Journal of Materials Processing Tech.

263 (2019) 256–265

Contents lists available at ScienceDirect

Journal of Materials Processing Tech.

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

Weld zone and residual stress development in AA7050 stationary shoulder T

friction stir T-joint weld
⁎
T. Suna, , M.J. Royb, D. Stronga, C. Simpsonc, P.J. Withersa, P.B. Prangnella
a
School of Materials, the University of Manchester, Manchester M13 9PL, UK
b
School of Mechanical, Aerospace and Civil Engineering, the University of Manchester, Manchester M13 9PL, UK
c
Now at the Department of Mechanical Engineering, University of Bristol, Bristol BS8 1TH, UK

A R T I C LE I N FO A B S T R A C T

Keywords: This paper shows the potential of stationary shoulder friction stir welding (SSFSW) for producing higher quality
SSFSW T- section joints relative to a conventional friction stir welding (FSW) approach. The residual stress distributions
AA7050 and their relationship to the weld zone microstructure and hardness distributions in SSFSW T-joints were in-
Neutron diffraction vestigated, as a function of welding travel speed. The final longitudinal residual stress fields were asymmetric,
Fillet weld
although individual weld zones showed similarities to those for conventional butt SSFSWs. The thermal ex-
Multi-pass
cursion and plastic strain arising from the second pass lowered the residual stresses seen from the first pass, so
that the largest tensile stresses (∼160 MPa) were found close to the nugget from the second weld pass. The
asymmetry in hardness distribution was caused by the thermal field of the second pass which thermally treated
material in the first pass and resulted in areas of age hardening and increased over ageing, depending on the
position of overlap of the thermal fields. The effects of the second weld pass on the first pass were more apparent
when a lower travel speed was used owing to the increase in heat input and duration of the thermal cycle.

1. Introduction
In transport applications, such as aircraft fuselages and railway
rolling stock, where high stiffness and low weight are required, stif-
fened panels are widely fabricated from aluminum alloys by attaching
extruded ‘stiffeners’ to a skin sheet. This requires the production of
mechanically efficient T-joints between the skin and stiffeners. For high
strength aluminum alloys, which are considered to be ‘un-weldable’ by
fusion processes, joining stiffened panels is usually achieved by riveting
the extruded stiffener to the skin, which requires a substantial overlap
reducing the joint efficiency. Recently, a series of studies by Fratini
et al. (2006), 2009; Donati et al. (2009) and Buffa et al. (2009) have
shown that friction stir welding (FSW) has potential as an alternative
approach for performing T-joints in high strength aluminum alloys,
thereby avoiding the additional overlap weight, as well as the oppor-
tunity for fatigue initiation associated with fasteners. These authors
investigated a FSW T-joint configuration where the welding tool pin
penetrated through the skin into a heavily constrained stiffener, to form
a joint; as shown in Fig. 1(a). However, geometrical issues associated
with a conventional FSW tool meant that in all cases only non-ideal
weld geometries could be considered. Despite heavy clamping either
side of the stiffener, to constrain the material, Cui et al. (2013) reported
that weld defects such as lack of full bonding and ‘tunnels’ were difficult
to eliminate with the tool through-skin configuration. This was attrib-
uted to the difficulty in ensuring sufficient material flow across the full
stiffener width near the tip of the pin. Particular additional issues re-
ported with this set-up included: thinning of the skin sheet, the high
precision required for the clamping and tool geometries to avoid de-
fects, and the limited fillet radius achievable. Thin-gauge sheet has also
proven to be very difficult to join with this weld configuration.
More recently, it has been demonstrated by Russell (2008) that it is
possible to produce FSWs with a non-rotating stationary shoulder. For
conventional butt welds, stationary shoulder FSW (SSFSW) has been
found by Avettand-Fènoël and Taillard (2016); Wu et al. (2015); Sun
et al. (2018a), 2017 and Sun et al. (2018b) to have a number of ad-
vantages including: a lower heat input, much higher surface quality, a
narrower and more uniform through thickness weld zone and reduced
residual stresses. A further advantage of a non-rotating shoulder is that
it can be modified to produce a range of weld bead profiles and fillet
welds. Martin and Way (2011) were the first researchers to realize the
potential for producing fillet welded T-joints using SSFSW by adopting
a 90° profiled tool shoulder (Fig. 1(b)). Subsequently, Maltin et al.
(2014), Martin (2014) and Li et al. (2015) have confirmed that superior
defect-free T-joints can be made by SSFSW, compared to the earlier

⁎
Corresponding author.
E-mail addresses: whhitstz@gmail.com, tianzhu.sun@postgrad.manchester.ac.uk (T. Sun), philip.prangnell@manchester.ac.uk (P.B. Prangnell).

https://doi.org/10.1016/j.jmatprotec.2018.08.022
Received 25 March 2018; Received in revised form 17 August 2018; Accepted 17 August 2018
Available online 23 August 2018
0924-0136/ Crown Copyright © 2018 Published by Elsevier B.V. All rights reserved.
T. Sun et al.
Fig. 1. Schematic diagrams of T-joints produced by (a) conventional FSW with the tool penetrating the skin configuration (Buffa et al., 2009; Donati et al., 2009;
Fratini et al., 2006, 2009) and (b) SSFSW fillet welding approach proposed by Martin and Way, (2011).
attempts based on a conventional FSW tool, with improved fatigue
resistance and mechanical performance. However, very little research
has yet been published on the effect of the process parameters on the
microstructure development in SSFSW T-joints. Furthermore, there
have been no reports on the associated residual stresses, without which
it is not possible to assess the joint’s structural integrity, nor optimize
the weld process parameters.
With the SSFSW process, it is currently not possible to simulta-
neously weld from both sides of a T-joint, as some degree of pin overlap
is required to avoid defects associated with undispersed segments of the
original join line between the flange and plate (see Martin and Way
(2011)). Thus, SSFSW fillet welding is an asymmetric sequential
welding process. In FSW, it has been noted by Barnes et al. (2008) that
sequential weld passes can influence the microstructure, hardness dis-
tribution and residual stress introduced by the previous pass. Further-
more, the effect of a subsequent weld pass on the first pass is position-
dependent due to the spatial variation in the local thermal field and
plastic deformation. For example, Brown et al. (2009) have investigated
repeating fully overlaid multi-pass AA7050 FSW butt welds, in the
context of a weld repair scenario, and found that the nugget grain size
and hardness were insensitive to the number of passes, while the
hardness and tensile strength in the heat affected zone (HAZ) cumula-
tively diminished. Subsequent modelling by Robson et al. (2010) has
demonstrated that the near constant properties of the WN can be ex-
plained by the re-deformation reaching steady state at the high strains
experienced in each pass and full solutionisation occurring during each
repeated pass, whereas the reduction in strength in the HAZ strength
was caused by the accumulative effect of the repeated thermal excur-
sion on coarsening of the strengthening precipitates. Other researchers,
such as Simoncini et al. (2016) and Cabibbo et al. (2014) have in-
vestigated double sided FSW and shown that where the thermal field
from the second pass overlaps the first, the second pass effectively re-
heat treats the first pass, leading to an asymmetric hardness distribu-
tion, with both zones of re-solution treatment and additional overaging
seen in the previous welding pass.
As well as affecting the final microstructural condition, the dual-
pass FSWs required by a T-joint configuration would be expected to
impact on the final residual stress state, but no data is currently
available in the literature. As described by Withers (2007a), residual
Fig. 2. Schematic diagram showing (a) the welding fixture employed for producing the T-joint fillet welds and (b) the SSFSW tool.
257
Journal of Materials Processing Tech. 263 (2019) 256–265
stresses are an important factor in determining the service life of
welded structural components. In single SSFSW butt welds, Sun et al.
(2017) have demonstrated that the residual stresses introduced can be
lower than in the conventional FSW, because the stationary shoulder
results in a more uniform and lower heat input. However, for a T-joint
welding configuration, the residual stress distribution will be more
complicated and also affected by the welding sequence, which involves
a second welding pass from the opposite side of the stiffener. For ex-
ample, Fu et al. (2014) have investigated the longitudinal residual
stress distribution in dual pass metal inert gas (MIG) steel T-joints and
determined a higher residual stress in the second weld pass, even
though the input power was kept the same. A similar residual stress
distribution has been reported by Ahn et al. (2018) in AA2024 laser
welded T-joints, where the lower level of residual stress seen in the
workpiece side with first weld was attributed to partial relief and re-
distribution of the residual stresses during the subsequent weld pass.
To fill the current knowledge gap reviewed above, we have in-
vestigated the residual stresses formed in a typical aerospace alloy
AA7050 T-joint by neutron diffraction and the associated weld zone
microstructure as a function of tool travel speed, which is known to be
the most important process variable (see Sun et al., 2018a). To un-
derstand the accumulative effect of the second weld on the micro-
structure and hardness distribution in the first pass, the behavior of the
individual welds was also investigated sequentially. However, due to
limited access to the neutron source, it was only possible to map the
residual stresses in completed T-joints and discuss the results with the
information concluded from the sequential microstructure and hardness
evolution.
2. Experimental details
2.1. Welding procedure
6 mm gauge hot rolled AA7050-T7651 plates (nominal composition
5.7–6.7%Zn, 1.9–2.6%Mg, 2.0–2.6%Cu, ∼0.11%Zr, < 0.15%
Fe, < 0.12%Si, balance Al) were used for this study. Plates were ma-
chined into 126 mm × 300 mm sample workpieces for the ‘skin’ and
60 mm × 300 mm for the stiffener. Prior to welding, material was skim
machined (0.15 mm) from all edges to provide surfaces free from
T. Sun et al. Journal of Materials Processing Tech. 263 (2019) 256–265

Table 1 weld used for strain measurement and analysed at the same positions as
Matrix of the T-joint SSFSW parameters and the measured torque and welding shown in Fig. 3 (a), as recommended by Withers et al. (2007). The
power. gauge volumes used for the stress-free lattice spacing measurement
Designation Pass No. Rotation Travel Torque Welding were 2(LD) × 2(TD) × 2(ND) mm3 in all three directions. The full time
speed speed (Nm) power of flight diffraction spectra were analysed using a multiple peak fitting
(rpm) (mm/min) (kW) Pawley refinement (Pawley, 1981) and procedures for the residual
stress calculation have been discussed elsewhere by Withers (2007b).
1100/100-1 1st 1100 100 18 2.1
1100/100-2 2nd 1100 100 15 1.7
1100/200-1 1st 1100 200 21 2.4 3. Results and discussion
1100/200-2 2nd 1100 200 16 1.9
1100/300-1 1st 1100 300 24 2.7
3.1. Weld process parameters
1100/300-2 2nd 1100 300 21 2.4
1100/300-S 1st 1100 300 24 2.7
The steady state torque values measured from the FSW machine for
both weld passes required to produce the double sided fillet-welded T-
significant oxide. Welding was performed in a 90° steel clamping fix- joints are plotted in Fig. 4 (a), as a function of travel speed, and the
ture, using a tool shoulder with a 90 ° profile to fully contact and re- corresponding specific weld energy (input power per unit length) is
strain the material, as illustrated in Fig. 2. The tool had a 6.8 mm long provided in Fig. 4 (b). After calibration against a dynamometer, Wu
conical threaded tri-flat probe with a diameter of 8 mm at the root and (2017) has shown the accuracy of the torque recorded by the FSW
5.6 mm at the tip, as shown in Fig. 2 (b). The welds were produced machine employed was within 1%. Unsurprisingly, as seen in Fig. 4 (a),
under force control mode with a 25 kN downforce and the welding a higher welding torque (and thus power) was recorded as the travel
configuration shown in Fig. 2 (a). Both the skin and stiffener were in- speed was increased. However, the specific weld energy (or heat input)
serted into the fixture prior to welding, with the tool aligned with the derived from the torque still decreased at greater travel speeds. It is also
joint line of the two parts. Once the first weld pass was completed, the clear from Fig. 4 that there was a distinctive reduction in the measured
welded plate was taken out, flipped over and reinserted into the fixture. torque and specific weld energy between the first and second weld
The second pass was then performed after an hour to allow natural passes, regardless of the travel speed, by about ∼ 20%. For alloys like
cooling. To simplify the experimental set-up, the welding start positions AA7050-T76, the material flow strength is controlled by the distribu-
were kept consistent so the welding directions between two the passes tion of strengthening precipitates (η' and η). The thermal cycle involved
were opposed. The matrix of welding parameters employed is shown in in the first weld pass can thus significantly soften the material across
Table 1. some of the volume traversed by the tool in the second weld pass. As a
result, the material flow could be driven by a lower torque during the
second weld pass even though the welding parameters were kept the
2.2. Characterization
same. Equally, the thermal field overlap from the second weld pass will
heat treat some of the material in the region of the first weld pass and
To determine the size and shape of the weld zones, the samples were
hence can change its microstructure. These sequencing effects have
sectioned normal to the welding direction (WD) in the transverse-
been studied here by analyzing partial T-joints containing a single fillet
normal (TD-ND) plane and prepared metallographically by polishing to
weld and will be discussed further below.
a 0.25μm finish, followed by etching in Keller’s reagent for optical
microscopy. Observation of the precipitation in the weld nuggets (WN)
3.2. Weld zones
was performed using a Zeiss Sigma VP SEM in BSE mode, on unetched
samples, and large area EBSD maps of ∼0.19 × 6 mm with a 0.8 μ m
Representative TD-ND cross-section macro views of typical T-joint
step size were obtained to determine the grain size distribution across
welds produced by SSFSW with a travel speed of 300 mm/min are
the two weld nuggets. Specimens for the EBSD study were lightly
displayed in Fig. 5 for the first pass only and after completing the
electropolished in 20% perchloric acid in ethanol (-15 °C, 20 V) to
second fillet required to fabricate the joint. With the current welding
produce a strain free surface. Micro-hardness maps were acquired on
configuration, the ‘advancing side’ (AS: as defined by Thomas (1991))
the weld cross sections using a Struers DuraScan 80 automated Vickers
for the two weld passes was located in the stiffener (top probe edge in
hardness tester, with an applied load of 0.5 kg for 10 s after post-weld
Fig. 5), while the ‘retreating sides’ (RS) for both welds were in the skin
natural ageing for four months. Each hardness map contained ∼2100
(bottom probe edge). It can be seen that the weld produced by the
data points with 0.5 mm spacing between indents.
single pass (Fig. 5 (a)) shows very similar microstructural character-
istics to that of a conventional SSFSW (see Wu et al. (2015)); com-
2.3. Residual stress determination by neutron diffraction prising a weld nugget zone (WNZ), a thermo-mechanically affected
zone (TMAZ) and a heat affected zone (HAZ). After the second weld
Neutron diffraction measurements were performed using ENGIN-X, pass a ‘weld nugget overlap zone’ (WNOZ) can be observed, where
a time-of-flight diffractometer (Santisteban et al., 2006) at the ISIS severe material flow has occurred, approximately at the position of
neutron spallation source at the Rutherford Appleton Laboratory, UK. intersection of the pin tip profiles, which contains the WNZ from the
Residual strains were measured in the longitudinal (LD), transverse first weld pass where it has been overwritten by the second pass. It can
(TD) and normal (ND) directions in step-wise scans across both the be noted that the shape of the weld nuggets are well matched to the tool
stiffener and skin. The measurement locations across the T-joint welds pin profile, as has been reported by Sun et al. (2017) for SSFSW butt
are shown in Fig. 3 (b). The gauge volumes were 2(LD) × 2(TD) × welding with similar parameters. No significant difference in WNZ size
2(ND) mm3 for the longitudinal strain measurement and 10(LD) × was observed between the two weld passes, despite the slightly dif-
2(TD) × 2(ND) mm3 for both transverse and normal residual strain ferent heat inputs shown in Fig. 4.
measurements. To minimize the beam attenuation through the sample,
the specimen was rotated 45 ° around its longitudinal axis during the LD 3.3. Grain size distributions
strain measurement, as advised by Martina et al. (2016). To determine
the stress-free lattice spacing, as a function of in-plane position, a The average grain sizes measured by EBSD across each weld nugget
‘comb’ sample, having teeth 4(LD) × 2.8(TD) × 2.8(ND) mm3, was down the stiffener into the bottom skin (along the location indicated in
machined using electro-discharge machining (EDM) from the same Fig. 5) are shown in Fig. 6, along with representative EBSD maps taken

258
T. Sun et al.
Fig. 4. (a) weld torque recorded by the FSW machine at steady state and (b) the specific weld energy calculated from the machine torque, as a function of travel speed
for the first pass and second welding pass, at a constant tool rotation rate of 1100 rpm.
at the positions indicated. In FSW the enormous strain rate reached in
the nugget region, which has been modelled by Arora et al. (2009) to be
in the range of 5 s−1 to 9 s−1, typically ensures a steady state is reached
in each welding pass, in terms of the fine nugget recrystallized grain
structure. As discussed by Robson and Campbell (2010), the grain size
in the nugget after a single pass is largely controlled by rapid post-
deformation grain growth, which is related to the peak weld
crostructure observation.
259
Journal of Materials Processing Tech. 263 (2019) 256–265
Fig. 3. Schematic diagrams of the residual
stress measurement positions used for neutron
diffraction, showing; (a) the cross sections
scanned in the welded plates, (b) the spatially
resolved neutron diffraction measurement po-
sitions across each cross section, and (c) the
stress-relaxed ‘comb’ specimen used to de-
termine the microstructurally-dependent neu-
tron diffraction stress-free lattice spacing’s, d0,
across the T joint.
temperature. In Fig. 6 (a) it is apparent that for the single weld and
double sided weld the grain sizes decrease from the weld crown to the
weld root, due to the through thickness temperature gradient experi-
enced for both weld passes, which is consistent with findings in con-
ventional SSFSW butt welds observed by Wu et al. (2015), although the
gradient is more severe here owing to the greater loss of heat at the
weld base due to geometry effects and the more substantial clamping
Fig. 5. TD-ND cross-section macro views
showing; (a) a single pass T-joint fillet SSFSW
weld and (b) a final double pass T-joint. Both
passes were produced with a rotation speed of
1100 rpm and travel speed of 300 mm/min.
The different weld zones are labelled, in-
cluding the WNZ (weld nugget zone), WNOZ
(weld nugget overlap zone), TMAZ (thermo-
mechanically affected zone affected zone) and
HAZ (heat affected zone). In each figure, the
first weld pass is on the right. The welding pin
profiles are indicated by the white dotted lines.
The dashed white rectangles and solid circles
indicate the positions used for EBSD grain size
measurements and higher magnification mi-
T. Sun et al.
Fig. 6. (a) Average grain size distributions as a function of the transverse position in the weld nugget of the single pass weld and double sided joint produced at a
travel speed of 300 mm/min (the measurement locations are indicated by the white boxes shown in Fig. 5, (b) representative EBSD Euler maps showing the grain
structures at the positions indicated by arrows in (a) for the single pass weld; (c) to (e) for the double sided weld.
required for fillet welding. Furthermore, where the welds are not
overlapped the grain size profile in the WNZ of the second weld pass is
similar to that for the first single pass weld (i.e. the left and right trends
are mirrored in Fig. 6 (a) for the double weld and single weld) sug-
gesting that the temperature histories between the individual two weld
passes were essentially similar. (Note: the average grain size is slightly
coarser in the second pass, possibly due to less efficient heat transfer to
the tooling due to first pass distortion). It can also be noted that in the
central nugget overlap region, where the fine nugget grain structure
from the original pass has again been severely deformed by the tip of
the welding tool, the grain size is dominated by the second welding pass
and behaves similarly to that in the tool tip region in the first pass (e.g.
the behavior is mirrored). In contrast, after the second weld the grain
size in the first weld to the right of the overlap region has coarsened
substantially (right arrow). This suggests that where the prior nugget
from the previous pass has not been severely re-deformed by the second
pass, the thermal excursion arising from the second pass has provided
an opportunity for additional grain growth to occur in the first weld
nugget, leading to an asymmetric final joint microstructure at the
nugget grain-scale. Rapid grain growth would be expected in a material
with such a fine starting grain structure, as has been previously noted
by Hassan et al. (2003)
3.4. Hardness distributions
Microhardness maps obtained from the TD-ND cross-section are
provided in Fig. 7, after natural ageing for four months and the corre-
sponding hardness line profiles at the positions shown along the stif-
fener and skin are plotted in Fig. 8. Representative weld nugget SEM
images taken from the different weld nugget positions indicated in
Fig. 5 are also displayed in Fig. 9, to help clarify the hardness results.
The rolled (T761 temper) parent plate displays an average hardness of
170–175 Hv, which is slightly lower at its mid plane due to macro
segregation. As discussed by Sullivan and Robson (2008), the hardness
distribution in FSWs of AA7050 is dominated by the size and density of
260
Journal of Materials Processing Tech. 263 (2019) 256–265
the η, and η precipitates, and is thus closely related to the peak weld
temperature reached locally within a weld zone.
It is apparent in Fig. 7 (a) that qualitatively the hardness distribu-
tion seen after the first weld pass is very similar to that found in con-
ventional AA7050 SSFSW butt welds reported by Sun et al. (2017)
despite the different boundary conditions for heat loss. This is pre-
sumably because the thermal gradients near the tool probe are similar,
irrespective of the far-field workpiece and clamping geometry. The
weld hardness minima are located in the HAZ close to the TMAZ
boundary, both either side and below the probe profile. It is now well
known that this is caused by coarsening and/or dissolution of the η’
strengthening precipitates, while the relatively higher hardness in the
weld nugget, which closely matches the pin profile, arises from strength
recovery during subsequent post-weld natural ageing. Strength re-
covery in the nugget occurs because the peak temperatures reached
during welding can approach, or exceed, the solvus temperature (∼
473 ℃ in AA7050 (Association, 1984)). As proposed by Sullivan and
Robson (2008), this causes almost full dissolution of the Zn/Mg con-
taining phases. It can also be seen in Fig. 9 (a), that this conclusion is
supported by the low volume fraction of coarse precipitates found at the
grain boundaries in the nugget region of the first pass, suggesting the
cooling rate was also sufficient to retain most of the solute in solution,
leading to a strong post-weld natural ageing response. Additionally, it is
apparent in Fig. 7 (a) that material that is subsequently deformed by the
tool during the second weld pass has been softened by the first weld
pass, as in the second pass a substantial proportion of the pin travels
through the low hardness region in the HAZ surrounding the probe
created by the first weld pass, which extends to the opposite side of the
stiffener wall.
Finally, it should further be noted that the second pass was per-
formed within one hour of the first weld, which would lead to far less
post-weld natural ageing occurring in the weld nugget region of the first
pass than is normally seen in post weld hardness profiles. For example,
from a fully solution treated state in 7050 Zheng et al. (2018) has
shown that on natural ageing the yield strength after 1 h is typically
T. Sun et al.
Fig. 7. Hardness contour maps from T-joint TD-ND cross sections, measured
after four months post-weld natural ageing, with a rotation speed of 1100 rpm;
showing (a) the first and (b) the second pass at 300 mm/min and (c) the ex-
panded HAZ for a double sided weld produced with a lower travel speed
(100 mm/min). The measurement grid used is indicated by the black dots and
black dashed lines represent the pin profiles. Solid circles in (b) and (c) indicate
positions for the HAZ microstructure observation.
only ∼180 MPa relative to ∼298 MPa after 24 h. Thus, the next weld
pass will be through material that is already greatly overaged, or has
been recently solution treated, both of which effects will reduce the
yield strength and welding torque, which also impacts on a reduced
heat input for the second pass. It can be further noted from Fig. 8(c)
that the size of the HAZ also increases with a lower weld travel speed
and this could potentially increase the material softening experienced
by the tool in the second pass, although a clear trend in this respect, in
terms of the percentage reduction in torque, is not revealed within the
261
Journal of Materials Processing Tech. 263 (2019) 256–265
Fig. 8. Microhardness profiles; measured in the stiffener along the mid-thick-
ness of the first and second passes for (a) single and double welds produced at
travel speed of 300 mm/min and rotation speed of 1100 rpm and (b) compar-
ison between double welds produced at a high (300 mm/min) and low travel
speeds (100 mm/min); (c) determined at a quarter thickness along the skin for
the slow and fast travel speed welds. Note: the hardness traverse positons are
indicated in the inserts and were at a depth of 1.5 mm below the surfaces in
each case.
limited data in Fig. 4.
Hardness maps determined after the second welding pass are com-
pared in Fig. 7 (b) and (c) for a fast (300 mm/min) and slow (100 mm/
min) travel speed. At face value, the hardness profiles for the second
weld look quite similar to that of the first pass shown in Fig. 7 (a), but
some subtle differences are apparent, that become accentuated when
the travel speed is reduced. These are also evident in the hardness
profiles shown in Fig. 8. The fact that similar hardness distributions are
seen in the weld nugget of the second pass is because the nugget region
T. Sun et al.
is re-solutionised in the second pass, erasing most of the memory of the
HAZ from the previous weld. However, in the surrounding HAZ regions
where the material was already overaged in the first pass overlap with
the new thermal field and of the second pass would be expected to
increase the extent of oveaging and reduce the hardness minima level,
as well as overage areas that were originally in a parent condition, or
solutionised in the first pass (e.g. the first pass nugget region). Evidence
for these effects can be seen by careful comparison of the maps in
Figs. 7 and 8. For example, in the high travel speed weld in the vertical
stiffener the HAZ has got larger and deeper after the second pass in
Fig. 7 (b) compared to in Fig. 7 (a) and with a slow travel speed there is
a more obvious reduction in the HAZ minima positon in the first pass
relative to the second pass in Fig. 8 (b). However, it should also be
noted that due to the material softening effect in the first pass, discussed
above, the heat input fell by about 20% in the second pass.
The above interpretation is supported by the SEM images of the
local precipitation states shown in Figs. 9 and 10. Although at this re-
solution the fine precipitates responsible for strengthening cannot be
observed (e.g. GPZs and η’), SEM can readily resolve coarser-scale
precipitates that reduce the local hardness by consuming solute and
through overaging. For the higher travel speed welds in Fig. 9 (d) the
second pass nugget can be seen to have a similar low level of coarse
second phase present to the first pass nugget in Fig. 9 (c), confirming it
has experienced a similar level of solutionisation. In the weld nugget
overlap zone the microstructure is again similar, but more coarse pre-
cipitation can be seen because the weld temperature at the base of the
pin in each case was similar and not sufficient to achieve full solution
treatment. However, where the second pass HAZ interacted with the
first pass weld nugget, evidence of greater overaging can be seen in the
first pass nugget (Fig. 9 (f)).
A lower travel speed (100 mm/min) corresponds to a higher heat
input (Fig. 4 (b)), longer time at temperature and a slower cooling rate,
as has been reported by Sun et al. (2017). As has been noted above in
the hardness maps, this resulted in stronger overlap effects between the
two weld passes. This greater asymmetry between the two weld passes
when a lower travel speed was used was also confirmed by SEM ana-
lysis. Examples are provided in Figs. 9 and 10. In Fig. 9 (i), at a similar
262
Journal of Materials Processing Tech. 263 (2019) 256–265
Fig. 9. SEM backscattered electron images
taken from the positions indicated in Fig. 6. (a)
- (c) shows the microstructure after the first
weld pass, (a) in the HAZ at the prior position
of the second pass nugget centre, (b) at the
base of the first pass nugget, which is subse-
quently in the pin overlap zone, and (c) the
weld nugget centre, which is not deformed by
the second pass. (d) - (f) provides the corre-
sponding microstructures at the same locations
after the second pass. Both welds were pro-
duced at a high travel speed (300 mm/min)
and a rotation rate of 1100 rpm. (g) – (i) shows
the microstructures of the weld produced with
a low travel speed (100 mm/min) and a rota-
tion rate of 1000 rpm at similar positions as in
(d) - (f).
positon to in Fig. 9 (f) where the second pass HAZ has overaged the
nugget in the first pass, with a lower travel speed, the level of overaging
of the previously solutionised nugget can be seen to be more extensive.
In Fig. 10 SEM images taken from the location of minimum HAZ
hardness in the vertical stiffener are also compared for low and high
travel speeds. They show that the reduced hardness minimum in welds
produced with a lower travel speed (top vs bottom images) can be at-
tributed to the greater extent of precipitate coarsening at this location.
As discussed by (Sullivan and Robson, 2008) this in turn is driven by the
longer accumulated time in the critical temperature range at which the
most rapid over ageing occurs (e.g. ∼ 350 ℃ for AA7050). In the T-
joint produced with the lower travel speed greater overaging from the
second pass is also evident at the location of the minima in the stiffener
caused by the first pass (Fig. 10 (c) compared to (d), relative to Fig. 10
(a) compared to (b)).
3.5. Residual stress distributions
A visualisation of the longitudinal residual stress distributions over
the weld cross-sections measured by neutron diffraction is shown in
Fig. 11, using the data measured at the positions indicated in Fig. 3 (b).
A more quantitative picture of the longitudinal residual stress profiles is
plotted in Fig. 12, plotted along traverses from the skin bottom at ∼
1.5 mm below both of the vertical stiffener’s surfaces (see the insert in
Fig. 12) which runs across the first and second weld passes at the same
depth. It can be seen in Fig. 11 that, irrespective of the welding speed,
significant tensile residual stresses are located in the weld zone, ba-
lanced by lower compressive residual stresses further away from the
weld area. The maximum tensile residual stresses measured lie in the
range 150–170 MPa (Fig. 12). It is notable that the peak tensile regions
are not symmetrically distributed, being largest in the vertical stiffener
near the position of the advancing side of the SSFSW pin in the second
weld pass. There is also a tensile stress peak area in the skin, which is
again off-set to the retreating side of the pin profile from the second
weld pass. From Figs. 11 and 12, it is apparent that the residual stresses
from the first pass are dominated by the second welding pass, with only
a weaker residual memory of the first pass in the skin and little evidence
T. Sun et al.
remaining of the residual stress field arising from the first pass in the
stiffener section.
In studies on butt welds in AA7050 by SSFSW Sun et al. (2017)
found an ‘M’ shaped residual stress profile across the welds, which was
similar to that measured in conventional FSWs by other workers such as
Peel et al. (2003) and Richards et al. (2008). However, the tensile re-
sidual stress field in the SSFSW process tended to be narrower and with
lower peak stresses than in FSW. This was attributed to the narrower
thermal field and lower heat input arising from having a stationary
shoulder. While affected by the different geometry, the current long-
itudinal residual stress distributions in the T-joint welds have simila-
rities to the ‘M shaped profiles seen in the SSFSW butt welds. For ex-
ample, in both cases, tensile residual stress peaks were determined at
the edge of the pin on both the advancing side and retreating side. This
feature has been explained using FE simulations by Richards et al.
(2008) and results from the diffuse nature of the thermal field in FSW,
relative to other fusion welding processes, combined with the extreme
softening behavior of high strength Al-alloys at high temperatures. This
gives rise to a central region with a very low relatively uniform flow
stress that readily relaxes to a low level of elastic strain, whereas there
is a steep gradient in temperature transverse to the welding direction
and, thus yield stress, where residual stresses will be ‘frozen in’ at a
higher level on cooling behind the tool. The zone of thermal plastic
misfit was predicted by Richards et al. (2008) to be wider than the
nugget region and the mechanical effects of the tool coupling with the
work piece were therefore assumed to be small and could be ignored to
263
Journal of Materials Processing Tech. 263 (2019) 256–265
Fig. 10. SEM backscattered electron images from the locations
of minimum hardness indicated in Fig. 7 in the HAZs in the
stiffener for; (a) the first weld pass, (b) the second weld pass in
the weld produced at a travel speed of 300 mm/min and (c)
the first weld pass, (d) second weld pass in the weld produced
at a travel speed of 100 mm/min.
a first approximation, greatly simplifying their FE analysis.
When producing dual pass opposing SSFSW fillet welds sequentially
with a T-joint geometry, two main factors need to be considered which
will control the strong asymmetry seen in the resultant residual stress
fields in Fig. 11. Firstly, substantial stress relaxation can occur in the
stress field introduced by the first pass, due to stress relief arising from
the thermal field associated with the second pass, which overlaps
strongly with the location of the residual stress field formed in the first
pass. For example, it has been reported by Pouget and Reynolds (2008)
that in a similar alloy FSW residual stresses can be substantially relaxed
by a post-weld artificial ageing treatment and it would be expected this
effect would be more evident in the weld produced at a lower travel
speed, where higher temperatures are reached over a larger area and for
a longer thermal cycle. However, given the location and large asym-
metry of the regions of final peak tensile stress it is clear that the zone of
local plastic misfit caused by the second pass largely erases the elastic
residual tensile field in the region near the second pass. The modelling
work by Richards et al. (2008) in simpler single butt geometries shows
that the compressive zone ahead of the tool is mainly controlled by the
thermal expansion field and relaxes plastically nearly fully close to the
tool, owing to the low material strength at temperature, to a bigger
width than the tool. In their butt weld simulations the compressive
plastic zone was also wider than the tensile zone of elastic misfit that
develops behind the tool as the weld cools down. If two welds were
directly overlaid, the new compressive plastic misfit strain introduced
would be far larger than the residual elastic misfit strain retained from
Fig. 11. Visualisation of the longitudinal residual stress
distribution found across the T-joint welds produced using
travel speeds of (a) 300 mm/min and (b) 100 mm/min, as
determined by neutron diffraction at ENGIN-X. Please note
the sparsity of the measurement grid as indicated by black
dots; the welding pin profile is indicated by black dashed
lines.
T. Sun et al.
the previous weld, and this would effectively wipe out the previous
residual stress field. Without modelling the much more complex SSFSW
T-joint geometry, it is difficult to estimate how much the thermal
plastic misfit associated with the second pass would overwrite the re-
sidual elastic strains from the previous weld, but given that the mi-
crostructural evidence suggests a substantial overlap of the thermal
fields this behavior seems likely to have had a large influence on the
highly one-sided residual stress distribution seen in the final T-joint.
Another factor that might be significant when repeating SSFSWs is
the mechanical effect of the down force from the second weld pass; for
example, it is known from work by Altenkirch et al. (2009) that locally
compressing a welded joint by post-weld rolling can greatly modify the
residual stresses in a weld. However, without a fully coupled model that
includes the material flow behavior it is difficult to quantify the sig-
nificance of such effects.
In addition, it is noteworthy that the effect of travel speed on the
peak residual stresses measured in the T-joint are consistent with pre-
vious results by Sun et al. (2018a) on SSFSW butt welds in the same
alloy. These results are summarized and compared in Fig. 12 (b) where
very similar trends can be seen, both in terms of the effect of travel
speed and the peak stress values, despite the different joint configura-
tions. Sun et al. (2018a) found that the peak longitudinal tensile re-
sidual stresses are most strongly influenced by the travel speed in FSW,
compared to the other process parameters such as rotation rate and tool
downforce. In FSW increasing the travel speed reduces the effective
heat input, but also strongly affects the width of the thermal field and
the cooling rate after the heat source has passed. At higher travel
speeds, the thermal field extends less far from the tool and this, com-
bined with higher cooling rates, locks-in a larger residual tensile stress
because the material’s yield strength will increase more quickly closer
to the tool with falling temperature and will, therefore, retain a greater
proportion of the tensile misfit generated on cooling. The agreement
between the butt weld and T joint results in Fig. 12 thus suggests the
peak residual stress is more dominated by the material’s yield strength
behavior with temperature, rather than the weld configuration, which
would be expected to be similar close to the probe for a given rotation
rate and travel speed.
Finally, the results discussed above have shown that the sequential
welding procedure required for the welding of T-joint by SSFSW results
in an asymmetrical distribution in residual stress and hardness, which
presents a challenge to optimization of the SSFSW processing para-
meters, compared to in single pass conventional butt welding. This is
due to the complex interaction between the effect of the overlapping
thermal fields on the residual stresses and weld zone microstructure and
the associated change in mechanical work between each welding pass.
In the present investigation, welding parameters for the two passes
were kept the same. However this need not be the case in practice and it
would be possible to tailor the combination of the travel speed, rotation
speed and tool downforce for each weld pass separately to optimize the
residual stress and hardness distributions for a specific application, as
264
Journal of Materials Processing Tech. 263 (2019) 256–265
Fig. 12. Plots of (a) longitudinal residual stress
distributions measured in the stiffener along
the mid-thickness of first pass and second pass
for welds produced at travel speed of 100 mm/
min and 300 mm/min and (b) summary of
peak longitudinal residual stress data in cur-
rent T-joint welds amalgamated with prior
work reported by Sun et al. (2018a) against
travel speed, rotation speed of each weld trial
in (b) was labeled.
proposed by Long et al. (2007) for traditional single pass butt welds. For
example, as the material has been softened during the first pass, less
energy (heat input) is required for the subsequent pass, which could be
achieved by a higher travel speed. This would not only mitigate further
over-ageing in the previous weld zone, but also increases the produc-
tion efficiency, which is often of more concern to a manufacturer in
practice. Other welding parameter combinations could also be used to
minimize weld distortion or the final asymmetry in hardness.
4. Conclusions
This study confirms that AA7050 T-joints can be successfully pro-
duced using a dual pass stationary shoulder friction stir fillet welding
process. The welded joints were characterized in terms of their weld
zone microstructures, hardness and residual stress distribution, as a
function of the tool travel speed. The main conclusions from this work
can be summarized as:
(1) Material softening induced by the first weld leads to a drop in the
torque required for the subsequent weld pass. This provides the
potential for utilizing a lower heat input in the second pass to
further reduce the overall thermal field and improve the joint
strength.
(2) The second pass introduces a thermomechanical treatment which is
broadly similar that for a single pass. This thermal and mechanical
excursion has a significant effect on the weld zone microstructure,
hardness and residual stress distribution arising from the first pass,
such that the stresses and hardness effects are dominated by the
second weld pass leading to asymmetric hardness and residual
stress fields.
(3) The microstructural interactions between the two welds arising
from the overlap of the thermal excursions are largest at lower
welding speeds, due to the higher heat input for each welding pass.
(4) A highly asymmetric final residual stress distribution was measured
in the dual pass T-joints, showing a high zone of local tensile re-
sidual stress close to the edge profile of the pin associated with the
second weld pass. This was attributed to the plastic thermal misfit
generated by the second weld pass largely overwriting the residual
stress field developed in the first weld pass.
(5) The travel speed had a strong effect on the peak residual stresses
and the T-joint showed a similar behavior to data from more
standard butt joint configurations, implying that the material yield
stress response at temperature dominates the peak stresses locked in
during welding irrespective of the joint configuration.
Acknowledgements
The Authors are grateful for the financial support provided by China
Scholarship Council (CSC) and the EPSRC through LightForm (EP/
R001715/1). Tianzhu Sun would like to thank Dr. Joe Kelleher for the
T. Sun et al.
enthusiastic support at Engin-X beamline at the ISIS neutron facility.
PJW is grateful to the European Research council for an advanced grant
No 695638 CORREL-CT and EPSRC under grant EP/M010619/.
References
Ahn, J., He, E., Chen, L., Pirling, T., Dear, J.P., Davies, C.M., 2018. Determination of
residual stresses in fibre laser welded AA2024-T3 T-joints by numerical simulation
and neutron diffraction. Mater. Sci. Eng. A 712, 685–703.
Altenkirch, J., Steuwer, A., Withers, P.J., Williams, S.W., Poad, M., Wen, S.W., 2009.
Residual stress engineering in friction stir welds by roller tensioning. Sci. Technol.
Weld. Join. 14, 185–192.
Arora, A., Zhang, Z., De, A., DebRoy, T., 2009. Strains and strain rates during friction stir
welding. Scr. Mater. 61, 863–866.
Association, A., 1984. Aluminum: Properties and Physical Metallurgy. ASM International.
Avettand-Fènoël, M.-N., Taillard, R., 2016. Effect of a pre or postweld heat treatment on
microstructure and mechanical properties of an AA2050 weld obtained by SSFSW.
Mater. Des. 89, 348–361.
Barnes, S.J., Steuwer, A., Mahawish, S., Johnson, R., Withers, P.J., 2008. Residual strains
and microstructure development in single and sequential double sided friction stir
welds in RQT-701 steel. Mater. Sci. Eng. A 492, 35–44.
Brown, R., Tang, W., Reynolds, A.P., 2009. Multi-pass friction stir welding in alloy 7050-
T7451: effects on weld response variables and on weld properties. Mater. Sci. Eng. A
513, 115–121.
Buffa, G., Fratini, L., Ruisi, V., 2009. Friction stir welding of tailored joints for industrial
applications. Int. J. Mater. Form. 2, 311.
Cabibbo, M., Forcellese, A., El Mehtedi, M., Simoncini, M., 2014. Double side friction stir
welding of AA6082 sheets: Microstructure and nanoindentation characterization.
Mater. Sci. Eng. A 590, 209–217.
Cui, L., Yang, X., Xie, Y., Hou, X., Song, Y., 2013. Process parameter influence on defects
and tensile properties of friction stir welded T-joints on AA6061-T4 sheets. Mater.
Des. 51, 161–174.
Donati, L., Tomesani, L., Morri, A., 2009. Structural T-joint produced by means of friction
stir welding (FSW) with filling material. Int. J. Mater. Form. 2, 295.
Fratini, L., Buffa, G., Filice, L., Gagliardi, F., 2006. Friction stir welding of AA6082-T6 T-
joints: process engineering and performance measurement. Proc. Inst. Mech. Eng.
Part B J. Eng. Manuf. 220, 669–676.
Fratini, L., Buffa, G., Micari, F., Shivpuri, R., 2009. On the material flow in FSW of T-
joints: Influence of geometrical and tecnological parameters. Int. J. Adv. Manuf.
Technol. 44, 570–578.
Fu, G., Lourenco, M.I., Duan, M., Estefen, S.F., 2014. Effect of boundary conditions on
residual stress and distortion in T-joint welds. J. Constr. Steel Res. 102, 121–135.
Hassan, K.A., Norman, A.F., Price, D.A., Prangnell, P.B., 2003. Stability of nugget zone
grain structures in high strength Al-alloy friction stir welds during solution treatment.
Acta Mater. 51, 1923–1936.
Li, D., Yang, X., Cui, L., He, F., Zhang, X., 2015. Fatigue property of stationary shoulder
friction stir welded additive and non-additive T joints. Sci. Technol. Weld. Join. 20,
650–654.
Long, T., Tang, W., Reynolds, A.P., 2007. Process response parameter relationships in
aluminum alloy friction stir welds. Sci. Technol. Weld. Join. 12, 311–317.
Maltin, C.A., Nolton, L.J., Scott, J.L., Toumpis, A.I., Galloway, A.M., 2014. The potential
adaptation of stationary shoulder friction stir welding technology to steel. Mater. Des.
64, 614–624.
Martin, J.P., 2014. Stationary shoulder friction stir welding. Proceedings of the 1st
International Joint Symposium on Joining and Welding 477–482.
265
Journal of Materials Processing Tech. 263 (2019) 256–265
Martin, J.P., Way, W., 2011. NOVEL TECHNIQUES FOR CORNER JOINTS USING
FRICTION STIR. Friction Stir Welding and Processing VI, 179.
Martina, F., Roy, M.J., Szost, B.A., Terzi, S., Colegrove, P.A., Williams, S.W., Withers, P.J.,
Meyer, J., Hofmann, M., 2016. Residual stress of as-deposited and rolled wire+ arc
additive manufacturing Ti–6Al–4V components. Mater. Sci. Technol. 32, 1439–1448.
Pawley, G.S., 1981. Unit-cell refinement from powder diffraction scans. J. Appl.
Crystallogr. 14, 357–361.
Peel, M., Steuwer, A., Preuss, M., Withers, P.J., 2003. Microstructure, mechanical prop-
erties and residual stresses as a function of welding speed in aluminium AA5083
friction stir welds. Acta Mater. 51, 4791–4801.
Pouget, G., Reynolds, A.P., 2008. Residual stress and microstructure effects on fatigue
crack growth in AA2050 friction stir welds. Int. J. Fatigue 30, 463–472.
Richards, D.G., Prangnell, P.B., Williams, S.W., Withers, P.J., 2008. Global mechanical
tensioning for the management of residual stresses in welds. Mater. Sci. Eng. A 489,
351–362.
Robson, J.D., Campbell, L., 2010. Model for grain evolution during friction stir welding of
aluminium alloys. Sci. Technol. Weld. Join. 15, 171–176.
Robson, J.D., Upadhyay, P., Reynolds, A.P., 2010. Modelling microstructural evolution
during multiple pass friction stir welding. Sci. Technol. Weld. Join. 15, 613–618.
Russell, M., 2008. Recent developments in the stationary shoulder FSW of titanium alloys.
7th International Friction Stir Welding Symposium.
Santisteban, J.R., Daymond, M.R., James, J.A., Edwards, L., 2006. ENGIN-X: a third-
generation neutron strain scanner. J. Appl. Crystallogr. 39, 812–825.
Simoncini, M., Cabibbo, M., Forcellese, A., 2016. Development of double-side friction stir
welding to improve post-welding formability of joints in AA6082 aluminium alloy.
Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 230, 807–817.
Sullivan, A., Robson, J.D., 2008. Microstructural properties of friction stir welded and
post-weld heat-treated 7449 aluminium alloy thick plate. Mater. Sci. Eng. A 478,
351–360.
Sun, T., Roy, M.J., Strong, D., Withers, P.J., Prangnell, P.B., 2017. Comparison of residual
stress distributions in conventional and stationary shoulder high-strength aluminum
alloy friction stir welds. J. Mater. Process. Technol. 242, 92–100.
Sun, T., Reynolds, A.P., Roy, M.J., Withers, P.J., Prangnell, P.B., 2018a. The effect of
shoulder coupling on the residual stress and hardness distribution in AA7050 friction
stir butt welds. Mater. Sci. Eng. A 735 (September), 218–227.
Sun, T., Tremsin, A., Roy, M.J., Hofmann, M., Prangnell, P.B., Withers, P.J., 2018b.
Investigation of residual stress distribution and texture evolution in AA7050 sta-
tionary shoulder friction stir welded joints. Mater. Sci. Eng. A 712 (January),
531–538.
Thomas, W.M., 1991. Friction stir butt welding. International Patent Application No.
PCT/GB92/0220.
Withers, P.J., 2007a. Residual stress and its role in failure. Rep. Prog. Phys. 70, 2211.
Withers, P.J., 2007b. Mapping residual and internal stress in materials by neutron dif-
fraction. C. R. Phys. 8, 806–820.
Withers, P.J., Preuss, M., Steuwer, A., Pang, J.W.L., 2007. Methods for obtaining the
strain-free lattice parameter when using diffraction to determine residual stress. J.
Appl. Crystallogr. 40, 891–904.
Wu, H., 2017. Systematic Analysis of the Advantages of Stationary Shoulder Friction Stir
Welding in Joining High Strength Aluminium Alloy AA7050-T7651, PhD Thesis. the
University of Manchester, Manchester.
Wu, H., Chen, Y.-C., Strong, D., Prangnell, P.B., 2015. Stationary shoulder FSW for joining
high strength aluminum alloys. J. Mater. Process. Technol. 221, 187–196.
Zheng, G., Li, H., Lei, C., Fu, J., Bian, T., Yang, J., 2018. Natural aging behaviors and
mechanisms of 7050 and 5A90 Al alloys: a comparative study. Mater. Sci. Eng. A 718
(March), 157–164.