See

discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/257519995

Implications of the lead crack philosophy

and the role of short cracks in combat
aircraft

ARTICLE in ENGINEERING FAILURE ANALYSIS JANUARY 2013

Impact Factor: 1.03 DOI: 10.1016/j.engfailanal.2012.10.023

CITATIONS READS

11 71

2 AUTHORS:

Rhys Jones Dinaz Tamboli

Monash University (Australia) Monash University (Australia)
372 PUBLICATIONS 3,908 CITATIONS 6 PUBLICATIONS 14 CITATIONS

SEE PROFILE SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Rhys Jones
letting you access and read them immediately. Retrieved on: 14 January 2016
 Engineering Failure Analysis 29 (2013) 149166

Contents lists available at SciVerse ScienceDirect

Engineering Failure Analysis

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

Implications of the lead crack philosophy and the role of short

cracks in combat aircraft
R. Jones , D. Tamboli
Centre of Expertise in Structural Mechanics, Department of Mechanical and Aeronautical Engineering, Monash University, Clayton, Vic 3800, Australia

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

Article history: The Australian Defence Science and Technology (DSTO)/Royal Australian Air Force (RAAF)
Received 3 August 2012 approach to the management of fatigue cracking in combat and trainer aircraft makes use
Received in revised form 21 September of the lead crack concept. In this approach crack growth is assumed to initiate from small
2012
naturally occurring defects/discontinuities with dimensions of approximately 10 lm and
Accepted 22 October 2012
Available online 2 December 2012
growth is assumed to commence from day one. As a result, for certication purposes, we
need to address the so called short crack anomaly, whereby for a given DK the crack growth
rate (da/dN) is signicantly greater for short cracks than it is for long cracks. In this paper
Keywords:
Lead cracks
we reveal that there are several instances where this anomaly vanishes if da/dN is repre-
Short cracks sented as a function of (DK  DKthr), where DKthr can be thought of as an apparent thresh-
R ratio old. We then show that for operational aircraft the true da/dN versus DK curve is an
Crack closure amalgam of the short and long crack growth curves. We next show that existing test pro-
Crack patching cedures used to establish the effect of composite repairs to cracks in eet aircraft overes-
timate their effect. We also show that the growth of cracks from small naturally occurring
defects exhibits little, if any, R ratio effects.
2012 Elsevier Ltd. All rights reserved.

1. Introduction

The DSTO-RAAF approach to the management of fatigue cracking in combat and trainer aircraft makes use of the lead
crack concept [1]. In this approach, the life of the eet is determined by lead fatigue cracks which in [1] were dened to
have the following features:

(a) Crack growth initiates from small naturally occurring defects or discontinuities, such as inclusions and pits, which
have dimensions that are equivalent to a fatigue crack-like size typically of about 10 lm in depth.
(b) Crack growth essentially starts from the day that the aircraft enters service. (This implies that the fatigue threshold
DKth is very small.)

As such understanding the growth of fatigue cracks from small naturally occurring defects is of fundamental importance
to managing the Australian eet. Indeed, the central role that small cracks play in understanding the durability of aircraft is
highlighted in [2]. Furthermore, as stated by Lados et al. [3]:
The use of long crack data can lead to signicantly non-conservative estimates of the fatigue response and serious design
errors.

Corresponding author. Tel.: +61 398786265.

E-mail address: rhys.jones@monash.edu (R. Jones).

1350-6307/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engfailanal.2012.10.023
150 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

The extent of these non-conservative estimates is aptly illustrated in [4] where it is shown that using FASTRAN together
with long crack da/dN versus DK data to predict the crack growth from a small 0.003 initial defect in a F/A-18 centre barrel
crack gave an estimate of the fatigue life that was more 300% greater than that seen in the test. Similarly Appendix C reveals
that the use FASTRAN together with short crack da/dN versus DK data can also lead to non-conservative estimates for the
fatigue life of a structure.
This means that it is important to address, and hopefully overcome, the so called short crack anomaly which is one of
the basic problems in materials science and particularly in fatigue crack growth prediction. Here it should be noted that as
stated in [1] and can be seen from the experimental data presented in [5] that typical small natural initiating defects in
military aircraft have a size that is of the order of 0.01 mm. This (initiating defect) size is consistent with the work of Merati
[6], where it was found that the size of initial defects in civil transport aircraft lie in the range 0.0090.029 mm and with the
paper by Schijve [2] where it was reported that the size of initial defects in civil transport aircraft lie in the range 0.007
0.030 mm. As such when discussing crack growth from small naturally occurring initial discontinuities in aircraft structures
we focus on initial defects that are typically 0.01 mm long/deep. (In this context it should be noted that Section 5.3.1 of the
USAF Damage Tolerant Design Handbook [7] suggests that cracks growing from initial defects greater than between 1.0 and
1.8 mm long behave like long cracks. A methodology for modelling crack growth from an arbitrary length initial notch is
proposed in [8] and will be discussed later in this paper.)
The short crack anomaly arises as experimental studies have shown that for a given DK, R ratio and specimen thickness
the increment in the crack length (or depth) per cycle (da/dN) seen in tests on short cracks is signicantly greater than that
seen in tests on long cracks. An example of this anomaly is shown in Fig. 1, from [9], which compared the growth of short
and long cracks in 2090-T8E41 at R = 0.1. A second example is shown in Fig. 2, from [10], which presents the da/dN versus DK
relationships obtained for AA7050-T7451, which is used in both the F/A-18 Hornet, the F/A-18 Super Hornet and the Joint

1.00E-05
Short cracks R = 0.1
Long cracks R = 0.1
1.00E-06
da/dN (m/cycle)

1.00E-07

Short crack
1.00E-08 anomaly

Long crack
1.00E-09 threshold

1.00E-10

1.00E-11
0.1 1 10 100
K MPa m

Fig. 1. Short and long crack growth in 2090-T8E41, from [9].

10 -5

10 -6
da/dN (m/cycle)

10 -7

10 -8 short crack effect

10 -9

10 -10 Long crack thresholds

10 -11
0.1 1 10 100
K MPa m
NASA Kmax DSTO CT R = 0.1
DSTO CT R = 0.1 NASA R = 0.7
NASA R = 0.1 DSTO CT R = 0.5
Short crack Test 2 R = 0.1 Short crack Test 2 R = 0.7
DSTO MT R = 0.2 Short crack Test 3 R = 0.1
Short crack Test 3 R = 0.7

Fig. 2. Comparison of the various da/dN versus DK test data for 7050-T7451, from [10,13].
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 151

1.0E-05

y = 7.00E-10 x2.00
1.0E-06

1.0E-07

da/dN (m/cycle)
1.0E-08

1.0E-09

1.0E-10

1.0E-11
0.1 1 10 100

[10] Short Crack Test 2 R = 0.1
[10] Short Crack R = -0.3
[11] FAA Long crack R = 0.7
[12] DSTO CT R = 0.5
NASA R = 0.7
[10] Short Crack Test3 R = 0.1
DSTO MT R = 0.2
[10] Short Crack Test 2 R = 0.7
[10] Short Crack R = 0.5
[12] DSTO CT R = 0.7
[12] DSTO CT R = 0.1
NASA Kmax
[10] Short Crack Test3 R = 0.7

Fig. 3. Long and small cracks in 7050-T7451, from [13].

Strike Fighter (JSF), for constant amplitude short crack tests [10] for R = 0.3, 0.1, 0.5 and 0.7 together with the long crack
growth data presented in [1113]. In the short crack tests detailed above cracks were grown from surface etch pits which
had initial depths of approximately 7 lm and the crack growth rate data was determined using quantitative fractography.
(It should be noted that the short crack da/dN versus DK data sets, which showed little R ratio dependency for crack growth
rates for 1  1010 < da/dN < 1  107 m/cycle, tended to merge with the high R ratio (=0.7) da/dN versus DK data at a crack
growth rate of approximately 1  108 m/cycle. However, the lengths at which this occurred varied from approximately 0.3
1.2 mm. Furthermore, as for the examples presented in Appendices B and D, the short crack test revealed that, allowing for
experimental error, the da/dN versus DK curves showed little if any R ratio effects at any stage of their life.) Jones et al. [13]
re-plotted the data as per following equation, viz:
p
da=dN DDK  DK thr = 1  K max =Ab 1
and the results are displayed in Fig. 3. Here b is a constant, that is frequently approximately 2, which is determined from the
p
slope of the da/dN versus DK  DK thr = 1  K max =A curve, D is a constant and the parameters A and DKthr are chosen so as
to best represent the experimental data. It should be noted that, as rst explained in [14], when seeking to express da/dN in
terms of (DK  DKthr) the term DKthr should not be confused with the term DKth, which the ASTM 647 test standard suggests
to be the value of DK at a crack growth rate da/dN of 1010 m/cycle. DKthr is chosen so as to ensure that Eq. (1) reproduces the
observed crack growth rates over the entire da/dN versus DK curve. Since da/dN is (in general) a function of the material,
thickness, R ratio and crack length so is DKthr. For long cracks the term DKthr is a function of the R ratio, whilst for small nat-
urally occurring cracks DKthr is a (very small) constant that is (essentially) independent of the R ratio, see Appendix D. Eq. (1)
has been shown to hold for a large cross-section of aluminium alloys, steels and titanium alloys [13]. p
The study reported in [13] revealed that when da/dN was plotted as a function of DK  DK thr = 1  K max =A the long
and small crack data associated with crack growth in 7050-T7451 essentially coincided. It was also shown [13] that, in this
p
instance, Eq. (1) with the value of the constants taken to be D = 7.0  1010, DKthr = 0.1 MPa m and b = 2 provided a good
p
representation of both the small and long crack data, see Fig. 3 from [13]. In these tests a value of DKthr = 0.1 MPa m was
used for all R ratios.
The values of the constants DKthr and A used in [13] for the various test data [1013] presented in Fig. 3 are given in
p
Table 1. Fig. 3 reveals that when da/dN is plotted against DK  DK thr = 1  K max =A the following data sets:

 the R = 0.7 test data obtained by NASA [13];

 the NASA Kmax test data;
 the DSTO CT test data, presented in [12] for three different R ratios;
 the R = 0.2 DSTO MT test data [13];
152 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

Table 1
Values of the constants used in the various tests.
p p p
Coupon R ratio or Kmax (MPa m) Kthr (MPa m) A (MPa m)
7050-T7451 [10,13] short crack R = 0.7 0.1 32
7050-T7451 [10,13] short crack R = 0.5 0.1 32
7050-T7451 [10,13] short crack R = 0.1 0.1 32
7050-T7451 [11] R = 0.7 0.98 38
7050-T7451 [12] R = 0.7 1.4 32
7050-T7451 [12] R = 0.5 1.6 32
7050-T7451 [12] R = 0.1 2.5 32
STOA Ti6Al4V [24] width = 76 mm R = 0.1 4.2 80
STOA Ti6Al4V [24] width = 51 mm R = 0.1 3.6 80
STOA Ti6Al4V [24] width = 25 mm R = 0.1 3.4 80
STOA Ti6Al4V [24] width = 76 mm R = 0.4 2.5 80
STOA Ti6Al4V [24] width = 51 mm R = 0.75 2.2 80
STOA Ti6Al4V [24] width = 25 mm R = 0.5 2.1 80
STOA Ti6Al4V [24] width = 25 mm R = 0.7 1.8 80
STOA Ti6Al4V [25] R = 0.1 4.2 80
STOA Ti6Al4V [25] R = 0.5 3.6 80
STOA Ti6Al4V [25] R = 0.8 3.4 80
STOA Ti6Al4V [25] Kmax = 26.5 2.5 80
STOA Ti6Al4V [25] Kmax = 36.5 2.2 80
STOA Ti6Al4V [25] Kmax = 56.5 2.1 80
STOA Ti6Al4V [26] Kmax = 10 3.7 80
STOA Ti6Al4V [26] Kmax = 20 2.9 80
STOA Ti6Al4V [26] Kmax = 40 2.4 80
STOA Ti6Al4V [27] Kmax = 24.5 1.5 80
STOA Ti6Al4V [28] short crack R = 0.5 0.6 80
STOA Ti6Al4V [28] short crack R = 0.1 0.6 80

 the FAA R = 0.7 test data [11];

 the short crack test data, obtained [10,13] for four different R ratios;

now all fall on the same curve. As a result, the short crack anomaly essentially vanishes. It was subsequently shown [15]
that this formulation unied the growth of short [16] and long cracks [17] in 4340 steel.
Having shown that under constant amplitude loading the growth of both small and long cracks in 7050-T7451 aluminium
alloy and 4340 steel can be unied by expressing da/dN as in terms of (DK  DKthr), as rst suggested by Hartman and Schi-
jve [18], rather than DK this paper rst reveals that this formulation also unies short and long crack growth in Ti6Al4V.
We subsequently show that the curve to be used in any damage tolerant analysis/assessment of operational aircraft struc-
tures/modications should be an amalgam of the short and long crack growth curves. From this study and from a comparison
between measured and predicted crack growth from small (near micron) etch pits we show that existing test procedures
used to establish the effect of composite repairs on eet aircraft overestimates their effect. We next show that the growth
of cracks from small naturally occurring defects exhibit little if any R ratio effects. We then build on this observation, and
from a comparison between measured and predicted crack growth from small (near micron) etch pits, to show that the
use of crack closure based growth laws to model crack growth from small naturally occurring defects in operational aircraft
and/or the effect of composite repairs on cracks, that have grown from small naturally occurring discontinuities in opera-
tional aircraft, is questionable.

2. Short and long cracks

The hypothesis that da/dN should be dependent on the amount by which DK exceeds the fatigue threshold DKth, i.e. on
DK  DKth, was rst proposed by Hartman and Schijve [18]. Zheng and Hirt [14] subsequently derived the crack growth
equation:

da=dN DDj2 2
where D is a constant and the crack driving force, Dj, is:
Dj DK  DK thr 3
where the value of the parameter DKthr depends on the R ratio. Eq. (2) was validated by examining crack growth data asso-
ciated with 28 different steels,1 see [14]. In this work [14] it was explained that the term DKthr should not be confused with the

1
Jones et al. [13] revealed that da/dN can be related to (DK  DKthr) and that this relationship held for twenty ve different aluminium and titanium alloys
and for a range of steels. Similarly the study by Ramsamooj revealed a relationship between da/dN and (DK  DKthr)2 that held for four steels, six aluminium
alloys and a super alloy.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 153

1.00E-05

1.00E-06

[24] R = 0.1 51 mm
[24] R = 0.1 76 mm
1.00E-07

da/dN (m/cycle)
[24] R = 0.4 51 mm
[24] R = 0.4 76 mm
[24] R = 0.7 76 mm
1.00E-08 [24] R = 0.1 25 mm
[24] R = 0.4 25 mm
[28] R = 0.5
[28] R = 0.1
1.00E-09 [26] Kmax = 276
[26] Kmax = 138
[26] Kmax = 68.9
1.00E-10 [25] Kmax = 26.5
[25] Kmax = 36.5
[25] Kmax = 56.5
[25] R = 0.8
1.00E-11
1 10 100
K MPa m

Fig. 4. Crack growth in STOA Ti6Al4V.

term DKth, which the ASTM 647 test standard suggests (somewhat arbitrarily2) to be the value of DK at a crack growth rate, da/
dN, of 1010 m/cycle. Here DKthr is chosen so as to ensure that Eq. (3) reproduces the observed crack growth rates over the entire
da/dN versus DK curve, for more details see [14].
Liu and Liu [19], Ramsamooj [20], Pandey and Chand [21], Wang [22] also derived crack growth equations that related da/
dN to (DK  DKthr)2. Ramsamooj [20] revealed that this relationship held for both long and short cracks. Ishihara et al. [23]
also revealed that for small near micron size defects da/dN was related to (DK  DKthr)2.
To continue this study we examined the growth of both long cracks [2427], where the da/dN versus DK data was ob-
tained as per the ASTM 647 load reducing test standard with C = 0.08 mm1, and short cracks [28] in STOA Ti6Al4V. Here
the crack growth rates were determined by taking replicas of the cracked surface. In [24] it was shown that for a given R ratio
changing the widths of the compact tension test specimens from 25 to 76 mm resulted in signicantly different da/dN versus
DK curves, see Fig. 4.
In [25] Boyce and Ritchie, where the da/dN versus DK data was obtained both in accordance with the ASTM 647 standard
load reducing test standard with C = 0.08 mm1and via constant Kmax tests, stated:
Crack closure is generally considered to be the primary reason for the effect of load ratio on the DKth fatigue threshold in
metallic materials. However, based on the closure argument, at load ratios where Kmin exceeds Kcl, (i.e. R > Rc) such that
the inuence of (global) closure is minimized, the value of DKth would be expected to become independent of R. This is
the generally anticipated behaviour, however, in the current Ti6Al4V alloy, we have observed a progressive reduction
in DKth from R = 0.5 (the closure-free condition) to R = 0.95.

This effect is shown in Fig. 5 together with the test results presented in [26] for tests performed at Kmax = 68.9, 137.8 and
p p
275.6 MPa m, which show a similar trend, and a Kmax = 24.5 MPa m test given in [27]. The Kmax tests were performed as
per the ASTM 647 standard. Boyce and Ritchie [25] also gave a range of steels and titanium alloys for which this, i.e. a reduc-
tion in DKth for R values above the closure free value of R (Rc), had been observed. Similar behaviours are reported in [2931].
Although the reason(s) for this phenomenon are unclear a brief discussion of two possible explanations is presented in
Appendix E. Fortunately, for operational aircraft it is unlikely to have a signicant effect since the initiating defects are gen-
erally so small, typically 0.01 mm, that any Kmax effects are also small. (As discussed above and shown in Appendix D there is
generally no signicant R ratio effect on crack growth at any stage in the life of cracks that grow from small naturally occur-
ring defects.) p
The data shown in Figs. 4 and 5 is re-plotted in Fig. 6 where da/dN is now plotted against DK  DK thr = 1  K max =A.
p
The values of DKthr and A, which as per [13] was kept xed at A = 80 MPa m, used for analysing these various tests are given
in Table 1. Here we see that by merely changing the apparent threshold (DKthr) all of the tests, including the short crack test

2
The precise wording used in ASTM E647 is: fatigue crack growth threshold, DKth [FL3/2]that asymptotic value of DK at which da/dN approaches zero. For
most materials an operational, though arbitrary, denition of DKth is given as that DK which corresponds to a fatigue crack growth rate of 1010 m/cycle. The
procedure for determining this operational DKth is given in 9.4.
154 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

1.00E-05

1.00E-06

Kmax = 276 [26]

1.00E-07

da/dN (m/cycle)
Kmax = 138 [26]
Kmax = 68.9 [26]
[25] Kmax = 26.5
1.00E-08 [25] Kmax = 36.5
[25] Kmax = 56.5
[25] R = 0.1
[25] R = 0.5
R = 0.94
1.00E-09 [25] R = 0.8
[27] Kmax = 24.5

1.00E-10
R = 0.62

R = 0.85
1.00E-11
1 10 100
K MPa m

Fig. 5. Crack growth in STOA Ti6Al4V, replotted.

[24] R = 0.1 51 mm
1.00E-05 [24] R = 0.1 76 mm
[24] R = 0.4 51 mm
[24] R = 0.4 76 mm
[24] R = 0.7 76 mm
1.00E-06 [24] R = 0.1 25 mm
Equation (4)
[24] R = 0.4 25 mm
[28] R = 0.5
1.00E-07 [28] R = 0.1
[26] Kmax = 40
[26] Kmax = 20
da/dN (m/cycle)

[26] Kmax = 10
[25] Kmax = 26.5
1.00E-08 [25] Kmax = 36.5
[25] Kmax = 56.5 y = 1.80E-10x2.00
[25] R = 0.1
[25] R = 0.5
1.00E-09 [25] R = 0.8
[27] Kmax =24.5

1.00E-10

1.00E-11

1.00E-12
0.1 1 10 100

Fig. 6. Summary of crack growth data for (STOA) Ti6Al4V.

data given in [28], essentially collapse onto a single curve, see Fig. 6. The value of D given in [13] for long cracks in this mate-
rial was 1.8  1010. As such the crack growth equation determined in [13] for long cracks in this material was:
p
da=dN 1:8  1010 DK  DK thr 2 = 1  K max =802 4
A plot of this equation is also shown in Fig. 6. From this gure we see that when plotted in this fashion the discrepancy
between the various tests now (essentially) vanishes and each of the test data sets presented in [2428] can be reasonably
well for by Eq. (4).
It thus appears that the spread of the data arises due to plotting da/dN against DK and that if as rst suggested by Hart-
man and Schijve [18] da/dN is plotted as a function of (DK  DKthr) then, allowing for experimental error, this discrepancy
essentially vanishes. Furthermore, since the data associated with [28] is for short sub mm cracks, which as stated in [28] had
little if any R ratio effect, we see that when plotted in this fashion the so called short crack anomaly also (essentially) van-
ishes for this material.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 155

1.0E-05 D

1.0E-06

da/dN (m/cycle)
x E
1.0E-07
Short crack data B
1.0E-08 F
x
1.0E-09
xG
1.0E-10
A C
1.0E-11
0.1 1 Kp KB 10 100

Fig. 7. Generic representation of long and short crack growth curves.

As such Eq. (1) may be thought of as collapsing the various R ratio da/dN versus DK curves onto the high R ratio da/dN
versus DK curve, which is generally thought of as the da/dN versus DKeff curve, which, by virtue of Eq. (1), can be expressed
as:
p
da=dN DDK eff  DK eff;thr = 1  K max =Ab 5
where DKeff (=Kmax  Kop) is the effective stress intensity factor range, Kop is the value of the stress intensity factor at which
the crack opens, DKeff,thr is a parameter that is chosen to ensure that Eq. (5) ts the experimental data, b is approximately 2
and D is a constant which, as shown in [13], can often be approximated as per Liu and Liu [19] by D = 0.036/Ery. Eq. (5)
resembles the crack growth equation developed by McEvily and co-workers [8,22,32,33], viz:

da=dN FDK eff  DK eff;th 2 6

where F is a constant, which can often be approximated as 0.04/Ery [34], which also collapses the various R ratio curves onto
a single curve and has also been shown to hold for both short and long cracks. Here it should be noted that, as shown in
[8,32,33], the term DKeff (=Kmax  Kop) can also be written as DKeff = DK  DKop where DKop = Kop  Kmin and that [8,32,33]
explained that DKop is an exponential function of the crack length so that for small cracks DKop becomes vanishingly small,
see Appendix A. Hence in the McEvily crack growth equation DKeff asymptotes to DK as the crack length reduces. Thus for
small naturally occurring initial defects the form of the (McEvily) crack growth equation, i.e. Eq. (6), resembles Eq. (1). The
difference is the term (1  Kmax/A) on the denominator of Eq. (1) which is used to account for the increase in the crack growth
rate as Kmax approaches the cyclic fracture toughness of the material. The similarity/equivalence of these two formulations is
illustrated by the fact that [33] gave the governing equation for the growth of both large and small cracks in Ti6Al4V to be:

da=dN 1:7  1010 DK eff  DK eff;th 2 1 DK eff =K c  K max 7

This equation is similar to Eq. (4). In [33] the term (1 + DKeff/(Kc  Kmax)) on the right hand side of Eq. (7), which is used to
account for the increase in crack growth as Kmax approaches Kc, was taken from [34]. In practice since this term only makes a
signicant contribution in the Region III and the upper portion of Region II and since for cracks that initiate from small nat-
urally occurring defects the time spent growing in these Regions is a small fraction of the total life Eqs. (4) and (7) are oper-
ationally equivalent. As such we will refer to Eq. (5) as the HartmanSchijveMcEvily (HSM) crack growth equation. A more
detailed discussion on this is given in Appendix A. It should also be noted that whereas there have been many papers that
have attempted to present a mechanics/materials explanation of the reason(s) for this short crack effect the authors concur
with the conclusion reached by Caton et al. [28], viz:
The mechanisms for this are not fully understood.

Section 5.3.1 of the USAF Damage Tolerance Design Handbook [2] states that crack growth data obtained for the growth of
short cracks from un-notched specimens cannot be used to predict the growth of short cracks from holes. This shortcoming
vanishes if Eq. (1), or equivalently Eq. (5) since for short cracks DKeff = DK, is used to predict crack growth, see Appendix B.
Indeed, as shown in [58] and in Appendix B the constants obtained from constant amplitude short crack tests can be used to
predict crack growth under operational load spectra without the need to further calibrate the input data, see [58] for more
details. An illustration of the application of the HartmanSchijveMcEvily crack growth equation, i.e. Eq. (5), to the growth of
long cracks under representative P3C operational loading [35] is illustrated in Appendix C where it is contrasted with results
obtained using FASTRAN.

3. Implications for crack growth in eet aircraft: which da/dN versus DK curve?

Despite the scatter in the experimental data, which is presented in Fig. 1, the paper by Rao, Yu and Ritchie [9] was one of
the rst papers to reveal that the growth of short cracks in aerospace quality aluminium and aluminiumlithium alloys often
took a form similar to that shown in Fig. 7. A similar behaviour was seen in the short and long crack growth data presented in
156 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

1.0E-06

Short crack R = 0.1

Long crack R = 0.1

da/dN (m/cycle)
1.0E-07

1.0E-08

1.0E-09
0.1 1 10 100

Fig. 8. Short and long crack growth data for 7075-T6511, from [37].

Fig. 2 for 7050-T7451, i.e. a traditional sigmoidal graph for the long crack data and a (near) power law relationship between
da/dN and DK for the short crack data. The data presented by Piascik [36] for crack growth in 7075-T6 also suggests the same
representation, i.e. a (near) power law relationship between da/dN and DK for the short crack data and a traditional sigmoi-
dal graph for the long crack data. Fig. 8 presents the data presented in [37] for the growth of cracks from small corrosion pits
as well as for the growth of long cracks in 7075-T76511. This Figure also reveals a similar behaviour. Similar behaviours are
shown in Appendix D for crack growth from small initial defects.
It would thus appear that for those aluminium alloys that do not exhibit a Marci effect, or indeed any effect whereby the
relationship between da/dN and DK is not one to one and onto, that the da/dN versus DK relationships for short and long
cracks often takes the generic form shown in Fig. 7.
At this stage it should be recalled that the equivalent da/dN and DK long crack data is generally obtained using ASTM E647
standard crack growth tests which involve cracks grown from articially induced starter notches that are fatigued so as to
subsequently produce a long, with a length that is in excess of 10 mm, sharp crack. In contrast cracks in operational aircraft
generally result from small naturally occurring defects/discontinuities [1,2,5,6] where the initiating defect is of the order of
0.01 mm, see Section 1. The question thus arises:
Given that the short and long crack da/dN versus DK relationships are very different which curve should be used in any
crack growth analysis?
The US Department of Defense Joint Structures Guidelines 2006 requires that all aircraft structures/modications be de-
signed/assessed in accordance with damage tolerance requirements [7]. In the associated analytical assessment it is required
that all structural hot spots be assumed to have an inherent initial aw, typically 1.27 mm (0.05 in.), and that these aws be
allowed to grow [7]. (This size corresponds to the minimum size that can be reliably detected in operational aircraft and
hence, from a design perspective, this size aw must be assumed to be present at all points in the airframe.) The life of
the structure is then determined by the time taken to grow the aw(s) to a size such that the residual strength of the aircraft
is reduced to limit load [7]. However, given the crack growth data presented in [1] for crack growth in both operational air-
craft and in full scale tests it is reasonable to assume, as per the lead crack philosophy [1] and [2,5,6], that in operational
aircraft the life of the eet is determined by cracks that grow from a small naturally occurring defect.3 We are now faced with
two possibilities:

(a) The value of DK corresponding to the initial aw size lies to the right of DKB, the value of DK associated with point B in
Fig. 7, so that the associated da/dN versus DK curve that should be used in the analysis/assessment is the curve BD;
(b) The initial value of DK lies to the left of DKB. In this instance since the defect has grown from a small naturally occur-
ring aw then the associated da/dN versus DK curve to be used in any damage tolerance analysis/assessment is the
curve ABD.

It thus follows that the curve to be used in any damage tolerant analysis/assessment of crack growth in operational air-
craft where the crack has initiated from small naturally occurring discontinuities should be the curve ABD. It is also clear that
in this instance using the curve CBD, which corresponds to that obtained using ASTM standard E647 tests, would result in a
non-conservative estimate of the fatigue life of the structure. From this it follows that for composite repairs [3840] to
cracked structures whereby the crack has grown from a small naturally occurring discontinuity and neither the crack nor

3
This does not detract from, nor undermine, the need in a damage tolerant design to ensure that the associated damage tolerant analysis assumes initial
defects as specied in the USAF Damage Tolerance Design Handbook [7]. Nor does it detract from the fact that such small naturally occurring defects cannot be
detected using currently available NDI tools. As such from a design perspective, this size aw must be assumed to be present at all points in the airframe.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 157

the material in front of the crack has been removed, i.e. the crack tip has not been stop drilled, then the curve to be used in
any analysis should be the curve ABD.
No discussion on the da/dN versus DK relationship associated with the growth of cracks from small naturally occurring
discontinuities in operational aircraft would be complete without discussing R ratio and crack closure effects and the clas-
sication of cracks into micro-structurally short, physically short and long cracks. In the present paper this is done in Appen-
dix D. From the examples presented in Appendices B and D it would appear that cracks in eet aircraft that grow from small
naturally occurring defects may well experience little if any R ratio effects at any stage of their life. Furthermore, the example
given in Appendix B illustrates that the use of crack closure based equations to predict crack growth from small initial defects
under operational loading can result in very un-conservative estimates of the crack length history. Even when the initial
crack depth (in this example) was taken to correspond to that mandated in the USAF damage Tolerance Design Handbook
[7], viz: 1.27 mm, the FASTRAN analysis still gave a very non-conservative estimate for the remaining life. From this it fol-
lows that the use of crack closure based crack growth equations to model the growth of small naturally occurring defects in
operational aircraft and therefore also the effect of composite repairs on cracks, that have grown from small naturally occur-
ring defects in operational aircraft, is questionable.
Having introduced the topic of composite repairs it should be noted that these ndings have another implication with
respect to the assessment of the use of composite patches [3840] to repair long cracks in eet aircraft where the crack
has grown from a small naturally occurring defect. (In such instances the crack is generally not removed and the tips of
the cracks are generally not stop drilled prior to repair/patching.) In such cases we have shown that the fatigue response
of the repaired structure will lie on curve ABD.
Let us assume that prior to patching the stress intensity factor associated with the (unpatched) crack lies at point E in
Fig. 7, which is to the right of DKB, on the da/dN versus DK curve. (This assumption may not be true for all loads in the ight
load spectrum.) The effect of patching long cracks, where the crack has grown from a small naturally occurring defect, is to
(generally) signicantly reduce the stress intensity factor, see [38]. In the case of the boron epoxy repair to Mirage III [38] DK
was reduced by approximately 90%. Hence it is to be expected that after patching long cracks the stress intensity factor range
DKp is such that the crack growth rate, at the corresponding point in the ight load cycle spectra, will be at point F on the
curve ABD, see Fig. 7. However, it is common practice [39] to use ASTM E647 like specimens, see Fig. 9 which presents a typ-
ical test specimen geometry [39,40], which generally have an initial (starter) notch that is greater than 5 mm [39,40] to eval-
uate the effect of a patch on crack growth. The use of such specimens, where the length of the initial notch from which the
crack grows is greater than 5 mm, will generate crack growth data where the crack will behave as if it is a long crack (see
Section 5.3.1 in the USAF Damage Tolerant Design Handbook [7]). This will result in a crack growth rate that corresponds to
point G, which has the same DK (=DKp) as point F, that lies on the curve CBD. From this it follows that tests performed using
specimens resembling those outlined in the ASTM standard, i.e. with cracks growth from an articially induced starter notch
greater than 5 mm, to evaluate the effect of a composite repair on cracks, that have grown from small naturally occurring
defects in operational aircraft, will produce crack growth data that lie on the curve CBD, rather than on curve ABD, and
as such will signicantly overestimate the effectiveness of the repair system to cracks in the eet.
A similar argument holds for fatigue tests performed to certify supersonic particle deposition (cold spray) repairs to eet
aircraft [41]. Indeed, [42] presented a related argument for specimen tests, and the associated analysis tools, performed to
assess the durability of adhesive bonds used in composite repairs to cracks in operational aircraft. (The effect of small nat-
urally occurring defects in the adhesive on composite repairs to eet aircraft was rst highlighted in [43].)

Single edge notch crack

in the underlying plate

y
panel
centreline
y

composite patch
bonded over the
crack

Fig. 9. Typical specimen geometry used to assess composite repairs to cracked metallic structures.
158 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

4. Conclusions and recommendations

This paper has shown that the experimentally obtained da/dN versus DK curves associated with short and long cracks lead
to the conclusion that the damage tolerance assessment of operational aircraft structures, where cracking has grown from
small naturally occurring defects, should use an amalgam of the experimentally obtained long and short crack growth data.
From this it follows that using the da/dN versus DK curves obtained via ASTM E647 like tests can result in a non-conservative
estimate of the fatigue life of the structure/modication. We have also shown that the existing test procedures used to eval-
uate the ability of composite patches to repair long cracks, that have grown from small naturally occurring defects, overes-
timate their effect. We also see that the cracks that grow from small naturally occurring defects sees little if any R ratio
effects and that the use of crack closure based crack growth equations to model crack growth from small etch pits can yield
very non-conservative estimates of the crack length history. As such given the nature of the initiating aws outlined in
[1,2,5,6] it follows that the use of crack closure based crack growth equations to model the growth of cracks from small nat-
urally occurring discontinuities in operational aircraft and the effect of composite repairs on cracks, that have grown from
small naturally occurring defects in operational aircraft, is questionable.

Acknowledgement

The authors would like to acknowledge the nancial assistance given by the Aircraft Structural Integrity Directorate
General Technical Airworthiness (ASI-DGTA) and in particular to Dr. Madabhushi Janardhana, Technologies ASIP Manager
(ASI3B), and Sqn. Ldr. Nathan Tate, OIC ASI1. The authors would also like to acknowledge the many fruitful and instructive
talks with Dr. Simon Barter and Lorrie Molent at DSTO and Dr. N. Iyyer TDA Inc. (USA) for his support and for supplying the
P3C crack growth data.

Appendix A. The HartmanSchijveMcevily crack growth equation

It is commonly thought that, for long cracks, the increment in the crack length per cycle, da/dN, can be related to DK and/
or the maximum stress intensity factor Kmax. This approach was rst suggested in 1961 by Paris, Gomez and Anderson [50],
who related (da/dN) to the maximum stress intensity factor Kmax. This work subsequently led to the well known Paris
equation:
da=dN CDKm A1
where the values of C and m are experimentally obtained, and are considered to be a constant for a particular material and
environment. Forman et al. [51] subsequently extended the Paris equation to account for Kmax approaching its fracture
toughness Kc, viz:
da CDKm
A2
dN 1  RK c  DK
where R = Kmin/Kmax. In 1970 Hartman and Schijve [18] suggested that da/dN should be dependent on the amount by which
DK exceeds the fatigue threshold DKth, i.e. on DK  DKth. This led them to develop a modication to the Forman equation,
viz:

da D0 DK  DK th b
A3
dN 1  RK c  DK
where D0 and b are constants. Liu and Liu [19] subsequently derived a crack growth equation which revealed that da/dN
should always be proportional to (DK  DKth)2 and suggested that the constant of proportionality was approximately
0.036/Ery, where E and ry are the Youngs modulus and the yield stress of the material respectively. An exponent b of
approximately 2 was also found for short crack growth in a weldable-structural CMn steel by Smith and Smith [52].
Ramsamooj [20], Pandey and Chand [21], Wang [22] and Chen et al. [53] also derived crack growth equations that related
da/dN to (DK  DKth)2. Ramsamooj [20] reported that this relationship held for both long and short cracks. Ishihara [23] also
revealed that for small near micron size defects da/dN was linearly related to (DK  DKth)2. The work of Wang and Zheng [54]
suggests that, for materials that exhibit crack closure effects [55], this formulation can also be rewritten in terms of
(DKeff  DKeff,th)2, where DKeff = Kmax  Kop and Kop is the value of the stress intensity factor at which the crack rst opens
and DKeff,th is the apparent threshold value associated with the da/dN versus DKeff curve. In such cases this formulation,
i.e. whereby da/dN is related to (DK  DKth)2, resembles that of McEvily and co-workers [8,32,33,57] which relates da/dN
to (DKeff  DKeff,th)2 and which has also been shown to hold for both short and long cracks. Here it should be noted that DKeff
can also be written as DKeff = DK  DKop where DKop = Kop  Kmin and that [8,32,33,56,57] have explained that DKop is an
exponential function of the crack length, viz:

DK op 1  eka DK opl A4
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 159

where DKopl is the long crack value of DKop and k is a material dependent constant (typically 10,000 m1). Thus for cracks
that initiate from small naturally occurring discontinuities DKop asymptotes to zero. Hence for cracks that initiate from small
naturally occurring discontinuities in operational aircraft DKeff asymptotes to DK. Consequently, for small naturally occur-
ring initial defects the form of the crack growth equations proposed in [8,32,33,56,57] resembles Eq. (1), viz:

da=dN DDjb A5

where the crack driving force, Dj, is taken to be

p
Dj DK  DK thr = 1  K max =A A6

where D and A are constants, DKthr is the apparent fatigue threshold and b is a constant which is generally approximately 2
[13,15]. (Note that for cracks growing from small naturally occurring defects DKthr is (essentially) independent of the R ratio,
p
see Appendix D and has a very small value, typically 0.1 MPa m.) This particular variant of the HartmanSchijve equation,
which has been shown to hold for a wide range of aerospace metals [13,15] as well as for delamination growth in composites
[15], the growth of cracks from etch pits at the bore of a fastener hole [58], and crack growth in Boeing wedge tests [42], has
the advantage that it allows the constant b to be determined directly from the slope of the da/dN versus Dj curve. It was also
shown [13] that the Liu and Liu [19] approximation for D, viz: D = 0.036/Ery, was a reasonable rst estimate for a wide range
of materials.
We have previously mentioned that for many aluminium and titanium alloys and steels the exponent b is a constant is
generally approximately 2 [1315]. In such cases for crack growth from small naturally occurring defects this equation has
only three parameters, viz: D, A and DKthr (which is essentially independent of the R ratio for small naturally occurring
cracks, see Appendix D). For long cracks in materials for which the formulation can be expressed as per Eq. (5) there are also
only three parameters, viz: D, A and DKthr,eff. If you allow the exponent b to vary there are then four parameters in both cases.
In this context it is informative to note that the crack growth equation used in FASTRAN to represent crack growth has ve
unknown parameters.
The ability of the HartmanSchijveMcEvily crack growth equation, i.e. Eq. (5) and Eq. (1) its short crack equivalent/sim-
plication, to model the growth of both short and long cracks is illustrated in Appendix B, for crack growth from an etch pit
at the bore of a fastener hole, and in Appendix C, for crack growth in a centre cracked panel under load spectra representative
of two fatigue critical locations (FCA-301 and FCA-351) in the P3C wing, where it is compared with predictions made using
FASTRAN. In each case the computed crack length histories are in better agreement with the experimental data that those
obtained using FASTRAN.
The crack growth program FASTRAN was chosen for this comparative study since, as stated in [35], of the three commer-
cially available crack growth programs AFGROW, NASGRO and FASTRAN it gave results that were the closest to the measured
crack growth data. It should also be noted that for the P3C problem both the FASTRAN and the HartmanSchijveMcEvily
crack growth equation, i.e. Eq. (5) which for small cracks reduces to Eq. (1), both needed to be calibrated to model the crack
length history for control point FCA-301. However, unlike the FASTRAN analysis presented in [35] which needed to be sep-
arately calibrated for control point FCA-351 the HartmanSchijveMcEvily crack growth equation gave predictions which
were in excellent agreement with the measured data without the need to further adjust any of the input parameters.
For the problem discussed in Appendix B, which involves crack growth from a small etch pit, the FASTRAN analysis gave
results that were highly non-conservative. This contrasts with the HartmanSchijveMcEvily crack growth equation which
gave results that were in good agreement with the measured data. Furthermore, even if the initial crack depth was taken to
correspond to that mandated in the USAF damage Tolerance Design Handbook [7], viz: 1.27 mm, the FASTRAN analysis still
gave very non-conservative estimates for the remaining life.
At this point it should be noted that although the present examples have used linear elastic fracture mechanics (LEFM) to
compute the stress intensity factors there may be instances when the plastic zone is sufciently large that elasticplastic
fracture mechanics characterisation will be needed. That said the ability of the present LEFM based formulation to accurately
predict the crack length history is aptly illustrated by the examples presented in Appendices B and C. Furthermore, its ability
to represent crack growth in a range of coupon tests is illustrated in Figs. 3, D1 and D2.

Appendix B. Crack growth from an etch pit at a hole under an operational load spectra

To illustrate how Eq. (1) can be used, in conjunction with the data obtained from short crack tests, to predict the growth of
small cracks under complex in-ight load spectra [58] studied the growth of surface cracks from small etch pits at a fastener
hole under a measured RAAF F/A-18 operational load spectrum with a peak (net section) stress of 155 MPa [5,59]. The spec-
imen geometry used in these tests is shown in Fig. B1. These coupons were surface etched to simulate the surface conditions
that are found under the ion vapour deposited (IVD) aluminium coating of the aluminium alloy 7050-T7451 F/A-18 compo-
nents. The etch solution was 50% (volume) nitric acid (HNO3, 70%) and 1% volume percent Hydrouoric acid (HF, 70% Tech-
nical grade), remainder tap water, with an etching time of 5 min at room temperature, see [59] for more details.
Fig. 2 presents the computed crack depth histories obtained in [58] using Eq. (1) with the values of the constants given in
p
Section 2 viz: b = 2 and D = 7.0  1010 and DKthr = 0.1 MPa m. Here we see good agreement between the measured crack
growth histories. In the case of specimen test KK1H353 the initial defect was approximately 0.018 mm deep, whilst in the
160 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

case of specimen test KK1H348 the initial defect was approximately 0.014 mm deep. The result obtained using FASTRAN,
taking the R = 0.7 small crack da/dN versus DK data as the da/dN versus DKeff curve, to compute the crack length history
for specimen test KK1H353 is also shown in Fig. B2. In this instance the computed crack length history is very nonconser-
vative. To address the question of what errors, if any, would be associated with the computed crack length history if the ini-
tial crack depth corresponded to that mandated in the USAF damage Tolerance Design Handbook [7], viz: 1.27 mm, the
FASTRAN analysis was repeated assuming a 1.27 mm deep initial crack. The results of this analysis are also presented in
Fig. B2 where we again see that the FASTRAN analysis gives a non-conservative estimate for the remaining life (from a depth
of 1.27 mm). It should be noted that this analysis again used the R = 0.7 small crack da/dN versus DK data as the da/dN versus
DKeff curve. Had the long crack da/dN versus DKeff been used the resultant estimate for the remaining life would have been
even more non-conservative since for any given DKeff the associated long crack growth rates are lower. This is because for
any given DK the R = 0.7 long crack curve gives a value of da/dN that is much less than the value associated with the R = 0.7
short crack curve, see Fig. 2.
It should be noted that the crack growth curves in Fig. B2 show a kink associated with a crack depth of approximately
2.3 mm. This kink corresponds to the transition from a three dimensional crack to a through-the-thickness crack. The large
(nonconservative) error resulting from the use of FASTRAN to predict crack growth reinforces the remark in Section 5.3.1 of
the USAF Damage Tolerant Design Handbook [7] that predicting the growth of small cracks from holes using data obtained
from small surface defects can, depending on the crack growth equation used, lead to large errors.

Fig. B1. The open hole specimen geometry, from [59].

10
Crack depth (mm)

KK1H353
KK1H348
Computed KK1H348
Computed KK1H353
0.1 FASTRAN KK1H353
FASTRAN 1.27 mm

0.01
0 25 50 75 100 125 150
Load Blocks

Fig. B2. Measured [5] and computed crack depth histories under a RAAF operational load spectrum where one block represents approximately 324 ight
hours.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 161

Appendix C. Crack growth under an operational load spectrum

This section compares predictions made using FASTRAN [35] and the HarmanSchijveMcEvily crack growth equation,
i.e. Eq. (5), for crack growth in a 177.8 mm  96.5 mm  2.29 mm (thick) 7075-T6 centre cracked specimen tested under
spectra supplied by the US Navy. The load spectra are associated with two fatigue critical locations, viz: FCA-351 and
FCA-301, in P-3C aircraft in service with the US Navy, Canada, New Zealand, the Netherlands and the Australian Air Force.
Location FCA-301 corresponds to a point at the lower front spar adjacent to the wing root of the P-3C aircraft and FCA-
351 represents a location at the lower front spar inboard area near the inboard engine, see [35] for more details. Location
FCA-301 experiences mainly tension-dominated loading with a mean R  0.11, whilst FCA-351 experiences a high degree
of compressive loading, see [35] for more details.
Ref. [35] presented predictions obtained using FASTRAN for spectra representing an 85th and a 50th percentile mission
prole usage. In the FASTAN analysis reported in [35] the da/dN versus DKeff relationship used was taken from NASGROs
7075-T6 extrusion material data, see [35]. FASTRAN has the ability to vary the constraint factor a during the crack growth
analysis and it is common practice for this to be done so as to calibrate the input data and thereby ensure that the analysis
best represents the experimental data [35]. The values of the constraint factors a1 and a2 used in [35] to estimate the crack
opening stress intensity factor Kop and the associated crack growth rates r1 and r2 are shown in Table C1. In [35] it was found
that in order to obtain a reasonable correlation with the experimental crack length versus cycles histories associated with
these tests the FASTRAN analysis needed to use constraint values of a1 and a2, and crack growth rates r1 and r2 for location
FCA-351 that differed from those used to analyse FCA-301. The FASTRAN analysis using the rst set of values was labelled TC
and the FASTRAN analysis using the second set of values was labelled TD, see [35] and Table C1.

1.0E-02
7075-T6 R = 0.7 [52]
1.0E-03
7075-T6 [35]

1.0E-04 NASGRO R = 0.5

da/dN (in/cycle)

NASGRO R = 0.5
1.0E-05

1.0E-06

1.0E-07

1.0E-08
y = 8.70E-08x2.05
R2 = 9.79E-01
1.0E-09

1.0E-10
0.1 1 10 100 1000
thr)/(1-Kmax/A)1/2 (ksi in)
(

Fig. C1. Crack growth in 7075-T6.

100

FASTRAN
Half Crack Length, mm

Test Data
Calibration

10

FCA 301, 85th Percentile

Variable Ampltidue, Tension Dominated (TD)
Spectrum
583,806 Cycles/15,000 Hours
Maximum Stress: 219.27 MPa
Minimum Stress: -46.31 MPa
1
0 20000 40000 60000 80000
Cyclic Test Hours

Fig. C2. Comparison of measured and predicted results for fatigue critical location 301 tested under an 85% spectrum.
162 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

The analysis performed in this paper used the HartmanSchijveMcEvily (HSM) crack growth equation, i.e. Eq. (5). As
such we rst needed to determine the associated constants. Examining the da/dN versus DK data presented in [35], and also

Table C1
Constraint factors and rates used in the FASTRAN analysis [35].

Values used for FCA-361 and FCA-351 (TC) Values used for FCA-3O1 (TD) Values used in the present analysis
a1 =1.8 for r1 < 2.54 10-4 mm/cycle a1 =1.8 for r1 < 7.62 10-4 mm/cycle a1 = 1.8 for r1 < 2.54 10-4 mm/cycle
a2 =1.2 for r2 > 1.78 10-2 mm/cycle a2 =1.2 for r2 > 3.81 10-2 mm/cycle a2 =1.2 for r2 > 1.78 10-2 mm/cycle

100.0
660,677 Cycles/15,000 Hours
Maximum Stress: 170.78 MPa
Minimum Stress: -91.19 MPa
Half Crack Length,mm

10.0

FASTRAN TD

FASTRAN TC

Test Data

HSM Prediction

1.0
0 5000 10000 15000 20000
Cyclic Test Hours

Fig. C3. Comparison of measured and predicted results for fatigue critical location 351 tested under an 85% spectrum.

100.0
Half Crack Length,mm

10.0 640,090 Cycles/15,000 Hours

Maximum Stress: 165.29 MPa
Minimum Stress: -84.93 MPa

FASTRAN TD

FASTRAN TC

Test Data

HSM Prediction

1.0
0 2000 4000 6000 8000 10000 12000 14000
Cyclic Test Hours

Fig. C4. Comparison of measured and predicted results for fatigue critical location 351 tested under a 50% spectrum.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 163

p p
in the NASGRO and FAA [51] data bases, gave values of D = 8.7  108, DKeff,thr = 0.65 ksi in and A = 101 ksi in, see Fig. C1.
(Note that imperial units have been used since these were the units used in [35].) Since the authors believed that the rela-
tionship used to determine the crack opening stress Kop should be the same regardless of whether the spectra is tension
dominated (TD) or a mixture of tension and compression loading (TC) the present study used a single representation for both
control points (i.e. locations). Given that the initial crack was only approximately 2 mm long an appropriate value of DKeff,thr
p
needed to be determined. A value of DKeff,thr = 0.18 ksi in was needed to match the experimental data for control point FCA-
301, see Fig. C2 where the associated computed curve is termed Calibration. The values of the constrain factors a1 and a2,
and rates r1 and r2 used in this calibration are given in Table C1. The resultant predicted crack length versus test hours curves
for control point FCA-351 are shown in Figs. C3 and C4. In both cases we see that the HartmanSchijveMcEvily crack growth
equation gave results that were in excellent agreement with the measured data. We also see that, in contrast to the FASTRAN
predictions presented in [35], Eq. (5) predicts a crack length histories that are in excellent agreement with the measured
data. Indeed, the results from the present study contrast with the results obtained in [35] using FASTRAN and the TD rep-
resentation, developed in [35] for control point FCA-301, to represent crack growth under load spectra representative of con-
trol point FCA-351. To overcome the initial discrepancy between the FASTRAN analysis and the measured crack growth
history associated with FCA-351 [35] changed the way in which the constraint factor a was varied, see Table C1. This
was not done in the present analysis. Nevertheless, as can be seen in Figs. C2C4 and C3, the HartmanSchijveMcEvily pre-
dictions, which use fewer adjustable constants, are in better agreement with the experimental measurements than the re-
sults obtained in [35] using FASTRAN and either the TC or the TD representation where the values of a were changed so as to
improve the t for each control point.

Appendix D. R Ratio effects and crack growth from small cracks in operational aircraft

The experimental results presented in [10,13] and shown in Fig. 2 reveal that the growth of cracks from small, in this case
approximately 7 lm deep, initial defects in 7050-T7451 have little (if any) R ratio dependency for any crack length. This phe-
nomena, i.e. that for materials that do not exhibit a Marci effect there is little R ratio effect or crack closure associated with
the growth of small fatigue cracks, has been reported in a number of independent investigations [3,10,28,4449,33]. For
example Caton et al. [28] who studied the growth of small cracks in Ti6Al4V, stated:
A systematic study of the effects of stress ratio on small fatigue crack growth in Ti6Al4V was conducted. . . .no discern-
ible effect (on the crack growth rate) of R is seen when plotting as a function of DK.

Cadario and Alfredsson [46], who studied the growth of small cracks in Ti-17, stated:
Fatigue growth rates were not inuenced by the R ratio during growth from short to long cracks with a = 0.053.6 mm.
Such R-independence would be expected if propagation was free from crack closure.

Cadario and Alfredsson [46] also stated that in this test program the cracks were short in a physical and micro-structural
sense. Thus, as in [10], in the tests discussed in [46] the cracks grew from being micro-structurally short to physically short
and nished as long cracks. However, allowing for experimental error, at no stage in these tests do you see a signicant R
ratio dependency develop, see Figs. 2 and D1. (In Fig. D1, which is taken from [46], the long crack closure free line is a long
crack curve obtained from (high) Kmax tests.) As a result Cadario and Alfredsson [46] stated that at no stage in these tests does
closure develop. Furthermore, as can be seen in Fig. D1, allowing for experimental error, the da/dN versus DK curves essen-
tially form a single straight line on this loglog graph, i.e. there is a near power law relationship between da/dN and DK

Fig. D1. Crack growth in Ti-17, from [46].

164 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

1.0E-02

1.0E-03

1.0E-04 R = 0.1
R = 0.3
1.0E-05 R = 0.5

da/dN (m/cycle)
R = -1
1.0E-06

1.0E-07

1.0E-08

1.0E-09

1.0E-10

1.0E-11
1 10 100
K MPa m

Fig. D2. Crack growth data, from [45].

which is (essentially) R ratio independent. Thus as the cracks grew and transitioned from being micro-structurally short to
physically short and end up as long cracks there is little change in the observed macroscopic relationship between da/dN
versus DK which shows little R ratio effects and hence little roughness or plastic wake induced crack closure.
To further illustrate this Fig. D2 presents the corresponding Figure from Alaoui et al. [45], who studied the growth of small
cracks in S355NL steel. As a result [45] stated:
These results conrm that load ratio does not have a real inuence on the short fatigue crack propagation, while check-
ing propagation rate versus DK. This result is fundamental to assume that short cracks are free from closure effects.

In this context it should be noted that: Lados et al. [3], who studied the growth of small cracks in AlSiMg cast alloys,
stated:
For small crack growth, both residual stress and microstructure related closure effects, when present, are insignicantly
small.

Peters et al. [44], who studied the growth of small cracks in AlSiMg cast alloys, in Ti6Al4V, stated:
Specically, comparable growth-rate data of microcracks at low (R = 0.1) and moderate (R = 0.5) load ratios indicate that
mean stresses play little role in small-crack propagation.

Hence from the above examples and the example presented in Appendix B it would thus appear that cracks in eet air-
craft that grow from small naturally occurring defects may well experience little if any R ratio effects and hence little rough-
ness or plastic wake induced crack closure. As such it follows that DKthr will see little if any R ratio dependency.

Appendix E. A brief discussion of the effect of Kmax on the fatigue threshold

Numerous researchers have shown that there is an inuence of Kmax that is separate from the plastic wake induced crack
closure mechanism, see [2529,6062]. Yamada and Newman [60] suggested that this was possibly due to roughness-in-
duced crack closure (RICC) and/or debris-induced crack closure (DICC). On the other hand Paris and co-workers [61,62] have
explained that experimental studies have revealed that the region behind the crack tip never closes so that cracks only ever
exhibit partial closure and the crack tip always has a nite radius. (This conclusion, i.e. that the region behind the tip al-
ways remaining open, is supported by the experimental work presented by Jones and Pitt [63].) If true then this could offer a
possible alternative explanation for the observed Kmax dependency.

References

[1] Molent L, Barter SA, Wanhill RJH. The lead crack fatigue ling framework. Int J Fatigue 2011;33:32331.
[2] Schijve J. Fatigue life until small cracks in aircraft structures: durability and damage tolerance. Advanced structural integrity methods for airframe
durability and damage tolerance. In: Harris CE, editors. Proceedings FAA/NASA international symposium, Hampton, 1994, NASA Conference Publication
3274; 1994. p. 66581.
 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166 165

[3] Lados DA, Apeliana D, Paris P, Donald JK. Closure mechanisms in AlSiMg cast alloys and long-crack to small-crack corrections. Int J Fatigue
2005;27:146372.
[4] Jones R, Molent L, Pitt S. Crack growth from small aws. Int J Fatigue 2007;29:165867.
[5] Molent L, Sun Q, Green AJ. The F/A-18 fatigue crack growth data compendium, DSTO-TR-1677, Melbourne, Australia; 2005.
[6] Merati A. A study of nucleation and fatigue behavior of an aerospace aluminum alloy 2024-T3. Int J Fatigue 2005;27:3344.
[7] Miedlar PC, Berens AP, Gunderson A, Gallagher JP. Analysis and support initiative for structural technology (ASIST), AFRL-VA-WP-TR-2003-3002.
[8] McEvily AJ, Eier D, Macherauch E. An analysis of the growth of short fatigue cracks. Eng Fract Mech 1991;40:57184.
[9] Venkateswara Rao KT, Yu W, Ritchie R. On the behavior of small fatigue cracks in commercial aluminumlithium alloys. Eng Fract Mech
1988;31(4):62335.
[10] Jones R, Barter S, Chen F. Experimental studies into short crack growth. Eng Fail Anal 2011;18(7):171122.
[11] Forman RG, Shivakumar V, Cardinal JW, Williams LC, McKeighan PC. Fatigue crack growth data base for damage tolerance analysis. DOT/FAA/AR-05/15,
Washington, USA; 2005.
[12] Sharp PK, Byrnes R, Clark G. Examination of 7050 fatigue crack growth data and its effect on life prediction, DSTO-TN-0729, Australia; 1998.
[13] Jones R, Molent L, Walker K. Fatigue crack growth in a diverse range of materials. Int J Fatigue 2012;40:4350.
[14] Zheng X, Hirt M. Fatigue crack propagation in steels. Eng Fract Mech 1983;18(5):96573.
[15] Jones R, Brunner AJ, Hui D, Pitt S. Application of the HartmanSchijve equation to represent Mode I and Mode II fatigue delamination growth in
composites. Compos Struct 2012;94(4):134351.
[16] Iyyer NS, Dowling NE. Fatigue growth and closure of short cracks. AFWAL-TR-89-3008; 1989.
[17] Forth SC, James MA, Newman JA, Everett RA. Jr., Johnston WM. Jr. Mechanical data for use in damage tolerance analyses. NASA/TM-2004-213503 and
ARL-TR-3375; 2004.
[18] Hartman A, Schijve J. The effects of environment and load frequency on the crack propagation law for macro fatigue crack growth in aluminium alloys.
Eng Fract Mech 1970;1(4):61531.
[19] Liu HW, Liu D. A quantitative analysis of structure sensitive fatigue crack growth in steels. Scripta Metall 1984;18:712.
[20] Ramsamooj DV. Analytical prediction of short to long fatigue crack growth rate using small- and large-scale yielding fracture mechanics. Int J Fatigue
2003;25:92333.
[21] Pandey KN, Chand S. Fatigue crack growth model for constant amplitude loading. Fatigue Fract Eng Mater Struct 2004;27(6):45972.
[22] Wang R. A fracture model of corrosion fatigue crack propagation of aluminum alloys based on the material elements fracture ahead of a crack tip. Int J
Fatigue 2008;30:137686.
[23] Ishihara S, Saka S, Nan ZY, Goshima T, Sunada S. Prediction of corrosion fatigue lives of aluminium alloy on the basis of corrosion pit growth law.
Fatigue Fract Eng Mater Struct 2006;29(6):47280.
[24] Newman Jr JC, Ruschau JJ, Hill MR. Improved test method for very low fatigue-crack-growth-rate data. Fatigue Fract Eng Mater Struct 2011;34:2709.
[25] Boyce BL, Ritchie RO. Effect of load ratio and maximum stress intensity on the small crack fatigue threshold in Ti6Al4V. Eng Fract Mech
2001;68:12947.
[26] John A, Newman, Mark A. James, William M, Johnston, Dy D. Le. Fatigue crack growth threshold testing of metallic rotorcraft materials, NASA/TM-
2008-215331 and ARL-TR-4472; July 2008.
[27] Sheldon JW, Bain KR, Donald JK. Investigation of the effects of shed-rate, initial Kmax, and geometric constraint on DKth in Ti6Al4V at room
temperature. Int J Fatigue 1999;21:73341.
[28] Caton MJ, John R, Porter WJ, Burba ME. Stress ratio effects on small fatigue crack growth in Ti6Al4V. Int J Fatigue 2012;38:3645.
[29] Tesch A, Pippan R, Trautmann KH, Doker H. Short cracks initiated in Al 6013-T6 with the focused ion beam (FIB)-technology. Int J Fatigue
2007;29:180311.
[30] Chan KS. Variability of large-crack fatigue-crack-growth thresholds in structural alloys. Metall Mater Trans A 2004;35A:372135.
[31] Vasudevan AK, Sadananda K, Louat N. Two critical stress intensities for threshold fatigue crack propagation. Scripta Metall Mater 1993;28:6570.
[32] Endo M, McEvily AJ. Prediction of the behaviour of small fatigue cracks. Mater Sci Eng A 2007;468470:518.
[33] Kondo Y, Endo M, McEvily AJ. Short crack growth behavior and its relation to notch sensitivity and VHCF. In: Proceeding 17th European conference on
fracture, 2nd5th September, Brno, Czech Republic; 2008. p.15967.
[34] Kanninen M, Atkinson C, McEvily AJ. The development of a general fatigue crack growth predictive model. Int J Fract 1977;13:88791.
[35] Iyyer NS, Kwon YS, Nam Phan, P-3C crack growth life predictions under spectrum loading, In: 2003 International committee on aeronautic fatigue,
Lucerne, Switzerland, May 59, 2003.
[36] Piascik RS. The growth of small corrosion fatigue cracks in alloy 7075; 2004 <http://www.nasa.gov>.
[37] Li D, Burns J, Bush R, Garratt M, Hinkle A, Mills T, et al. Characterization of corrosionfatigue interaction for aluminum alloy substitution. In: Young RD,
editor. Proceedings of the 8th joint NASA/FAA/DoD conference on aging aircraft; 2005. <http://www.JCAA.US/AA_Conference>.
[38] Baker AA, Callinan RJ, Davis MJ, Jones R, Williams JF. Application of B.F.R.P. patching to Mirage III aircraft. Theor Appl Fract Mech 1984;2(1):116.
[39] Baker AA, Jones R. Bonded repair of aircraft structure. The Hague: Martinus Nijhoff Publishers; 1988.
[40] Jones R. A scientic evaluation of the approximate 2D theories for composite repairs to cracked metallic components. Compos Struct
2009;87(2):15160.
[41] Jones R, Matthews N, Rodopoulos CA, Cairns K, Pitt S. On the use of supersonic particle deposition to restore the structural integrity of damaged aircraft
structures. Int J Fatigue 2011;33(9):125767.
[42] Jones R, Pitt S, Peng D. Characterization of the durability of adhesive bonds. Fatigue Fract Eng Mater Struct 2012. http://dx.doi.org/10.1111/j.1460-
2695.2012.01688.x.
[43] Chalkley P, Geddes R. Service history of the F-111 wing pivot tting upper surface boron/epoxy doublers. DSTO-TN-0168. Melbourne
(Australia): Department of Defence; 1998.
[44] Peters JO, Boyce BL, Chen X, McNaney JM, Hutchinson JW, Ritchie RO. On the application of the KitagawaTakahashi diagram to foreign-object damage
and high-cycle fatigue. Eng Fract Mech 2002;69:142546.
[45] El Malki AlaouiA, Thevenet D, Zeghloul A. Short surface fatigue cracks growth under constant and variable amplitude loading. Eng Fract Mech
2009;76:235970.
[46] Cadario A, Alfredsson B. Fatigue growth of short cracks in Ti-17: experiments and simulations. Eng Fract Mech 2007;74:2293310.
[47] Shyama A, Lara-Curzio E. A model for the formation of fatigue striations and its relationship with small fatigue crack growth in an aluminum alloy. Int J
Fatigue 2010;32:184352.
[48] Ishihara S, Yoshifuji S, Mcevily Aj, Kawamoto M, Sawai M, Takata M. Study of the fatigue lifetimes and crack propagation behaviour of a high speed
steel as a function of the R value. Fatigue Fract Eng Mater Struct 2010;33:294302.
[49] Hudak SJ, Davidson DL. The dependence of crack closure on fatigue loading variables, mechanics of crack closure, ASTM STP 982. In: Newman JC, Elber
W, editors. Philadelphia (USA): American Society for Testing of Materials; 1988. p. 12158. ISBN 0-8031-0996-2.
[50] Paris PC, Erdogan F. Critical analysis of crack growth propagation laws. ASME Trans J Basic Eng 1963;85D(4):52834.
[51] Forman RG, Kearney VE, Engle RM. Numerical analysis of crack propagation in cyclic loaded structures. ASTM A. Mtg 1966. Paper no. 66 WA/Met. 4;
1966.
[52] Smith IFC, Smith RA. Fatigue crack growth in a llet welded joint. Eng Fract Mech 1983;18(4):8619.
[53] Liu Yan-Ping, Chen Chuan-Yao, Li Guo-Qing. Fatigue crack growth and control of 14MnNbq welding plates used for bridges. ASCE J Eng Mech
2012;138:305.
166 R. Jones, D. Tamboli / Engineering Failure Analysis 29 (2013) 149166

[54] Wang R, Zheng X. Corrosion fatigue crack propagation of an aluminum alloy under periodic overloads. Fatigue Fract Eng Mater Struct
2012;35(5):38998.
[55] Elber W. The signicance of fatigue crack closure, damage tolerance of aircraft structures. ASTM STP-486; 1971. p. 23042.
[56] Endo M, McEvily AJ. Prediction of the behaviour of small fatigue cracks. Mater Sci Eng A 2007;468470:518.
[57] Ishihara S, Yoshifuji S, Mc.Evily Aj, Kawamoto M, Sawai M, Takata M. Study of the fatigue lifetimes and crack propagation behaviour of a high speed
steel as a function of the R value. Fatigue Fract Eng Mater Struct 2010;33:294302.
[58] Jones R, Pitt S. Crack growth from etch pits in 7050-T7451 under combat aircraft spectra, paper 168. In: Proceedings corrosion & prevention 2012,
Australasian Corrosion Society, Crown Conference Centre, Melbourne, 11th14th November; 2012.
[59] Huynh J, Molent L, Barter S. Experimentally derived crack growth models for different stress concentration factors. Int J Fatigue 2008;30:176686.
[60] Yamada Y, Newman JC. Crack closure under high load-ratio conditions for Inconel-718 near threshold behaviour. Eng Fract Mech 2009;76:20920.
[61] Paris PC, Lados D, Tada H. Reections on identifying the real Keffective in the threshold region and beyond. Eng Fract Mech 2008;75:299305.
[62] Paris PC, Tada H, Donald KJ. Service load fatigue damage a historical perspective. Int J Fatigue 1999;21:S3546.
[63] Jones R, Pitt S. An experimental evaluation of crack face energy dissipation. Int J Fatigue 2006;28(12):171624.