Fatigue & Fracture of
Engineering Materials & Structures
doi: 10.1111/j.1460-2695.2008.01234.x
Fatigue behaviour of friction stir welded AA2024-T3 alloy: longitudinal
and transverse crack growth
M. T. MILAN, W. W. BOSE FILHO, C. O. F. T. RUCKERT and J. R. TARPANI∗
Department of Materials, Aeronautics and Automotive Engineering, Engineering School of São Carlos, University of São Paulo, Av. Trabalhador
São-Carlense, 400, Centro, CEP. 13.566-590, São Carlos-SP, Brazil
Received in final form 20 April 2007
A B S T R A C T The fatigue crack growth properties of friction stir welded joints of 2024-T3 aluminium
alloy have been studied under constant load amplitude (increasing-K), with special
emphasis on the residual stress (inverse weight function) effects on longitudinal and
transverse crack growth rate predictions (Glinka’s method). In general, welded joints
were more resistant to longitudinally growing fatigue cracks than the parent material at
threshold K values, when beneficial thermal residual stresses decelerated crack growth
rate, while the opposite behaviour was observed next to K C instability, basically due to
monotonic fracture modes intercepting fatigue crack growth in weld microstructures.
As a result, fatigue crack growth rate (FCGR) predictions were conservative at lower
propagation rates and non-conservative for faster cracks. Regarding transverse cracks,
intense compressive residual stresses rendered welded plates more fatigue resistant than
neat parent plate. However, once the crack tip entered the more brittle weld region
substantial acceleration of FCGR occurred due to operative monotonic tensile modes of
fracture, leading to non-conservative crack growth rate predictions next to K C instability.
At threshold K values non-conservative predictions values resulted from residual stress
relaxation. Improvements on predicted FCGR values were strongly dependent on how
the progressive plastic relaxation of the residual stress field was considered.
Keywords aluminium alloy; crack growth rate prediction; fatigue; friction stir welding;
residual stress.
NOMENCLATURES
∗
a = crack length
AA2024-T3 = high-strength aluminium alloy grade
d = slot aperture
da/dN = crack growth rate
E = plane-stress Young’s modulus
E′ = plane-strain Young’s modulus
EL = elongation at fracture
FCGR = fatigue crack growth rate(s)
FSW = friction stir welding
h(x,a) = weight function
HAZ = heat-affected zone
K C = critical stress intensity factor
K MAX = maximum applied stress intensity in fatigue
K Ir = residual stress intensity factor in mode I of crack opening
Correspondence: J. R. Tarpani. E-mail: jrpan@sc.usp.br
526
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
527
K rx , K ry = residual stress intensity factor distribution on x and y directions
L = length of generic test piece
L 0 = original gage length
M = location of strain-gage
R = stress ratio
R′ = effective stress ratio
RA = reduction in area at fracture
S = engineering, nominal, remote or gross stress
S, L(x), T(y) = three main orthogonal metallographic axes or directions
TP = test piece
TMAZ = thermo-mechanically affected zone
UTS = ultimate tensile strength
W = width of generic test piece
WEDM = wire electro-discharge machine
YS = yield strength
Z(a) = influence function
K (th) = range of stress intensity factor in fatigue (threshold value)
ε = strain
ε M = strain measured at M position
ν = Poisson’s coefficient
σ rx , σ ry = residual stress distribution on x and y directions
INTRODUCTION
In recent years, friction stir welding (FSW), a solid-state
joining technique, has been considered a potential technique to replace conventional riveting operations and fusion welding methods (e.g. laser and electron beam) in
aircraft manufacture. However, the weld still results in
a continuous medium for crack propagation and hence,
knowledge of the fatigue and fracture properties of such
classes of materials is vital if a damage-tolerant design is
adopted.
Advantages of FSW process include: design simplification (easy periodical inspection, less macro stressconcentrators), low distortion, poreless welding process
(less micro-stress concentrators), static strength as high
as 80–100% of parent material, improved fatigue performance and better load distribution. FSW disadvantages
comprise: lack of extensive data on mechanical properties, continuous crack propagation medium, mismatching
between plastic properties of weld and parent metals and
residual stress effects.
Residual stresses are invariably present in welded structures after fabrication. They are likely to affect mechanical
and corrosion properties of the materials and therefore
influence the in-service performance of structural components. The effects of residual stresses on fatigue crack
propagation have been reported by several authors such
as Itoh et al.,1 Bussu and Irving2 and Milan and Bowen.3,4
Based on Parker’s superposition principle,5 they concluded that tensile residual stresses increase the crack
growth rate due to increasing effective stress ratio (R′ ).
On the other hand, compressive residual stresses reduce
the fatigue crack growth rate (FCGR) by decreasing the
effective stress ratio. Additionally, residual stresses were
found to affect initiation fracture toughness values (K C )
of aluminium alloys.6,7
Transverse cracks present an even more complex situation, inasmuch as the defect propagates from the parent
material towards the weld region. The crack tip intersects different microstructures and distinct intrinsic fatigue cracking behaviours can be expected.
This paper presents data obtained from studying the fatigue crack resistance of FSW joints of aeronautical grade
AA2024-T3 high-strength alloy containing either longitudinal or transverse cracks. The effective stress ratio
method (Glinka’s R′ )8 was employed to predict FCGR in
the welded alloy taking into account the residual stress intensity factor calculated by the slitting or cut compliance
method9 and parent material fatigue properties. Predicted
crack growth rates are then compared to experimental
values.
The authors expect to contribute to the still limited
body of knowledge regarding FSW materials by providing useful fatigue crack growth data for both safe-life and
damage-tolerant designs.
INCREMENTAL SLITTING AND WEIGHT
FUNCTION METHODS
The cut compliance or incremental slitting method is a
helpful technique to determine both the near surface and
through thickness residual stress profiles. It is based on
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
528
M . T . M I L A N et al.
the fact that when a cut, simulating a crack, is incrementally introduced into a part, the residual stresses are relieved, causing the part to deform. Such deformation can
be sensed by strain gauges attached at specific positions
of the part (Fig. 1) and the residual stress intensity factor
profile can be derived.9,10 Assuming a sufficiently narrow
slot (d ≪ a), linear elastic fracture mechanics (LEFM) can
be employed to establish a relationship (Eq. 1) between
the measured strains, ε, and the corresponding residual
stress intensity factor in opening mode, K Ir :9
K Ir (a) =
E ′ d εM
,
Z(a) d a
(1)
where ε M is the measured strain at the back face position M during the cutting procedure, a is the slot length,
E′ , the generalized form of the Youngs modulus (E′ = E
for plane-stress, and E′ = E/1 − ν 2 for plane-strain conditions), and Z(a) the ‘influence function’ that depends
on the test-piece geometry, cut plane location and strain
measurement position. Here Z(a) is considered as being
independent of the residual stress profile.
For a rectangular plate, where L > 2W , and taking strain
measurements at the back face (position M), Z(a) is given
as follows:11
- for a/W < 0.2 (shallow crack):
Z(a) =
−2,532
(W−a)1.5
1 − 25 ·
5.926 · 0.2 −
a
W
a
W
2
2
− 0.2 ·
− 0.288 · 0.2 −
a
W
+1
stress distribution in the flawless test specimen, σ r :12
K Ir (a) =
a
h(x, a) · σr (x) · d x
(4)
0
Insofar as the weight function h(x,a) is universal for
a given crack geometry, one can determine the inverse
weight function, so that σ r can be determined from K Ir
(i.e. inverse path).
Therefore, by employing the slitting technique and the
incremental stress method (inverse weight function) it is
possible to obtain the original residual stress profile in
the uncracked body, as well as the residual stress profile
redistribution due to crack growth.
MATERIAL AND TEST SPECIMENS
AA2024-T3 3.2 mm-thick plates were friction stir welded
along the rolling direction (Fig. 2). Welding parameters are proprietary to the Brazilian aircraft manufacturer,
Embraer S/A. Basic tensile properties of the parent alloy
are provided in Table 1 for both longitudinal and transverse directions. Tri-dimensional views of, respectively,
chemically etched and unetched microstructures of the
material tested are shown in Fig. 3. Full-plate-thickness
test pieces with in-plane dimensions of 60 × 120 mm2
were wire electro-discharge machined according Figs. 4
and 5 for, respectively, longitudinal and transverse crack
growth specimens.
No post-weld heat treatments were applied to the FSW
plates, so that all experiments were carried out after all
natural aging had ceased.
(2)
EXPERIMENTS AND ANALYSIS
- for 0.2<a/W < 1 (deep crack):
Z(a) =
−2.532
(W − a)1.5
Residual stress intensity factor evaluation
(3)
Equation (4) dictates the relation between the residual
stress concentration factor, K Ir (Eq. 1), and the residual
Fig. 1 Rectangular plate containing a slot along the central plane,
with strain measurements taken at the location M.
A strain gauge was attached to the back face of each test
piece, as shown in Figs. 4 and 5. The strain gauge and connection wires were covered with a plastic resin in order
to avoid reaction with the environment. Shielded wires
Fig. 2 Freshly friction stir welded plates tested in this study.
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
529
Table 1 Mechanical properties of the Al-alloy tested at ambient temperature
Orientation/Property
E (GPa)
UTS (MPa)
YS (MPa)
ELa (%)
RA (%)
AA2024-T3 longitudinal
Standard deviation
AA2024-T3 transverse
Standard deviation
79
4.6
81
2.0
477
4.2
463
3.8
350
6.0
308
4.5
21.1
0.36
21.4
0.31
20.8
2.46
23.0
1.45
Average of three test pieces for each specimen orientation.
a L = 25 mm.
0
Fig. 3 Typical microstructures of the rolled A1-2024-T3 alloy: (a) Keller’s etching; (b) Simply polished. Note different image
magnifications (approximately 2:1).
Fig. 4 T-L testpiece configurations for measuring transverse residual stress intensity factor on y direction, K ry , and for fatigue crack
propagation tests along the rolling or longitudinal (x) direction: (a) Slot at the centre of the weld line (0 mm), and (b) 5 mm displaced from
zero position. Tool shoulder width is 16 mm. FSW tools travelling from the bottom to the top of the page and rotating in counter-clockwise
direction.
were used to connect the strain gauges to the strain data
acquisition apparatus in order to minimize electromagnetic interferences. Plate cutting was performed by wire
electro-discharge machining (WEDM) either longitudi-
nally or transversally to the weldline. For the former case,
as depicted in Fig. 4, the slot was introduced in two different positions: 0 mm and 5 mm distant from the weld
nugget centre line (on the FSW tool advancing side), to
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
530
M . T . M I L A N et al.
Fig. 5 L-T testpiece configuration for measuring longitudinal residual stress intensity factor on x direction, K rx , and for fatigue crack
propagation tests along transverse (y) direction. Strain measurments are taken at the point M. Tool shoulder width is 16 mm.
simulate a crack growing in the weld and in the heat affected zone, respectively. Recent work by Milan et al.13
has found WEDM more practical, precise and less likely
to introduce additional stresses compared to abrasive saw
method. In all experiments, cutting increments of 0.5 mm
were chosen. Readings were taken 5 min after each single
slotting procedure had finished.
After the strain data were obtained as a function of the
slot length, the secant method was used to derive dε/da
data, which were then employed in Eq. (1) to determine
corresponding K ry and K rx values, for longitudinal and
transverse cracks, respectively.
Fatigue crack growth tests
A 5-mm-long notch was introduced at the edge of the
welded test pieces, in the positions indicated in Figs. 4 and
5. Once a 5-mm-long fatigue pre-crack emanated from
the notch tip, mode I fatigue loading tests were conducted
at ambient temperature under constant amplitude loading
condition (increasing-K testing). Sinusoidal waveform
under a frequency of 60 Hz was applied to all the test
specimens. The parent metal was tested under stress ratios
(R) of 0, 0.3 and 0.5, respectively, in order to provide data
for the Glinca’s R′ method predictions, while the weld
metals where tested at R = 0.5 only. A detailed description
of the R′ method can be found in Milan and Bowen.4
RESULTS AND DISCUSSION
Residual stress intensity factor profile
Longitudinal cracking
Figure 6a presents transverse residual stress intensity factor profiles, K ry , obtained for longitudinal slot test pieces.
For both cases, that is, for the slot introduced at 0 mm and
5 mm from the weld centre line, K ry values remain low
and ‘negative’ in the entire range of acquired data. Rigorously speaking, a ‘negative K r value’ does not exist for a
closed crack, but it means that when an external load is applied, the thermal residual stress profile shields the crack,
that is, effectively reduces the locally applied K factor,
and hence decelerates FCGR. Clearly, this effect is a welcome safety margin against failure in damage-tolerant approaches. It is observed that K ry profiles vary only slightly
from one test piece to another, so that no systematic trend
can be established regarding the slot position in relation
to the weldline, which means that nugget and thermomechanically affected zone (TMAZ) microstructures are
similar to some extent with regard to fatigue properties.
Using the inverse weight function method, as detailed
by Schindler,9 it was possible to obtain the initial residual
stress profile present in the material before the slot was
introduced, as depicted in Fig. 6b. Tensile residual stress
in the centre of the test piece and balancing compressive residual stresses near the edges can be noticed. From
Fig. 6, it can be observed that compressive residual stresses
may delay crack nucleation at the edge of a pristine component. On the other hand, tensile residual stresses may
favour nucleation at the centre of the piece. These findings have important implications in safe-life approaches.
Transverse cracking
Figure 7a shows longitudinal residual stress intensity factor profile, K rx , for test pieces with the slot axis perpendicular to the weldline. A ‘negative’ peak value develops
at approximately 16 mm from the edge of the test piece,
so that K rx remains negative up to the weld centre line
(30 mm position from the specimen edge). Therefore,
the FCGR of a growing crack approaching perpendicularly to the weldline is expected to be slowed down.
However, after the crack crosses the weld region (i.e. positive K rx values in Fig. 7a) the FCGR will accelerate as
compared to the crack growth rate in the parent material
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
531
Fig. 6 (a) Transverse residual stress intensity factor profiles, and (b) Initial residual stress profiles derived for longitudinal slot testpieces.
under identically applied stress intensity factor range and
R ratio.
In terms of non-destructive inspection programs of
damage-tolerant structures, the above provided results
point out the need for a much more complex surveillance
schedule for welded structures as compared to monolithic
(non-welded) ones in order to guarantee structural integrity for both short- and long-fatigue growing cracks.
Figure 7b depicts the original residual stress profile in the
flawless material, that is, before the slot introduction, as
derived through the inverse weight function method.9 It is
possible to observe the tensile residual stress approaching
80% of the yield strength of the parent plate in the weld
region (in the advancing side of the tool), while near the
edges the residual stresses remain compressive. A likely
crack nucleation at the TMAZ/HAZ zone can then be
predicted.
FCGR curves (experimental vs. predicted)
Longitudinal cracking
Fatigue crack propagation rate curves of longitudinally
slotted test pieces are presented in Fig. 8. For the same
applied K level, a crack positioned on the centre of
the weld nugget (0 mm position) exhibits fatigue growth
rates slightly lower than a crack located along the TMAZ
at about 4 mm from the weld centre line. Compared to
the parent material however, all the welded test pieces
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
532
M . T . M I L A N et al.
Fig. 7 (a) Residual stress intensity factor, and (b) Corresponding initial residual stress profiles for transverse slot testpieces.
presented higher fatigue crack growth resistance at low
applied K values. This is consistent with the presence
of a ‘negative’ peak of K ry at a slot length of order of
5 mm (Fig. 6a). Nonetheless for higher applied K val√
ues (above 10 MPa. m), that is, longer slots (cracks) and
faster crack growth rate, although K ry remains negative
during the entire range of testing, the welded material presented higher values of FCGR. This leads to the assumption that whereas microstructure and residual stresses act
simultaneously in the threshold region, at K values approaching K C , monotonic fracture modes (dimple and/or
cleavage micromechanisms) intercept fatigue crack propagation, that is, they accelerate FCGR as compared to
the parent plate behaviour. It is worth mentioning that
at these crack tip positions only microstructural effects
can play a role, because K ry values approach zero (refer to
Fig. 6a).
Figure 9 shows the fatigue crack propagation rates estimated via the so-called R′ method, for a test piece containing a crack in the weld centre line. For K values
√
below 8 MPa. m, predicted values are conservative compared to the experimental ones, whereas for K values
√
above 10 MPa. m, predicted values are otherwise nonconservative. Because the prediction model takes into
account only the parent metal fatigue properties and the
K ry profile of the welded material, it is possible that
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
Fig. 8 Fatigue crack propagation curves of parent and welded
materials, with cracks propagating along the weld centre line (0
mm) and TMAZ (5 mm), respectively.
Fig. 9 Experimental and predicted FCGR of the welded material,
with the crack propagating along the weld centre line (0 mm).
the microstructure of the weld nugget is responsible for
the differences observed between predicted and experimentally measured figures. The nugget region is formed
by fine equiaxed dynamically recrystallized grains,2 while
the parent metal presents elongated grains due to the
rolling process, as can be seen in Fig. 10. Therefore, it is
533
likely that the intrinsic fatigue crack growth resistance of
the nugget microstructure differs significantly from that
of the parent plate. Such a hypothesis is supported by
the fact that the discrepancies between predicted and experimental values are higher at low K values and near
the final fracture region, where FCGR is much more dependent on microstructural features. Conversely, in the
Paris region, where microstructure plays a minor role in
FCGR, there is good correlation between experimental
and predicted values.
Briefly, recalling that at the threshold region (i.e. next
to the K th value) conservative FCGR predictions were
derived for the weld microstructure on the basis of the
intrinsic fatigue behaviour of the parent plate, the results
strongly suggest that at low applied K values the former
microstructure is more damage tolerant than the parent
metal. On the other hand, near the final catastrophic
fracture region (i.e. approaching K C value), the nonconservatism of FCGR predictions for the weld nugget
indicates that monotonic fracture modes (i.e. K MAX controlled mechanisms) are not precisely accounted for
by Glinka’s model, and evidences the parent metal as the
toughest microstructure at high applied K values.
For cracks positioned on the TMAZ and subjected to
low K values, predicted and experimental FCGR are
in good agreement (Fig. 11), indicating that the TMAZ
microstructure exhibits the same intrinsic fatigue crack
growth resistance as the parent material. It should be
mentioned that as the crack grew during the increasingK tests, it deflected towards the interface between the
TMAZ and the heat affected zone (HAZ), as seen in
Fig. 12. As a result, FCGR predictions at higher K values were underestimated (i.e. non-conservative approach).
Assuming that residual stress profiles for both TMAZ
and HAZ regions are similar, the underestimated prediction suggests that the microstructure of the TMAZ/HAZ
interface offers lower resistance to fatigue crack propagation than the parent plate material probably due to a
lower fracture toughness value, which would contribute
to stronger K MAX effect on fatigue crack growth resistance. This point will be revisited in the next section for
transverse cracks. On the basis of Fig. 6a data, one could
argue that a less compressive stress field would have also
contributed to higher crack growth rates. Regardless of
fact that the accelerated FCGR was generated by either
a less tough microstructure or a less compressive stress
field, or even by both effects simultaneously, the truth is
that the experiment did fail to test Glinka’s model.
Transverse cracking
Fatigue crack growth curves for cracks running perpendicularly to the weldline are presented in Fig. 13. Up
to a position corresponding to the border of the tool
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
534
M . T . M I L A N et al.
Fig. 10 Typical microstructures developed in FSW aluminium alloy. FSW tool advances in the right hand side region. The nugget consists
of recrystallized equiaxial grains, TMAZ of stretched grains due to rotational tool movement, and HAZ possesses similar appearance to the
parent metal.
Fig. 12 Longitudinal fatigue crack growth specimen showing crack
deflection towards the TMAZ/HAZ interface distant
approximately 8 mm from weld centre line. Penetrant-dye testing
was performed in order to facilitate crack visualization.
Fig. 11 Experimental and predicted FCGR of welded material,
with the longitudinal crack propagating 5 mm from the weld centre
line.
shoulder, results clearly show that the welded joint
presents higher fatigue crack growth resistance than the
parent material, for the same applied K value, or K MAX
value since the nominal R value is the same. This is mainly
attributed to the strong ‘negative’ K rx values shown in
Fig. 7a, in a position still located in the parent metal. A
‘negative K rx value’ reduces both the stress ratio, R, and
the effective K values, so decreasing the net crack tip
driving force. However, when the crack tip indeed enters the weld region, FCGR are found to be higher for
the welded joint than for the parent material, although
the K rx values still remain considerably negative (Fig. 7a).
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
Fig. 13 Fatigue crack propagation curves of welded joint and
parent material, with crack growing transverse to the weld line.
This fact suggests that the weld microstructure is brittle than the parent metal because at this position residual
stresses approach zero.
Figure 14 provides Vickers microhardness profile transverse to the weldline at the mid-thickness position of the
welded joint. Although it agrees quite well with literature
data for 2024-T3 Al-alloys,14 the relatively soft nugget
could lead to the conclusion of a tougher microstructure
if compared to the parent plate. However, recent study
by Kamp et al.15 has shown that alloys containing precipitate distributions that are unstable at elevated temperature, such as aerospace aluminium alloys, are quite prone
to profound microstructural changes when friction stir
535
welded. These include significant intermetallic particles
coarsening at grain boundaries in the HAZ and nugget
regions, hence leading to poor fracture properties eventually associated to only modest bulk hardness values.
Furthermore, non-conservative Glinka’s predictions
given in Fig. 15 endorse the assumption of brittle nugget
microstructure. Inasmuch as the R′ method takes into
account the fatigue properties of the parent material,
it is possible, in the final fracture region (i.e. K MAX →
K C ), that the low-initiation fracture toughness (K C value)
of the nugget microstructure enhances the FCGR by
strengthening monotonic tensile modes of fracture, resulting in somewhat underestimated FCGR predictions.
This was verified as well during longitudinal cracking
analysis, and it seems again that Glinka’s model is not
able to cope with this particular condition. At low applied
K values, however, a possible microstructural effect is
discarded because the crack tip is still in the parent material. Thus, in this case, it is more likely that residual stress
relaxation is taking place due to the cyclic load application, which would result in non-conservative predictions.
The R′ method is only valid if linear elastic conditions are
maintained during the test, that is, if crack tip plasticity is
small. If this condition is broken, the obtained K rx profile
is no longer the best choice to carry out the prediction.
Moreover, it deserves to be emphasized that FCGR prediction is based on initial K Ir profile, which presents high
‘negative’ values as shown in Fig. 7a. In order to verify
whether there was plastic relaxation or not, both K rx and
corresponding σ rx profiles were again derived from the
broken halves of a fatigued test piece.
Results exhibited in Fig. 16 indicate that the material indeed suffered significant stress relaxation by plastic flow.
This is in line with recent findings from Liljedahl et al.,16
in which good prediction of residual stress redistribution was achieved for growing cracks at constant stress
Fig. 14 Vickers hardness profile transversal
to the friction stir weld line.
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
536
M . T . M I L A N et al.
inasmuch as the relaxation takes place progressively as
loading cycles are applied to the test samples. This can be
confirmed in Fig. 16b, when a single load cycle is enough
to cause a detectable stress relaxation and hence a rearranged residual stress profile.
Last but not least, it should be noted that the transverse crack growth case illustrates very well the limitation of the presently adopted approach in identifying
the main controlling factor of FCGR and drawing conclusive facts about it. Aimed at separating the effects of
the magnitude of the residual stresses and the weld microstructure on fatigue behaviour of the FSW plates,
FCGR experiments employing pre-strained test pieces
are being carefully planned. FCGR testing of these stressrelieved specimens certainly will clarify current uncertainties and allow the role of the secondary stresses to be fully
understood.
CONCLUDING REMARKS
Fig. 15 FCGR predictions for the welded joint compared to
experimental values, with crack growing transverse to the weld line.
intensity range, a scenario farther from situations in practice than that studied in this work (i.e. increasing-K
testing).
The maximum nominal stress initially applied to the test
piece was approximately 80 MPa. By adding this value
to the peak tensile residual stress observed on a position
corresponding to the advancing side of the rotating tool
(Fig. 7b), an effective tensile stress approaching 300 MPa
is obtained (Fig. 16b). Such stress level is certainly high
enough to cause plastic deformation as the elastic limit
of the material is around 250 MPa. Plastic deformation reduces both the peak tensile residual stress and
the compressive residual stress near the edges of the test
piece because a zero-stress global balance must be maintained. Thus, the initial K rx values calculated considering the non-relaxed residual stress profile are likely to be
overestimated, resulting in non-conservative prediction of
FCGR.
A new FCGR prediction of the welded joint based on the
relaxed K rx profile is given in Fig. 17. Results indicate that
even considering the residual stress relaxation, forecast
values for the final fracture region within the weld region
are still non-conservative, so confirming that monotonic
fracture modes do prevail at higher K values. However,
√
for K values below 17 MPa. m it is now observed that
predicted values are invariably higher than experimental
ones, evidencing that the relaxed K rx profile is now underestimated. Therefore, the real K rx profile during fatigue
testing must range from that obtained with the intact
test pieces to that derived from the test specimens halves,
Longitudinal and transverse fatigue crack growth properties of friction stir welded joints of thin rolled plates of
2024-T3 aluminium alloy have been studied under constant load amplitude conditions (increasing-K).
Corresponding residual stress intensity factors in mode I
(K Ir ) profiles were determined using a fracture mechanics
approach, and the inverse weight function method was
utilized to obtain the residual stress profiles.
Fatigue properties of the parent metal were employed
alongside K Ir profiles and Glinka’s R′ method in order to
predict FCGR of the welded joints.
Main conclusions that may apply to the weld region, but
not to necessarily a welded component as a whole, are as
follows:
1. Welded joints are more resistant to longitudinally growing fatigue cracks than the parent metal at threshold
K values probably due to beneficial thermal residual stresses arresting crack propagation, but less resistant next to the final instability point probably owing
to monotonic tensile fracture modes intercepting subcritical crack growth in weld microstructures. Consequently, FCGR predictions trend to be conservative at
lower propagation rates and non-conservative for faster
cracks;
2. Intense compressive residual stresses render welded
plates more fatigue resistant to transverse cracks than
unmixed parent plate. Nonetheless, once the crack
tip enters the brittle weld region FCGR accelerates
due to operative monotonic tensile modes of fracture,
leading to non-conservative crack growth rate predictions next to K C instability. At threshold K values
non-conservative predictions values result from residual
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
FATIGUE BEHAVIOUR OF FRICTION STIR WELDED AA2024-T3 ALLOY
537
Fig. 16 Comparison between K rx profiles
obtained from intact (TPs 1 e 2) and already
fatigued testpieces (TPs 3 e 4).
stress relaxation. Improvements on predicted growth
rate values are strongly dependent on how the progressive plastic relaxation of the residual stress field is
considered;
3. With regard to design approaches broadly utilized in the
aviation industry, the following findings should be noted:
(i) If a damage-tolerant approach (stress intensity factor,
K ) is applied, a growing crack is likely to be less detrimental when advancing towards, rather than along the
weldline; (ii) If a safe-life approach (engineering stress,
S) is employed, a crack is more likely to initiate transversally to the weldline in a pristine structure, rather than
longitudinally to it;
4. Only complementary FCGR testing of pre-strained (i.e.
stress-relieved) test pieces will permit the full understanding of the role of the secondary stresses on the fatigue
behaviour of longitudinal and transverse FSW joints.
Acknowledgements
M. T. MILAN would like to thank Fapesp (Fundação para
o Amparo à Pesquisa do Estado de São Paulo, Brazil) for
Grants 02/09027-4 and 03/11059-4. Thanks are also to
Embraer S/A (Brazil) for providing the materials tested in
this work.
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538
538
M . T . M I L A N et al.
5
6
7
8
9
10
11
Fig. 17 FCGR predictions for the welded joint (crack growth
transverse to the weld line) compared to the experimental values,
considering the K rx profiles obtained, respectively, from intact
testpieces and broken halves of fatigue specimens.
12
13
REFERENCES
1
2
3
4
Itoh, Y. Z., Suruga, S. and Kashiwaya, H. (1989) Prediction of
fatigue crack-growth rate in welding residual-stress field.
Engng. Fract. Mech. 33, 397–407.
Bussu, G. and Irving, P. E. (2003) The role of residual stress
and affected zone properties on fatigue crack propagation in
friction stir welded 2024-T351 aluminium joints. Intl. Jnl.
Fatigue 25, 77–88.
Milan, M. T. and Bowen, P. (2002) Effects of particle size,
particle volume fraction and matrix composition on the
fatigue crack growth resistance of Al alloy / Al alloy + SiC
bimaterials. Proceedings of the Institute of Mechanical
Engineering L. Jnl Mater. Des. Appl. 216, 245–255.
Milan, M. T. and Bowen, P. (2003) Experimental and
predicted fatigue crack growth resistance in Al2124/Al2124 +
35%SiC biomaterial. Intl. Jnl. Fatigue 25, 649–659.
14
15
16
Parker, A. P. (1984) An overview of the mechanics of fracture
and fatigue in the presence of residual stress. J. Mech. Work.
Tech. 10, 165–174.
Milan, M. T. and Bowen, P. (2004a) Fracture toughness of
selectively reinforced Al2124 alloy: precrack tip in the
composite side. Metal. Mater. Trans. A 35A, 1393–1401.
Milan, M. T. and Bowen, P. (2004b) Fracture toughness of
selectively reinforced Al2124 alloy: precrack tip in the
aluminium alloy side. Mater. Sci. Tech. 20, 783–789.
Glinka, G. (1979) Effect of Residual Stresses on Fatigue Crack
Growth in Steel Weldments under Constant and Variable
Amplitude Loads, ASTM Special Technical Publication (STP)
677, American Society for Testing and Materials,
Philadelphia, PA, USA,2 pp. 198–214.
Schindler, H. J. (1996) Determination of residual stress
distributions from measured stress intensity factors. Intl. Jnl.
Fract. 74, R23–R30.
Prime, M. (1999) Residual stress measurement by successive
extension of a slot: the crack compliance method. Appl. Mech.
Rev. 52, 75–96.
Schindler, H. J. and Bertschinger, P. (1997) Some step
towards automation of the crack compliance method to
measure residual stress distributions. In: Proceedings of the 5th
International Conference on Residual Stresses ICRS-5. (Edited by
T. Ericsson, M. Oden and A. Andersson), Linkoping,
Sweden, Vol. 1.
Fett, T. and Munz, D. (1997) Stress Intensity Factors and
Weight Functions, Computational Mechanics Publications,
Southampton, UK.
Milan, M. T., Bose Filho, W. W., Malafaia, A. M. S., Silva, C.
P. O. and Pellizer, B. C. (2006) Slot machining effects on
residual stress measurements using the crack compliance
method. Jnl. Test. Eval. 34, 149–152.
Khaled, T. (2005) An Outsider Looks at Friction Stir Welding.
Report #: ANM-112N-05-06. Federal Aviation
Administration, Lakewood, USA.
Kamp, N., Sullivan, A., Tomasi, R. and Robson, J. D. (2006)
Modelling of heterogeneous precipitate evolution during
friction stir welding process. Acta Mater. 54, 2003–2014.
Liljedahl C. D. M., Brouard, J., Zanellato, O., Lin, J., Tan, J.
F., Ganguly, S., Irving, P. E., Fitzpatrick, M. E., Zhang, X.
and Edwards, L. (2007) Weld residual stress effects on fatigue
crack growth behaviour of aluminium alloy 2024-T3. In:
Proceedings of the First International Conference on Damage
Tolerance of Aircraft Structures. (Edited by R. Benedictus, J.
Schijve, R. C. Alderliesten and J. J. Homan), Delft, The
Netherlands.
c 2008 The Authors. Journal Compilation
c 2008 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct . 31, 526–538