International Journal of Fatigue 123 (2019) 157–167

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Fatigue crack growth in a laser shock peened residual stress field T

a b a c a
M. Pavan , D. Furfari , B. Ahmad , M.A. Gharghouri , M.E. Fitzpatrick
a
Coventry University, Priory Street, Coventry CV1 5FB, United Kingdom
b
Airbus Operations GmbH, Kreetslag 10, Hamburg 21129, Germany
c
Canadian Nuclear Laboratories, Chalk River, Ontario K0J 1J0, Canada

ARTICLE INFO ABSTRACT

Keywords: Laser Shock Peening is a surface treatment technique used in the aerospace sector to increase fatigue life, as well
Fatigue as resistance to fretting fatigue and stress corrosion cracking. In this study, laser shock peening was applied to a
Crack growth 6-mm-thick middle-crack tension specimen made of aluminium 2524-T351. Residual stress was measured with
Laser peening neutron diffraction and the contour method, along the predicted crack path prior to fatigue testing. Fatigue crack
Residual stress
growth test results showed that fatigue life improved by a factor of 4 compared to an untreated component,
Crack growth modelling
owing to a significant crack growth rate reduction inside the laser peened area. A linear-elastic finite-element
crack growth prediction model was also developed, obtaining predicted results in excellent agreement with the
experimental data.

1. Introduction initiation and the early stages of crack propagation [1,6]. Its application
to airframe components, characterised by potentially longer cracks, has
The use of laser shock peening has become increasingly popular not yet been thoroughly investigated.
owing to the decreasing cost of laser systems. The process has been The effect of residual stress on fatigue life and fatigue crack pro-
shown to be more effective and controllable than similar techniques pagation is of great importance in any engineering application and has
such as shot peening and deep rolling, especially in terms of residual been therefore widely studied since the development of Linear-Elastic
stress introduced, providing furthermore a better surface roughness. A Fracture Mechanics (LEFM). This approach allows the stress state near
complete review of laser shock peening is provided in [1], while the the crack tip to be described using linear equations, simplifying the
fatigue life performance improvements achieved using this surface effort to understand and predict crack propagation. Several techniques
treatment for different metals are reported in [2–5]. In the laser shock based on LEFM have been developed to deal with residual stress at the
peening process, a high energy laser pulse (1–100+ J) is fired at a crack tip, such as the superposition and the modified superposition
metal target surface, usually covered with an ablative layer, also called methods [7]. According to the superposition principle, the stress state
a sacrificial layer. As the pulse reaches the target surface, because of the due to two or more loads acting together is equal to the sum of each
high energy involved (several GW/cm2), the ablative layer vaporizes, load acting separately, provided that the material behaviour is linear-
creating a high-pressure plasma. The interaction of the high-pressure elastic [8]. In a cyclically-loaded residual-stress-bearing body, the
plasma with the metal surface generates a pressure pulse of several GPa. stresses due to applied mechanical loads are therefore linearly super-
In order to maximize the energy transferred in the process, the target imposed on the residual stresses. The superposition technique involves
surface is usually covered by a thin layer of flowing water, which the calculation of two separate stress intensity factors, one associated
confines the plasma between it and the metal surface. The pressure with the applied load, KL, and one associated with the residual stress,
pulse produced propagates as a shockwave within the metal, locally KRes [7]. These two values are then summed to obtain the total stress
overcoming the Hugoniot Elastic Limit and causing plastic deformation. intensity factor at the crack tip:
As a result, a compressive residual stress is generated to a depth that Kmax , TOT = Kmax , L + KRes (1)
depends on the process parameters but which can be several milli-
metres. Kmin, TOT = Kmin, L + KRes (2)
In the aeronautical sector today, laser shock peening is applied to
where Kmax,L and Kmin,L are the stress intensity factors at the maximum
specific critical engine components subjected to high loads, foreign
and minimum applied loads, respectively. The total stress intensity
object damage, and fretting fatigue, mainly to prevent or retard crack
factor range is calculated as follows:

E-mail address: [email protected] (M. Pavan).

https://doi.org/10.1016/j.ijfatigue.2019.01.020
Received 8 November 2018; Received in revised form 24 January 2019; Accepted 29 January 2019
Available online 31 January 2019
0142-1123/ © 2019 Elsevier Ltd. All rights reserved.
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

KTOT = Kmax, TOT Kmin, TOT = Kmax, L + KRes Kmin, L KRes Table 1
AA2524-T351 minimum tensile properties [20].
= Kmax, L Kmin, L (3)
Orientation Tensile Strength Yield Strength at 0.2% Elongation at failure
Eq. (3) shows that the stress intensity factor range does not depend [MPa] Offset [MPa] [%]
on the residual stress, since the KRes contribution cancels out. However,
L-T 427 276 15
the total stress ratio, RTOT, does depend on the residual stress:
Kmin, TOT Kmin, L + KRes
RTOT = =
Kmax , TOT Kmax, L + KRes (4)
and is therefore representative of the true conditions at the crack
tip.
The effect of crack face contact, not considered in the superposition
method, can be included in the stress intensity factor calculation using a
modified superposition approach, in which Kmin,TOT is set to 0 in Eq. (2)
whenever its value becomes negative, simulating crack closure. The
total stress intensity factor range in this formulation becomes:
KTOT = Kmax , TOT Kmin, TOT if Kmin, TOT > 0
KTOT = Kmax , TOT if Kmin, TOT 0 (5)
The total stress ratio is then:
Kmin, TOT
RTOT = Kmax , TOT
if Kmin, TOT > 0
RTOT = 0 if Kmin, TOT 0 (6)
Superposition-based techniques are commonly used owing to their
simplicity, but they have been criticised as being incapable of ac-
counting for the redistribution and relaxation of residual stress that
accompanies crack propagation [9–12]. Even though ([13]) the redis-
tribution and relaxation of stress associated with the presence of a crack
does not invalidate the superposition principle, the use of FE models to
calculate the residual stress intensity factor KRes is often favoured since
they naturally include these effects during crack growth. Fig. 1. Specimen dimensions in mm.
An alternative approach to superposition-based techniques, referred
to as Plasticity-Induced Crack Closure (PICC), was introduced in 1970 experimental results for propagation outside of the peened areas, where
by Elber [14], who noticed that cracks in fatigue specimens are fully the crack surfaces were only partially closed during the loading cycle.
open only for a part of the loading cycle, even when the loading cycle is A comparison between linear-elastic and elastic-plastic finite ele-
fully tensile. Elber argued that the closure phenomenon is an effect of ment modelling, with and without crack closure, was carried out by
the plastic deformation developed in the wake of the propagating crack, Garcia et al. in [18]. The authors studied the fatigue crack growth be-
resulting in a compressive residual stress field which in turn causes the haviour on a rectangular specimen made of AA7050-T7451 with re-
crack to close before the minimum load is reached, at a value of Kop sidual stress introduced by four point bending, testing with both posi-
instead of Kmin (Kop > Kmin), reducing the stress intensity factor range tive and negative stress ratios. The crack closure behaviour was
from K to K eff according to the following equation: simulated with a rigid surface which did not allow a negative dis-
K eff = Kmax , TOT K op (7) placement of the crack faces, i.e., the faces could not inter-penetrate.
Results highlighted how for a positive stress ratio the models with
where Kop represents the stress intensity factor at which the crack starts closure better approximated the test outcome in the compressive areas.
to open. Below this threshold, the crack faces are closed, therefore the The finite-element elastic-plastic model provided crack growth rate
part of the cycle for which the stress intensity factor is lower than Kop results that were closer to the experimental data but consistently un-
does not contribute to fatigue crack growth. The PICC approach re- derestimated, resulting in a non-conservative fatigue-life estimation.
quires the calculation of the opening stress intensity factor Kop, which is For the negative stress ratio, the linear-elastic approach without crack
generally achieved through elastic-plastic finite element analysis or by closure presented the best results.
using simpler analytical models such as Newman’s [15,16]. A recent study by Kashaev et al. [19] investigated the effect of laser
Both superposition-based and crack-closure-based techniques have shock peening on the microstructure and fatigue crack propagation
been successfully applied in the literature to predict crack growth rates behaviour of thin AA2024 CT specimens. The authors correlated the
in the presence of residual stress fields. A completely linear-elastic effect of the residual stress originated by the treatment with the Crack
approach was used by Parker [13], who achieved good agreement with Opening Displacement (COD) at different crack lengths. Results showed
experimental data using the modified superposition method for dif- how, compared to the baseline material, COD was increased or de-
ferent geometries and loading conditions; and by Hill and Kim [17], creased depending whether the residual stress field encountered was
who compared superposition, modified superposition, superposition tensile or compressive. Compressive residual stress caused the crack to
contact and a newly-developed superposition contact method. The au- open less, generating a beneficial crack closure effect.
thors considered a standard Compact-Tension (CT) specimen treated by In the current work, several techniques for predicting crack propa-
laser shock peening over a square area surrounding the initial notch, gation are compared: superposition, modified-superposition and plas-
calculating stress intensity factors and crack opening using weight ticity-induced crack closure based on Newman’s analytical equations
functions. The different methods yielded very similar predictions for (henceforth referred to as the Newman method) [15,16].
crack propagation within the laser peened area, where the effect of the
compressive residual stress was predominant. However, the newly-de-
veloped superposition contact method gave better agreement with the

158
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Table 2
Laser Shock Peening parameters.
LSP Power Density [GW/cm2] Energy[J] Spot Size [mm2] Pulse Duration [ns] Spot Offset Spot Overlap Layers

3–18-4 3 16 5.5 × 5.5 18 50% 5% 4

Table 3 Table 2.
Specimen designation, laser peening and loading conditions (AM = As Each layer of treatment was applied sequentially on both sides, to
Manufactured, LSP = Laser Shock Peened); BL signifies the baseline samples generate a uniform compression through the thickness and limit the
that were tested without peening. amount of geometric distortion. The full list of specimens tested is
Designation Condition RL σmax [MPa] presented in Table 3.

BL-1 AM 0.1 113

2.2. Residual stress measurements
BL-2 AM 0.1 113
LSP-1 LSP 0.1 113
LSP-2 LSP 0.1 113 The residual stress along the predicted crack path was determined
LSP-3 LSP with neutron diffraction [2] and with the contour method [21] on the
specimen LSP-3. Neutron diffraction measurements were carried out
using the L3 diffractometer of the Canadian Neutron Beam Centre
2. Experimental methods (CNBC) located in the NRU reactor of Canadian Nuclear Laboratories.
The aluminium 311 reflection was used for the lattice strain measure-
The five specimens used in this study were supplied by Airbus ments to obtain neutrons with a nominal wavelength of 1.727 Å, giving
Operations GmbH, Germany. They were made of a 2524-T351 clad a scattering angle close to 90°. The x-, y-, and z- components of lattice
aluminium alloy, commonly used for aircraft structural components strain were measured at 23 depths at three locations (Points 1–3) using
such as wing lower covers. Each specimen was extracted from a rolled a step size of 0.250 mm as shown in Fig. 2. In addition to these depth
aluminium panel, such that the longitudinal (stress) axis was parallel to profiles, thickness-averaged measurements of lattice strain were ob-
the rolling direction. Typical mechanical properties for this aluminium tained along Line 1 (Fig. 2) by oscillating the gauge volume con-
alloy are reported in Table 1. tinuously over the whole specimen thickness (6 mm) during each
Each specimen was 200 mm wide, 400 mm long and 6 mm thick, as measurement.
shown in Fig. 1. Only three of the five specimens were laser shock The contour method measurements were carried out at Coventry
peened, using the processing parameters described in the next section. University on the plane shown in Fig. 3. The specimen was cut in two
halves by EDM cutting and the surface displacements originated from
2.1. Laser Shock Peening the stress relief were measured with a Zeiss Contra G2 Coordinate
Measuring Machine. A point spacing of 0.2 mm was used in both di-
The Laser Shock Peening (LSP) treatment was carried out by Metal rections on the cut surface, with a spherical probe of 3 mm diameter.
Improvement Company, Earby, UK, using the parameters listed in The major benefit of the contour method is the production of a

Fig. 2. Neutron Diffraction measurement point locations, dimensions in mm, not to scale.

159
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

carried out prior to notch creation. All tests were carried out in la-
boratory air at ambient temperature (20 °C) according to ASTM stan-
dard E647 [22]. The fatigue loading conditions were the following:
σmax = 113 MPa, RL = 0.1, frequency = 10 Hz, with an initial crack
length including pre-cracking of 4 mm.

3. Numerical modelling

The numerical modelling strategy consisted of three different steps

as shown in Fig. 4. In the first step, the residual stress is introduced into
the specimen based on the neutron diffraction and contour method
results presented in Section 2. In the second step, fatigue crack growth
is simulated by advancing the crack and extracting the forces and dis-
placements at the crack front from the model. In the final step, the
extracted force and displacement information is used to calculate the
stress intensity factors and the fatigue life according to the crack growth
prediction models.
The first and second steps were performed using the Finite Element
(FE) software ABAQUS 6.14-1 [23], while the third step was performed
using MATLAB 2013b [24]. A 3D rather than a 2D Finite Element model
was preferred because (1) the specimen thickness (6 mm) is sufficient to
develop a 3D state of stress, and (2) it allowed better control over the
residual stress distribution introduced during the first step. To reduce
the number of elements and the computation time, only a quarter of the
overall specimen was simulated. Full integration eight-node linear brick
elements were used (C3D8), with a linear elastic material behaviour
based on the following parameters: Young’s modulus E = 73 GPa,
Fig. 3. Contour Method measurement plane.
Poisson’s ratio ν = 0.33. A mesh sensitivity analysis was carried out,
resulting in the mesh presented in Fig. 5.
complete 2D map of residual stress over the cut surface. The drawback,
however, is that only the component of residual stress normal to the 3.1. Residual stress input
cutting plane (Y direction in Fig. 3) is obtained. The contour method is
normally less accurate for the near surface results due to the errors The residual stress was introduced into the FE model through the
arising from the wire EDM cutting and data smoothing. In this study, an Abaqus user subroutine SIGINI, a FORTRAN program which initialises
effort was made to keep the power settings to a minimum value ob- the six components of stress for each 3D element to a predefined value.
taining therefore a good surface finish, which in turn decreased the With an incomplete set of measured input values, it is unlikely that
EDM wire entry and exit artefacts improving the results. Due to the force and moment balance will be satisfied. As a result, the Finite
destructive nature of the method, the test was carried out after the Element software generates a balanced solution which deviates from
neutron diffraction measurements on the same specimen. the measured values. Since it is not possible to measure a complete
residual stress distribution, an iterative process was used instead. The
initial residual stress distribution entered into the model was adjusted
2.3. Fatigue crack growth tests
until the relative error between the measured values and the Abaqus
solution at specified control points, defined at different locations within
A central notch was created by Electro-Discharge Machining (EDM)
the laser peened area and through the thickness, was smaller than 1%.
in both the laser peened and unpeened specimens to prepare them for
This allowed to achieve the desired residual stress distribution not just
fatigue crack growth testing, except for LSP-3 which was used only for
over the model surface, but also on each layer of elements through the
residual stress measurements. The laser shock peening treatment was
thickness. A uniform in-depth residual stress profile was chosen for this
study, which according to the residual stress results presented in the
next chapter provided a good approximation of the real residual stress
distribution. Finally, only the σy and σx components of the measured
compressive residual stress were introduced into the model, leaving the
software free to calculate the balancing tensile residual stresses. The
residual stress target values were obtained by averaging the neutron
diffraction and contour method measurements averaged through the
thickness at Point 1 (Fig. 2): σy = −146 MPa and σx = −98 MPa.
Thickness-averaged residual stresses have been successfully applied in
the literature when modelling fatigue crack growth, as reported in [25].

3.2. Crack growth modelling

After the residual stress was introduced, external load was applied
with a pressure boundary condition at the top surface to simulate fa-
tigue loading (Fig. 6). For each crack length, both the maximum and
minimum loading conditions (σmax and σmin) were applied to calculate
the corresponding maximum and minimum cyclic stress intensity fac-
Fig. 4. Numerical modelling steps. tors (Kmax,TOT and Kmin,TOT), according to the Modified Virtual Crack

160
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Closure Technique [26]. The forces and displacements extracted from

the FE software were then processed with MATLAB to generate fatigue
crack growth and fatigue life predictions.
Crack propagation was simulated by progressively releasing the
symmetry boundary condition applied on the nodes ahead of the crack
tip. Using this scheme, the crack length was increased in 1 mm incre-
ments from 4 mm – the initial crack length after precracking – to
70 mm. The final crack length was determined by averaging the crack
length at failure on the fractured specimens. This approach was pre-
ferred to using a failure condition since all the measured crack lengths
at failure were similar. The crack front shape was assumed to be
straight and the crack was uniformly extended during propagation, in
agreement with the work of many other researchers when modelling
through-cracks [27,28].

3.3. Crack growth predictions

Three fatigue crack growth prediction methods were compared in

this work: the Superposition method, the Modified Superposition
method, and the Newman method.

3.3.1. Superposition method

The total stress intensity factor range and total stress ratio were
calculated according to Eq. (3) and Eq. (4), respectively. An effective Fig. 6. Fatigue crack growth model boundary conditions.
stress intensity factor range, which is a function of the total stress ratio,
was then calculated according to the Walker equation [29]:
Table 4
KTOT Walker’s lookup table coefficients.
K eff =
(1 RTOT )1 (8) Keff [MPa√m] da/dN [m/cycle]

where γ is a fitting parameter that is experimentally determined [29]. 6.37 1.59 × 10–8
13.65 3.72 × 10–7
29.51 1.30 × 10–6
3.3.2. Modified superposition method 63.81 3.47 × 10–5
The total stress intensity factor range and total stress ratio were
calculated according to Eq. (5) and Eq. (6), respectively. The crack
closure mechanism was simulated using a hard contact between the

Fig. 5. Finite Element Mesh used to model a quarter of the M(T) specimen.

161
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Table 5 A1 = (0.415 0.071 ) max

Newman’s lookup table coefficients. 0 (13)

Keff [MPa√m] da/dN [m/cycle] A2 = 1 A0 A1 A3 (14)
4.67 1.94 × 10–8
A3 = 2A 0 + A1 1 (15)
9.78 3.71 × 10–7
20.48 1.20 × 10–6 and
42.87 2.43 × 10–5
YS + UTS
=
0
2 (16)
crack surface and a defined 3D rigid analytical shell lying on the Y For each method, the crack growth rate da/dN was determined from
symmetry pane and constrained to a reference point which was fixed the calculated effective stress intensity factor range Keff using a table
with encastre boundary conditions. These constraints prevented crack lookup approach as suggested in [17]. The coefficients, presented in
surfaces from overlapping and therefore the generation of negative Tables 4 and 5, were obtained from fatigue tests on several baseline
values for the stress intensity factor range. As in the Superposition (unpeened) specimens using different stress ratios (R = 0.1, 0.3 and
method, the effective stress intensity factor range was calculated ac- 0.6).
cording to Eq. (8).
4. Results and discussion
3.3.3. Newman method
The total stress intensity factor range and total stress ratio were 4.1. Residual stress results
calculated according to Eq. (3) and Eq. (4), respectively. The effective
stress intensity factor range was then calculated according to the The results obtained using neutron diffraction and the contour
Newman analytical equations [16]: method for the σy component of residual stress at Points 1, 2, 3 (depth
profiles) and along Line 1 (through thickness averages), are presented
(1 op / max )
K eff = KTOT in Fig. 7 and 8, respectively.
1 RTOT (9) The two methods are in generally good agreement. The through-
thickness residual stress distributions at Points 1 and 3 (Fig. 7) show a
op
= A0 + A1 RTOT + A2 RTOT 2 + A3 RTOT 3 for RTOT 0 fully compressive residual stress profile resulting from the laser peening
(10)
process. Its variation with depth, furthermore, shows a fairly constant
max

op residual stress profile which is matching the through-thickness aver-

= A0 + A1 RTOT for RTOT < 0
max (11) aged values, with the exception of the near-surface areas.
The through-thickness averaged residual stress distribution along
where Line 1 (Fig. 8) exhibits compression within the laser shock peened areas
1 with values up to −146 MPa, balanced by tension (up to 60 MPa) near
A0 = (0.825 0.34 + 0.05 2) cos max
the centre of the specimen between the laser shock peening patches
2 0 (12) (–25 mm < X < 25 mm). A small discrepancy between the two

Fig. 7. Comparison between Neutron Diffraction (ND) and Contour Method (CM) residual stress results: σy at Points 1, 2 and 3 (Fig. 2).

162
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Fig. 8. Comparison between Neutron Diffraction (ND) and Contour Method (CM) residual stress results: σy averaged through the thickness along Line 1 (Fig. 2).

Fig. 9. Comparison between initial Finite Element residual stress (un-cracked body), and the neutron diffraction and contour method results for σy, averaged through
the thickness.

techniques in Fig. 8 is visible in the transition zone (15 mm < X < conditions with the same loading, showing little scatter in the fatigue
25 mm) where the neutron diffraction measurements show a somewhat results. This gives increased confidence in the crack growth data col-
lower level of compression, possibly because of microstructural changes lected, despite the small number of specimens available. The fatigue life
which can lead to changes in the lattice parameter which are not due to of the treated components was found to be about 4× that of the pristine
residual stress. unpeened material, as shown in Fig. 10.
A comparison between the initial FE residual stress solution (un- The figure also shows that crack propagation in the laser peened
cracked body), the neutron diffraction results, and the contour method samples (LSP-1 and LSP-2) was slightly faster than in the untreated
results is presented in Fig. 9 for the Y component of residual stress (σy) components (BL-1 and BL-2) in the region between the laser peened
averaged through the thickness along Line 1. areas, due to the balancing tensile residual stress from the peening
Fig. 9 shows very good agreement between the FE solution and the treatment. This effect is more clearly revealed in the zoomed view in
experimental measurements, not only in terms of the compressive re- Fig. 11, where only the early part of crack propagation (up to a crack
sidual stress inside the laser peened area, but also in terms of the bal- length a of 40 mm) is shown. Once the crack entered the peened region,
ancing tensile residual stress in the rest of the section. the decrease in crack growth rate more than compensated for the crack
acceleration in the unpeened area, resulting in the four-fold increase in
4.2. Fatigue crack growth test results fatigue life.
The crack growth rate is plotted as a function of K in Fig. 12. The
Two samples were tested for each of the unpeened and laser peened data show that the crack growth rate is significantly lower in the LSP

163
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Fig. 10. Fatigue crack growth rate results, a vs N for the baseline (BL) and laser-
peened (LSP) specimens. The peened area spanned a distance from the centre of
Fig. 12. Fatigue crack growth rate, da/dN vs K. The crack growth rate de-
the sample equivalent to crack lengths between 25 and 55 mm.
creases significantly in the laser peened samples (LSP) once the crack enters the
peened area. The baseline (BL) samples show steady acceleration of the crack as
it grows.

stress distribution, but with a smoother profile. Fig. 13(b) shows the
variation of effective stress intensity factor range Keff with crack
length obtained using the three modelling approaches, together with
the nominal K from the loading conditions, KL. The reduction in K
due to the compressive residual stress in the peened area is greatest
using the Modified Superposition method, slightly less pronounced
using the Newman approach, and much weaker for the Superposition
approach. Fig. 13(c) shows how the stress intensity factor varies with
crack length for the three approaches. The three approaches show
identical values for Kmax,TOT, with no crack closure occurring at the
maximum applied load. In contrast, while the Superposition and
Newman approaches predict identical values for Kmin,TOT which be-
comes strongly negative in the peened region, Kmin,TOT cannot become
negative in the Modified Superposition approach as this approach im-
poses rigid crack closure. Similar behaviour can be observed in
Fig. 13(d): the total stress ratio for all three approaches is identical up
Fig. 11. Fatigue crack growth acceleration due to tensile residual stress. The
to the start of the peened region, at which point it becomes negative for
laser peened (LSP) samples show faster growth than the baseline (BL) samples
before the crack intersects the laser peened region at a crack length the Superposition and Newman approaches, but is constrained to be
of ∼ 25 mm. zero for the Modified Superposition approach. Though the nominal
applied stress ratio was 0.1, the tensile residual stress acting between
the specimen centre and the laser peened area pushes the predicted
specimens that in the unpeened baseline specimens when the crack tip
value up to 0.36 in this region.
is inside the laser shock peened area owing to the compressive residual
Fig. 13(e) and (f) present the predicted crack growth rates together
stress acting along the crack opening direction (σy) in this region. The
with the experimental data. The baseline test results for BL-1 and BL-2
crack growth rate was lowest at the end of the treated area
are also included, showing good agreement with predictions (BL-W)
(a = 55 mm), where da/dN was 13% of its value just before the crack
apart from a small underestimation of the growth rate in the very early
entered the compressive section of the residual stress field due to pe-
stages of crack propagation. The Modified Superposition method (cir-
ening.
cular empty markers on the figure) gives the best predictions, with most
of the predicted values falling within a factor of 1.25 of the measured
4.3. Fatigue crack growth predictions values. The maximum deviation (within a factor of 2) occurs towards
the end of the crack propagation, and therefore corresponds to very
The crack growth rates predicted using the three modelling ap- long cracks approaching the critical length. Under these conditions, the
proaches are presented in Fig. 13, which is divided into 7 sub-figures model assumptions may not be valid, since region III of crack propa-
showing KRes vs x and the Residual Stress vs a in (a); K vs a in (b); the gation is approached. Since these later stages of crack propagation ac-
Stress Intensity Factor vs a in (c); the Stress Ratio vs a in (d); da/dN vs a in count for a small portion of the fatigue life, the large discrepancy be-
(e); da/dN vs K in (f) and σop/σmax vs a according to Eq. (10) and Eq. tween experimental results and model predictions do not significantly
(11) in (g). affect the fatigue life prediction.
The residual stress intensity factor KRes highlighted in Fig. 13(a) The crack growth rates predicted using the Newman and Modified
(black solid line) follows the main trend shown by the initial residual

164
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Fig. 13. (a) KRes vs x and Residual Stress vs a; (b) predicted K vs a as a result of the applied KL; (c) predicted Stress Intensity Factor vs a, (d) Predicted Stress Ratio vs
a; (e) da/dN vs a for the experimental samples, along with the predicted crack growth rates from the three models used; (f) da/dN vs K for the experimental samples,
along with the predicted crack growth rates from the three models used; (g) Predicted σop/σmax vs a from Eq. (10) and Eq. (11); S = Superposition method,
MS = Modified Superposition method, N = Newman method, BL-W = Predicted baseline with Walker equation.

Superposition methods are very similar up to a crack length of 5. Conclusions

∼35 mm. Beyond this value, the Newman method predicts higher
growth rates, eventually converging again with the Modified The application of laser shock peening as a method to achieve fa-
Superposition results for crack lengths greater than ∼60 mm. The tigue crack growth retardation for long cracks in an aerospace alumi-
Superposition method heavily overestimates the growth rate within the nium alloy was studied by placing two laser-peened patches onto a
laser shock peened area, giving the worst results. Fig. 13(g) presents the middle-crack tension specimen made of AA2524-T351, with a thickness
normalised crack opening stresses σop/σmax as a function of the crack of 6 mm. The benefits introduced by the laser peening treatment in
length a, calculated with the Newman method. Results follow the total terms of fatigue performance, fatigue life, and crack growth were stu-
stress ratio trend shown in Fig. 13(d), since this is the only parameter died with both experimental testing and numerical modelling. The re-
affected by the residual stress in Eq. (10) and Eq. (11). The a vs N curves sidual stress resulting from the laser peening process was determined
presented in Fig. 14 show that the Modified Superposition method using neutron diffraction and the contour method. This information was
provides the most accurate fatigue life estimation with a prediction of incorporated into a linear-elastic fracture mechanics finite-element (FE)
128,520 cycles, 7% more conservative than LSP-2 (137565 cycles). model to accurately simulate fatigue crack growth in the presence of
Results from the Newman model are even more conservative, with a the residual stress field. Several fatigue crack growth prediction models
predicted life of 94,246 cycles, 31% less than the experimental value. were used, namely the Superposition, the Modified Superposition, and
The Superposition method shows almost no improvement in fatigue life the Newman methods. The following conclusions can be drawn:
from the laser peening process, with a predicted life of 50,997 cycles,
only slightly higher than the baseline results. 1. The residual stress generated by the peening along the crack
opening direction, σy, was compressive through the thickness inside

165
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

Fig. 13. (continued)

the laser peened areas, with through-thickness averaged values up mechanism, generated slightly more conservative estimates of fa-
to −146 MPa. The compression introduced by the laser peening tigue life.
treatment was balanced by a tensile residual stress field between the
laser peened patches, with through-thickness averaged values up to In summary, we have demonstrated that laser peening can be used
60 MPa. successfully to extend the fatigue life of aerospace aluminium in a
2. The laser peened samples had fatigue lives about 4× those of section thickness relevant to wing cover applications, leading poten-
unpeened specimens. The compressive residual stress introduced by tially to reduced maintenance costs or to weight savings. Furthermore,
the laser peening treatment caused a dramatic decrease in fatigue if the residual stress distribution from the laser peening can be fully
crack growth rate when the crack tip entered the peened region, so characterized by experiment, as was the case here, then a relatively
extending the overall fatigue life. A small fatigue crack growth rate simple FE model can provide a suitably accurate and conservative
increase was observed at the beginning of crack propagation, when prediction of the fatigue life.
the crack propagated within the tensile residual stress field that
balances the compressive stress in the peened region.
3. Fatigue crack growth predictions showed very good agreement with Acknowledgments
experimental data as long as a crack closure mechanism was con-
sidered, such as in the Modified Superposition and Newman The authors would like to thank Airbus Operations GmbH for the
methods. The former, which was realised in the FE model by pre- financial support and material provided. MEF is grateful for funding
venting crack surfaces from overlapping, provided the best results, from the Lloyd’s Register Foundation, a charitable foundation helping
predicting a fatigue life only 7% shorter than observed experimen- to protect life and property by supporting engineering-related educa-
tally. The latter, based on the plasticity-induced crack closure tion, public engagement and the application of research.

166
M. Pavan, et al. International Journal of Fatigue 123 (2019) 157–167

[8] Bueckner HF. The propagation of cracks and the energy of elastic deformation.
Trans Am Soc Mech Eng 1958;80:1225-1230.
[9] Underwood JH, Pook LP, Sharples JK. Fatigue-Crack Propagation Through a
Measured Residual Stress Field in Alloy Steel. Flaw Growth Fract; 1977;ASTM STP
6:402–15.
[10] Fukuda S, Yasuyuki T. An Experimental Study of Redistribution of Welding Residual
Stress with Fattgue Crack Extension 1978:67–72.
[11] Chandawanich N, Sharpe WN. An experimental study of fatigue crack initiation and
growth from coldworked holes. Eng Fract Mech 1979;11:609–20. https://doi.org/
10.1016/0013-7944(79)90122-X.
[12] Lam YC, Lian KS. The effect of residual stress and its redistribution of fatigue crack
growth. Theor Appl Fract Mech 1989;12:59–66. https://doi.org/10.1016/0167-
8442(89)90015-3.
[13] Stress Parker AP. Intensity factors, crack profiles, and fatigue crack growth rates in
residual stress fields. Residual Stress Eff Fatigue 1982;STP;776:13–31. https://doi.
org/10.1520/STP30095S.
[14] Wolf E. Fatigue crack closure under cyclic tension. Eng Fract Mech 1970;2:37–45.
https://doi.org/10.1016/0013-7944(70)90028-7.
[15] Newman JC. A crack-closure model for predicting fatigue crack growth under air-
craft spectrum loading. Methods Model Predict Fatigue Crack Growth under
Random Load 1981;748:53–84. http://doi.org/10.1520/STP28334S.
[16] Newman JC. A crack opening stress equation for fatigue crack growth. Int J Fract
1984;24.. https://doi.org/10.1007/BF00020751.
[17] Hill MR, Kim J. Fatigue crack closure in residual stress bearing materials. J ASTM
Int 2012;9:65–86. https://doi.org/10.1520/JAI104071.
[18] Garcia C, Lotz T, Martinez M, Artemev A, Alderliesten R, Benedictus R. Fatigue
crack growth in residual stress fields. Int J Fatigue 2016;87:326–38. https://doi.
org/10.1016/j.ijfatigue.2016.02.020.
[19] Kashaev N, Ventzke V, Horstmann M, Chupakhin S, Riekehr S, Falck R, et al. Effects
Fig. 14. a vs N; S = Superposition method, MS = Modified Superposition of laser shock peening on the microstructure and fatigue crack propagation beha-
method, N = Newman method, BL-W = Predicted baseline with Walker equa- viour of thin AA2024 specimens. Int J Fatigue 2017;98:223–33. https://doi.org/10.
1016/j.ijfatigue.2017.01.042.
tion, BL-N = Predicted baseline with Newman equation.
[20] SAE International. AMS4296B 2012.
[21] Prime MB, Gonzales AR. the Contour Method: Simple 2-D Mapping of Residual
Stresses. Sixth Int Conf Residual Stress Int Conf Residual Stress July 10-12,
References 2000;836:617–24.
[22] ASTM Standard E647 − 13a. Standard Test Method for Measurement of Fatigue
[1] Montross CS, Wei T, Ye L, Clark G, Mai YW. Laser shock processing and its effects on Crack Growth Rates. Am Soc Test Mater 2014:1–50. http://doi.org/10.1520/
microstructure and properties of metal alloys: a review. Int J Fatigue E0647-13A.2.
2002;24:1021–36. https://doi.org/10.1016/S0142-1123(02)00022-1. [23] Dasssult Systemes. ABAQUS User’s Manual, version 6.14 2014.
[2] Fitzpatrick ME, Lodini A. Analysis of Residual Stress by Diffraction using Neutron [24] MathWorks. MATLAB User’s Manual, version R2013b 2013.
and Synchrotron Radiation. 2003. [25] Liljedahl CDM, Zanellato O, Fitzpatrick ME, Lin J, Edwards L. The effect of weld
[3] Macherauch E, Kloss KH. Proceedings of the International Conference on residual residual stresses and their re-distribution with crack growth during fatigue under
Stresses. Garmisch-Partenkirchen: FRG; 1986. p. 167–74. constant amplitude loading. Int J Fatigue 2010;32:735–43. https://doi.org/10.
[4] Withers PJ, Bhadeshia HKDH. Residual stress. Part 1–measurement techniques. 1016/j.ijfatigue.2009.10.012.
MaterSci Technol 2001;17:355–65. https://doi.org/10.1179/ [26] Krueger R, Krueger R. The Virtual Crack Closure Technique : History. Hampton:
026708301101509980. Approach and Applications; 2002.
[5] Reimers W, Pyzalla AR, Schreyer A, Clemens H. Neutrons and Synchrotron [27] Solanki K, Daniewicz SR, Newman JC. Finite element analysis of plasticity-induced
Radiation in Engineering Materials Science. Weinheim: WILEY-VCH Verlag GmbH & fatigue crack closure: an overview. Eng Fract Mech 2004;71:149–71. https://doi.
Co. KGaA; 2008. org/10.1016/S0013-7944(03)00099-7.
[6] Achintha M, Nowell D. Eigenstrain modelling of residual stresses generated by laser [28] Gozin MH, Aghaie-Khafri M. 2D and 3D finite element analysis of crack growth
shock peening. J Mater Process Technol 2011;211:1091–101. https://doi.org/10. under compressive residual stress field. Int J Solids Struct 2012;49:3316–22.
1016/j.jmatprotec.2011.01.011. https://doi.org/10.1016/j.ijsolstr.2012.07.014.
[7] LaRue JE, Daniewicz SR. Predicting the effect of residual stress on fatigue crack [29] Walker EK. The effect of stress ratio during crack propagation and fatigue for
growth. Int J Fatigue 2007;29:508–15. https://doi.org/10.1016/j.ijfatigue.2006.05. 2024–T3 and 7075–T6 aluminum ASTM STP. Am Soc Test Mater 1970:1–14.
008. https://doi.org/10.1520/stp32032s.

167