Engineering Fracture Mechanics 99 (2013) 48–61

Contents lists available at SciVerse ScienceDirect

Engineering Fracture Mechanics

journal homepage: www.elsevier.com/locate/engfracmech

Assessment of equivalent initial flaw size estimation methods in

fatigue life prediction using compact tension specimen tests
A.R. Shahani ⇑, H. Moayeri Kashani
Fracture Mechanics Research Laboratory, Department of Applied Mechanics, Faculty of Mechanical Engineering, K.N. Toosi University of Technology, P.O. Box
19395-1999, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: To assess the efficiency of the Equivalent Initial Flaw Size (EIFS) concept in the life estima-
Received 12 February 2012 tion of mechanical components, a novel approach is outlined in this paper. For this pur-
Received in revised form 14 January 2013 pose, experimental tests are conducted on the compact tension specimens made of 4340
Accepted 16 January 2013
steel and the number of cycles required for the crack to grow from the end of the notch
up to the fracture of the specimen is counted. In fact, the fatigue cycling for the pre-crack
initiation is a part of cycle count procedure and it is assumed that an initial micro-crack
Keywords:
exists at the end of the notch, which grows due to fatigue loading, the length of which is
EIFS
Life prediction
estimated using the EIFS method. Three methods of back extrapolation, Kitagawa–Takah-
Fatigue crack growth ashi diagram, and time to crack initiation are used in order to estimate EIFS and their
Threshold value of stress intensity factor results are compared. For estimating EIFS by Kitagawa–Takahashi diagram, the threshold
value of the stress intensity factor (DKth) is required. In this paper, DKth is estimated by
both K-increasing and K-decreasing methods. The experimental results show that the value
of DKth estimated by the K-increasing method is lower than that estimated by K-decreasing
method. It is observed that the estimated EIFS by back extrapolation method and TTCI
method is dependent on the loading amplitude, while the estimated EIFS by Kitagawa–
Takahashi diagram is identical for all loading amplitudes and it can be considered as the
material property. The predicted life based on the Kitagawa–Takahashi method is in rela-
tively good agreement with the experimental results. However, the TTCI method does not
have sufficient accuracy especially in low amplitude fatigue.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Fracture mechanics based life prediction is an applicable method in industry especially in aerospace industries. It has
been a research topic of many international organizations such as AGARD, ASTM, NASA, and CAE from 1992 up to present
[1]. In this concept, it is always assumed that an initial crack (similar to the crack which forms and initiates from voids) exists
in the specimen. Of course, this assumption is not far away from reality for many industrial components which experience
defects during their manufacturing processes.
The numbers of cycles are usually determined according to the available fatigue models, such as the Paris model (1).
Z af
da
Nf ¼ ð1Þ
ai CðDKÞm

⇑ Corresponding author. Tel.: +98 21 84063221; fax: +98 21 88677273.

E-mail address: shahani@kntu.ac.ir (A.R. Shahani).

0013-7944/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engfracmech.2013.01.007
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 49

Nomenclature

a crack length
ac critical crack length
af final crack length
ai initial crack length
B specimen thickness
C coefficient of Paris equation
E Young’s modulus
K stress intensity factor
KIC fracture toughness
Kmax/Kmin maximum/minimum stress intensity factor
m exponent of Paris equation
N number of cycles
Ne experimental measured life
Nf predicted life
P load
Pmax/Pmin maximum/minimum load
V crack mouth opening displacement
W specimen width
da/dN fatigue crack growth rate
gs(a) small crack growth curve
gL(a) large crack growth curve
DK cyclic stress intensity factor
DKth threshold value of cyclic stress intensity factor
DP load amplitude
DPf load limit corresponding to fatigue limit
Drf fatigue limit
EIFS equivalent initial flaw size
IFS actual initial flaw size

In this relation, the upper limit of the integral can be calculated according to fracture toughness of the material. However,
calculation of the initial crack length (ai) is one of the problems of this method. The initial crack length can be measured by
NonDestructive Tests (NDTs). However, the initial flaw size (IFS) can be below the current detection capability of the NDT
technique. If the NDT detection limit is chosen as the initial flaw size, it will result in a very conservative design [2]. In addi-
tion, the behavior of a small crack growth is complicated and is dependent on the microstructures of the material. The crack
growth behavior of small and large crack growth rate curves have been compared in Fig. 1. It is useful to distinguish between
small cracks and short cracks. For a small crack, all of its dimensions are similar to or smaller than the dimension of greatest
microstructural significance, such as the average crystal grain size or the average reinforcement particle spacing. However, a
short crack has a dimension that is large compared with the microstructure. The behavior of a small crack can be profoundly
affected by the microstructure. For example, while the crack is within a single crystal grain in a metal, the growth rate is
much higher than expected from the usual da/dN versus DK curve, as illustrated by Fig. 1. Upon encountering a grain bound-

Fig. 1. Comparison between (a) small and (b) large crack growth rate curves [2].
50 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

ary, the growth is temporarily retarded. Until the crack becomes several times larger than the grain size, the average growth
rate, as affected by lattice planes within grains and grain boundaries, is considerably above the usual da/dN versus DK curve
[3].
Thus, two cases can be considered in life prediction based on fracture mechanics concept. In the first case, the lower limit
of the integral is assumed to be the actual initial flaw size (IFS), and the small crack growth curve gs(a) should be used
according to relation (2).
Z ac
1
NIFS ¼ da ð2Þ
IFS g s ðaÞ
In the second case, the usual large crack growth curve gL(a) is used instead of complicated small crack growth curve. In
this case, the lower limit of the integral must be replaced by an Equivalent Initial Flaw Size (EIFS), as presented in following
equation:
Z ac
1
NEIFS ¼ da ð3Þ
EIFS g L ðaÞ

Newman et al. [4] predicted the life of various notched specimens based on the small crack growth curve with the aid of
Fastran software [5]. The main problem of this method is the complexity of small crack theory. However, the life prediction
based on EIFS concept is a very applicable method considered by lots of industries and scientists [6–8].
In order to estimate the EIFS value, the underlying areas below gs(a) and gL(a) must be equal, as shown in Fig. 2. On the
other hand, the EIFS concept, which avoids small-crack growth analysis, uses an equivalent initial flaw size in large-crack
growth analysis and estimates the fatigue life to match the experimental one [2]. In the other words, instead of matching
life predictions obtained from long crack growth and small crack growth analyses, EIFS is calculated by matching the life
predicted by long crack growth analysis with that observed in fatigue S–N testing [2].
The most applicable methods to estimate EIFS are the following:

I. Back extrapolation method.

II. Kitagawa–Takahashi diagram.
III. Time To Crack Initiation (TTCI) method.

The first method estimates EIFS by comparing the measured life in laboratory and the predicted life with the aid of fatigue
crack growth rate analysis considering one assumed initial crack length (and crack shape). It is obvious that this method re-
quires trial and error procedure. The second one considers the EIFS as a material property which describes its initial quality
[2]. The last looks for crack length corresponding to N = 0 in the experimental curves of crack length versus number of cycles
(a–N).
To assess the efficiency of the EIFS concept in the life estimation of mechanical components, a novel approach is outlined
in this paper. For this purpose, experimental tests are conducted on the C(T) specimens made of 4340 steel and the number
of cycles required for the crack to grow from the end of the notch up to its final length at the onset of fracture is counted. In
fact, the fatigue cycling for the pre-crack initiation is a part of cycle count procedure. This is because a small crack can always
exist at the end of the notch the length of which is to be computed using the EIFS techniques. Three methods of back extrap-
olation, time to crack initiation and Kitagawa–Takahashi diagram are used in order to estimate EIFS and their results are
compared.
For estimating EIFS by back extrapolation method, the coefficients of the Paris equation and fracture toughness of 4340
steel are measured according to ASTM-E647 [9] and ASTM-E399 [10] standards, respectively. Also, the threshold value of
stress intensity factor (DKth) is needed to estimate EIFS using Kitagawa–Takahashi diagram. The methods of estimation of

Fig. 2. Equivalent initial flaw size concepts [9].

 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 51

DKth in laboratory are described in the ASTM-E647 standard [9]. Precise techniques are required in order to conduct this test.
These requirements make this test to be known as one of the most difficult fracture mechanics tests. In this paper, DKth is
estimated according to the decreasing cyclic stress intensity factor method by test equipment available in the fracture
mechanics research laboratory of K.N. Toosi University of Technology.

2. Experimental study

The geometry of tested Compact Tension (CT) specimens is shown in Fig. 3. Thickness, B, and width, W, of the specimens
are equal to 12.5 mm and 50 mm, respectively. The specimens are made of 4340 steel with the mechanical properties as gi-
ven in Table 1. The tested sheet specimens made by wire-cut electric discharge machine are shown in Fig. 4.
The specimens were tested by test equipment available in the fracture mechanics research laboratory of the mechanical
engineering faculty of K.N. Toosi University of Technology.
There are various methods for the crack length measurement in the laboratory such as potential drop, compliance, or vi-
sual methods. In this study, the crack length is measured by compliance method according to Eqs. (4) and (5) which are given
in ASTM-E647 standard [9].

Fig. 3. Geometry of Compact Tension specimens (CT).

Table 1
Mechanical properties of 4340 steel.

Elastic modulus (GPa) 200

Poisson ratio 0.3
Yield stress (MPa) 1135
Ultimate stress (MPa) 1400

Fig. 4. Compact Tensional (CT) tested specimens.

52 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

Fig. 5. Measurement of the crack length by compliance method with the aid of clip gage.

Table 2
Load amplitudes of tested specimens.

SPC Pmax (kN) Pmin (kN) DP (kN) Nf (cycles) R-ratio

1 13.5 9.45 4.05 37253 0.7
2 11.5 8.05 3.45 85274 0.7
3 9.5 6.65 2.85 225146 0.7
4 8 5.6 2.4 452056 0.7
5 6.5 4.55 1.95 – 0.7

Fig. 6. Load amplitude versus life of tested specimens.

a
¼ 1:002  4:0632U x þ 11:242U 2x  106:04U 3x þ 464U 4x  650:66U 5x ð4Þ
w

( 1=2 )1
EVB a
Ux ¼ þ1 ; 0:2 6 6 0:975 ð5Þ
P W

in which V is the Crack Mouth Opening Displacement (CMOD) along the load line measured by clip gage as shown in Fig. 5.

2.1. Life measurement and calculations of the Paris coefficients

The fatigue lives of the tested specimens were measured under different load amplitudes and stress ratio of 0.7. Also, the
Paris coefficients of 4340 steel are calculated using the linear region of the crack growth rate versus cyclic stress intensity
factor diagram according to ASTM-E647 standard [9]. The amplitude of the load applied to the first specimen is 4.05 kN,
and it is subsequently reduced for the other tested specimens, as provided in Table 2. The load amplitudes with respect
to the fatigue lives are plotted in Fig. 6. It is evident that the life increases as the load amplitude decreases.
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 53

Fig. 7. Crack growth rate versus cyclic stress intensity factor.

Fig. 8. Linear area/zone of the crack growth rate diagram versus cyclic stress intensity factor.

It is observed that the crack in the fifth specimen has only grown up to 0.1 mm after subjecting to 550,000 cycles. There-
fore, the test was stopped regarding the fact that the life of this specimen would be greater than 106 cycles. So, the related
result must be excluded.
Fig. 7 depicts the crack growth rate versus cyclic stress intensity factor diagram for the tested specimens (4340 steel).
Fitting a line to the linear part of the diagram, the Paris coefficients are obtained as C = 3.14  1011 and m = 2.5, which
are shown in Fig. 8.

2.2. Fracture toughness measurement

The fracture toughness of the material is measured according to ASTM-E399 standard [10]. A sharp pre-crack (a0/W = 0.6)
has been generated in the specimens by fatigue loading. The load versus crack extension diagram for 4340 steel is illustrated
in Fig. 9. According to the procedure outlined in ASTM-E399, the fracture toughness of the tested specimens is equal to
pffiffiffiffiffi
K IC ¼ 85 MPa m.

2.3. Estimation of threshold cyclic stress intensity factor

The experimental methods for estimating DKth are described in the ASTM-E647 standard [9]. In this standard, two meth-
ods are suggested for measuring da/dN  DK diagram. The first method is K-increasing and the second one is K-decreasing
method. The first method holds in the case that da/dN P 108 m/cycles; otherwise, the second method is suitable. For precise
industrial applications, however, the K-decreasing method is recommended for measuring DKth.
In K-increasing method, the value of the stress intensity factor value increases during the test while in the K-decreasing
method, the test begins with a determined DK and then the load level is decreased during the test. The load decreasing can
be applied either step by step or continuously. Precise techniques are required in order to conduct FCGR tests in near thresh-
54 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

Fig. 9. Load versus extension diagram.

Table 3
Loading values in K-decreasing method.

a (mm) Pmin (kN) Pmax (kN) Da (mm)

From To
BLOCK-01 30 31.2 8.2 5.74 1.2
BLOCK-02 31.2 32.3 6.97 4.879 1.1
BLOCK-03 32.3 33.3 5.9245 4.14715 1
BLOCK-04 33.3 34.1 5.0358 3.5251 0.8
BLOCK-05 34.1 34.8 4.2805 2.9963 0.7
BLOCK-06 34.8 35.4 3.6384 2.5469 0.6
BLOCK-07 35.4 35.9 3.0926 2.1648 0.5
BLOCK-08 35.9 36.3 2.6287 1.8401 0.4
BLOCK-09 36.3 36.6 2.2344 1.5641 0.3
BLOCK-10 36.6 36.9 1.8993 1.3295 0.3
BLOCK-11 36.9 37.1 1.6144 1.1301 0.2
BLOCK-12 37.1 37.3 1.3372 0.93606 0.2
BLOCK-13 37.3 37.5 1.1366 0.7956 0.2
BLOCK-14 37.5 37.7 1.0798 0.7558 0.2

old region. These requirements make this test to be known as one of the most difficult fracture mechanics tests. In recent
years, some researchers have been studied different methods of conducting fatigue crack growth test to calculate the thresh-
old value of the cyclic stress intensity factor [11–16].
Newman et al. [11,12] performed several tests on aluminum, steel, titanium, and nickel-based super alloy by three meth-
ods: (a) compression constant-amplitude, (b) compression load-reduction, and (c) the ASTM-E647 load-reduction test meth-
ods. They showed that the threshold value of stress intensity factor in some steel and aluminum alloys is not dependent on
the test method. However, the test method affects the results obtained for nickel-based super alloy specimens.
Sivaprasad et al. [13] showed that for the prevention of crack retardation, the rate of decrease in cyclic stress intensity
factor must be in proportion with the rate of increase in crack tip plastic region radius.
Kranenburg et al. [14] also considered load decreasing method such that maximum stress intensity factor remains con-
stant while its minimum value increases. The main disadvantage of this method is that the stress ratio is variable during a
single test and it is not clear that the determined DKth relates to which value of the stress ratio [15,16].
The K-decreasing method has been adopted in the present study. A sharp pre-crack (a0/W = 0.6) has been generated in the
specimens by fatigue loading. In order to overcome the crack tip blunting in the pre-cracking stage, the test is started with a
specific DK followed by smaller values of DK. For this purpose, the load is decreased in a block wise manner to obtain at least
five points of fatigue crack growth rate in the region between 1010 and 109 m/cycle. Therefore, the loading has been di-
vided into 14th blocks according to Table 3. The maximum load level in each block is 15% lower than that in the previous
block (except the last block in which the load decrease is 5% due to its closeness to the threshold zone). In fact, the load
was continuously decreased until the experimental data locate in the region 109 6 da=dN 6 1010 m=cycle. The stress ratio
is considered to be R = 0.7 during the test.
The load reduction occurs in determined intervals of crack growth, which has been specified in the standard [9]. In each
block, the loading amplitude also remains constant.
Variation of cyclic stress intensity factor in each load block is depicted in Fig. 10. The value of DK increases in each block
because the load amplitude remains constant in the corresponding block. However, the value of DK generally decreases dur-
ing the entire loading.
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 55

Fig. 10. Cyclic stress intensity factor versus crack length in each load block.

Fig. 11. Fatigue crack growth rate against cyclic stress intensity factor in each load block.

Fatigue crack growth rate diagram against cyclic stress intensity factor is also illustrated for each load block in Fig. 11. As
this figure reveals, only points which are related to the thirteenth load block have been located in the region
109 6 da=dN 6 1010 m=cycle.
According to ASTM-E647 standard [9], in order to obtain the threshold value of the cyclic stress intensity factor, a line is
fitted to the points which are located in the region of 109 6 da=dN 6 1010 m=cycle in the fatigue crack growth rate versus
cyclic stress intensity factor diagram. Consequently with regards to the equation of this line, DKth is equal to DK value cor-
responding to da/dN = 1010. This way, the threshold value of the cyclic stress intensity factor is calculated which is equal to
pffiffiffiffiffi
DK th ¼ 3:14895 MPa m. It is necessary to mention that this test took two and half days. Noteworthy is that the cyclic stress
pffiffiffiffiffi
intensity factor in the 14th load block is equal to DK ¼ 2:99 MPa m which is 5% lower than that in the thirteenth block. In
this case after spending more than 288,000 cycles (8 h), the crack did not extend. Clearly, this is because the corresponding
pffiffiffiffiffi
DK, that is, DK ¼ 2:99 MPa m is greater than DKth.
In addition, after the 14th load block, the test was followed by the K-increasing method and DKth was also measured by
pffiffiffiffiffi
this method which was obtained equal to DK th ¼ 2:763 MPa m. It is observed that the value of DKth estimated by the K-
increasing method is lower than that estimated by the K-decreasing method.
In other words, the K-increasing method gives conservative result. However, for the 14th block, for example, the corre-
sponding DK-value is greater than the threshold value obtained by the K-increasing method, while crack extension does not
pffiffiffiffiffi
occur. This means that the DKth obtained by the K-decrease method, that is, (DK th ¼ 3:15 MPa m in this case) is more reli-
able than that obtained by the K-increase method.

2.4. Estimation of loading amplitude corresponding to the fatigue limit

The concept of fatigue endurance limit is traditionally used within the safe-life design approach, which defines the ampli-
tude of a stress under which no failure occurs (N > 106 cycles). According to ASTM-E466 standard, the endurance limit Drf is
obtained using S–N diagram [17]. The amplitude of the load under which no failure occurs (N > 106 cycles) is estimated in
this section. For this purpose, a logarithmic curve is fitted to the experimental data of DP–N diagram as shown in Fig. 7.
56 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

The load amplitude value corresponding to N = 5  106 cycles is considered as DPF. Using the curve fitted to the experimental
data, the value of DPF is obtained equal to 0.751433 kN.

3. EIFS estimation and life prediction based on fracture mechanics

3.1. Back extrapolation method

The algorithm of this method has been illustrated in Fig. 12. According to this method, at first the life is measured by con-
ducting fatigue tests on the C(T) specimens. Assuming a pre-existing crack with a specified initial length, the life of the spec-
imen is predicted using Paris model. Finally, the life of the specimen which is obtained in laboratory is compared with the
one predicted. Afterward, the assumed initial crack length is modified in the subsequent analyses until an appropriate error
achieves.
The cyclic stress intensity factor for a C(T) specimen is calculated using Eq. (6), which is provided in ASTM-E647 standard
[9].
DP ð2 þ aÞ
DK ¼ pffiffiffiffiffiffi 3
ð0:886 þ 4:64a  13:32a2 þ 14:72a3  5:6a4 Þ ð6Þ
B W ð1  aÞ2

Fig. 12. Algorithm of back-extrapolation method.

Fig. 13. Final crack length estimation (upper integral limit) with the help of the fracture toughness.
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 57

Table 4
Comparison between the estimated values for af using KIC with the final crack length measured in the laboratory.

SPC af (mm) (measured in laboratory) af (mm) (predicted using KIC) Error (%)
1 32.89 32.97 0.24
2 35.07 34.60 1.34
3 35.67 36.34 1.88
4 37.2 37.73 1.43

Fig. 14. EIFS estimation by the back extrapolation method.

in which a = a/w. Thus, by substituting Eq. (6) in the Paris relation, the life of a C(T) specimen can be calculated from the
following equation:
Z af
da
N¼  m ð7Þ
EIFS
c DPffiffiffiffi ð2þa=wÞ
p
B W ð1a=wÞ32
ð0:886 þ 4:64a=w  13:32ða=wÞ2 þ 14:72ða=wÞ3  5:6ða=wÞ4 Þ

Owing to the fact that fast fracture happens once the value of Kmax reaches the fracture toughness, KIC, the upper limit of
this integral (af) can be estimated. Hence, af can be calculated for each tested specimen according to Fig. 13 and Eq. (8) if the
pffiffiffiffiffi
line K max ¼ K IC ¼ 85 MPa m intersects with Kmax–a curves.

Pmax ð2 þ aÞ
K max ¼ pffiffiffiffiffiffi 3
ð0:886 þ 4:64a  13:32a2 þ 14:72a3  5:6a4 Þ ð8Þ
B W ð1  aÞ2

The values for final crack length, af, estimated using KIC and the same parameter as measured in the laboratory are com-
pared in Table 4. It is observed that the estimated values of the final crack length have a good agreement with the experi-
mental results.
Using the Paris coefficients obtained for 4340 steel and the final crack length calculated for the C(T) specimens, the life of
the specimens is estimated according to Eq. (7). The amount of the EIFS for each specimen can be obtained from the com-
parison of the estimated life with the life measured in the laboratory.
Fig. 14 illustrates the variations of the EIFS with respect to the loading amplitude for R = 0.7. According to this figure, it
can be noted that the EIFS value increases with increase in the load amplitude. The dependency of the EIFS to the load ampli-
tude is one of the main drawbacks of the back extrapolation method.
The fatigue lives predicted based on the estimated EIFS by the back extrapolation method are compared with the lives
measured in the laboratory in Table 5. As it is expected, they are in very good agreement.

3.2. Kitagawa–Takahashi diagram method

The concept of fatigue crack threshold is used within the damage tolerance design approach, which defines a loading cri-
terion under which the cracks will not grow significantly. The concept of fatigue limit is used within the safe-life design ap-
proach, which defines a loading criterion under which no failure occurs (N > 106 cycles) [2]. These two approaches have been
considered for estimation of EIFS by the Kitagawa–Takahashi diagram method.
58 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

Table 5
Comparison between the fatigue lives predicted based on the estimated EIFS by the back
extrapolation method and the lives measured in the laboratory.

SPC N ef ðcyclesÞ (experimental) N pf ðcyclesÞ (predicted) Error (%)

1 37253 37247 0.016

2 85274 85252 0.026
3 225146 225110 0.016
4 452056 452020 0.008

Fig. 15. Algorithm for EIFS calculation by Kitagawa–Takahashi diagram.

Fig. 16. Comparison of predicted life based on fracture mechanics and estimated EIFS by Kitagawa–Takahashi diagram with the measured life in laboratory.

As mentioned earlier, the EIFS determined by this method can be considered as a material property. The algorithm of EIFS
calculation according to this method is shown in Fig. 15.
In this method, EIFS estimation requires cyclic stress intensity factor relation in terms of loading, geometry, and crack
length (in this case, Eq. (6)). According to this method, DKth and DPf are substituted instead of DK and DP, respectively,
in Eq. (6) and thereafter the crack length is calculated. This crack length can be considered as EIFS. Because both DKth
and DPf are material properties, EIFS is also a material property and does not depend the load amplitude.
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 59

Fig. 17. EIFS estimation for the crack length curves versus life for the tested specimens.

Fig. 18. EIFS estimation for tested specimens from TTCI method.

For the tested specimens, both DKth and DPf have been calculated in Sections 2.3 and 2.4, respectively. Therefore, substi-
tuting DKth and DPF instead of DK and DP in Eq. (6), the initial crack length is obtained as EIFS = 0.4345 mm. This estimated
value is identical for all specimens.
In Fig. 16, the predicted life based on the above EIFS value is compared with the life measured in laboratory. It is observed
that a good agreement exists between the results.
60 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61

Fig. 19. Comparison of the predicted life based on fracture mechanics and estimated EIFS by TTCI method with the measured life in laboratory.

Fig. 20. Comparison of predicted life based on fracture mechanics using estimated EIFS by the three methods with the measured life in laboratory.

3.3. Time To Crack Initiation (TTCI) method

In this method, EIFS is estimated by the crack length-life (a–N) or crack length–time (a–t) curves which have been ob-
tained from the experiments. Fitting a curve (such as a second order polynomial) to the data of the large crack growths zone
(linear region of da/dN  DK diagram), the crack length corresponding to N = 0 (or t = 0) in this curve is considered as EIFS.
Fig. 17 shows a–N curves of SPC-1 to SPC-2, respectively, in which a second order polynomial is fitted to the data. Then,
EIFS is calculated substituting N = 0 in the obtained equation.
The calculated values for EIFS under different loading conditions are shown in Fig. 18. It is observed that the EIFS varies
with the different load amplitude values.
The predicted life based on the fracture mechanics and estimated EIFS by TTCI method is compared with the measured
life in laboratory in Fig. 19. As it is observed, this method has no sufficient accuracy especially in low amplitude fatigue.

4. Conclusion

In this paper, EIFS of C(T) specimens made of 4340 Steel has been estimated by three methods of back extrapolation,
Kitagawa–Takahashi diagram, and Time To Crack Initiation (TTCI). Then, the fatigue life of the C(T) specimens has been pre-
dicted based on the fracture mechanics using the estimated EIFS by each method. The main results of this investigation are as
follows:

(1) The threshold value of the stress intensity factor for 4340 steel is estimated by both K-increasing and K-decreasing
methods. The experimental results show that the value of DKth estimated by the K-increasing method is 12.5% lower
than that estimated by K-decreasing method.
 A.R. Shahani, H. Moayeri Kashani / Engineering Fracture Mechanics 99 (2013) 48–61 61

(2) The estimated EIFS by back extrapolation method is dependent on the loading amplitude. In the other words, EIFS
increases with the increase in loading amplitude. The predicted life based on the fracture mechanics using estimated
EIFS by back extrapolation method and the measured life in laboratory are in very good agreement, as it is expected.
(3) The estimated EIFS by Kitagawa–Takahashi diagram can be considered as material property. Thus, the EIFS is identical
for all loading amplitudes. The predicted life based on the Paris model and this estimated EIFS is in good agreement
with the life measured in the laboratory especially in the region of small loading amplitudes. This is because that the
estimated EIFS by this method is based on DKth and DPF and both of these material properties are also measured based
on experimental data in the region of small loading amplitudes.
(4) The estimated EIFS by TTCI method is also dependent on the loading amplitude. The predicted life based on the frac-
ture mechanics using the estimated EIFS by TTCI method has no sufficient accuracy especially in low amplitude fati-
gue. This is because that the rate of fatigue crack growth (slope of the a–N curve) is small in the region where the
loading amplitude is also small. Therefore, fitting a curve in this region yields to a large value of EIFS.

The predicted life based on the fracture mechanics using estimated EIFS by the three methods is compared with the life
measured in laboratory in Fig. 20.

References

[1] Newman Jr JC. The merging of fatigue and fracture mechanics concepts. Prog Aerosp Sci 1998;34:347–90.
[2] Liu Y, Mahadevan S. Probabilistic fatigue life prediction using an equivalent initial flaw size distribution. Int J Fatigue 2009;31:476–87.
[3] Norman E, Dowling. Mechanical behavior of materials. 3rd ed. New Jersey: Prentice Hall; 1999.
[4] Newman Jr JC, Phillips EP, Swain MH. Fatigue-life prediction methodology using small crack theory. Int J Fatigue 1999;21:109–19.
[5] Newman Jr JC. Crack closure model for predicting fatigue crack-growth under aircraft spectrum loading. Am Soc Test Mater 1981:53–84. ASTM STP
748.
[6] Pólo FEUP. Equivalent initial flaw size on open hole INASCO specimens. ADMIRE PROJECT. Portugal: University of Porto; 2004.
[7] Molent L, Sun Q, Green AJ. Characterization of equivalent initial flaw sizes in 7050 aluminum alloy. Fatigue Fract Engng Mater Struct 2006;29:916–37.
[8] Makeev A, Nikishkov Y, Armanios E. A concept for quantifying equivalent initial flaw size distribution in fracture mechanics based life prediction
models. Int J Fatigue 2007;29:141–5.
[9] ASTM code E647. Standard test method for measurement of fatigue crack growth rates. Am Soc Test Mater 1995.
[10] ASTM code E399. Standard test method for plane-strain fracture toughness of metallic materials. Am Soc Test Mater 1997.
[11] Newman Jr JC, Yamada Y. Compression pre-cracking methods to generate near-threshold fatigue crack growth rate data. Int J Fatigue 2010;32:79–885.
[12] Newman Jr JC, Schneider J, Daniel A, McKnight D. Compression pre-cracking to generate near threshold fatigue crack growth rates in two aluminum
alloys. Int J Fatigue 2005;27:1432–40.
[13] Sivaprasad S, Tarafder S, Tarafder M, Ray KK. An alternative method of decreasing DK FCGR testing. Int J Fatigue 2000;22:593–600.
[14] Kranenburg CV, He W, Zuidema J, Veer F. Influence of measurement method on fatigue crack growth threshold. In: 15th European conference on
fracture; 2004.
[15] McEvily AJ, Groeger J. On the threshold for fatigue crack growth. Int Conf Fract 1977;2:1293–8.
[16] Schijve J. Four lectures on fatigue crack growth. Engng Fract Mech 1979;11:167–221.
[17] ASTM code E466. Standard practice for conducting force controlled constant amplitude axial fatigue tests of metallic materials. Am Soc Test Mater
1991.