Available online at www.sciencedirect.

com
International
Journalof
International Journal of Fatigue 30 (2008) 1169–1180
Fatigue
www.elsevier.com/locate/ijfatigue

A study of fatigue crack growth of 7075-T651 aluminum alloy

Tianwen Zhao, Jixi Zhang, Yanyao Jiang *

Department of Mechanical Engineering (312), University of Nevada, Reno, NV 89557, USA

Received 8 March 2007; received in revised form 8 September 2007; accepted 16 September 2007
Available online 26 September 2007

Abstract

Both standard and non-standard compact specimens were employed to experimentally study the crack growth behavior of 7075-T651
aluminum alloy in ambient air. The effects of the stress ratio (R), overloading, underloading, and high–low sequence loading on fatigue
crack growth rate were investigated. Significant R-ratio effect was identified. At the same R-ratio, the influence of specimen geometry on
the relationship between crack growth rate and stress intensity factor range was insignificant. A single overload retarded the crack
growth rate significantly. A slight acceleration of crack growth rate was identified after a single underload. The crack growth rate
resumed after the crack propagated out of the influencing plastic zone created by the overload or underload. A parameter combining
the stress intensity factor range and the maximum stress intensity factor can correlate the crack growth at different stress ratios well when
the R-ratio ranged from 2 to 0.5. The parameter multiplied by a correction factor can be used to predict the crack growth with the
influence of the R-ratio, overloading, underloading, and high–low sequence loading. Wheeler’s model cannot describe the variation
of fatigue crack growth with the crack length being in the overload influencing zone. A modified Wheeler’s model based on the evolution
of the remaining affected plastic zone was found to predict well the influence of the overload and sequence loading on the crack growth.
 2007 Elsevier Ltd. All rights reserved.

Keywords: 7075-T651 Aluminum alloy; Fatigue crack driving force; Overloading; Sequence loading; Stress ratio effect; Underloading

1. Introduction the microstructural characteristics, the R-ratio (the mini-

Most structural components are subjected to cyclic
loading and fatigue fracture is the most common form of
failure. In general, fatigue process consists of three stages:
initiation and early crack propagation, subsequent crack
growth, and final fracture. The fatigue crack growth rate,
da/dN, which determines the fatigue life of the components
after crack initiation, has been extensively investigated
experimentally and theoretically.
Under constant amplitude loading, the fatigue crack
growth curve, the crack growth rate (da/dN) versus the
stress intensity factor range (DK) in the log–log scale, typ-
ically includes three regions [1]. Region I represents the
early crack growth where the growth rates are typically
in the 6106 mm/cycle range. Region I is influenced by
*
Corresponding author. Tel.: +1 775 784 4510; fax: +1 775 784 1701.
E-mail address: yjiang@unr.edu (Y. Jiang).
mum applied stress over the maximum applied stress),
and the environmental conditions. It is generally accepted
that there exists a stress intensity factor threshold (DKth)
below which the fatigue crack growth does not occur.
Region III of the crack growth curve involves rapid and
unstable fatigue crack growth prior to fracture. A crack
growth rate higher than 103 mm/cycle is common in
metallic materials when the crack growth curve asymptot-
ically approaches the fracture toughness (Kc) of the mate-
rial. Although largely influenced by the R-ratio, the
material microstructure, and the component thickness,
Region III crack growth is often ignored in practice due
to the insignificant fatigue life remaining upon entering
the region. Region II of the fatigue crack growth curve
involves stable crack growth with a rate typically between
106 mm/cycle and 103 mm/cycle for metallic materials.
A number of equations have been developed to describe
the sigmoidal da/dN–DK relationship. Paris and Erdogan
[2] were apparently the first to discover the power law

0142-1123/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2007.09.006
1170 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180
relationship to describe the stable crack growth in Region
II. Many variations based on the Paris law have been
developed to consider the R-ratio effect, the threshold value
of the stress intensity factor range (DKth), and the facture
toughness of the material (Kc) [3–6]. Since the fatigue crack
growth rate is mainly controlled by the stress intensity fac-
tor range, DK, and the maximum stress intensity factor,
Kmax, Vasudevan, Sadananda and co-workers [7–12] devel-
oped a two-parameter unified approach for the fatigue
crack growth description. Kujawski [13,14] proposed a
two-parameter model combining DK and Kmax to form
the crack driving force. The driving force does not invoke
disputable and non-repeatable crack closure data. Instead,
it is determined using the positive part of the range of the
applied stress intensity factor, DK+, and the corresponding
maximum value, Kmax. It unifies the overall crack rate pre-
diction methodology regarding the load ratio effects for
both the long- and short-crack growth behavior. The
two-parameter driving force parameter yields a fairly good
correlation with the experiments with regard to the R-ratio
effect and the threshold conditions for a number of engi-
neering materials. For a positive R-ratio, the fatigue driv-
ing force proposed by Kujawski [13,14] is mathematically
identical to the effective stress intensity factor proposed
by Walker [15]. It is noticed that these models are limited
to constant amplitude loading. Additional considerations
are needed to account for the influences of overloading,
underloading, and sequence loading.
It is well-known that a single overload results in a retar-
dation in the crack propagation [16–24]. Crack growth
retardation increases exponentially with the increasing
overload ratio (the magnitude of the overload over the
maximum load in the constant amplitude loading). Full
crack arrest may occur for overload ratios between 2.0
and 2.7 [25]. Periodic overloads separated by a certain
number of constant amplitude cycles can lead to the max-
imum interaction [26]. On the other hand, application of a
compressive underload tends to accelerate crack growth
[16,22,24,27]. The acceleration from a compressive over-
load is typically less than the retardation resulting from
an equivalent tensile overload. Makabe et al. [28] found
that the fatigue crack growth rate decelerated after being
overloaded at a positive R-ratio. However, the crack
growth rate accelerated rather than decelerated under cer-
tain stress conditions with a negative R value. Therefore,
the residual fatigue life after overloading could become
longer or shorter, depending on the overload level and
the baseline stress conditions.
The retardation phenomenon due to overloading is
commonly explained by compressive residual stresses in
the material ahead of the crack tip introduced by the appli-
cation of the overload. Many models have been developed
to describe the fatigue crack growth behavior after over-
loading [1,29–36]. Elber [29] employed the crack closure
concept to explain the crack growth retardation after over-
loading. With respect to the constant amplitude loading,
the overload caused a large compressive stresses at the
crack tip resulting in crack closure. A large portion of
the subsequent tensile load had to be used to ‘‘open’’ the
crack, leading to a much smaller ‘‘effective’’ load and the
crack growth retardation. The model of Matsuoka et al.
[30] was based on the crack closure concept, with a consid-
eration of crack tip blunting. Wheeler [31] emphasized the
enlarged plastic zone ahead of the crack tip caused by over-
loading. After unloading, large compressive stresses were
developed in the affected zone. The crack tip must grow
beyond the affected zone created by overloading before
the crack growth rate can return to the stable values. Sheu
et al. [32] defined the Wheeler retardation parameter as ‘‘a
power function of the ratio of the present plastic zone size
(i.e., diameter of the plastic zone) to the distance from the
crack tip to the border of the plastic zone caused by the
overload’’. Lui [33] provided a further explanation of this
concept. Lu and Li [34] argued that several co-existing
and mutually competitive mechanisms were involved in
the fatigue crack growth after overloading. In addition to
crack closure and compressive residual stresses, other
mechanisms such as crack tip blunting and strain harden-
ing may accelerate the crack growth rate. A semi-empirical
model was proposed to consider the fatigue crack growth
after a single overload. Rushton and Taheri [1] modified
Wheeler’s model to consider a case in which an underload
was followed by an overload. However, few models are
available to consider the sole effect of underload
[1,27,36]. Kim et al. [37] developed a model to consider
the multiple overloads. Markabe et al. [38] presented a
waveform shape to consider the crack growth under the
negative R-ratio conditions.
Despite the extensive work on crack growth, there is still
a need for a satisfactory and generally applicable method
to predict the fatigue crack propagation rate to consider
various effects. In the current investigation, systematic
crack growth experiments were conducted on an aluminum
alloy. The effects of stress ratio, overload, underload, and
sequence loading on fatigue crack growth were studied.
Several existing models were evaluated critically based on
the experimental results.
2. Crack growth experiments
Both standard and non-standard compact specimens of
7075-T651 aluminum alloy were used in the fatigue crack
growth fatigue experiments. The dimensions of the stan-
dard specimen are shown in Fig. 1a. Notches with four dif-
ferent sizes were machined using either the EDM (Electro-
discharging machining) or a diamond saw. Notch 1 was a
sharp slot with a length of 1.0 mm cut by a diamond saw
and a width of 0.325 mm. Notch 2 contained a hole with
a radius of 0.64 mm. The holes in Notch 3 and Notch 4
had a radius of 1.75 mm and 3.18 mm, respectively. The
dimensions of the non-standard compact specimens are
shown in Fig. 1b. The initial notch has a radius of
0.16 mm. These specimens were designed to study the fati-
gue crack growth behavior with changing loading direc-
 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180 1171

The compact specimens were cut from cold rolled 7075-

T651 plates. Experiments were conducted with the cracks
φ12.7X2
orientated to the rolling direction (0) and the direction
r=0.16, 0.64, 1.75, 3.18
perpendicular to the rolling direction (90). The experimen-
r tal loading conditions are summarized in Tables 1 and 2.
60.96 28.04 The first part of the specimen designation (the first column
an in the tables) specifies the crack orientation of the testing
a specimen with respect to the rolling direction of the cold-
rolled aluminum plate.
Seventeen specimens were tested under the constant
amplitude loading condition with five different R-ratios
(0.1, 0.5, 0.75, 1, and 2) (Table 1). Three specimens,
W=50.80 0_C11, 0_C10 and 0_C27, listed in Table 1 were non-stan-
63.50 dard compact specimen (Fig. 1b). All the rest of the exper-
iments reported in the current study employed the standard
40.411 compact specimen (Fig. 1a). Most of the experiments were
tested from crack initiation till a significant long crack
32.156
length. ‘‘Number of cycles when a* = 0.5 mm’’ in Table 1
o 10.16 reflects the crack ‘‘initiation’’ life of the specimen corre-
30
sponding to a fatigue crack size of 0.5 mm measured from
the root of the notch. The fatigue life, Nf, is the number of
the loading cycles from the start of the experiment till the
termination of the experiment when the crack size was af.
It should be noted that af was not the crack size before
r = 0.16 the final fracture of the specimen.
38.10 63.50 Three specimens were subjected to a single overload dur-
a ing the constant-amplitude loading crack growth experi-
ments and two specimens were tested with a single
31.75 underload (Table 2). One specimen was subjected to
high–low sequence loading. An illustration of the single
φ12.7X3 overload, underload, and high–low sequence loading is
shown in Fig. 2.
W = 66.04
Among the three specimens for overloading experi-
ments, two specimens were tested with a stress ratio of
76.20 R = 0.1 and load amplitude of DP/2 = 1.35 kN. With a
load magnitude of 6.0 kN, the overload ratio (overload
Fig. 1. Compact tension specimens.
over the maximum load in the constant amplitude loading)
was 2.0. The overload was applied when the crack length
was 8.45 mm and 14.96 mm, respectively, on these two
tion. The specimens were first subjected to Mode I loading. specimens. A third specimen was subjected to a constant
After the crack reached a certain length, the external load- amplitude loading with R = 1 and DP/2 = 2 kN. The
ing direction was changed 30 from the original loading overload ratio was 3 and the overload was applied when
direction. The results of Mode I loading experiments are the crack length was 16.54 mm. After the application of
reported together with those obtained from the standard an overload, the specimen underwent constant amplitude
compact specimens. loading identical to that before the overload.
Both EDM and diamond saw cut result in minimal Two specimens were tested with a single underload
residual stress near the notch due to the very slow material during the crack propagation experiments with the under-
removal rate by the two methods. After the notches were loads being 6 kN and 10 kN, respectively (Table 2).
machined, one side of the specimens was finely polished The underloads were applied when the crack lengths
to facilitate crack growth measurement using an optical were 12.14 mm and 12.74 mm, respectively for the two
reading micrometer. The experiments were carried out specimens.
using an Instron 8872 load frame with a 25 kN load capac- One experiment was conducted under high–low
ity and 8800 electronics with computer control and data sequence loading (Fig. 2c). The specimen had a notch of
acquisition. Depending on the load magnitude, the loading 3.18 in radius. The notch depth was 8.813 mm measured
frequencies ranged from 1 to 10 Hz. The crack length was from the line of action of the externally applied load. Both
measured using an optical reading microscope with a mag- loading steps had a R-ratio of 0.1. The loading amplitude
nification of 40. in the first step loading was 1.8 kN. After the crack length
1172 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180

Table 1
Experimental conditions for fatigue crack growth under constant amplitude loading
Spec# R-ratio DP/2 f (Hz) Notch radius Notch depth/precrack an Number of cycles when Nf af (mm)
(kN) (mm) (mm) a* = 0.5 mm (cycle) (cycle)
0_C01 0.1 2.700 1–10 0.80 3.54 15,910 24,787 18.73
90_C01 0.1 1.350 1–10 0.10 4.78 24,270 109,000 30.34
0_C02 0.1 1.125 1–10 0.10 4.46 21,140 177,830 29.89
*
0_C03 0.1 0.720 1–10 7.30 205,300 498,900 34.55
0_C12 0.1 1.350 1–10 0.10 6.75 21,200 84,303 28.53
*
90_C02 0.5 0.575 1–10 5.11 143,800 881,650 33.11
*
0_C04 0.5 0.575 1–10 6.00 57,780 480,240 30.39
**
90_C19 0.5 0.750 1–10 6.79 34,790 175,520 27.39
*
0_C05 0.75 0.575 1–10 6.18 35,480 240,800 23.98
*
0_C07 0.75 0.575 1–10 5.33 48,830 298,583 25.44
90_C18 0.75 0.625 1–10 0.14 4.83 99,470 360,958 22.36
0_C06 1 1.600 1–10 0.10 5.38 113,400 519,548 33.44
0_C20 1 3.500 1 3.56 6.48 1372 18,968 16.62
90_C13 2 2.400 1–10 1.74 7.42 73,220 439,100 33.89
0_C10 0.1 1.350 1–10 0.17 0.142 29,390 123,188 11.00
0_C11 0.1 1.350 1–10 0.17 0.327 16,560 142,876 19.28
0_C27 0.1 1.125 1–10 0.17 0.457 31,700 230,358 14.75
R-ratio, minimum load over the maximum load in a loading cycle; DP/2, loading amplitude; an, distance between the notch root and the line of action of
the externally applied load; a*, distance between the crack front and the root of the notch; f, loading frequency; Nf, total number of cycles; af, final crack
length measured from the line of action of the externally applied load.
*
Pre-cracked with cyclic load of R = 0.1, DP/2 = 1.35 kN.
**
Pre-cracked with cyclic load of R = 0.5, DP/2 = 1.25 kN.

Table 2
Constant amplitude loading with a single overload or underload
Spec# R-ratio DP/2 (kN) f (Hz) Notch radius (mm) Notch depth an (mm) POL NOL (cycle) aOL (mm) Nf (cycle) af (mm)
90_C03 0.1 1.35 1–10 0.10 4.23 6.0 86,542 8.45 135,296 32.34
90_C16 0.1 1.35 1–10 0.17 6.74 6.0 51,400 14.96 87,828 26.02
90_C14 1 2.00 1–10 0.64 6.64 6.0 140,336 16.54 478,927 32.66
90_C15 0.1 1.50 1–10 1.74 6.91 6.0 108,668 12.14 125,015 24.46
90_C22 0.1 1.125 1–10 0.64 6.61 10.0 168,801 12.74 198,109 21.82
POL, magnitude of the overload/underload; aOL, crack length at overloading/underloading; NOL, number of cycles of the constant amplitude loading
before overloading/underloading.

reached 22.47 mm, the loading amplitude was switched to The stress intensity factor for the compact specimen was
1.125 kN (Fig. 2c). calculated using the following formula:
Generally, a compact specimen is not recommended for P ð2 þ nÞ
tension-compression loading (R < 0). In the current investi- K ¼ pffiffiffiffiffi 3=2
ð0:886 þ 4:64n  13:32n2
gation, the two loading holes in the compact specimen were B W ð1  nÞ
machined to have a tight tolerance so that the gap between þ 14:72n3  5:6n4 Þ ð1Þ
the pin in the loading fixture and the hole in the specimen
was minimal. For the range of the testing frequencies used where
in the experiments, the maximum and minimum loads dic- a
n¼ ð2Þ
tate the fatigue behavior. The experiments indicated that W
the maximum and minimum loads can be controlled accu- The symbol B in Eq. (1) denotes the thickness of the com-
rately with the compact specimen for the R < 0 loading pact specimen and W is the distance between the applied
cases. force P and the left edge of the specimen (refer to Fig. 1).
A parabolic curve was adopted to best fit a set of five The symbol a in Eq. (2) is the crack length measured from
successive data points in the experimentally obtained rela- the line of the application of the external load, P. Accord-
tionship between the crack length and the number of load- ing the ASTM standard E647 [39], the equation is valid
ing cycles. The crack growth rate at the middle point (the when n > 0.2. The finite element (FE) method was used
third point) was determined from the derivative of the to determine the stress intensity factor when n < 0.2. The
parabola. results of the FE calculation were compared with those
 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180 1173

0.04
P (kN) Overload
0.03

K/(P/B)
0.02

0.01

0.00
0
0 5 10 15 20
Time
a (mm), Crack length

Fig. 4. Stress intensity factor for the non-standard compact specimen

P (kN) obtained from the FE method.

tor. For the given geometry and dimensions shown in

0 Fig. 1b, the relationship between the normalized stress
intensity factor and the crack length is shown in Fig. 4.

Underload 3. Experimental results

Time
3.1. Constant amplitude loading

P (kN) Fig. 5 summarizes the experimental crack growth results

under constant amplitude loading in the traditional form of
ΔP/2 = 1.8kN, N1=105,974 cycles
a1= 22.471 mm crack growth rate versus the stress intensity factor range.
For a given R-ratio, the fatigue rate curves with different
ΔP/2 = 1.125kN, N2=117,467 cycles
a2= 31.835 mm loading amplitudes were practically coincided. Since speci-
mens with two different orientations with respect to the
rolling direction were used, the results also suggest that
the cold rolling process of the material does not practically
0 influence the fatigue crack growth results. Previous
Time research [40] on the same material revealed that the plate
Fig. 2. Illustration of loading histories: (a) overloading; (b) underloading; specimens with two different orientations yielded very sim-
(c) high–low sequence loading. ilar deformation and fatigue crack initiation results.
The material displays a significant R-ratio effect. With
an identical stress intensity factor range, a higher R-ratio
0.10
a/W=0.2
results in a higher crack growth rate. Although the thresh-
old was not experimentally measured, the tendency indi-
cates that the threshold value of the stress intensity factor
range increases as the R-ratio decreases. It should be noted
K/(P/B)

0.05 that among the 14 standard specimens tested under con-

stant amplitude loading, six specimens started with a pre-
crack and the rest eight specimens started with the notches
FE formed either by EDM or diamond saw cut (refer to Table
Eq. (1)
0.00 1). The early crack growth also reflects the notch effect. The
0 10 20 30 40 experimental results of the three non-standard specimens
a (mm), Crack length are shown together in Fig. 5. At the same R-ratio, the crack
growth curves of the non-standard specimens (marked with
Fig. 3. Stress intensity factor obtained from using Eq. (1) and that from
‘‘*’’ in the legend of Fig. 5) are almost coincident with
the FE method for the standard compact specimen.
those of the standard specimens. It is obvious that the spec-
imen geometry has no significant influence on the crack
obtained from using Eq. (1) in Fig. 3. Clearly, the FE re- growth curve as long as the stress intensity factor range
sults and those obtained from using Eq. (1) are identical and the R-ratio are kept identical.
when n > 0.2. An observation of the crack profiles on both the front
For the non-standard compact specimen (Fig. 1b), the and back surfaces of a specimen indicates that the crack
FE method was used to determine the stress intensity fac- profile on the front surface was almost identical to that
1174 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180

-2 R=0.1
10
7075-T651
Compact specimens
R=-1
R=0.75 R=0.5
R=-2

-3
10 R=0.1
ΔP/2=2.7kN
ΔP/2=1.35kN
da/dN (mm/cycle), Crack Growth Rate

ΔP/2=1.125kN
ΔP/2=0.72kN
ΔP/2=1.35kN
*ΔP/2=1.35kN
-4
*ΔP/2=1.35kN
10 *ΔP/2=1.125kN
R=0.5
ΔP/2=0.575kN
ΔP/2=0.575kN
ΔP/2=0.75kN
R=0.75
ΔP/2=0.575kN
-5
ΔP/2=0.575kN
10 ΔP/2=0.625kN
R=-1
ΔP/2=1.6kN
ΔP/2=3.5kN
R=-2
ΔP/2=2.4kN

-6
10
3 4 5 6 7 89 2 3 4 5 6 7 89 2 3
1 2
10 10
ΔK (MPa√m), Stress Intensity Factor Range

Fig. 5. Crack propagation under constant amplitude loading with the effect of the R-ratio.

on the back surface. The crack plane is approximately flat model predictions which will be discussed in a later section
and perpendicular to the external loading direction. are presented together in Fig. 6.

3.2. Overload effect 3.3. Underload effect

Three specimens were tested with a single overload and
the experimental results are shown in Fig. 6. After over-
loading, the growth rate decreased drastically to a mini-
mum value. The crack growth rate recovered rapidly and
then gradually approached the stable growth of the con-
stant amplitude loading. This tendency of the crack growth
rate after overloading was consistent with that observed by
von Euw et al. [41]. When the R-ratio kept the same, the
size of the transient zone was larger when the overloading
occurred at a longer crack length. It is apparent that the
transient zone size is determined by the plastic zone caused
by the overloading. The influence of the overload dimin-
ished after the crack propagated out of the influencing zone
created by overloading. To reduce the redundancy, the
Fig. 7 shows the experimentally obtained fatigue crack
growth rates for two specimens subjected to underload in
constant amplitude loading. The effect of the underload
on the specimen subjected to a lower constant amplitude
(DP/2 = 1.125 kN, Punderload = 10 kN) was not apparent.
For the specimen subjected to a higher constant amplitude
(DP/2 = 1.5 kN, Punderload = 6 kN), the underload
resulted in a identifiable but insignificant acceleration in
crack growth rate. Right after underloading, the fatigue
crack growth rate increased immediately to a higher value.
The subsequent crack growth rate was approximately con-
stant before the crack propagated through the influencing
zone created by the underload. The growth rate resumed
the normal level expected for the constant amplitude load-
 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180 1175

-3 -3
10 10 90_C16
90_C03

da/dN (mm/cycle), Crack Growth Rate

da/dN (mm/cycle), Crack Growth Rate

-4
10 -4
10

-5
10
-5
10
-6
10

-6
10 Overload, POL=6kN
-7 Overload, POL=6kN
10 ΔP/2=1.35kN, R=0.1
ΔP/2=1.35kN, R=0.1 Experiment
Experiment
Prediction
Prediction
-8 -7
10 10
8 9
10 12 13 14 15 16 17 18
ΔK (MPa√m), Stress Intensity Factor Range ΔK (MPa√m), Stress Intensity Factor Range

-3
10
90_C14
da/dN (mm/cycle), Crack Growth Rate

-4
10

-5
10

-6
10
Overload, POL=6kN
ΔP/2=2kN, R=-1
Experiment
Prediction
-7
10
1 3 4
2x10
ΔK (MPa√m), Stress Intensity Factor Range

Fig. 6. Overload effect on crack growth.

ing after the crack penetrated the influencing zone. Com- result clearly, the crack length instead of the stress intensity
paring with the overloading effect, the underloading effect factor range was adopted for the horizontal axis in Fig. 8.
was short-lived and insignificant. The crack growth behavior in the second loading step
resembles that observed from the overloading effect. Upon
3.4. High–low loading sequence effect the reduction in loading amplitude after the completion of
the first step, the crack growth rate decreased drastically
Fig. 8 shows the experimental crack growth results from until a minimum value was reached. Thereafter, the crack
the two-step high–low sequence experiments. The first growth rate increased rapidly as the crack length extended.
loading step had a load amplitude of 1.8 kN and the second The increase in the crack growth rate with the crack exten-
step had a loading amplitude of 1.125 kN. The load ratio in sion slowed down as the crack approached the boundary of
both steps was 0.1. After switching the load amplitude the plastic zone of approximately 1.00 mm created by the
from a high value to a low value, the stress intensity factor previous loading step. Stable crack growth was resumed
was reduced greatly. In order to represent the experimental thereafter.
1176 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180

3 where Kmax is the maximum stress intensity factor in a

loading cycle, and m is a material constant. Considering
2
that the compressive external load has minimal contribu-
tion to the fatigue crack growth, Kujawski [13,14] modified
da/dN (mm/cycle), Crack Growth Rate

-3 Eq. (3) using the following form,

10 Punderload=-6kN
j ¼ ðK max Þa ðDK þ Þ1a ð4Þ
6
5 where DK+ is the positive part of the range of the stress
4 intensity factor, and a is a material constant. For R > 0,
3 Eqs. (3) and (4) are identical and the following relationship
exists: m = 1  a. For a number of engineering materials,
2
Punderload=-10kN Eq. (4) can model the R-ratio effect well for both short-
and large-crack growth [13]. For aluminum alloys tested
10
-4 in ambient environment, a was found to be a @ 0.5 [14].
The consideration of the R-ratio effect using Eq. (4) with
6 a = 0.35 is shown in Fig. 9 for the aluminum alloy tested
5 ΔP/2 = 1.5kN, R=0.1 under constant amplitude loading. Except for the experi-
4 ΔP/2 = 1.125kN, R=0.1 mental data for R = 0.75, a good correlation is achieved
3
for the constant amplitude loading experiments with differ-
2 ent R-ratios. The model especially works well in Region II.
7 8 9 2 For a wide range of j in Region II, the relationship
10 between da/dN and j follows Paris type power law,
ΔK (MPa√m), Stress Intensity Factor Range

Fig. 7. Underload effect on crack growth. da=dN ¼ Cjn ð5Þ

where C and n are material constants. Fitting the

-2
10 90_C17 experimental data in Fig. 9 using the least-squared method
High-low sequence loading gives C = 6.0 · 108 and n = 3.32 when the crack growth
rate p
is ffiffiffiin
ffi mm/cycle and the stress intensity factor is in
MPa m.
da/dN (mm/cycle), Crack Growth Rate

-3
10

-2
10

-4
10
da/dN (mm/cycle), Crack Growth Rate

-3
10
-5
10 R=0.1, ΔP/2=1.8kN R=0.1, ΔP/2=1.125kN

Experiment -4
Prediction 10
-6
10
18 20 22 24 26 28
α=0.35
a (mm), Crack Length
R=0.1
Fig. 8. Crack propagation under two-step high–low sequence loading. -5 R=0.5
10 R=0.75
R=-1
4. Crack growth modeling R=-2
Eq. (5)

4.1. Constant amplitude loading -6

10
3 4 5 6 7 8 2 3 4
Walker [15] proposed an effective stress intensity factor 10
range, DKeff, to consider the R-ratio effect with the follow- α + (1−α)
Kmax (ΔK ) (MPa√m)
ing form,
Fig. 9. Constant amplitude crack propagation with the effect of the R-
DK eff ¼ K max ð1  RÞm ð3Þ ratio using Eq. (5).
 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180 1177

4.2. Overloading, underloading, and sequence loading effects -3

da/dN, (mm/Cycle), Crack Growth Rate

10 8
aR
6 aOL
For a given material under constant amplitude loading, 4
Eq. (5) can be used to estimate the fatigue crack growth
rate. However, Eq. (5) cannot be directly used to consider 2

the effects of overloading, underloading, and sequence -4

10
loading. As a practice method, a correction factor, /R, 8
6
can be added to the right side of Eq. (5) to consider differ- 4
ent loading effect, Overload, POL=6kN
Experimental data
n 2
da=dN ¼ /R Cj ð6Þ (ΔP/2=1.35kN, R=0.1)
Wheeler's Model
-5
It is clear that /R = 1 is for constant amplitude loading. 10
Wheeler’s model is widely used to consider the overloading 13.5 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0
effect due to its simplicity. In Wheeler’s model [31], /R is Crack length, a (mm)
defined as,
(h im Fig. 10. Prediction of crack growth rate after overloading using Wheeler’s
rpi model.
aOL þrp;OL ai
; ai þ rpi 6 aOL þ rp;OL
/R ¼ ð7Þ
1; ai þ rpi P aOL þ rp;OL ting the experimental data related to the minimum crack
growth rate after overloading. According to Wheeler’s
where ai represents the current crack length at the ith load-
model, the crack growth rate decreases to reach a minimum
ing cycle, rpi denotes the current plastic zone size due to the
value after overloading. It increases slowly at first and then
ith loading cycle, aOL is the crack length at which the over-
increases rapidly until it resumes the normal crack growth
load was applied, rp,OL is the plastic zone created by the
rate. However, the experimental results indicate that once
overload, and m is the Wheeler’s empirically adjustable
the minimum growth rate after overloading is reached,
shaping exponent which can be simply taken as the value
the growth rate increases sharply and then asymptotically
that best fits the data. Typical value for Wheeler’s shaping
approaches the normal rate. It is clear that Wheeler’s
exponent, m, ranges from 1 to 3. According to Wheeler
model cannot predict correctly the details of the crack
[31], the crack growth retardation will occur as long as
growth after overloading. It is necessary to make a modifi-
the current plastic zone lies within the plastic zone created
cation to Wheeler’s model in order to give a better predic-
by the overload. As soon as the boundary of the current
tion of the crack growth in the retardation zone.
plastic zone reaches the boundary of the plastic zone cre-
In the current investigation, the correction factor, /R, in
ated by the overload, the crack growth retardation will
Wheeler’s model is modified and defined with the following
cease (i.e., /R = 1).
form:
The plastic zone sizes in Eq. (7) are often calculated by  m
using Irwin’s method [42]. The plastic zone diameter under rpi
/R ¼ ð10Þ
cyclic loading is estimated as [43], rpmr
 2
1 DK i where rpi denotes the current plastic zone size and rpmr rep-
rpi ¼ ð8Þ
bp 2ry resents the effective residual plastic zone size with the com-
bined consideration of overloading and the crack
while the monotonic overload plastic zone size is calculated propagation. Immediately after the overloading, rpmr
as [1], reaches its maximum value to reflect the instantaneous ef-
 2
1 K OL fect of overloading. During the subsequent crack propaga-
rp;OL ¼ ð9Þ tion, rpmr reduces rapidly until it recovers to the value of
bp ry
the current plastic zone size. rpmr should satisfy the follow-
In Eqs. (8) and (9), b = 1 and 3 for plane stress and plane ing conditions: when ai = aOL, rpmr = rpOL; when ai = aR,
strain conditions, respectively. DKi is the stress intensity rpmr = rpi, where aR presents the crack length at which
factor range at the ith loading cycle and KOL is the stress the crack growth rate resumes to its stable level of the con-
intensity factor corresponding to the overload. ry is the stant amplitude loading. Considering the trend of the crack
yield stress of the material. In the current investigation, growth rate in the transient zone, many expressions can be
the thickness of specimens is less than 5 mm and plane chosen, such as,
stress conditions are expected to prevail. Therefore b = 1  p  p
is used in Eqs. (8) and (9). For 7075-T651 aluminum alloy ai  aOL rpmr  rpi
þ ¼1 ð11Þ
under consideration, the yield stress (ry) is estimated to be aR  aOL rpOL  rpi
420.3 MPa. and,
A prediction for the crack growth rate of specimen   p
90_C16 after overloading by Wheeler’s model is shown in ai  aOL
rpmr ¼ rpi þ ðrpOL  rpi Þ exp  ð12Þ
Fig. 10. In Fig. 10, m = 1.056, which is determined by fit- aR  ai
1178 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180
The two expressions give very close results. In the current
investigation, Eq. (12) is employed.
The value of aR in Eq. (12) can be estimated by,
aR ¼ aOL þ rpOL  rpR ð13Þ
where rpR represents the plastic zone size at the crack
length of aR and can be determined by using Eq. (8), i.e.,
 2
1 DK R
rpR ¼ ð14Þ
bp 2ry
DKR is the stress intensity factor range at the crack length
of aR and can be calculated by using Eqs. (1) and (2). It is
clear that rpR in Eq. (13) in turn is a function of aR. By
using the iteration method, the value of aR can be easily
determined. The estimated transient zone size (aR  aOL)
agrees with experimentally measured values very well.
The fatigue crack growth results predicted by using Eqs.
(5), (6), (8), (9), (10), and (12), (13), (14) are shown in solid
lines in Figs. 6 and 8 for the specimens subjected to the
overloading and high–low sequential loading conditions.
For the retardation zone of the crack growth rate after
overloading or after switching the loading amplitude to a
lower value, Eqs. (6), (8), (9), (10), and (12), (13), (14) are
employed to predict the crack growth rate. The value of
m in Eq. (10) and the value of p in Eq. (12) are adjusted
to best fit the experimental data of specimen 90_C16. The
fitting result gives that m = 2.6 and p = 5.3. The same val-
ues of m and p are applied to the predictions of the other
three overloading and high–low sequential loading speci-
mens. The normal crack growth rates before and after
the retardation zone are calculated using Eq. (5). The mod-
ified model can predict the crack growth well for the influ-
ence of overloading and high–low sequence loading.
According to the modified Wheeler’s model, the mini-
mum value of the crack growth rate occurs immediately
after overloading or switching the loading amplitude from
a higher value to a lower value. The crack growth rate
increases sharply with the crack propagation. When the
crack approaches the boundary of the transient zone, the
crack growth rate slows down and gradually resumes to
the normal level under constant amplitude loading. Com-
pared to Wheeler’s model, the modified Wheeler’s model
can predict qualitatively the details of the crack growth
within the retardation zone.
5. Discussion
In the current investigation, both standard ASTM com-
pact specimens and non-standard specimens were used to
study the crack growth properties of 7075T651 aluminum
alloy. The observed relationship between da/dN and DKare
practically independent of the geometry and size of the
specimens (Fig. 5). This confirms that the non-standard
specimens can be used reliably for some special purpose,
such as the investigation of the effect of changing the load-
ing direction on the crack propagation behavior. Since the
microstructure of the material under investigation is not
very homogenous, the crack growth path sometimes devi-
ated a little up or down the strictly defined straight line per-
pendicular to the external loading.
When the R-ratio effect under constant amplitude load-
ing is considered, Kujawski [13,14] suggested that a = 0.5
in Eq. (4) can correlate the fatigue crack growth rate of
7075-T651 aluminum alloy well. The experimental data in
the current investigation indicates that when the R-ratio
is between 2 and 0.5, the correlation is acceptable with
a = 0.5 but an apparent deviation occurs for the R = 0.75
cases. It was found that a = 0.35 results in a better correla-
tion for the overall experimental data, as shown in Fig. 9.
Noroozi et al. [44] suggested that a modification of Eq.
(4) into the following form can get a better correlation
for 7075-T651 alloy:
p 0:5
j ¼ ðK max Þ ðDK þ Þ ð15Þ
where p is a material constant. However, it is difficult to
identify the unit of j with such an expression. In fact, no
improvement can be achieved in crack growth prediction
by using Eq. (15). For example, after adopting Eq. (15)
for j, Eq. (5) can be rewritten as,
da=dN ¼ Cjn ¼ C½K pmax ðDK þ Þ0:5 n
h p 0:5
inðpþ0:5Þ
þ pþ0:5
¼ C K pþ0:5
max ðDK Þ
0
¼ C½K amax ðDK þ Þ1a n ð16Þ
0
p
As long as a ¼ and n = n (p + 0.5), Eqs. (15) and (4)
pþ0:5
are identical.
Different materials exhibit different fatigue cracking
behavior in the transient zone after the application of an
overload. For most materials, the crack growth rate first
decreases to a minimum value and then increases rapidly
to the stable value under constant amplitude loading. Some
materials exhibit an acceleration of the crack growth rate
immediately after overloading [28,45]. A high overload
can result in the arrest of crack growth. Wheeler’s model
and the modified Wheeler’s model proposed in this investi-
gation can predict the tendency of the crack growth rate in
the transient zone only after the minimum growth rate is
reached. However, they cannot account for the complex
phenomena immediately following the overloading, such
as overshooting, arrest, and drastic decrease in the growth
rate. In addition, these models cannot account for under-
loading effect. In order to develop a general model to con-
sider the variable amplitude loading cases, more
mechanisms should be incorporated into the model.
Another major limitation of Wheeler’s model lies in the
empirical nature of its shaping exponent (m). This param-
eter bears no physical significance and is typically evalu-
ated based on a trial-and-error approach to ‘best fit’ the
experimental data. More importantly, the value of m in
Wheeler’s model is not a material constant. Usually, m
depends on the stress ratio (R) and the overload ratio
(OLR) [1]. As a result, Wheeler’s model has the potential
 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180
for requiring a large database for the shaping exponents,
each corresponding to a particular loading history. One
remedy is to represent the Wheeler shaping exponent (m)
as a function of R and OLR [1]. However, even for the
same R and OLR, the value of m for the different crack
length at which the overloading occurs can be different.
Consequently, Wheeler’s model cannot predict the crack
growth rate for a more general loading case. In the modi-
fied Wheeler’s model, the values of m and p are obtained
from best fitting the results of one overloading experiment.
The constants are used to predict the crack growth of the
rest three overloading cases and the high–low sequence
loading cases.
Although Wheeler’s model and its modifications take
the plastic deformation at the crack tip after overloading
into consideration, they are still mainly based on the stress
intensity factor concept and are essentially empirical in nat-
ure. The stress intensity factor reflects the elastic stress field
near the crack tip and is only related to the external stress
and the geometry of specimen. It is difficult for such a con-
cept to consider the physical mechanisms governing the
local plastic deformation and fracture at the crack tip
which is crucial for the crack growth. As a result, the
parameters in the models lack of physical meaning and
can only be determined by fitting the crack growth experi-
mental data.
A rational model should make use of least experiments
possible to determine its parameters and can be applicable
to a general loading case. Such a model generally takes into
consideration the main physical mechanisms governing the
local elastic–plastic deformation and damage behavior.
The parameters in the model possess physical meaning
and can be determined from the results of basic experi-
ments. Once the parameters in the model are reliably deter-
mined, the model can be used to predict the crack growth
under a general loading case condition. It is noticed that
Jiang and co-workers [46–49] have made an effort to
develop such an approach based upon cyclic plasticity
and accumulation of fatigue damage. One difficulty involv-
ing in using a plasticity based model for fatigue prediction
of an aluminum alloy is the accurate description of the cyc-
lic plasticity material behavior.
6. Conclusions
The following conclusions can be drawn from the cur-
rent study,
(1) Comprehensive crack growth experiments were con-
ducted on 7075T651 aluminum alloy. The influences
of the R-ratio, overloading, underloading, and high–
low sequence loading were experimentally studied.
(2) An effective fatigue crack driving force,
(Kmax)a(DK+)1a defined by Kujawski [13,14] can cor-
relate well the fatigue growth rate for the effect of dif-
ferent R-ratio when the R-ratio was in the range of
2 to 0.5.
1179
(3) The retardation of fatigue crack growth rate after a
single overloading or after reducing the loading
amplitude in a high–low sequence loading was very
significant. However, the acceleration of crack
growth rate after a single underloading was not
significant.
(4) A modified Wheeler’s model based on the evolution
of remaining influencing plastic zone can very well
predict the influence of overloading and high–low
sequential loading on the fatigue crack growth.
Acknowledgements
The Office of Naval Research (N000140510777) and the
Missile Defense Agency (F49620-03-1-342) sponsored this
work. The views and conclusions contained herein are
those of the authors and should not be interpreted as nec-
essarily representing the official policies or endorsements,
either expressed or implied, of the Missile Defense Agency,
the Air Force Office of Scientific Research, the Office of
Naval Research, or the US government.
References
[1] Rushton PA, Taheri F. Prediction of crack growth in 350WTsteel
subjected to constant amplitude with over- and under-loads using a
modified wheeler approach. Marine Struct 2003;16:517–39.
[2] Paris PC, Erdogan F. A critical analysis of crack propagation laws. J
Basic Eng 1960;85:528–34.
[3] Foreman RG, Keary VE, Engle RM. Numerical analysis of crack
propagation in cyclic-loaded structures. J Basic Eng 1967;89:459–64.
[4] Weertman J. Rate of growth of fatigue cracks calculated from the
theory of indefinitesimal dislocations distributed on a plane. Int J
Fract Mech 1966;2:460–7.
[5] Klesnil M, Lukas P. Influence of strength and stress history on
growth and stabilization of fatigue cracks. Eng Fract Mech
1972;4:77–92.
[6] McEvily AJ. On closure in fatigue crack growth. ASTM STP 982.
Philadelphia: American Society for Testing and Materials; 1988. p.
35–43.
[7] Sadananda K, Vasudevan AK. In: Lutjering G, Nowack H, editors.
A unified framework for fatigue crack growth. Fatigue 96, vol. 1.
Berlin: Pergamon; 1996. p. 375–80.
[8] Vasudevan AK, Sadananda K, Louat NA. A review of crack closure,
fatigue crack threshold and related phenomena. Mater Sci Eng
1994;A188:1–22.
[9] Vasudevan AK, Sadananda K. Application of unified fatigue damage
approach to compression-tension region. Int J Fatigue 1999;21:
S263–73.
[10] Sadananda K, Vasudevan AK, Holtz RL. Extension of the unified
approach to fatigue crack growth to environmental interactions. Int J
Fatigue 2001;23:S277–86.
[11] Vasudevan AK, Sadananda K, Glinka G. Critical parameters for
fatigue damage. Int J Fatigue 2001;23S:S39–53.
[12] Sadananda K, Vasudevan AK. Crack tip driving forces and crack
growth representation under fatigue. Int J Fatigue 2004;26:39–47.
[13] Kujawski D. A fatigue crack driving force parameter with load ratio
effects. Int J Fatigue 2001;23:S239–46.
[14] Kujawski D. A new (DK+Kmax)0.5 driving force parameter for crack
growth in aluminum alloys. Int J Fatigue 2001;23:733–40.
[15] Walker K. The effect of stress ratio during crack propagation
and fatigue for 2024-T3 and 7075-T6 aluminum. In: Effects of
1180 T. Zhao et al. / International Journal of Fatigue 30 (2008) 1169–1180
environment and complex loading history on fatigue life. ASTM STP
462. Philadelphia (PA): American Society for Testing and Materials.
p. 1–14.
[16] Hunag HL, Ho NJ. The observation and analysis of the dislocation
morphology of fatigue crack tips at steady state propagation rates
subject to a single peak load. Mater Sci Eng 2000;A298(1–2):251–61.
[17] McEvily AJ, Yang Z. The nature of the two opening levels following
an overloading in fatigue crack growth. Metal Trans 1990;21A(10):
2717–27.
[18] Lee EW, Chakrabortty SB, Starke Jr EA. The effect of overload on
the fatigue crack propagation in metastable Beta Ti–V alloys. Metal
Trans 1984;15A(3):511–7.
[19] Gan D, Weertman J. Fatigue crack closure after overload. Eng Fract
Mech 1983;18(3):155–60.
[20] Chaki TK, Li JCM. Overload effect and fatigue crack propagation in
amorphous metallic alloys. Scripta Metal 1984;18(7):703–8.
[21] Higo Y, Nunomura S. The effect of a single peak overload on fatigue
crack propagation rate and crystallographic path in Fe–3%Si single
crystals. Acta Metal 1984;32(7):1029–34.
[22] Schijve J, Broke D. The result of a test program based on a gust
spectrum with variable amplitude loading. Aircraft Eng 1962;34:
314–6.
[23] Corbly DM, Packman PF. Influence of single- and multiple-peak
overloads on fatigue crack propagation in 7075-T6511 Al. Eng Fract
Mech 1973;5(2):479–97.
[24] Topper TH, Yu M T. The effect of overloads on threshold and crack
closure. Int J Fatigue 1985;7(3):159–64.
[25] Nelson DV. Review of fatigue crack-growth prediction under
irregular loading. Chicago, IL: Spring Meet, Society for Experimental
Stress Analysis; 1975. p. 11–6.
[26] Yildirim N, Vardar O. Study of periodic overloads at a fixed
overload-ratio. Eng Fract Mech 1990;36(1):71–6.
[27] Masuda H, Matsuoka S. Effect of Young’s modulus on fatigue crack
propagation under variable loading. Advances in fracture research.
ICF9, vol. 3. Fatigue of metallic and non-metallic materials and
structures. Sydney, Australia: 1–5 April 1997. p. 1557–62.
[28] Makabe C, Purnowidodo A, Miyazaki T, McEvily AJ. Deceleration
and acceleration of crack propagation after an overload under
negative baseline stress ratio. ASTM J Test Eval 2005;33(3):181–7.
[29] Elber W. Fatigue crack closure under cyclic tension. Eng Fract Mech
1970;2:37–45.
[30] Matsuoka S, Tanaka K, Kawahara M. The retardation phenomenon
of fatigue crack growth in HT80 steel. Eng Fract Mech 1976;8:
507–23.
[31] Wheeler OE. Spectrum loading and crack growth. J Basic Eng Trans
ASME, D94 1972;1:181–6.
[32] Sheu BC, Song PS, Hwang S. Shaping exponent in Wheeler model
under a single overload. Eng Fract Mech 1995;51(1):135–43.
[33] Lui AF. Structural life assessment methods. Materials Park, OH:
ASM International; 1998.
[34] Lu Y, Li K. A new model for fatigue crack growth after a single
overload. Eng Fract Mech 1993;46(5):49–856.
[35] Bolotin VV, Lebedev VL. Analytical model of fatigue crack growth
retardation due to overloading. Int J Solids Struct 1996;33(9):
1229–42.
[36] Fuhring H. Practical application of a model for fatigue damage
with irregular cyclic loading. Advances in fracture research (frac-
ture 81), vol. 4. Cannes, France: 29 March–3 April 1981. p. 1823–
1832.
[37] Kim KS, Cho DI, Ahn JW, Choi SB. An experimental study on
fatigue crack propagation under cyclic loading with multiple over-
loads. Key Eng Mater 2005;297–300:2495–500.
[38] Makabe C, Purnowidodo A, Sueyoshi T, Utsunomiya T. Detecting
overload from strain information during fatigue crack propagation
under negative stress ratio. ASTM J Test Eval 2004;32(1):56–61.
[39] ASTM E647-05. Standard test method for measurement of fatigue
crack growth rates. ASTM International.
[40] Zhao T, Jiang Y. Fatigue of 7075-T651 aluminum alloy. Int J Fatigue
2008;30:834–49.
[41] von Euw EFJ, Hertzberg RW, Robert R. Delay effects in fatigue-
crack propagation. ASTM STP 513. American Society for Testing
and Materials, Philadelphia, 1972. p. 230–59.
[42] Irwin GR. Analysis of stresses and strains near the end of a crack
traversing a plate. Trans ASME J Appl Mech 1957;E24:361.
[43] Antunes FV, Borrego LFP, Costa JD, Ferreira JM. A numerical
study of fatigue crack closure induced by plasticity. Fatigue Fract
Eng Struct 2004;27:825–35.
[44] Noroozi AH, Glinka G, Lambert S. Prediction of fatigue crack
growth under variable amplitude loading based on the elasto-
plastic crack tip stresses and strains. Oral Reference: FT463.
2006.
[45] Wang X, Jiang Y, Ding F, Zhao T, Gao Z. An experimental
investigation of crack growth behavior of a pressure vessel steel. In:
Proceedings of PVP2006-ICPVT-11 2006; PVP2006-ICPVT-11-
93471, ASME pressure vessels and piping division conference, July
23–27, 2006. Vancouver, BC, Canada.
[46] Feng M, Ding F, Jiang Y. A study of loading path influence on
fatigue crack growth under combined loading. Int J Fatigue 2006;28:
19–27.
[47] Jiang Y, Ding F, Feng M. An approach for fatigue life prediction.
ASME J Eng Mater Tech 2007;129(2):182–9.
[48] Feng M, Ding F, Jiang Y. A study of crack growth retardation due to
artificially induced crack surface contact. Int J Fatigue 2005;27:
1319–27.
[49] Jiang Y, Feng M. Modeling of fatigue crack propagation. ASME J
Eng Mater Tech 2004;126:77–86.