
applied.
The crack growth analysis permits more reasonable fatigue damage assessment. For example, the
crack growth analysis may derive a longer fatigue life for thermal fatigue than the assessment using the
CUF. The thermal fatigue, which is caused by a fluid temperature fluctuation, is accompanied by a steep
stress gradient in the depth direction [12-14] and the stress intensity factor (SIF) used for the crack growth
prediction can incorporate the stress gradient in the growth analysis [15,16]. It should be noted that most
cyclic loadings considered for nuclear power plant component designs are caused by a thermal transient
such as plant startup and shutdown [17]. The stress gradient also appears at the notch root, where the peak
stress is used for calculating the CUF. The crack growth analysis also allows the critical crack size to be
taken into account. The critical crack size for component failure depends on geometrical and loading
conditions of each component. The allowable cycles are extended if the component can tolerate a large
crack.
The crack growth analysis for fatigue damage assessment has already been implemented in the flaw
tolerance analysis prescribed in Section XI Appendix L of the ASME (the American Society of
Mechanical Engineers) Boiler Pressure Vessel Code (hereafter, ASME code) [18,19]. The flaw tolerance
analysis is performed if the CUF exceeds unity. The residual allowable cycle is calculated by a crack
growth analysis according to the fitness-for-service (FFS) assessment procedure prescribed in Section XI
for a postulated crack when no crack is found by an inspection. If the crack growth analysis can be made
for a small crack, which corresponds not to CUF =1 but to CUF <<1, it is possible to assess the fatigue
damage by the crack growth analysis.
Treatments of loading conditions and material properties in the crack growth analysis made for the
FFS assessment [18,20] are different from those for the CUF calculation for design, although both CUF
calculations and crack growth analysis are made for the same plant components and loading conditions.
For example, the strain range is used as the fatigue damage driving force in the design, while the stress
range (the SIF range) is used for the crack growth prediction. The crack growth rates used for the FFS
assessment have been obtained under the small scale yielding condition [21], while strain-controlled
fatigue test results were used for determining the DFC [22,23]. The effect of plastic strain was considered
using a correction factor (Ke factor) [24] in design assessments, whereas no correction is made for
general yielding in the crack growth analysis for FFS assessments. As for an influence of temperature, the
growth rate used for FFS assessments depends on the temperature [20], although the same DFC is used
for various temperature conditions [23,25]. The mean stress effect is considered according to the modified
Goodmans diagram in the DFC [23], although the stress ratio is used for the crack growth analysis [20].

2
The structural factor (safety margin) is considered in the DFC by a fixed ratio [23], whereas no explicit
structural factor is given for the growth rate in FFS assessments. In order to assess the fatigue damage by
the crack growth analysis and correlate the CUF to crack size, the gaps between the CUF calculation and
crack growth analysis for FFS assessments should be closed.
This study aimed at adopting crack growth analysis to the design and maintenance of stainless steel
components. First, the fatigue crack initiation and growth on specimen surface was observed by
periodical replica investigations in Chapter 2. Then, in Chapter 3, fatigue life was estimated by predicting
crack growth with an incubation period before a small crack initiation. Furthermore, allowable cycles
prescribed in the DFC were simulated using the crack growth rate for FFS assessments. In Chapter 4, the
procedure for predicting crack growth for thermal stress was developed and it was shown that the fatigue
damage was reasonably assessed by the crack growth prediction. Further discussions were made in
Chapter 5 for treatments of mean stress, environmental effect and safety margin.

2. OBSERVATIONS OF CRACK INITIATION AND GROWTH

2.1 Fatigue test procedure

The material used for the fatigue test was solution heat-treated Type 316 austenitic stainless steel. Its
chemical composition (in mass %) was: C, 0.06; Si, 0.50; Mn, 1.30; P, 0.031; S, 0.027; Ni, 10.18; Cr,
16.94; Mo, 2.02 and balance Fe. The average 0.2% proof strength, tensile strength, elongation and
reduction of area from two tensile tests were 297 MPa, 611 MPa, 0.58 and 0.79, respectively.
Figure 1 shows the specimen geometry. The diameter was 10 mm at the minimum cross section and
it was changed gradually in order to localize the crack initiation to the center in the axial direction. The
stress and strain were derived assuming that the specimen had a parallel gage section. The 12.5 mm gage
length extensometer was used for strain measurements. The surface of the specimen was polished using
up to 3 m diamond paste.
The pull-push axial strain-controlled fatigue test was conducted in a room temperature laboratory
environment using a servo-electric test machine. Strain range was controlled to 0.6% under a constant
strain rate of 0.4%/s and the number of cycles to failure (fatigue life) Nf was 41,500. The crack growth
during the tests was monitored by taking replicas of the specimen surface using acetyl cellulose films.
The interval of the replica investigations was 2,000 cycles at the beginning of the test and it was reduced
to 500 cycles before the specimen failure.

2.2 Test results

Figure 2 shows the change in the measured surface length (2c) of a primary crack, which caused
specimen failure. The primary crack was first observed at the number of cycles of N = 4,000. The initial
length was 12.5 m and it grew continuously. The crack coalescence was not observed for the primary
crack, although many cracks were observed on the specimen surface.
The number of cycles was normalized by the fatigue life Nf in Fig. 3. Test results for = 1.2% [26]
and 2.0% [27] obtained using the same material are also shown. The number of cycles when the first
cracks were observed normalized by the fatigue life N/Nf was 0.096, 0.085 or 0.478 for = 0.6%, 1.2%
or 2.0%, respectively. It should be noted that it was difficult to distinguish the crack from slip steps for
= 2.0% due to the roughened surface. Although the initial crack length was 130.6 m for = 2.0%, the
smaller crack was deduced to be initiated earlier. The changes in crack length with N/Nf were almost the
same regardless of .

3
 Figure 4 shows the crack growth rate obtained from the change in crack length. The figure includes
the growth rates identified by the replica technique obtained in previous studies [26][27] for stress or
strain controlled tests using the same material. A simple point-to-point method was employed to obtain
the crack growth rates. The growth rate to the depth direction da/dN was derived from that to the surface
direction by assuming the ratio of the depth a to the surface length 2c to be a/2c = 0.5. The ratio a/2c
measured from the fractured surface was approximately 0.5 [28] and it was the same for the other
conditions [26,27].
The SIF range K was calculated by:

K f a (1)

where is the range of stress and the measured value at each step was used. The geometrical correction
factor f was obtained by [26]:

3 2
a a a
f 0.8379 0.6486 0.4128 0.6103 (2)
R R R

where R is the radius of the gage section of the specimen, which was 5 mm. The growth rates showed
significant scatter, although they increased with K. The larger strain or stress range tended to cause
faster growth rate under the same K. The fatigue limit of this material was a stress amplitude a of 240
MPa [29] and significant inelastic strain was observed even for a = 250 MPa [30]. Therefore, K could
not represent the driving force of the crack growths even for the high-cycle fatigue regime [27].
Figure 4b shows the relationship between the crack growth rates da/dN and the equivalent stress
intensity factor (Keq), which is defined by [11]:

Keq f E 25C a (3)

where E(25C) is the Youngs modulus at 25C and E(25C) = 195 GPa was used [30]. The crack growth rate
obtained for various loading conditions collapsed into a single curve. The Keq was derived from the
stress intensity factor by replacing the stress range with the strain range multiplied by the Youngs
modulus. Since the fatigue life correlated well with the strain range and the incubation period before the
small crack initiation was relatively short, it was reasonable that the crack growth rates correlated well
with the strain range rather than stress range. It is noteworthy that E(25C) used in Eq. (3) is equivalent to
the stress intensity used for the ordinate of the DFC [1,2]. The regression of all data shown in Fig. 4b was
obtained as:

5.06 1012 Keq

da 2.76
(4)
dN

where growth rate is given in m/cycle and Keq in MPa m0.5.

3. CORRELATION BETWEEN CRACK GROWTH AND FATIGUE LIFE

4
3.1 Fatigue life estimation by crack initiation and growth analysis

Since the incubation period before the initiation of a small crack of several tens of micrometers was
relatively short, the fatigue life should correspond to the cycles for the small crack to grow to the critical
crack size for specimen failure. The crack depth when the round-bar specimen failed was about 3 mm.
Then, the number of cycles necessary for a crack growing from its initial depth ai to 3 mm was calculated
using the growth rate given by Eq. (4) and denoted as Np. Figure 5 compares the Np and fatigue lives
obtained by the strain-controlled fatigue test using the same material [29]. The Np agreed well with the
test results particularly for the initial depth of ai = 50 m for relatively large strain range. The growth
predictions tended to give a shorter fatigue life than the test results for the smaller strain range. It was
deduced that the incubation period before the crack initiation was longer for the smaller strain range and it
was more than 106 cycles when was less than 0.4%.
In order to adjust the difference between the fatigue life obtained by the tests and Np, the incubation
period before the initiation of a crack of 100 m depth Ni(ai=100m) was assumed as:

Ni ai 100m 4811 %
3.922
. (5)

Equation (5) was derived by inverse analyses in which the sum of Ni(ai=100m) and Np(a=100m-3mm), which is
the number of cycles necessary for a crack to grow from 100 m to 3 mm in depth, gave identical number
of cycles in the fatigue life obtained by the test. Figure 6 shows the estimated fatigue life and Ni(ai=100m)
obtained by Eq. (5) together with the test results corresponding to Ni(ai=100m), for which the number of
cycles at a = 100 m was calculated by linear interpolation of results of the periodical replica
investigations. The test results were not always agree with the prediction according to Eq.(5). As shown
in Fig. 5, the Np agreed well with fatigue life of low-cycle fatigue regime even if incubation period was
not considered. The Ni(ai=100m) given by Eq.(5) can be regarded as the fitting parameter.
Thus, it was shown that the fatigue life can be estimated by predicting the crack growth by assuming
an incubation period before the small crack initiation. The equivalent stress intensity factor enabled
prediction of the crack growth for a given strain range even in the low-cycle fatigue regime. Similar
approach has been made for multiaxial fatigue for stainless and carbon steels [31].

3.2 Correlation between design fatigue curve and crack growth

In this section, the flaw tolerance analysis procedure was extended to that for a small crack
corresponding to CUF << 1. The crack growth prediction was made using the crack growth rate for
stainless steels in air environment prescribed in JSME (the Japan Society of Mechanical Engineers) FFS
code [20], which is referred to as FFS growth rate. The FFS growth rate for fully-reversed load is given
by:

da
10H 18.61103 K
3.3
(6)
dN
H 9.984 1.337 103 T 3.344 106 T 2 5.949 109 T 3 (7)

where T is the temperature given in C. The growth rate and K are given in m/cycle and MPa m0.5,
respectively. Figure 7 shows the growth rate given by Eqs. (6) and (4) together with the test results for
= 0.6%. The crack growth rates obtained using compact tension (CT) specimens made from the same

5
material [27] are also shown in the figure. The test was conducted with the stress ratio of R = 0.1 at 5 Hz
by controlling the applied load so that the small scale yielding condition was satisfied. The crack size was
monitored by the compliance method using a clip gage.
The crack growth rates obtained by the CT specimens agreed well with that of the FFS growth rate.
The growth rates for Keq (Eq. (4)) obtained by the round-bar specimens were also close to the FSS
growth rate, although some deviation was found for relatively large K. Since the growth rates used for
obtaining Eq. (4) included the results for the high-cycle fatigue regime, it is reasonable that Eq. (4) is
comparable to the FFS growth rate. As the fatigue life showed little dependency on test temperature under
the same strain range [22,23], it is deduced that the growth rate for the same Keq does not depend on the
temperature. On the other hand, the FFS growth rate varies with the temperature. The temperature
dependency of the growth rate might be brought about by that of Youngs modulus. The K and Keq are
not the same under a different temperature even for a fully elastic condition because the strain range
calculated from K for the different temperature depends on the Youngs modulus, whereas the fixed
Youngs modulus E(25C) is used for calculating Keq. In order to correct the temperature dependency of
the Youngs modulus and to make K equivalent to Keq, the following correction was made for the K:

E 25C
Kcorrected K (8)
E

where E is Youngs modulus for the test temperature and E = 174 GPa [32] was adopted for 325C in Fig.
7. The corrected K gave almost the same growth rate as that obtained for 25C. The FFS growth rate for
25C was used in the following analyses.
The DFC prescribed in the ASME code section III [1,23] was examined in this study. The DFC was
derived by a best-fit regression of numerous test results for stainless steels, which is expressed by Eq. (9)
[23].

ln Nf =6.891 1.920ln a [%] 0.112 (9)

The DFC of stainless steels analyzed in this study was drawn by considering margins in order to
incorporate effects of data scatter, component size, surface finish, loading history and so on. The details of
factors are given in Table 1 [23]. The conclusion made in Ref. [23] was that the DFC should be drawn by
reducing the best-fit regression by a factor of 2 on stress (strain) and 12 on cycles, whichever is more
conservative. Figure 8 shows the obtained DFC without considering the mean stress effect. As discussed
in section 5.3, the mean stress effect was not considered in the current study.
In order to divide the DFC into the incubation period and Np, the margin corresponding to 2 and 12
has to be considered in the crack initiation and growth analysis. Then, the margin for each factor listed in
Table 1 was assigned for initiation and growth as follows. Since the surface finish affects the incubation
period or initial crack size and has little influence on the crack growth, the margin for the surface finish
should be considered for the initiation. The size effect was also assigned to the initiation because the
effect was considered for incorporating difference in the risk volume for crack initiation. A chance for
larger crack initiation is deduced to be larger for actual components than the small specimens used in
testing. On the other hand, the loading history effect was assigned to the growth because it takes account
of the change in growth rate due to variable loading and the influence of growth for a small strain range
less than the fatigue limit. The material variability and data scatter were considered for both initiation and
growth. Then, as summarized in Table 1, the factor of 12 was divided into 5 for initiation and 2.4 for

6
growth. Accordingly, the crack growth rate for the crack growth analysis was given by:

2.4 109.95 18.61103 Keq .

da 3.3
(10)
dN

where constant H (= -9.95) was determined assuming T = 25C.

The initial depth was empirically determined to be 500 m in order to consider the margin for the
crack initiation. The growth prediction from a small initial depth such as 100 m did not reach 3 mm
within the cycles given by the DFC. The incubation periods were determined so that the sum of the
incubation and Np(a=500m-3mm) would be identical to the DFC for a given strain amplitude, and then, from a
regression of the incubations periods, Ni(ai=500m) was obtained as:

ln Ni ai=500m =3.794 2.202ln a [%] 0.056 . (11)

The fatigue limit was reduced from 0.112% to 0.056% in order to consider the factor of 2 reduction in the
strain. Figure 9 shows the incubation periods obtained and fatigue life. The predicted fatigue life
Ni(ai=500m) + Np(a=500m-3mm) was almost identical to that by the DFC. This means that the fatigue damage
assessment for component designs can be performed by the crack growth analysis together with the
incubation period before a crack initiation of 500 m in depth.

4. CRACK GROWTH UNDER THERMAL STRESS

Most of the fatigue damage considered in design of nuclear power plant components is brought
about by a thermal transient due to a change in operating mode. The cyclic stress and strain caused by the
fluid temperature fluctuation on a component surface has a gradient in the depth direction. Namely, the
stress fluctuation is largest at the surface and becomes smaller in the depth direction. The conventional
fatigue damage assessment using the CUF only considers the maximum stress on the surface. Retardation
in the crack growth due to the stress gradient may extend the fatigue life. In this chapter, a crack growth
analysis procedure was developed for the thermal stress and the influence of the stress gradient on the
fatigue life was discussed.

4.1 Thermal stress analysis

Stress response to ramp change of fluid temperature was analyzed in this study. The temperature was
assumed to change linearly from Tm to TmTf with a transient time of ts as shown in Fig. 10. The change
in temperature at depth z and time t, T(z,t), was derived by superposing step responses T(i)(z,t) [28] as
follows:

n
T z, t T i z, t i (12)
i 1

2sin k cos k z w k2 F

T i z, t Tfi 1 e (13)
k sin k cos k
k 1

7
where w is the thickness of the body. F =t/w2 (: thermal diffusivity) and k is the positive solution of:

Hw
tan Bi . (14)

H and are the heat transfer coefficient and heat conduction coefficient, respectively.
Once the temperature change T(z,t) is obtained, the thermal stress can be derived by dividing
T(z,t) into three components: the membrane component, Tm(t), bending component, Tb(t) and peak
component, Tp(z, t). Each component is obtained by [33]:

2z
T z, t Tm t Tb t 1 Tp z , t , (15)
w
1 w
Tm t T z, t dz ,
w 0
(16)

6 w
Tb t
w
T z , t dz . (17)
w2 0
2

The peak component is the remainder of the membrane and bending components and varies with z.
The membrane temperature component yields little thermal stress when there is no external deformation
constraint [34]. On the other hand, the bending and peak components are fully converted to thermal stress.
Accordingly, the stress change in the surface direction (z, t) caused by the temperature change T(z,t) is
obtained as:

E 2z
z, t Tb t 1 Tp z, t (18)
1 w

where and are the coefficient of linear expansion and Poissons ratio, respectively.
The analyses were performed for a stainless steel plate of thickness of w = 20 mm. The constants
used for the analyses are summarized in Table 2. The maximum order of k for k in Eq. (13) was 5 and the
time step in Eq. (12) was kept less than 0.05ts.
Figure 11 shows stress (z, t) calculated for a conditions of Tf = 100 K and H = 5,000 W/m2K. A
relatively small transient time ts = 10 s resulted in a large peak stress on the surface. The time of the peak
stress was about t = 12.7 s whereas it was almost identical to ts in the case of ts = 100 s. No clear peak was
observed for ts = 1,000 s. The stress change in the depth direction is shown in Fig. 11b for the time when
the stress on the surface showed the peak. The stress was the maximum on the surface and decreased in
the depth direction and it became compressive at deeper than 0.45w regardless of ts.

4.2 Stress intensity factor calculation

The weight function method [35-37] was applied for the SIF calculation in order to consider the
stress gradient in the depth direction. If the stress (z, t) is given by Eq. (18), SIF can be derived by:
a
K t z, t h z dz (19)
0

where K(t) is the SIF at time t, a is the depth of crack. h(z) is the weight and the value for a

8
two-dimensional crack has been obtained as [33]:

1 1 3

2 z 2 z 2 z 2
h z ,
a a
1 m1 1 m2 1 (20)
a a

2 6
a a
m1 0.6147 17.1844 8.7822 , (21)
w w

2 6
a a
m2 0.2502 3.2889 70.0444 . (22)

w w

It is noted that the weight derived by Eq. (20) is valid only when a/w < 0.5. In this study, initiation and
growth of a semi-elliptical surface crack was analyzed. The weight for the semi-elliptical crack was
obtained from reference [38].
Figure 12 shows SIF calculated for a two-dimensional crack of a = 3 mm using the stress
distribution obtained for Fig. 11. The change in SIF was similar to that of the stress at the surface.
However, the time for the peak of the SIF was larger than that for the surface stress; e.g. the peak of SIF
was observed at about 15.0 s and 100.3 s for ts = 10 s and 100 s, respectively, whereas it was not observed
clearly for ts = 1,000 s.
In order to validate the thermal stress and SIF analyses, finite element analyses were performed for a
two-dimensional model of plane strain using Abaqus. The geometry and boundary conditions of the
analyzed model are shown in Fig. 13. Only a half of the cracked plate was modeled by plane strain
elements owing to the symmetry of the model. Applied material and thermal properties were the same as
those used for Figs. 11 and 12. The SIFs obtained were shown in Fig. 12 for the same condition; Tf =
100 K, H = 5,000 W/m2K and a = 3 mm. The agreement with the finite element analysis results validated
the thermal stress and SIF calculation procedures.

4.3 Stress intensity factor for thermal stress

Figure 14 shows change in SIF with crack depth obtained for a conditions of Tf = 100 K and H =
5,000 W/m2K. The crack was assumed to have a semi-elliptical shape of which the aspect ratio was a/c =
0.5. The maximum SIF, which was denoted as Kmax, during the transient was calculated for the deepest
and surface points. Kmax at the deepest point became the maximum when the crack depth was about a =
0.2t and decreased to zero when a = 0.8t. The decrease in Kmax was brought about by the stress gradient in
the depth direction. On the contrary, Kmax at the surface was increased monotonically with the crack depth.
The smaller ts caused larger stress and made the Kmax larger.
Figure 15 shows the Kmax normalized by Ko, which is defined by:

Ko f max z 0 a (23)

where max(z=0) is the stress at the surface. Ko corresponds to the SIF obtained for a uniform stress, of
which the magnitude is the same as that at the surface. Kmax/Ko exhibited almost the same change for a/t
regardless of ts. The change in Kmax/Ko was also insensitive to the heat transfer coefficients H as shown in
Fig. 16.
The condition of ts = 1,000 s and H = 5,000 W/m2K showed the maximum Kmax/Ko in the analyzed
cases and these values were applied in Fig. 17 for various a/c. The Kmax/Ko depended on a/c at the deepest

9
point, while it showed less dependency at the surface. The Kmax/Ko increased with the increase in a/c and
the values for a/c =0.1 were almost the same as those calculated for the two-dimensional crack. In order
to give the change in Kmax/Ko with crack depth, the following equations are proposed.

0.8
K max a a
1 0.41 0.98 (For the deepest point) (24)
Ko c w

1.2
K max a
1 0.55 (For the surface point) (25)
Ko w

The solid lines in Fig. 17 indicate the prediction made by Eqs. (24) and (25). These equations give almost
the upper bound of the analyzed value. By using Eqs. (24) and (25), once max(z=0) is obtained, the change
in Kmax with the crack depth can be estimated without performing complex stress and SIF calculations
considering the stress gradient in the depth direction. It should be noted that max(z=0) has been calculated
for deriving CUF in component designs.

4.4 Crack growth under thermal stress

Growth of a surface crack was simulated under the thermal stress for H = 5,000 W/m2K. Tf was
determined so that the strain amplitude at the surface would be a = 0.5%. The growth rate was
determined by Eq. (10) and Keq was calculated as Keq = 2Kmax considering a reversed load
corresponding to a trapezoidal waveform of the fluid temperature fluctuation. It should be noted that the
environmental effect is not considered in Eq. (10). The crack depth of the (i+1)th step a(i+1) was calculated
from the depth of the (i)th step a(i) by:

da
a
i 1
a N .
i
(26)
dN

The same calculation was made for the surface length. N was determined so that the growth in the depth
and surface directions, respectively, became less than 0.05a and 0.05c at each step. The initial depth and
aspect ratio were assumed to be 500 m and a/c = 1.0, respectively.
Figure 18 shows the change in crack depth and aspect ratio. The growth under the constant stress,
which SIF was calculated by Eq. (23), is also shown in the figure. The crack growth under the thermal
stress was much less than that obtained for the uniform stress and hardly depended on ts for the same
strain amplitude. The aspect ratio was decreased monotonically due to the stress gradient. The prediction
made using the simplified SIF of Eqs. (24) and (25) showed similar trends.

5. DISCUSSION

5.1 Fatigue life for thermal stress

Figure 19 shows fatigue lives predicted as Ni(ai=500m) + Np(a=500m-3mm) for the thermal stress and
uniform stress obtained by the procedure described in Section 4. The initial depth and aspect ratio were
assumed to be 500 m and a/c = 1.0, respectively. The predicted fatigue lives for the uniform stress was
almost identical to the DFC. This indicates the validity of the fatigue life estimation by the crack growth
analysis for a surface crack on a plate considering a variable aspect ratio. The crack growth for the

10
thermal stress made the fatigue life longer than that obtained for the uniform stress, although the
difference was not significant on the log scale.
The fatigue life for thermal stress was predicted for various critical depths af and normalized by the
life obtained for the uniform stress condition of af = 3 mm, which is denoted as Nfo(af=3mm) in Fig. 20. The
normalized fatigue life Nf/Nfo(af=3mm) was more than 1.2 times at the maximum case for af = 3 mm. If the
crack was allowed to grow to 10 mm in depth, the fatigue life was extended more than 1.6 times.
Although Nf/Nfo(af=3mm) was small for the relatively small strain amplitude less than a = 0.1%, the smaller
strain amplitude has little influence on the design because Nfo(af=3mm) for a = 0.1% is about 100,000
cycles.

5.2 Relationship between crack size and fatigue damage

Figure 21 shows the relationship between crack depth and the number of cycles normalized by
Nf(af=10mm), which corresponded to Ni(ai=500m) + Np(a=500m-10mm). If the number of allowable cycles in the
design is given by Nf(af=10mm), the abscissa is equivalent to the CUF. The relationship between the crack
size and CUF allows the degree of fatigue damage accumulated in components to be quantified by
inspection results. For example, if a crack of 2 mm depth is detected in a component, the fatigue damage
is estimated as CUF 0.7 for thermal stress. If no crack is found by the inspection in which the minimum
detectable crack depth is 2 mm, the CUF is deduced to be less than 0.7. In other words, a 2 mm crack has
to be detected in order to ensure the CUF is less than 0.7. Improvement of the inspection technique for
detecting smaller cracks would allow fatigue damage to be detected at an earlier stage. On the contrary, it
is difficult to detect the small fatigue damage caused by cyclic internal pressure because the CUF for 2
mm in depth is about 0.9 for the uniform stress condition.
Thus, the crack initiation and growth analysis allows the degree of fatigue damage to be quantified
using the crack size. The flaw tolerance assessment can be performed not only for CUF = 1, but also for
CUF < 1. The time for the next inspection can be determined from the inspection results.

5.3 Gaps between CUF calculation and crack growth analysis

As mentioned in the introduction, there are many gaps between the fatigue damage assessment made
for component design using the CUF and that made for FFS assessments. In this study, by closing the
gaps, the crack initiation and growth analysis was tried as a replacement for the CUF calculation. The
detailed procedure for applying the crack growth analysis for fatigue design and remaining problems to be
solved are summarized as follows.
The fact that the Keq was applicable not only for elastic conditions but also for elastic-plastic
conditions enabled the crack growth analysis to be done for a given strain range using the FFS growth
rate, which was obtained under the small scale yielding condition. In the fatigue damage assessment using
the CUF, the plastic strain is derived by multiplying the plastic correction factor Ke by the stress
calculated in an elastic analysis [24]. Similarly, the Keq for the low-cycle fatigue regime can be derived
by multiplying Ke by the SIF evaluated in the elastic analysis.
As for the temperature effect on the crack growth rate, as discussed in section 3.2, the gap between
the CUF calculation and crack growth analysis was closed by using the corrected SIF Kcorrected defined
by Eq. (8) or by using the Keq for the strain range.
The mean stress effect was not treated in this study. The modified Goodmans diagram is applied to
consider the influence of the mean stress on the DFC [23]. However, the correction made for the stress
amplitude using the ultimate strength is not consistent with the fact that the fatigue life is correlated to the

11
strain range, although it may be valid from a practical viewpoint [39]. Some experimental results [40,41]
showed that the mean stress brought about a beneficial effect on the fatigue life of stainless steels under
the same stress amplitude. On the other hand, in the growth analysis for FFS assessment, change in the
crack growth rate due to the mean stress is considered using the stress ratio. It is well known that the
growth rate correlates with the effective SIF range rather than the SIF range. The mean stress effect in the
growth rate can be considered as the change in the effective SIF range. However, it is not clear whether
the effective strain range plays the same role in the crack growth. Further consideration is necessary to
incorporate the mean stress effect in the crack growth analysis using Keq.
The environmental effect of high-temperature primary water was not considered in the assessments
performed in this study. It has been shown that the fatigue life reduction due to environmental effect was
brought about by the crack growth acceleration [42]. Therefore, the reduction in fatigue life due to the
environmental effect can be considered by accelerating the growth rate in the crack growth analysis.
Since the safety margin is not considered explicitly in the growth rate prescribed in the FFS code, 2.4
times faster growth rate was determined from the margin considered for the DFC. As mentioned, the
margin considered for the DFC was determined empirically and might be excessively conservative. On
the other hand, by reconsidering the appropriate margin for crack initiation and growth, it is possible to
optimize the margin for the design. The crack initiation and growth analysis makes it easy to quantify the
influence of the factors on the fatigue life as done for obtaining Table 1.

CONCLUSIONS

In this study, fatigue crack initiation and growth were observed during the strain-controlled fatigue
test using Type 316 stainless steel in order to assess the fatigue damage by crack growth analysis for
design of plant components. Then, the fatigue life was estimated by the crack initiation and growth
analysis. The growth analysis was also performed for thermal stress considering the stress gradient in the
depth direction. Considerations were made to replace the fatigue damage assessment using the CUF with
the crack initiation and growth analyses. Finally, gaps between the design and FFS assessment were
discussed for the plastic strain, temperature, mean stress effect, environment effect and the margin. The
conclusions are summarized as follows.

(1) Fatigue cracks of several tens of micrometers were initiated in the early stage of the fatigue test,
which was less than 0.1Nf, even for = 0.6%. The growth rates correlated well with the
equivalent stress intensity factor, which is calculated using the strain range.
(2) The fatigue life could be estimated as a sum of the number of cycles for crack initiation of 100
m in depth and its growth to 3 mm.
(3) The temperature dependency of the crack growth rate for K was deduced to be brought about
by temperature dependency of the Youngs modulus. The similar crack growth rate was obtained
under the same strain range regardless of the temperature.
(4) By assuming 2.4 times faster crack growth and the initiation of a crack of 500 m depth with the
incubation period given by Eq. (11), the DFC was successfully divided into the cycles for crack
initiation and growth.
(5) The SIF for a thermal transient considering the stress gradient in the depth direction can be
estimated using Eqs. (24) and (25) based on the stress amplitude at the surface.
(6) The estimated fatigue life was extended by considering the stress gradient in the depth direction.
The estimated fatigue life became longer by allowing a deeper crack.
(7) The relationship between the CUF and crack size was shown. The crack depth was deduced to

12
 be 2 mm in depth at CUF = 0.7. By using this relation, it is possible to estimate the fatigue
damage accumulated in the component from an identified crack size, or possible damage can be
determined even if no crack was detected by the inspection.

ACKNOWLEGEMENT

The author wishes to acknowledge the experimental support provided by Dr. M. Kawakubo. The
replica investigations performed in this study were completed through his efforts.

REFERENCES

[1] ASME, Rules for Construction of Nuclear Facility Components ASME Boiler and Pressure
Vessel Code Section III, ASME, New York, 2013.
[2] Japan Society Mechanical Engineers, Codes for Nuclear Power Generation Facilities: Rules on
Design and Construction for Nuclear Power Plants, JSME S NC1-2012, 2012.
[3] Weitze WF, Gilman TD, Drenth L, Application of common basis stress evaluation methodology
for environmentally assisted fatigue for a pressurized water reactor, Proceedings of the ASME
2013 Pressure Vessels and Piping Conference (2013) paper no. 97597.
[4] Seichter J, Reese SH, Kluche D, Sophisticated procedures for fatigue evaluation in the
framework of German KTA codes, Proceedings of the ASME 2013 Pressure Vessels and Piping
Conference (2013) paper no. 97265.
[5] Higuchi M, Comparison of environmental fatigue evaluation methods in LWR water,
Proceedings of ASME 2008 Pressure Vessels and Piping Conference (2008) paper no. 61087.
[6] Japan Society Mechanical Engineers, Codes for Nuclear Power Generation Facilities:
Environmental Fatigue Evaluation Method for Nuclear Power Plants, JSME S NF1-2006, 2006.
[7] Higuchi M, Sakaguchi K, Hirano A, Nomura Y, Final proposal of environmental fatigue life
correction factor (Fen) for structural materials in LWR water environment, Proceedings of ASME
2007 Pressure Vessels and Piping Conference (2007) paper no. 26100.
[8] ASME, Rules for Construction of Nuclear Facility Components ASME Boiler and Pressure
Vessel Code Section III, Code Case N-792: Fatigue Evaluations Including Environmental Effects,
ASME, 2010.
[9] Murakami Y, Miller KJ, What is fatigue damage? A view point from the observation of low cycle
fatigue process. International Journal of Fatigue. 2005; 27: 991-1005.
[10] Xue Y. Modeling fatigue small-crack growth with confidence A multistage approach.
International Journal of Fatigue. 2010; 32: 1210-1219.
[11] Kamaya M, Nakamura T, A flaw tolerance concept for plant maintenance using virtual fatigue
crack growth curve, Proceedings of the ASME 2013 Pressure Vessels and Piping Conference
(2013) paper no. 97851.
[12] Chellapandi P, Chetal SC, Raj B. Thermal striping limits for components of sodium cooled fast
spectrum reactors. Nuclear Engineering and Design. 2009; 239: 2754-2765.
[13] Kamaya M, Taheri S, A study on the evolution of crack networks under thermal fatigue loading,
Nuclear Engineering and Design. 2008; 238: 2147-2154.
[14] Le HN, Gardin C. Analytical prediction of crack propagation under thermal cyclic loading
inducing a thermal gradient in the specimen thickness Comparison with experiments and
numerical approach. Engineering Fracture Mechanics. 2011; 78: 638-652.
[15] Kim YW, Kim JI, Chang MH. An explicit integral expression for the stress intensity factor of a

13
 semi-elliptic surface crack subjected to thermal transient loading. International Journal of
Pressure Vessels and Piping. 1999; 76: 631-639.
[16] Kamaya M, Assessment of thermal fatigue damage caused by local fluid temperature fluctuation
(Part II: Crack growth under thermal stress), Nuclear Engineering and Design. 2014; 268:
139-150.
[17] Iwasaki M, Takada Y, Nakamura T, Evaluation of environmental fatigue in PWR PLM activities,
Proceedings of the ASME 2005 Pressure Vessels and Piping Conference (2005) paper no. 71509.
[18] ASME, Rules for Inservice Inspection of Nuclear Power Plant Components ASME Boiler and
Pressure Vessel Code Section XI, ASME, New York, 2013.
[19] Gosselin SR, Simonen FA, Heasler PG, Doctor SR, Fatigue crack flaw tolerance in nuclear power
plant piping - A basis for improvements to ASME code section XI Appendix L,
NUREG/CR-6934 (2007).
[20] Japan Society Mechanical Engineers, Codes for Nuclear Power Generation Facilities: Rules of
Fitness-for-Service for Nuclear Power Plants, JSME S NA1-2012, 2012.
[21] Standard Test Method for Measurement of Fatigue Crack Growth Rates, ASTM E647-00, 2000.
[22] Jaske CE, O'Donnell WJ, Fatigue design criteria for pressure vessel alloys, ASME Journal of
Pressure Vessel Technology. 1997; 99: 584-592.
[23] Chopra OK, Shack WJ, Effect of LWR coolant environments on the fatigue life of reactor
materials - Final Report, NUREG/CR-6909, ANL-06/08 (2007).
[24] Lang H, Rudolph J, Ziegler R. Performance study of Ke factors in simplified elastic plastic
fatigue analyses with emphasis on thermal cyclic loading. International Journal of Pressure
Vessels and Piping. 2011; 88: 330-347.
[25] Kanasaki H, Higuchi M, Asada S, Yasuda M, Sera T, Proposal of fatigue life equations for carbon
& low-alloy steels and austenitic stainless steels as a function of tensile strength, Proceedings of
the ASME 2013 Pressure Vessels and Piping Conference (2013) paper no. 97770.
[26] Kamaya M, Kawakubo M, Damage assessment of low-cycle fatigue by crack growth prediction
(development of growth prediction model and its application), Transactions of the Japan Society
of Mechanical Engineers. 2012; A78: 1518-1533.
[27] Kamaya M, Kawakubo M, Strain-based modeling of fatigue crack growth An experimental
approach for stainless steel, International Journal of Fatigue. 2012; 44: 131-140.
[28] Kamaya M, Damage assessment of low-cycle fatigue by crack growth prediction (fatigue life
under cyclic thermal stress), Transactions of the Japan Society of Mechanical Engineers. 2013;
A79: 1530-1544.
[29] Kawakubo M, Kamaya M, Fatigue life prediction of stainless steel under variable loading
(damage factors determining fatigue life and damage evaluation for two-step test), Journal of the
Society of Materials Science, Japan. 2011; 60: 871-878.
[30] Kawakubo M, Kamaya M, Effect of mean stress on fatigue strength of Type 316 stainless steel,
Journal of the Society of Materials Science, Japan. 2012; 61: 635-641.
[31] Takahashi Y, Multiaxial fatigue failure criterion considering formation and growth of small
cracks, Proceedings of the ASME 2010 Pressure Vessels and Piping Conference (2010) paper no.
26015.
[32] ASME, Materials ASME Boiler and Pressure Vessel Code Section II, ASME, New York, 2013.
[33] Jones IS, Lewis WJ. The effect of various constraint conditions in the frequency response model
of thermal striping. Fatigue Fract. Engng Mater. Struct. 1995; 18: 489-502.
[34] Kamaya M, Assessment of thermal fatigue damage caused by local fluid temperature fluctuation
(Part I: Characteristics of constraint and stress caused by thermal striation and stratification),

14
 Nuclear Engineering and Design. 2014; 268: 121-138.
[35] Glinka G, Shen G, Universal features of weight functions for cracks in mode I. Engineering
Fracture Mechanics. 1991; 40: 1135-1146.
[36] Kim YW, Lee JH, Yoo B. An analysis of stress intensity factor for thermal transient problems
based on green's function. Engineering Fracture Mechanics. 1994; 49: 393-403.
[37] Kim YW, Kim JI, Chang MH. An explicit integral expression for the stress intensity factor of a
semi-elliptic surface crack subjected to thermal transient loading. International Journal of
Pressure Vessels and Piping. 1999; 76: 631-639.
[38] Shen G, Glinka G. Weight functions for a surface semi-elliptical crack in a finite thickness plate.
Theoretical and Applied Fracture Mechanics. 1991; 15: 247-255.
[39] M. Kamaya, M. Kawakubo, Influence of mean stress on fatigue strength of stainless steel,
Transactions of the JSME (in Japanese). 811 (2013) SMM0037.
[40] Colin J, Fatemi A, Taheri S, Fatigue behavior of stainless steel 304L including strain hardening,
prestraining, and mean stress effects, Journal of Engineering Materials and Technology. 2010;
132: no.021008.
[41] Miura N, Takahashi Y, High-cycle fatigue behavior of type 316 stainless steel at 288 C including
mean stress effect, International Journal of Fatigue. 2006; 28: 1618-1625.
[42] Kamaya M. Environmental effect on fatigue strength of stainless steel in PWR primary water
Role of crack growth acceleration in fatigue life reduction, International Journal of Fatigue. 2013;
55: 102-111.

15
 24 12 24
150

Fig. 1 Geometry of test specimen (unit: mm).

 7

Surface length 2c, mm

5

0
0 10,000 20,000 30,000 40,000 50,000
Number of cycles, cycle

Fig. 2 Change in crack surface length obtained by replica investigation

( = 0.6%).

9
=0.6% (This study)
8 = 1.2% [26]
Surface length 2c, mm

7 = 2.0% [27]
1.0
6
Surface length 2c, mm

0.8

5
0.6

4 0.4

3 0.2

2 0.0
0.0 0.2 0.4 0.6 0.8 1.0

1 N/Nf

0
0.0 0.2 0.4 0.6 0.8 1.0
N/Nf

Fig. 3 Change in crack surface length with normalized cycles obtained

by replica investigations.
 1.E-04

Crack growth rate da/dN, m/cycle

1.E-05

1.E-06

1.E-07

1.E-08 = 0.6% (This study)

= 1.2%
1.E-09 = 2.0%
a = 250 MPa(1)
1.E-10 a = 250 MPa(2)
a = 270 MPa
1.E-11
1 10 100 1000
DK , MPa m1/2
(a) Correlation with stress intensity factor range (K)

da

2.76
5.06 1012 DKeq
dN

1.E-04
Crack growth rate da/dN, m/cycle

1.E-05

1.E-06

1.E-07

1.E-08 = 0.6% (This study)

= 1.2%
1.E-09 = 2.0%
a = 250 MPa(1)
a = 250 MPa(2)
1.E-10
a = 270 MPa
Eq. (4)
1.E-11
1 10 100 1000
DKeq , MPa m1/2

(b) Correlation with equivalent stress intensity factor range (Keq)

Fig. 4 Relationship between the stress intensify factor ranges and

crack growth rate obtained under various loading conditions.
 2.0

1.5
Strain range , %

1.0

This study
0.5 Fatigue test [29]
Np (a=20m-3mm
Run out
Np (a=50m-3mm
Np (a=100m-3mm
0.0
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

Number of cycles to failure Nf

Fig. 5 Fatigue lives estimated by crack growth analysis and those

obtained by the low-cycle fatigue tests.
 2.0

1.5
Strain range , %

1.0

0.5
Fatigue test [29]
Ni(ai=100m+Np(a=100m-3mm)
test (a = 100 m)
Ni(ai=100m) by Eq.(5)
0.0
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

Number of cycles to failure Nf

Fig. 6 Incubation period before initiation of 100 m depth crack (Ni(ai=100m)) and
fatigue life predicted as Ni(ai=100m) + Np(a=100m-3mm).
 Eq.(4)
FFS (T = 25 C)
FFS (T = 325 C)
= 0.6% (This study)
CT specimen
FFS (T = 325 C) (corrected by Eq.(8))

Fig. 7 Crack growth rates obtained using cylindrical and compact

tension specimens together with those prescribed in the FFS code.
 1.0
Best-fit by Eq. (9)
0.8 N/12

Strain amplitude a, %
N for a2
DFC
0.6

0.4

0.2

0.0
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
N

Fig. 8 Design fatigue curves obtained from Eq. (9) without

considering mean stress effect.

1.0
Ni(ai=500m)+Np(a=500-3mm)
DFC
0.8 Ni by inverse analysis (ai=500m
Strain amplitude a , %

Ni(ai=500m) (Eq.(11))

0.6

0.4

0.2

0.0
1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Number of cycles to failure Nf

Fig. 9 Incubation period before initiation of 500 m depth crack (Ni(ai=500m)) and fatigue
life predicted as Ni(ai=500m) + Np(a=500m-3mm) for simulating the design fatigue curve.
 Tm
Tf(1)(t1) Time t
Fluid temperature
Tf(2)(t2)
Tf(3)(t3)

Tm Tf

ts

Fluid temperature

z
w = 20 mm

Fig. 10 Analyzed model for thermal stress caused by fluid temperature change.
 250
ts = 10s
ts = 100s
200 ts = 1000s

Stress, MPa
150

100

50

0
0 200 400 600 800 1000 1200

Time t, s

(a) Stress on the surface

400
ts=10s
300 ts=100s
ts=1000s

200
Stress, MPa

100

-100

-200
0.0 0.2 0.4 0.6 0.8 1.0

Depth z/w
(b) Stress in the depth direction

Fig. 11 Thermal stress obtained for ramp change of fluid temperature ( Tf

= 100 K, H = 5,000 W/m2K).
 30
ts=10s
Stress intensity factor, MPa m0.5

ts=100s
ts=1000s
20

10

0
0 200 400 600 800 1,000 1,200

Time t , s

Fig. 12 Change in stress intensity factor of two-dimensional crack of 3 mm depth caused by

ramp change of fluid temperature ( Tf = 100 K, H = 5,000 W/m2K). Broken lines are values
obtained by finite element analyses.
 a
w = 20 mm

2L = 200 mm

Fig. 13 Analyzed model for two-dimensional cracked plate subjected to

thermal transient.
 16
ts=10s
14 ts=100s
ts=1000s
12

Kmax, MPa m0.5

10

0
0.0 0.2 0.4 0.6 0.8
Crack depth a/w

(a) Deepest point

35
ts=10s
30 ts=100s
ts=1000s
25
Kmax, MPa m0.5

20

15

10

0
0.0 0.2 0.4 0.6 0.8
Crack depth a/w
(b) Surface point

Fig. 14 Change in maximum stress intensity factor of surface crack caused by ramp
change of fluid temperature ( Tf = 100 K, H = 5,000 W/m2K, a/c = 0.5).
 1.0
ts=10s
ts=100s
0.8 ts=1000s

0.6
Kmax /Ko

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8

Crack depth a/w

(a) Deepest point

1.0

0.8

0.6
Kmax /Ko

0.4

0.2 ts=10s
ts=100s
ts=1000s
0.0
0.0 0.2 0.4 0.6 0.8

Crack depth a/w

(b) Surface point

Fig. 15 Change in the maximum stress intensity factor normalized by Ko with

normalized crack depth a/w (Ko: SIF obtained for maximum peak stress (Eq.23),
Tf = 100 K, H = 5,000 W/m2K, a/c = 0.5).
 1.0

0.8

0.6
Kmax /Ko

0.4
H = 40 kW/(m2K)
0.2 H = 20 kW/(m2K)
H = 10 kW/(m2K)
H = 5 kW/(m2K)
0.0
0.0 0.2 0.4 0.6 0.8

Crack depth a/w

(a) Deepest point

1.0

0.8

0.6
Kmax /Ko

0.4

H = 40 kW/(m2K)
0.2 H = 20 kW/(m2K)
H = 10 kW/(m2K)
H = 5 kW/(m2K)
0.0
0.0 0.2 0.4 0.6 0.8

Crack depth a/w

(b) Surface point

Fig. 16 Change in the maximum stress intensity factor normalized by Ko for various heat transfer
coefficients (Ko: SIF obtained for maximum peak stress (Eq.23), Tf = 100 K, a/c = 0.5, ts = 10 s).
 1.0
a/c = 1
a/c = 0.7
0.8 a/c = 0.5
a/c = 0.3
a/c = 0.1
0.6 2D crack
Kmax /Ko Eq.(24)

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8
Crack depth a/w
(a) Deepest point

1.0

0.8

0.6
Kmax /Ko

0.4 a/c = 1
a/c = 0.7
a/c = 0.5
0.2 a/c = 0.3
a/c = 0.1
Eq.(25)
0.0
0.0 0.2 0.4 0.6 0.8

Crack depth a/w

(b) Surface point

Fig. 17 Change in the maximum stress intensity factor normalized by Ko for various a/c and values
predicted by Eqs. (24) and (25) (Ko: SIF obtained for maximum peak stress (Eq.23), Tf = 100 K,
H = 5,000 W/m2K, ts = 1,000 s).
 10
Uniform stress
ts=10s
8 ts=100s
ts=1000s

Crack depth a, mm
Eqs. (24) and (25)
6

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6

Number of cycles x103, cycle

(a) Change in crack depth

1.0

0.8
Aspect ratio a/c

0.6

0.4
Uniform stress
ts=10s
0.2 ts=100s
ts=1000s
Eqs. (24) and (25)
0.0
0 2 4 6 8 10

Crack depth a, mm

(b) Change in aspect ratio

Fig. 18 Crack growth prediction results for a thermal transient ( a = 0.5%, H = 5,000 W/m2K,
initial conditions: a = 0.5 mm, a/c = 1.0).
 1.0
DFC
Uniform stress
0.8
Thermal stresss (Eqs.(24) and (25))

Strain amplitude a , %
0.6

0.4

0.2

0.0
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Number of cycles

Fig. 19 Fatigue lives obtained by DFC, crack growth analysis for uniform stress and crack growth
analysis for thermal stress (Ni(ai = 500m) is given by Eq. (11), initial a/c = 1.0, af = 3 mm).

1.8
af = 3 mm
af = 5 mm
1.6 af = 10 mm
Nf/Nfo(af=3mm)

1.4

1.2

1.0
0.0 0.2 0.4 0.6 0.8 1.0
Strain amplitude a ,%

Fig. 20 Fatigue lives for various critical depth af under thermal stress normalized by those obtained
for af = 3mm under uniform stress condition (Ni(ai = 500m) is given by Eq. (11), initial a/c = 1.0).
 10
a=a =0.1%
0.1%
a=a =0.5%
0.2%
8 a=a =0.2%
0.5%
a=a =0.5% (uniform
0.5% stresss)
(uniform stress)
Crack depth a, mm

0
0.0 0.2 0.4 0.6 0.8 1.0
N/Nf(af=10mm)

Fig. 21 Change in crack depth with normalized cycles obtained for various strain amplitudes under
uniform and thermal stress conditions (Ni(ai = 500m) is given by Eq. (11), initial a/c = 1.0).
 Table 1 Subfactors considered on fatigue life of stainless steels.

Margin
Factors Ref. [23]
Initiation Growth
Variation Log mean
Material variability and
2.1-2.8 2.4 (2.4)0.5 (2.4)0.5
data scatter

Size effect 1.2-1.4 1.3 1.3 1

Surface finish 2.0-3.5 2.6 2.6 1

Loading history 1.2-2.0 1.5 1 1.5

Total 12 5 2.4
 Table 2 Constants used for thermal stress analyses

Density Specific heat: c 3.85 106 [J/m3K]

Heat conduction coefficient: 15.86 [W/mK]
Youngs modulus: E 195000 [MPa]
Poisons ratio: 0.3
Heat expansion coefficient: 1.64 10-5 [1/K]