International

Journalof
Fatigue

International Journal of Fatigue 29 (2007) 15311545

www.elsevier.com/locate/ijfatigue

The eect of controlled shot peening on the fatigue behaviour

of 2024-T3 aluminium friction stir welds
A. Ali a, X. An b, C.A. Rodopoulos b,*, M.W. Brown a, P. OHara c,
A. Levers d, S. Gardiner d
b

a
Department of Mechanical Engineering, The University of Sheeld, Mappin Street, Sheeld S1 1JD, United Kingdom
Structural Materials and Integrity Research Centre, Materials and Engineering Research Institute (MERI), Sheeld Hallam University,
City Campus, Howard Street, Sheeld S1 1WB, United Kingdom
c
Metal Improvement Company, Newsbury, Berkshire RG14 5TU, United Kingdom
d
Airbus UK, Chester Road, Broughton, Chester CH4 0DR, United Kingdom

Received 5 April 2006; received in revised form 18 September 2006; accepted 16 October 2006
Available online 29 December 2006

Abstract
The work examines the microstructural and fatigue properties of friction stir welds made of 2024-T3 aluminium alloy and provides
extensive information towards their cyclic stressstrain behaviour, residual stress distribution and crack initiation sites. To eliminate the
cost associated with the removal of the ow arm by milling and other costs associated with the quality control of the welding process
(residual stress distribution, micro-hardness prole, welding scar, etc.), controlled shot peening is introduced. Tensile residual stresses
introduced in the thermomechanical aected zone during welding are found to become compressive after peening. The eect can be held
responsible for increasing the fatigue resistance of the weld beyond the values of the bare (parent) material.
2006 Elsevier Ltd. All rights reserved.
Keywords: Friction-stir welding; Controlled shot peening; Residual stresses; Fatigue life; Cyclic yield stress; Rening of precipitates

1. Introduction
Commercial transport airplanes generally consist of a
built-up structure where the skin-to-stringer, skin-to-clip
and clip-frame joints are riveted, bolted or bonded. Such
joints for many years have been the subject of extensive
research, especially in terms of multiple site damage, widespread damage, fretting fatigue, etc. The proceedings of the
International Conference of Aeronautical Fatigue provide
an excellent source for referencing.
Friction stir welding (FSW) is a relatively new process
patented by The Welding Institute (Cambridge, UK) in
1992 [1]. A friction stir butt weld is produced by plunging
a rotating tool into the facing surfaces of the two plates.
The tool consists of a shoulder and a proled pin emerging

Corresponding author. Tel.: +44 114 2254 257; fax: +44 114 2253501.
E-mail address: C.Rodopoulos@shu.ac.uk (C.A. Rodopoulos).

0142-1123/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2006.10.032

from it. As the rotating pin moves along the weld line, the
material is heated up by the friction generated by the shoulder and stirred by the rotating pin in a process similar to an
extrusion. Since the temperatures are well below the melting point, problems associated with the liquid/solid phase
transformation are avoided.
Besides the attractive mechanical properties, especially
in fatigue and load bearing capacity strength, FSW integral
structures are claimed to oer cost and weight savings [2,3].
Therefore, FSW was recently identied by leading aircraft
manufacturers as key technology for fuselage and wing
manufacturing [4,5]. Yet, problems associated with the fatigue behaviour of FSW are numerous and not well
established.
Generally FSW produces ve distinct microstructural
zones [6], namely the weld nugget (N), the shoulder contact
zone or ow arm region, the thermomechanical aected
zone (TMAZ), the heat aected zone (HAZ) and unaected
zone or parent plate (PP). Consequently, the fatigue

1532

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

strength of FSW joints varies for each zone of the weld.

The FSW weld zone is V-shaped and widens near the top
surface due to the close contact between the shoulder of
the tool and the upper surface [7]. The above indicates
potential discontinuities in the strength and fatigue volume
properties. Sato et al. [8] pointed out that the shape of the
weld zone depends on the welding parameters and the
material used. Dalle Donne and Biallas [9] show that with
proper FSW tooling and welding parameter control, a
reduction of only 20% compared to the base material values for the joint ultimate strength and fatigue endurance
can be achieved. In addition, the zones have also been considered responsible for variations in the fatigue failure initiation sites. Booth and Sinclair [6] identied two forms of
failure in the 2024-T351 FSW: (a) failure occurred from
within the actual weld material (Nugget) and (b) failure
occurred outside of the actual weld, either in the TMAZ
or HAZ. Failure within the nugget region was associated
with discontinuities in the material ow pattern at the surface. With no obvious defects being seen, the exact origins

of crack initiation within this region were not clearly

identiable, whilst the failure in TMAZ and HAZ initiated
by decohesion of large S-phase particles or transgranular
failure. They suggest that heterogeneous precipitation at
particle interfaces may inuence the decohesion strength
of the intermetallics at a specic location.
Dierences in the fatigue behaviour are also manifested
by the hardness characterisation in relation to the ve
microstructural zones. Jata et al. [11] reported for the
7050 Al alloy that the hardness of the top side is lower than
the bottom side of the weld. They suggested that this is due
to the fact that the top side is in full contact with the tool
shoulder, and thus, experiences direct heat. The bottom
side, on the other hand, is in indirect contact with a back
plate that acts as a heat sink. Comparing the hardness
between the zones, the hardness within the nugget varies
depending on the alloy and its initial heat treatment. For
2024-T351, 7050-7745 and 6061-T6 alloys hardness proles
in the weld nugget show a local maximum value at the plate
joint line or centre of the nugget [9,6,11]. Hardness proles

Table 1
Chemical composition of 2024-T351 in wt% based on EDX measurements
Alloy
2024-T351

Minimum
Maximum

Si

Fe

Cu

Mn

Mg

Cr

Zn

Ti

Zr

Al

0.50

0.50

3.8
4.9

0.30
0.90

1.2
1.8

0.10

0.25

0.15

Balance

Table 2
Basic mechanical properties of 2024-T351 according to ASTM E8m-94a
Mechanical properties

Mean

99% Condence interval

Standard deviation

Standard error

0.2% Yield strength (MPa)

Tensile strength (MPa)
Elongation (%)
Fracture toughness plane stress (MPa m1/2) for thickness 1.6 mm
Strain energy density (MJ/m3)

347.4
484.6
15.0
135.5
70.5

5.13
2.6
0.2
8.7
0.6

4.6
2.3
0.2
3.00
0.6

1.5
0.8
0.1
1.5
0.2

Fig. 1. Grain structure distribution exhibited within the ve zones. (a) Flow arm; (b) nugget; (c) TMAZ; (d) HAZ; and (e) parent.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

near surface region and surface roughening [2629]. Residual stresses are likely to benet the fatigue resistance of
high strength materials. Softer materials on the other hand
are likely to experience fatigue resistance improvement
owning mostly to strain hardening, since partial or even
complete relaxation of the residual stresses may occur
depending on the type of loading, stress level and the residual stress distribution prole [3032]. Strain hardening is
likely to increase the ow resistance of the material to plastic deformation.
Roughening of the surface is the major detrimental eect
of CSP. Surface roughness, owing to the local intensication of the far-eld stress, can account for the premature
initiation and propagation of short fatigue cracks [33].
Rodopoulos et al. [34] suggested that a portion of residual
stresses is consumed in order to counteract the detrimental
eects of surface roughening. In brief, the elastic stress concentration provided by the surface roughness will increase
the surface stress and hence the near-surface crack growth
rate.
In this work, the use of the CSP technology to provide
an improvement of the fatigue resistance of FSW has
been selected on the grounds that: (a) the technology

180
170
160

Hardness (HV)

taken from nugget zone of 6063 Al alloy showed a minimum value among other regions. These dierences in hardness value within the nugget have been correlated with the
size of the precipitates present in the region [10,1214].
Investigations in the 2xxx and 7xxx aluminium series
showed hardness minima within the TMAZ zone
[6,8,15,16]. The eect has been attributed to overaging [10].
Residual stress elds are widely believed to signicantly
eect catastrophic crack nucleation and growth. In [8,17
19], residual stress distribution was reported to vary along
the zones of the weld. Webster et al. [20] measured the
residual stresses using Synchrotron X-ray technique and
reported tensile residual stress in the nugget zone of
7108-T79. Similar nding were also reported by Bussu
and Irving [10], Oosterkam et al. [21] for 2024-T351 and
AA7108-T79, respectively. Nevertheless, Jata et al. [11]
and Dalle Donne et al. [22] found a small compressive
residual stress located at the centre of the nugget zone for
7050-T7451, AlLiCu and 6013-T6.
Defects associated with the FSW process are strongly
associated with fatigue resistance. In [13,15,23] it was
reported that voids, inclusions and surface cracks dominate the nugget and represent potential sites for crack initiation. The above makes clear that quality process control
and quality fatigue damage tolerance control over FSW
joints is a complex requirement demanding extensive and
well organised international research. Yet, driven from
todays market and societal needs for prompt innovation,
cost and pollutant emission reduction [24], the fatigue
behaviour of FSW joints needs to be improved and safeguarded without the need for an in-depth research. Such
solution can be sought in terms of controlled shot peening
(CSP).
CSP is a well established surface engineering treatment
in the area of aeronautical and automotive engineering
[25]. Pellets made of steel, ceramic or even dry ice, accelerated by either pneumatic or mechanical means are directed
through controllable ow conditions onto the surface of
the target material. The above results into the development
of compressive residual stresses, strain hardening of the

1533

150
140
130
120
Hardness on the top weld surface

110

Hardness on the bottom weld surface

100
-100

-80

-60

-40

-20

20

40

60

80

100

Distance from PJL (mm)

Fig. 3. Vickers hardness distribution across the weld surface. A 10 kg load

was used.

Fig. 2. 2-D mapping of the weld. The plate joint line (PJL) corresponds to the centre weld section and is being used as scaling origin. The map also reveals
the non-symmetrical distribution of the zones left or right to the PJL with the right TMAZ showing smaller area compared to its left counterpart. Such
discrepancy is attributed to contact pressure dierence in the disc shoulder (possibly during rotation).

1534

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

has been previously applied to aluminium tungsten-inertgas (TIG) and metal-inert-gas (MAG) welds with exceptional results [35]; (b) FSW does not create softening
eects typical to TIG which would prevent the complete
development of residual stresses [35]; (c) CSP has the
potential of altering the state and magnitude of residual
stresses (from tension to compression) [36]; and (d) the
strain rate from CSP is low enough not to aect microstructural properties like the explosive hardening treatment [36] which would have complicated the case. This
work presents a detailed analysis towards the potential

of CSP to improve the fatigue resistance of 2024-T3 aluminium alloy friction stir welds.
2. Experimental procedures
2.1. Weld micromechanical properties and zone mapping
The investigation was performed on 13 mm thick 2024T351 FSW joint. The joints were provided by Airbus UK,
Broughton. Plates 13 mm thick have been welded along
their long edge with the weld direction parallel to the

150

145
Experimental PJL
Regression

Experimental 3mm from PJL

Regression

140

Microhardness, HV 0.1

Microhardenss, HV 0.1

140

130

120

110

100

135

130

125

120

90

115
0

10

12

14

2000

Depth from top surface (mm)

140

8000

10000

12000

14000

150
Experimental 9mm from PJL
Regression

135

145

Microhardness, HV0.1

Microhardness HV0.1

6000

(b) Section 2 - 3mm from PJL

(a) Section 1 - PJL

130
125
120
115

140

135

130

125

110
105

Experimental 14mm from PJL

Regression

120
0

10

12

14

Depth from top surface (mm)

10

12

14

12

14

Depth from top surface (mm)

(c) Section 3 - 9mm from PJL

(d) Section 4 - 14mm from PJL

155

150

Experimental 22mm from PJL

Regression

150

Microhardness, HV0.1

140

Microhardness, HV0.1

4000

Depth from the top surface (mm)

130

120

110

145

140

135

130

Experimental 16mm from PJL

Regression

125

100

10

Depth from surface (mm)

(e) Section 5 - 16mm from PJL

12

14

10

Depth from Surface (mm)

(f) Section 6 - 22 mm from PJL

Fig. 4. Microhardness measurements along the six selected sections. To better acknowledge the tendencies, regression analysis was performed.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

1535

Fig. 5. Back scattered images taken from GDOES prepared spots at dierent locations as well as from the left ( ) or right (+) side of the weld.

1536

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

longitudinal (rolling) orientation. The welds have been produced using a spindle speed of 200 rpm and a feed rate of
120 mm/min. The chemical composition and basic static
mechanical properties of the parent material are shown in
Tables 1 and 2, respectively.
For identication and measurement of the ve zones
microstructure, the weld was cross-sectioned using a Beuhler Isomet 4000 precision diamond saw with rotational disk
speed of 4100 rpm and a feed rate of 1.6 mm/min. The sections, undergone mechanical polishing using a series of
emery papers and diamond pastes until reaching a 1/
4 lm nish. To reveal the microstructure the sections were
etched using Kellers reagent in accordance with ASTM
E340-00. Fig. 1 shows the grain size distribution of the different zones.
The nugget exhibits the nest grain with sizes in the
range between 5 and 10 lm. The ow arm zone was found
having a ne equiaxed grain structure of approximately 15.
The TMAZ exhibits a non-uniform and severely elongated
microstructure. The elongated grain size is found to range
between 200 and 400 lm representing dierent degrees of
thermomechanical plastic deformation which governed
the amount of material drawn into the weld zone. The
grain structure in HAZ and in the parent plate appears
have similar size in the order of 150200 lm. The process
also allowed the detailed 2-D mapping of all the dierent
zones as shown in Fig. 2.
Fig. 3 shows hardness proles measured along the top
and bottom surface. Generally the hardness of the bottom
surface is lower than that at the top. This is particularly the
case for the nugget zone (ow arm for the top surface). The
lowest hardness value from the top surface, approximately
118 Hv1, was found in the TMAZ and especially closer to
the interface with the HAZ (13 mm from the plate joint line
(PJL)). In contrast, the bottom surface exhibited its lowest
microhardness value within the nugget and TMAZ zone
(approximately 105118 Hv1). The highest top surface
microhardness value of 167 Hv1 was found within the
HAZ (approximately 16 mm from the PJL). Similar results
can also be conrmed for the bottom surface.
In order to investigate the eect of the gradient of the
thermomechanical plastic deformation taking place during
the welding process, which, as previously shown is manifested by dierences between the top and bottom weld surface, through thickness microhardness measurements were
taken from six selected sections (see Fig. 2). The measurements are depicted in Fig. 4.
The analysis reveals that: (a) section 1 (nugget) exhibits
degradation of its microhardness value with depth at a rate
of 1.4 HV/mm; (b) section 2 exhibits similar behaviour to
section 1 at a rate of 0.78 HV/mm; (c) section 3 (TMAZ +
HAZ) demonstrates minimum microhardness growth at a
rate of 0.42 HV/mm. Herein it is important to note that the
transition from TMAZ to HAZ at a depth of approximately
3.2 mm did not register any signicant change; (d) section 4
(HAZ close to TMAZ) shows signicant increase in the
microhardness with a growth rate of 0.85 HV/mm; (e)

section 5 (clear HAZ) shows signicant increase at a rate of

1.35 HV/mm; and nally (f) section 6 (clear parent) shows
trivial increase at a rate of 0.21 HV/mm.
To investigate possible alterations in the nature of Cu
precipitates due to the welding process, etching of the top
surface at dierent locations was prepared by sputtering erosion (using argon ion) using a glow discharge optical emission spectrometer (GDOES), LECO GDS-750 QDP,
operating at voltage 600 V and current 25 mA. The etched
spots (4 mm in diameter) were later used for back scattered
scanning electron microscopy. Fig. 5a shows an image taken
from the PJL zone. It reveals that Cu precipitates exhibit a

Fig. 6. (a) 3-D distribution of residual stresses from the top (b) and
bottom (c) surface of the weld.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

non-uniform size distribution and are concentrated at the

interfaces between ner and coarser grains (being the result
of unsuitable cooling). Fig. 5b shows that at the right-hand
side of the weld (+) and at a distance of 2 mm, precipitates
are still agglomerated. Precipitate agglomeration is a typical
feature of the rolling process the plate underwent prior to
friction stir welding. Such feature disappears at a distance
of 5 mm. Signicant grain size variations and strong
agglomeration of precipitates have also been found in the
TMAZ, as shown in Fig. 5c. The black spots have been identied by EDS as K, Si and Cl oxides. The HAZ, Fig. 5d,
shows signicant coarsening and more uniform distribution
of the precipitates as well as elongated microstructure. It is
worth noting that K and Mn oxides were traced only on
the left-hand side of the weld. The above changes are better
appreciated by comparison to the parent material shown in
Fig. 5e.
Residual stress measurements were performed using
Incremental Hole-Drilling according to ASTM E837-99.
Eighteen drilling points, shown in Fig. 6a have been
selected. Prior to drilling, a 0.5 mm layer was milled to
remove the onion ring scar. The results shown in Fig. 6b
(top surface of the weld) indicate that: (a) the nugget is in
compression with values reaching maxima at the near surface layer; (b) the TMAZ is also in compression but with a
tendency to achieve tension with depth; (c) the HAZ is until
a depth of 1 mm is in compression after which signicant
tension (especially on the left part of the weld) is built up.
Fig. 6c, showing the results from the bottom surface reveals
that: (a) the nugget surface is under negligible residual stress
but goes into compression with depth; (b) the TMAZ and
HAZ exhibit similar behaviour to the top surface with residual stresses reaching maxima at the surface but gradually
relax with depth. Yet again, the TMAZ at the left part of
the weld exhibited steep transition into tension.
2.2. Weld cyclic properties
Four sections of the weld have been selected to provide
testing material for the evaluation of cyclic yield stress and

1537

Table 3
The result of cyclic tests in pure bending
Section

Youngs Strain
Hardening Cyclic
modulus hardening constant
yield
(GPa)
exponent (MPa)
stress
(MPa)

Parent material
Nugget
HAZ
TMAZ
(35% TMAZ + 65%
HAZ)

68
68
68
68

0.086
0.12
0.05546
0.413

770
720
719
800

340
270
410
250

cyclic hardening in pure bending. These, according to

Fig. 6a, are 514, 413, 211, 110. The obtained results,
in terms of the RambergOsgood equation are shown in
Table 3. Information on the testing procedure can be found
in [37].
The results indicate that the TMAZ and nugget section
exhibits a cyclic yield stress value signicantly lower to
that of the parent material while at the same time a higher
strain hardening exponent. In contrast, the cyclic yield
stress of the HAZ section demonstrates an enhanced
value. Considering the strong presence of oxides in both
zones, such dierence could be attributed to the irregular
grain size distribution of the former. Of course it is important to note that the section named TMAZ, in Table 3,
contains 65% of the HAZ section and thus, if quite simplistically a rule of mixtures is used, the actual cyclic yield
stress of a pure TMAZ could be expected to be even
lower.
Fatigue endurance tests were carried out in a four point
bending conguration test in accordance with ASTM
D6272 and with the weld longitudinal to the loading axis.
The design of the specimen and experimental set-up is
shown in Fig. 7. The testing frequency was set at 20 Hz following a sinusoidal waveform and with a stress ratio of 0.1.
Mirror polished (Ra < 0.8 lm) and as-welded specimens
were used to evaluate the eect of the FSW onion ring scar.

Fig. 7. Specimen and loading geometry.

1538

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

250

For comparison, identical specimens machined from the

parent material were also tested. The results are shown in
Fig. 8.
The results indicate that in both conditions, the weld
underperformed the parent metal with the as-welded registering lowest lives. Yet, in the area of low cycle fatigue
(N < 105), both the polished and the as-welded exhibited
similar fatigue lives, with the former demonstrating significant scatter.

200

2.3. Fractographic analysis of as-received state

150

Fractographic analysis of the broken specimens was performed using a Philips XL-40 scanning electron microscope. Fig. 9a and b shows the crack nucleation sites
from the as-weld and polished condition tested at
300 MPa. The crack nuclei for the as-welded case are

500
Parent
As-Welded
Polished

Maximum Stress (MPa)

450
400
350
300

100
10 4

10 5

10 6

10 7

10 8

No. of cycles to failure

Fig. 8. SN curves of as-welded; mirror polished and parent specimens.

Fig. 9. Fractographs from the broken surface of (a) as-welded and (b) polished tested at a maximum stress of 300 MPa.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

Fig. 10. Fractographs taken at stress levels near to life run-outs (high cycle fatigue).

1539

1540

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

0

150
100

Residual Stress (MPa)

Residual Stress (MPa)

-50
-100
-150
-200

-300
0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

TMAZ As-welded
TMAZ Peened as-welded
TMAZ Peened after 3mm skimming

0
-50
-100
-150
-200

PJL Peened as Welded

PJL as Welded
PJL Peened after skimming

-250

50

-250
-300
0.0

1.6

0.2

0.4

Depth (mm)

1.0

1.2

1.4

1.6

50
0

Residual Stress (MPa)

Residual Stress (MPa)

0.8

Depth (mm)

50

-50

-100

-150

-200
0.0

0.6

HAZ as Welded
HAZ peened after skimming
HAZ peened as Welded

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

-50
-100
-150
-200

1.8

-250
0.0

Parent as welded
Parent peened as welded
Parent peened after 3mm skimming

0.2

0.4

Depth (mm)

0.6

0.8

1.0

1.2

1.4

1.6

Depth (mm)

Fig. 11. Residual stress proles of each zone for the states of as-welded, peened as-welded and peened after 3 mm skimming.

190

190
PJL As-welded
PJL Peened as-welded
PJL Peened after 3mm Skimming

170
160
150
140
130
120
110
0.0

TMAZ as-welded
TMAZ Peened as-welded
TMAZ Peened after 3mm Skimming

180

Microhardness, HV0.1

Microhardness, HV0.1

180

170
160
150
140
130
120

0.2

0.4

0.6

0.8

110
0.0

1.0

0.2

Depth from the top surface (mm)

170

0.8

1.0

Parent As-welded
Parent Peened as-welded
Parent Peened after 3mm skimming

170

Microhardness, HV0.1

Microhardness, HV0.1

0.6

180
HAZ As-welded
HAZ Peened as-welded
HAZ Peened after 3mm skimming

160
150
140
130
120
110
100
0.0

0.4

Depth from the top surface (mm)

160
150
140
130
120

0.2

0.4

0.6

0.8

1.0

Depth from the top surface (mm)

1.2

110
0.0

0.2

0.4

0.6

0.8

1.0

Depth from the top surface (mm)

Fig. 12. Microhardness proles of each zone for the states of as-welded, peened as-welded and peened after 3 mm skimming.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

2.4. Controlled shot peening

In the previous section, it was shown that in the case
of the mirror-polished specimens, crack initiation under

500

400

300
250

150
100
10 4

10 5

10 6

10 7

10 8

Number of cycles to failure

Fig. 13. SN curves from the testing matrix. The arrows indicate run-outs.

low cycle fatigue conditions took place at the interface

between ow arm and nugget owning to debonding. Debonding of the ow arm has two detrimental eects: (a) is
likely to introduce local stress concentrations and (b) will
increase the number of cracks within the nugget area and
hence the probability for premature crack coalescence.
Yet, mechanical removal of the ow arm (skiing) to a

0
TMAZ Prior to loading
TMAZ After failure

PJL Prior to loading

PJL After failure

Residual Stress (MPa)

Residual Stress (MPa)

350

200

-100
-120

Parent
Peened as-welded
Peened after 3mm skimming
As-welded
Mirror Polished

450

Peak Stress (MPa)

located at the HAZ and the nugget zone. In contrast, the

mirror-polished surface produced the same number of initial cracks, except at locations within the nugget zone. Such
discrepancy can be attributed to the stress concentration
eect of the FSW onion ring scar onto the HAZ. Herein,
it is important to note that in the case of the mirror polished specimen, the fractographs indicate debonding of
the ow arm from the nugget and the possibility of creating
a free surface discontinuity.
In the stress case approaching life run-outs, 161 MPa
for as-welded and 270 MPa for the mirror-polished condition, the as-welded case exhibited similar crack nucleation sites as in the case of the 300 MPa. In contrast,
the polished condition exhibited crack nucleation only
within the nugget zone, see Fig. 10. Comparisons with
Fig. 9 reveals that: (a) the nugget zone is less resistive
to fatigue crack initiation primarily due to the lower
hardness or ow resistance characterising the zone; and
(b) the potential for crack nucleation in the nugget seems
to be invariant to local stress concentrations (onion ring
scar).

1541

-140
-160
-180
-200
-220

-50

-100

-150

-200
-240
-260
0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

-250
0.0

1.6

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Depth from the top surface (mm)

Depth from the top surface(mm)

0

Residual Stress (MPa)

-20
-40
-60
-80
-100
-120
-140

HAZ After failure

HAZ Prior to loading

-160
-180

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Depth from the top surface (mm)

Fig. 14. Residual stress proles prior and after failure at a maximum stress level of 310 MPa. Measurements were taken from the peened after 3 mm
skimming state.

1542

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

safe depth would also aect the distribution of benecial

compressive residual stresses found close to the surface as
well as increasing the cost. For example, compressive
residual stresses characterising the rst 0.4 mm from the
surface of the TMAZ and HAZ zones will disappear
leaving the surfaces of those zones being characterised
by tension. To assess the potential of using CSP as a
solution towards such problem, peening was performed
on 3 mm skimmed and un-skimmed specimens. CSP
was performed by Metal Improvement Company using
shot S230, 200% coverage and an Almen Intensity of
14 A. Fig. 11 shows the residual stress prole of the different zones after CSP, after incremental hole drilling. It
is worth reminding that direct comparison between the
peened as-welded and the peened after skimming state
is only valid in terms of fatigue analysis. This is because
the nal residual stress prole (after CSP) is strongly
aected by their pre-treatment state making such comparison impractical.
Comparison in the residual stress prole between the aswelded and the peened as-welded state reveals the potential
of CSP to aect the initial residual stress state (see PZL) to
such an extend as to reverse its direction (HAZ and
TMAZ). In terms of fatigue damage, direct comparison
of all three cases clearly shows the generated potential for
improvement. Similar potential is appreciated by examining the cross-section microhardness proles of each zone
as shown in Fig. 12.
The results indicate that each zone experiences some
degree of strain hardening with maximum benet being
collected by the PJL, TMAZ and the parent material both
in terms of magnitude and depth. Strain hardening on the
HAZ zone appears to be substantial in terms of magnitude
close to the surface but disappears at a relatively short
depth (50100 lm). Similarly to the above, the 3 mm skimming only allows comparison of the three states for fatigue
analysis.
Fatigue testing was performed in a way identical to
that described in the previous section. The results shown
in Fig. 13 clearly indicate the improvement provided by
the CSP. Skimming was found to be benecial to the
overall fatigue life considering that it registers values
higher than the parent material in the HCF region.
Yet, CSP on the as-welded state has been found to provide superior results compared to those taken by the aswelded and the mirror-polished state. Such nding can
support decision towards replacing machining of the scar
with CSP.
The residual stress prole is expected to relax with the
number of loading cycles at a rate and pattern depending
on the maximum stress level, the strain hardening exponent of the material and the gradients constitute the original prole as a function of depth [38]. To examine the
above, incremental hole drilling was performed after failure or run-out. Typical results are depicted in Fig. 14.
The highest relaxation rate exhibited by the HAZ (lowest
strain hardening exponent, see previous section) conrms

Fig. 15. Fractographs taken from (a) peened as-welded and (b) peened
after 3 mm skimming tested at 412 MPa (low cycle fatigue).

James [38] micro-plasticity relaxation theory. The fact that

fatigue life improvement, after CSP, is manifested in both
the low and high cycle fatigue regions and in conjunction
to the exhibited minimum residual stress relaxation patterns, points towards attributing fatigue life enhancement
mostly to the modied residual stress proles rather than
to strain hardening (strain hardening eects in LCF are
negligible).
2.5. Fractographic analysis of peened state
Fig. 15 shows the fracture surface of (a) peened as-welded
and (b) peened after 3 mm skimming tested at 320 MPa
(close to life run-out). In both cases cracks were found to
initiate at the nugget zone. Similar tendency has been
observed in the low cycle fatigue region, see Fig. 16. Multiple cracking can be attributed to the surface roughness
induced by the CSP. In general CSP, delivers a change in
the crack initiation sites concentrating them into the nugget
zone.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

1543

Fig. 16. Fractographs taken from (a) peened as-welded and (b) peened after 3 mm skimming tested at 320 MPa (high cycle fatigue).

3. Discussion and conclusions

Micromechanical and cyclic loading investigation was
performed on 2024-T351 friction stir welds. Owing to tensile residual stresses especially in the TMAZ, microstructural irregularities and oxides, the welds underperformed

the fatigue resistance of the parent material either in the

as-welded or the as-polished state. The presence or not,
of the onion ring scar has been shown to play an important
role in the high cycle fatigue region. Yet, at high stress levels such feature is found to be of small importance in terms
of fatigue life. Herein, it is important to say that the

1544

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

presence of such stress concentration could be held responsible for shifting the positions of the crack nuclei, by inducing or not crack growth from the TMAZ/HAZ interface.
Once such feature is removed, the crack initiation is positioned within the nugget and TMAZ. Herein, it is important to note that the debonding of the ow arm from the
main nugget body is of crucial importance considering
the resulting changes in the stress gradient. Yet, both cases
exhibited multi-cracking in LCF. Such tendency, changes
to single crack nucleation and growth under HCF conditions. In addition, of out most importance is the value relation between the TMAZ cyclic yield stress and that of the
endurance limit found in the as-polished case. Considering
that no other area is characterised by such low cyclic
elasto-plastic transition value, it is rational to assume a
physical relation in terms of assuming LCF conditions
for the TMAZ even when the external stress denotes
HCF conditions for all other areas. Agglomeration of precipitates, variations in the grain size distribution and the
presence of K, Si oxides (inclusions) can be held responsible for such behaviour. Of course dierences in the cyclic
yield stress value were also measured in every other zone,
resulting in a highly irregular accumulation of straining
during cyclic loading. The above results into strain irregularity across the interfaces between the zones and could be
held responsible for crack nucleation, mainly, at these locations. Crack initiation from the debonding of the ow arm
from the nugget raises concerns over whether such feature
should be retained.
In general, CSP delivers a signicant improvement in
terms of fatigue life both in the region of low and high cycle
fatigue. The improvement against the as-welded state is evident for both the peened as-welded and peened after 3 mm
skimming state, with the rst showing further improvement. In all cases, CSP concetrated the crack initiation sites
into the nugget zone. Multiple cracking in this zone could
be attributed to surface stress concentration (roughness)
being particular detrimental since the material is exhibiting
ne grained structure. Considering that removal of the
onion ring scar, either by polishing, skimming or peening
(relieved some stress concentration), resulted into the
change of the crack initiation sites into the nugget zone it
is dicult to conclude on the direct impact that CSP has
on fatigue crack initiation. Being speculative, it would
not be irrational to consider that the most fragile zone in
terms of fatigue resistance is the TMAZ. The zone exhibits
irregular grain size distribution, contains tensile residual
stresses and its cyclic yield stress is particularly low. Yet,
only in the as-welded case tested in high cycle fatigue conditions, crack initiation was found to emanate within. On
the other hand, the nugget zone characterised by low cyclic
yield stress, ne grained microstructure, agglomeration and
rening of Cu precipitates and an overall compressive
residual stress nature, exhibited high density of crack initiation. Based on the above it is rational to conclude that
CSP does not alter the crack initiation tendency of the
FSW and hence the enhanced fatigue life is the result of

a slow near surface crack propagation, possible due to

the simultaneous application of the increased microhardness and compressive residual stresses found in the nugget
area. Such potential can be exploited in cases where costs
related to the quality control of FSW are in question.
Acknowledgements
The authors thank Airbus UK, Metal Improvement
Company (Mr. Peter OHara) and the Malaysian Ministry
of Science for a scholarship to one of the authors (Ali).
Special thanks to Dr. C. Dalle-Donne from the Corporate
Research Centre of EADS in Munich for his very useful
comments. The work does not represent ocial views of
Airbus UK or EADS.
References
[1] Thomas WM, Nicholas ED, Needham JC, Murch MG, Templesmith
P, Dawes CJ. Improvements relating to friction welding. European
Patent EP 0 615 480 B1; 1992.
[2] Hansen M. A cooler weld, mechanical engineering, design supplement; 2003. Available from: www.memagazine.org/medes03/coolweld/coolweld.html.
[3] Pettit RG, Wang JJ, Toh C. Validated feasibility study of integrally
stiened metallic fuselage panels for reducing manufacturing costs.
NASA 2000; CR 2000-209342.
[4] Talwar R, Bolser D, Lederich R, Baumann J. Friction stir welding of
airframe structures. In: Proceedings of the 2nd international symposium on friction stir welding, TWI, UK; 2000.
[5] Lohwasser D. Welding of airframes by friction stir. In: Proceedings of
the 3rd international symposium on friction stir welding, TWI, UK;
2001.
[6] Booth D, Sinclair I. Fatigue of friction stir welded 2024-T351
aluminium alloy. Mater Sci Forum 2002;396402:16716.
[7] Heinz B, Skrotzki B, Eggeler G. Microstructural and mechanical
characterization of a friction stir welded Al-alloy. Mater Sci Forum
2000;331337:175762.
[8] Sato YS, Park SHC, Kokawa H. Microstructure factors governing
hardness in friction stir welds of solid solution hardened Al alloys.
Metall Mater Trans A 2001;32A(12):303342.
[9] Dalle Donne C, Biallas G. Fatigue and fracture performance of
friction stir welded 2024-T3 joints. In: Proceeding European conference on spacecraft structure, material and mechanical testing,
Braunschweig, Germany; 1999 (428). pp. 30914.
[10] Bussu G, Irving PE. The role of residual stress and heat aected zone
properties on fatigue crack propagation in friction stir welded 2024T351 aluminium joints. Int J Fatigue 2003;25:7788.
[11] Jata KV, Sankaran KS, Ruschau JJ. Friction stir welding eects on
microstructure and fatigue of aluminium alloy 7050-T7451. Metall
Mater Trans A 2000;31(9):218192.
[12] Murr LE, Li Y, Trillo EA, Flores RD, McClure JC. Microstructure in
friction stir welded metals. J Mater Process Manuf Sci 1998;7:14561.
[13] Sato YS, Kokawa H. Distribution of tensile property and microstructure in friction stir weld of 6063 aluminium. Metall Mater Trans
A 2001;32A(12):302331.
[14] Flores OV, Kennedy C, Murr LE, Brown D, Pappu S, Nowak BM,
et al. Microstructural issues in a friction stir welded aluminium alloy.
Scripta Mater 1998;38(5):7038.
[15] Bussu G. Damage tolerance of welded aluminium aircraft structure.
PhD. Thesis, Craneld University; 2000.
[16] Rhodes CG, Mahoney MW, Bingel WH. Eect of friction stir
welding on microstructure of 7075 aluminium. Scripta Mater
1997;36(1):6975.

A. Ali et al. / International Journal of Fatigue 29 (2007) 15311545

[17] Ericsson M, Sandstrom R. Fatigue of friction stir welded AlMgSialloy 6082. Mater Sci Forum 2000;331337:178792.
[18] Esparza JA, Davis WC, Trillo EA, Murr LE. Friction stir welding of
magnesium alloy AZ31B. J Mater Sci Lett 2002;21:91720.
[19] Mahoney MW, Rhodes CG, Flinto JG, Spurling RA, Bingel WH.
Properties of friction stir welded 7075 T651 aluminium. Metall Mater
Trans A 1998;29A:195564.
[20] Webster PJ, Oosterkam LD, Browne PA, Hughes DJ, Kang WP,
Withers PJ, et al. Synchrotron X-ray residual strain scanning of a
friction stir weld. J Strain Anal 2001;36(1):6170.
[21] Oosterkam LJ, Withers PJ, Browne PA, Vaughan GBM. Residual
stress eld in a friction stir welding aluminium extrusion. Mater Sci
Forum 2000;347349:67883.
[22] Dalle Donne C, Lima E, Wegener J, Pyzalla A, Buslaps T.
Investigations on residual stress in friction stir welds. In: Proceedings
of the 3rd international symposium on friction stir welding, Kobe,
Japan, 27 and 28 September 2001, TWI (UK); 2001.
[23] Dalle Donne C, Biallas G, Ghidini T, Raimbeaux G. Eect of weld
imperfections and residual stresses on the fatigue crack propagation
in friction stir welded joints. In: Proceedings of the second international conference on friction stir welding, 2628 June, Gorthenburg,
Sweeden; 2000.
[24] Strategic research agenda, Advisory council for aeronautics research
in Europe, vol. 2, October; 2004.
[25] Shot peening applications, Metal Improvement Company Inc., 7th
ed.; 1988.
[26] Vohringer O. Changes in the state of the material by shot peening. In:
Wohlfahrt H, Kopp R, Vohringer O, editors. International conference on shot peening, vol. 3; 1987. pp. 185204.
[27] Altenberger I, Scholtes B. Improvement of fatigue lifetime of
mechanically surface treated materials in the low cycle fatigue regime.
In: Brebbia CA, Kenny JM, editors. Surface treatments IV. London:
WIT Press; 1999. p. 28190.
[28] Dorr T, Wagner L. Fatigue response of various titanium alloys to
shot peening. In: Brebbia CA, Kenny JM, editors. Surface treatments
IV. London: WIT Press; 1999. p. 34957.

1545

[29] Romero JS. Ph.D. Thesis. The study of residual stresses due to shot
peening on aluminium alloys 2024 and 7150, The University of
Sheeld; 2002.
[30] Zhuang WZ, Halford GR. Investigation of residual stress relaxation
under cyclic load. Int J Fatigue 2001;23:S317.
[31] Curtis S, Rodopoulos CA, de los Rios ER, Levers A.
Analysis of the eects of controlled shot peening on fatigue
damage of high strength aluminium alloys. Int J Fatigue
2003;25:5966.
[32] Ordieres JM. Investigating the stability of residual stresses induced
by controlled shot peening on 7150-T651 aluminium alloy
subjected to cyclic loading, M.Sc. Thesis, The University of
Sheeld; 2003.
[33] Rodopoulos CA, Edwards R, Curtis SA, Romero JS, Choi J, de los
Rios ER, et al. In: Wagner L, editor. Theoretical analysis of
benecial and detrimental eects of controlled shot peening in high
strength aluminium alloys, 8th ICSP. New York: WileyVCH; 2003.
p. 54753.
[34] Rodopoulos CA, Curtis SA, de los Rios ER, Romero JS. Optimisation of the fatigue resistance of 2024-T351 aluminium alloys by
controlled shot peening methodology, results and analysis. Int J
Fatigue 2004;26:84956.
[35] Nitschecke-Pagel T, Wohlfahrt H. In: Wagner L, editor. Fatigue
strength improvement of welded aluminium alloys by dierent postweld treatment methods, 8th ICSP. New York: WileyVCH; 2003. p.
3606.
[36] Altenberger I. In: Wagner L, editor. Alternative mechanical surface
treatments: microstructures, residual stresses and fatigue behaviour,
8th ICSP. New York: WileyVCH; 2003. p. 42134.
[37] Ali A. Improving the fatigue life of aircraft components by using
surface engineering, Ph.D. Thesis, The University of Sheeld;
2005.
[38] James MR. The relaxation of residual stresses during fatigue. In:
Kula E, Volker Weiss, editors. Proceedings of the 28th Sagamore
army materials research conference. New York: Plenum Press; 1981.
p. 297314.