Please cite this article in press as: Buffa, G., et al.

, Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
Journal of Materials Processing Technology xxx (2014) xxxxxx
Contents lists available at ScienceDirect
Journal of Materials Processing Technology
j our nal home page: www. el sevi er . com/ l ocat e/ j mat pr ot ec
Analytical bonding criteria for joint integrity prediction in friction stir
welding of aluminum alloys
Gianluca Buffa

, Sergio Pellegrino, Livan Fratini

Department of Chemical, Management, Computer Science and Mechanical Engineering, University of Palermo, Viale delle Scienze, 90128 Palermo, Italy
a r t i c l e i n f o
Article history:
Received 22 July 2013
Received in revised form
23 December 2013
Accepted 7 February 2014
Available online xxx
Keywords:
Friction stir welding
Aluminumalloys
FEM
Bonding criterion
a b s t r a c t
In this study, two bonding criteria, previously used for porthole die extrusion, are applied to FSW starting
from the local value of the main eld variables calculated through a specically developed 3D numerical
model of the process. Their applicability and effectiveness have been assessed through an experimen-
tal and numerical campaign carried out with the main process parameters varying in a wide range.
The pressuretimeow criterion was demonstrated to be better suited for FSW processes when large
welding speed is used.
2014 Elsevier B.V. All rights reserved.
1. Introduction
Friction Stir Welding (FSW) is a solid-state welding process par-
ticularly suited for joining both light alloys and highly resistant
materials that are difcult to be welded by conventional fusion
welding. In FSW a non-consumable rotating tool with a specially
designed pin and shoulder is inserted into the abutting edges of
the sheets to be joined which are rigidly positioned one against the
other. It is possible to incline the tool respect to the normal working
axis by a 25

angle inthe opposite directionof the welding one (tilt

angle). The base of the pin must be plunged nearly to the bottom
of the joint so that the pin is in contact with all the welding pieces
along the total thickness. Once the plunge phase is nished, the
simple translation of the rotating tool along the welding line deter-
mines the formation of the joint. The weld seamwill present nely
striped bands, and it will be slightly lowered and not so irregular
and protruding as the conventional welding seams.
The relative motion between the tool and the workpiece causes,
due to the friction, the generation of the needed heat: maximum
temperature is found between the pin and the workpiece but the
melting temperature of the alloy is not reached. The generated heat
and the high temperature cause a decrease of the yield stress of the
metal that caneasilyowaroundthetool. Beingtemperaturebelow

Corresponding author. Tel.: +39 091 23861869; fax: +39 091 7099973.
E-mail addresses: gianluca.buffa@unipa.it (G. Buffa), sergio.pellegrino@unipa.it
(S. Pellegrino), livan.fratini@unipa.it (L. Fratini).
the melting one, there is no change in volume during the process
and this entails lowvalues of residual stresses after cooling. Mishra
andMa (2005) reportedthe mechanisms responsible for the forma-
tion of seam welds in FSWprocesses and also the microstructural
renement andnal mechanical properties of the weldedjoint. The
authors highlighted also that the maximumtemperature reached,
incorrespondence of the central part of the welding seam, is a func-
tion of the ratio between the tool rotation speed R [rpm] and
tool feed rate v [mm/min] along the weld seam: in general, with
the increase of this ratio, maximumtemperature increases.
Based on the description of the process, it is immediately noted
that the plastic ow of the material generated by the action of
the tool gives origin to asymmetric joints. Guerra et al. (2002)
focusedtheir attentionona fewkey aspects of the process as micro-
structural issues and material ow analysis. They observed, in a
cross section of the joint, an advancing side and a retreating
side: in the rst case the vectors peripheral speed of rotation and
feed rate of the tool have the same direction; on the contrary, in
the retreating side the two vectors are opposite each other.
The mechanism of formation of the weld through FSW can be
assimilated to an extrusion and forging of the plates to be welded.
Buffa et al. (2006a,c) investigated the operating parameters of the
FSW process, as optimal tool geometry and welding speed, for
improving nugget integrity of aluminumalloys as well as the bond-
ing mechanics. Due to the rotation, the pin moves the material in
contact with it towards the area by creeping, namely the metal is
extruded around the pin and immediately forged by the shoulder.
The longitudinal axis of the tool will then be the forging axis, while
http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
0924-0136/ 2014 Elsevier B.V. All rights reserved.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
2 G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx
the welding direction will be the extrusion axis. The mixed mate-
rial is usually dragged many times around the tool before being
deposited. Reynolds et al. (2000) and Heinz et al. (2000) investi-
gatedthe mechanical properties andthe microstructure of a FSWed
joint showing the characteristics of the weldedjoints. They demon-
strated that the material ows also vertically, in an almost circular
pattern as seen in a longitudinal section.
The most common welding defects that occur in a FS welded
joint are interruptions in the physical structure of the junction due
to incorrect material ow(tunnel defects). Additionally, at the bot-
tomof the joint swirl phenomena may be observed, thus resulting
in an ineffective material owand the possible insurgence of inter-
nal folding defects due to the geometrical discontinuity induced
by the tool pin shape. Chen et al. (2006) pointed out the condi-
tions that cause the typical defects of the FSW and the inuence
on the mechanical properties of the joint. They found that the low
heat input leads to the generation of tunnel and/or kissing bond.
The latter defect is particularly difcult to be spotted with non-
disruptive tests. Finally, besides tool rotation and welding speed,
the selection of the correct tilt angle inuences the heat input.
The material ow induced by the tool, which determines the
effectiveness of the weld, is dramatically inuenced by material
properties, such as yield strength, tool design, and FSW process
parameters. Once the material to be welded is xed and the tool
is designed, process parameters as tool feed rate and rotation are
responsible for the distribution of the main eld variables that gov-
ern the solid bonding phenomenon. Proper values of temperature,
strain and strain rate are needed in order to get an effective solid
bonding. Balasubramanian (2008) studied how the FSW process
parameters inuence the weld quality. It is worth noticing that, at
the moment, only empirical relationships can be used to predict
the FSWprocess parameters to produce defect free joints. A weld-
ing criterion, depending on the local values of the above cited eld
variables and embedded in a proper process model, represents an
important tool for sound joints design. As far as the authors know,
no solid bonding criterion has been used for FSW.
A few researches can be found in literature characterizing the
solid bonding phenomena for processes as Roll Bonding (RB) or
Porthole Die Extrusion (PDE), also including the use of a proper
bonding criterion. Donati and Tomesani (2005) studied the rela-
tion between product quality and die design in extruded aluminum
proles with seam welds. The effectiveness of the obtained joint
strongly depends on several operating parameters, both geomet-
rical and technological. Using different combinations of process
parameters different weld quality is obtained for the joints (defec-
tive or sound). Donati et al. (2007) also analyzed the process
parameters and the different die geometries for the extrusion of
a hollow prole with a seamweld. Ceretti et al. (2009a,b) showed
how it is possible to correctly simulate the bonding phenomena
occurring in extrusion porthole dies. They investigated the solid
bonding phenomenon through modular matrices in which it was
possible to change the geometric parameters of the reference
matrix, inorder tovarytheweldingconditions. Inthis way, thevari-
ables having a major inuence on the process could be identied.
The analysis was conducted froma purely phenomenological point
of view, identifying one by one of the critical parameters for the
obtainment of a sound bond, in order to derive general design rules
for theprocess. Donati andTomesani (2004) andDonati (2004) have
developed numerical campaigns of extrusion processes in order
to obtain the local conditions of the welds experimentally pro-
duced deriving, as a result, a bonding criterion valid under different
welding conditions. In fact, once the welding surface is identied,
it is possible to calculate, through a numerical analysis, the eld
variables involvedinthe bonding process, as contact pressure, tem-
perature, local owstress, strain, nodal speed. In the following the
three most utilized bonding criteria are briey described.
1.1. Maximumpressure criterion
This criterion considers as a discriminating parameter only the
maximum pressure inside the welding chamber. Ackeret (1972)
stated that when this value exceeds a critical threshold, which
depends only on the local and instantaneous conditions of the
material, the weld can be considered sound. This criterion is by far
the most applied in practice because of its simplicity. The effective-
ness of this criterion for RB processes has been demonstrated by
Azushima et al. (2008) which produced lightweight parts by using
high strength metal for the safety and reliability of micro-parts. It
is worth noticing that, due to its extreme simplicity, this criterion is
not particularly suited for manufacturing processes characterized
by more complex material ow, as PDE and FSW. Consequently,
this criterion has not been considered in the present study.
1.2. Pressuretime criterion
This criterion, proposed by Plata and Piwnik (2000), is based on
the integral in time of the ratio between the contact pressure and
the ow stress of the material. The value obtained must exceed a
critical threshold. Jo et al. (2003), in their paper, studied the inu-
ence of bearing length, product thickness and billet temperature in
portholedieextrusionof hollowsectiontubes onthepressureat the
interface of the welding plane. The best mechanical properties are
obtained when the pressure at the welding plane is approximately
3.55.8 times the average ow stress. Donati et al. (2007) investi-
gated the effectiveness of the criterion in the PDE process through
both mechanical and metallurgical surveys in order to validate
the numerical approach. The obtained results demonstrated that
it is possible to correctly simulate the solid bonding phenomena
occurring during the porthole die extrusion. Ceretti et al. (2009a)
determined the critical value of the criterion through roll bonding
(RB) experiments on AA6061. The threshold curve obtained was
used to model the extrusion of complex hollow proles. Finally,
DUrso et al. (2012), through a coupled experimental-simulative
strategy on aluminum alloy AA6082, implemented a new proce-
dure for the identication of the pressuretime bonding criterion
as a functionof the temperature. Flat rolling experimental tests and
FEMsimulations of the rolling process for the same conditions were
carried out and an exponential limit curve was identied.
Although the above cited authors found satisfying results,
Donati and Tomesani (2004) pointed that this approach is some-
times unable to give correct predictions when large gradients of
the node velocity are observed.
1.3. Pressuretimeow criterion
As previously observed, Donati and Tomesani (2004) showed
that the pressuretime criterion over emphasizes the role dead
zones of the material in a PDE matrix, in which the residence times
tend to innity. Consequently, they introduced the speed correc-
tion factor, pointing out that the material ow passing through
a generic point should be considered. The criterion validity was
demonstrated starting from experimental PDE tests. In particular,
special H proles were produced by changing the geometry of the
leg and the width of the central section, so as to create a variety of
operatingconditions inorder toobtainjoints withdifferent welding
quality. The obtained results showed that it is possible to pre-
dict the solid bonding phenomena occurring during PDE using the
pressuretimeow approach. The effectiveness of this criterion
has not been tested on industrial case studies yet.
In this paper, two solid bonding criteria are selected, prop-
erly applied to FSW and nally compared. Experimental welding
tests have been performed with varying tool feed rate and rota-
tion in order to obtain both defective and sound welds. Then, the
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx 3
Table 1
Experimental case studies.
Test ID R [rpm] V [mm/min]
A 500 400
B 500 200
C 500 100
D 1000 400
E 1000 200
F 1000 100
experimental tests havebeensimulatedthroughapre-existingFEM
model for FSW. The objective of this work is to assess the reli-
ability and the effectiveness of bonding criteria, already present
in literature and developed for other processes, for FSW through
experimental test, microstructural investigation and FEManalysis.
Additionally, the two criteria are compared and the most suited for
FSWprocess is identied.
2. Methods
2.1. Experiments
Butt joints were obtained out of aluminum alloy AA6061-T6
100mm200mm sheets, 2.4mm in thickness. The base material
was characterized by a yield stress of 260MPa and an ultimate ten-
sile stress (UTS) of 305MPa. A Mazak Nexus 410A milling machine
properly equipped with clamping xture for FSW was used. The
utilized tool was made in H13 steel quenched at 1020

C, charac-
terized by a 52 HRc hardness. A conical pin was designed, with
conical angle of 30

, major diameters equal to 3.5mm and height

equal to 2mm. The shoulder diameter was equal to 10mm.
A wide range of variation was selected for both the rotational
speed, varyingform500rpmto1000rpm, andthe feedrate, varying
form 100mm/min to 400mm/min. A constant tilt angle of 2

was
adopted for all the tests. In Table 1 the analyzed case studies are
reported:
Each test was repeated three times and from each joint
specimens were cross-sectioned perpendicularly to the welding
direction. Macrographs were used to analyze the material area
involved in the process mechanics; furthermore, the presence of
ow defects was investigated through micrographs. In order to
obtain such results the specimens were hot mounted, polished and
nally etched with Keller reagent and observed by a LM.
2.2. Numerical simulations
The commercial FEA software DEFORM-3D
TM
, Lagrangian
implicit code designedfor metal forming processes, was used. Buffa
et al. (2006b) proposed a continuumbased FEMmodel for friction
stir welding process. This model was rst calibrated by comparing
calculated force and temperature distribution with experimental
results. Then it was used to investigate the distribution of the main
eld variables in the heat affected zone and the weld nugget. A
coupled thermo-mechanical analysis with rigid-viscoplastic strain,
strain rate and temperature dependent material behavior was per-
formed. The material data were taken from the ASM Handbook
(1990) and in house experiments. The tool was modeled as a rigid
object. The FSWmodeling was divided into two stages: the plunge
stage and the welding stage. In the rst phase the tool, which has
a tilt angle of 2

, moves down vertically with the assigned rotating

speed. In this way the temperature eld needed for the beginning
of the actual welding is reached. In the next step the rotating tool
moves along the welding line with proper feed rate.
The tool was meshed, for the thermal analysis, with about 3000
tetrahedral elements. The two sheets to be welded were modeled
as a single-block. This is because in a double block model instability
Fig. 1. The gure shows the utilized mesh for tool and sheet.
problems may arise due to the sheet-sheet contact. However, Buffa
et al. (2006b,c) have demonstrated that this approximation does
not signicantly affect the distribution of the main eld variables.
In other words, Friction Stir Processing (FSP) is simulated instead
of FSW.
For the plate material, i.e. the aluminumalloy AA6061, the fol-
lowingvalues of thermal conductivityk =180[N/(s

C)] andthermal
capacity c =2.4[N/(mm
2
C)], taken fromliterature, were used. No
variation of k and c with temperature was taken into account.
This assumption makes the thermal problem linear, speeding up
the numerical solution at each time increment. A constant inter-
face heat exchange coefcient of 11 [N/(mms

C)] was utilized for

the tool sheet contact surface. The sheet blank, 2.4mm thickness,
was meshed with about 12,000 tetrahedral elements. Because of
the large gradients of the calculated variables, it was necessary to
introduce a ner discretization along the welding line. Addition-
ally, a re-meshing referring volume was identied all along the tool
feed movement. In this area, each tetrahedral element had mini-
mum single edge of about 0.5mm; in this way, about ve elements
were placed along the sheet thickness. An interface penalty con-
stant equal to 1E9 was used to penalize the penetration velocity
of the nodes of the sheets through the tool master surface. In this
way, elements folding, especially in the contact area between the
tool pin and the workpiece, where the material closes behind the
tool, was avoided reducing the risk of unexpected simulation stop
due to failure in convergence. A constant shear friction factor of
0.46 was used for the tool-sheet interface on the basis of a previous
experimental thermal characterization and of a numerical sensi-
tivity analysis for the shear friction factor m conducted by Fratini
et al. (2005). In particular, the shear factor was optimized based
on experimental temperature measurements performed with both
thermo-camera andthermocouples. Fig. 1shows the meshedsingle
block and tool.
At the end of the simulation, the material owwas investigated
through the analysis of the nodes movement and the main eld
variables history that they experience. The node tracking option
of the software DEFORM-3D
TM
was utilized, highlighting, for a set
of nodes initially placed along the sheets separation line in a trans-
verse section, their nal position after deformation. The identied
points were monitored throughout the process: six points equally
spaced along the joint thickness were identied as shown in Fig. 2.
The reference transverse section was taken after 40mmof weld
length, when the process has already entered a steady state and the
obtained data are free fromtransient effects.
3. Results
3.1. Welding criteria implementation
In this study two welding criteria were considered, namely
the pressuretime (W) and the pressuretimeow (W

). Both the
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
4 G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx
Fig. 2. The identied points for the node tracking option.
criteria are based on integrals in time, that, in order to perform
the calculation, have been approximated with sums over small
time intervals. Eqs. (1) and (2) represent the pressuretime and
the pressuretimeowcriteria, respectively.
W =

t
0
P

dt

=

j
p
j
j
t
j
(1)
W

t
0
P

vdt

=

j
p
j
j
v
j
t
j
(2)
where p is the contact pressure at the interface, is the owstress
of the material in the given temperature, strain and strain rate con-
ditions and v is the velocity of the considered node. As it can be
noticed, Wis measured in s, while W

in mm. Hence, Wcan be con-

sidered as the equivalent time, calculated with proper weights of
pressure and temperature (i.e. the ow stress) for the occurrence
of solid bonding. Similarly, W

can be considered as the equiva-

lent length of material ow. In other words, a particle of material
must possess a certain minimum velocity for a proper time once
appropriate temperature and pressure are reached.
In order to calculate, for each of the observation points high-
lighted in Fig. 2, the value of the welding criterion, the material
owoccurring during the FSWprocess must be properly predicted.
Buffa et al. (2013) proved the effectiveness of the developed model
for the prediction of the material owby comparing the calculated
results with experimental measurements of the zig-zag line due
to the oxides particles dispersed in the transverse section. Follow-
ing the approach proposed in the above cited paper, the reference
points were tracked during the process. It is worth noticing that the
observation time interval begins as the points experience non-zero
pressure values, i.e. before the tool reaches the reference transverse
section. By the same token, the observation period ends after the
tool leaves the reference section and pressure drops to zero. This
behavior is due to the peculiar FSWprocess mechanics: the mate-
rial close to the sheets separationline rotates together withthe tool
for a certain number of rounds, depending on the selected combi-
nation of process parameters; the rotation begins in the retreating
side. Then the material is left at the advancing side of the joint. In
fact, the latter is the area at which fracture occurs during tensile
tests of sound joints. Fig. 3 shows the evolution of the tracking line,
initially laying on the ideal separation surface between the sheets,
till the nal position reached by the six reference points (test C).
Process time and the x coordinate (welding direction) are provided
for each gure. The origin x =0 is set when the tool axis correspond
to the considered transverse section. It is necessary to observe that
at the beginning the reference line is in the leading edge of the joint
while at the end it lays on the trailing edge.
The local value of the owstress for a given reference point and
time was calculated starting fromthe local values of temperature,
strain and strain rate. Figs. 4a, 5a and 6a show, for case study C,
the distribution of the above cited eld variables in the reference
transverse section, right after the material closed the crack behind
the tool; Figs. 4b, 5b and 6b show the eld variables histories of
point P5, as indicated in Fig. 3.
The eld variables distribution highlights some of the charac-
teristics of the FSWprocess: temperature is symmetric respect to
the welding line; in turn, strain and strain rate peaks are shifted
towards the advancing side. Looking at the eld variables variation
with time, it arises that the strain rate assumes non-zero values
from about t =18 to t =26. When the material is rotating around
the tool, the accumulated strain increases quickly till a maximum
value of about 26. Temperature start increasing as the welding tool
approaches the consideredtransverse sectionandstarts decreasing
during the last part of the deformation, i.e. whenthe tool pinleft the
reference section and only the effect of the tool shoulder is present.
It should be observed that, as the analyzed case study is character-
ized by a welding speed of 100mm/min, during this time interval
the tool covers a distance of about 12mm, corresponding to a few
mmmore than the shoulder diameter. Then, for the P5 point, pos-
itioned in the upper half of the joint thickness, it can be stated that
the tool shoulder (and not just the pin) has a signicant inuence
on the evolution in time of the main eld variables. Smaller time
intervals are observed for points located closer to the bottomof the
joints indicating that, in those areas, the main mechanical effect is
due to the pin. Besides, also the nal values of the accumulated
strain is lower for points closer to the bottomof the joints, starting
fromabout 3.5 (see Fig. 6a).
As far as the contact pressure is considered, it is worth notic-
ing that the separation line between the sheets (see again the red
line in Fig. 3) changes both in shape and direction during the weld-
ing process. Consequently, the contact pressure must be calculated
considering the change in the direction normal to this line for
each of the considered points. It is worth noticing that the utilized
software does not allow the extrapolation of the contact pressure
between two moving ows of the same deformable object. For
this reason, the values of stress in the x and y directions were used
to obtain the contact pressure value. The nal results of this proce-
dure are shown in Fig. 7a in which, consistently with the previous
Figs. 46, the evolution of the normal pressure of P5 is considered
for test C. In Fig. 7b the evolution of the nodal velocity, needed for
the calculation of the W

parameter, is illustrated. In Fig. 7c the ow

stress characterizing P5 during the bonding process and deriving
fromthe simultaneous effect of temperature, strain and strain rate,
is shown.
Large values of pressure are observed during the deformation.
At the same time, the ow stress is below 100MPa. Starting from
t =18s, the owstress decreases withincreasing temperature; then
a maximum, corresponding to the strainrate peak, anda minimum,
when temperature reaches the highest peak, are observed. Finally,
it starts increasing again after strain rate drops to zero and tem-
perature starts decreasing. In this way, the welding parameters W
and W

can assume values larger than 1. Finally, as expected, nodal

velocity shows a trend consistent with the strain rate one.
3.2. Welding criteria comparison
In order to assess the applicability of the considered bonding
criteria to FSWwelding experiments were carried out as indicated
inTable 1. Fig. 8shows the macrographs of the etchedcross sections
for all the analyzed case studies.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx 5
Fig. 3. The identied points for the node tracking option test C.
Fig. 4. (a) Temperature distribution in the transverse section and (b) evolution in time for P5 test C.
Fig. 5. (a) Strain distribution in the transverse section and (b) evolution in time for P5 test C.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
6 G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx
Fig. 6. (a) Strain rate distribution in the transverse section and (b) evolution in time for P5 test C.
As it canbe seen, the three welds characterizedbyrotationequal
to500rpmarecharacterizedbyowdefects. Inparticular, test Ahas
a very large void area at the bottomof the pin and a crack is visible
starting fromthe bottomof the joint and reaching the top surface.
Looking at test B a tunnel defect is visible at the bottomof the pin,
whileaverysmall voidcanbeobservedinthesameareafor test C. In
turn, the three welds produced with rotation equal to 1000rpmare
defect free. InFig. 9the micrographs corresponding to the a, b, andc
red dotted squares highlighted in Fig. 8 are reported. The small void
observed at the bottomof the pin for test C is due to an insufcient
heat conferred to the weld. As a consequence the material below
the pin was too cold and was not properly mixed by the tool. As
the weld becomes colder, i.e. with increasing welding speed, the
void becomes larger assuming the name of tunnel or wormhole.
Finally a sound weld, corresponding to test E, is shown in Fig. 9c.
The produced weds underwent tensile tests in order to assess
the effectiveness of the bonding obtained during the process. In
Fig. 10 the obtained results are reported as percentage of the ratio
Fig. 7. Evolution in time of (a) pressure, (b) nodal velocity and (c) owstress for P5 test C.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx 7
Fig. 8. Etched cross section of the analyzed case studies.
Fig. 9. Micrographs of the red dotted squares highlighted in Fig. 8.
betweenthe UTS of the testedjoint andthe one of the base material.
The three defect free welds, i.e. the ones obtained with rotation
equal to 1000rpm, have a similar resistance, ranging between 72%
and 74%. On the other hand, a decreasing trend is found for the
case studies characterized by rotation of 500rpm. In particular, the
resistance of test C is only slightly lower than the one of the sound
joints. As the welding speed increases, the defect becomes larger
and a dramatic drop of the UTS is observed till an almost zero value
is observed for test A.
The bonding parameters Wand W

were calculated, for the six

observation points highlighted in Figs. 2 and 3, using the numerical
results shown in the previous paragraph. As far as the Wparameter
is regarded, i.e. the one based on the Pivnik and Plata criterion, a
threshold value was taken from literature. In particular, as briey
discussed in the introduction paragraph, Ceretti et al. (2009a) used
roll bonding tests on AA6061 to determine the critical value as a
function of temperature. A regression was carried out obtaining
the following analytical expression:
W
lim
= 4.9063e
0.0017T
T >320

C (3)
Fig. 11 shows the calculated W parameter and the threshold
curve for the 36 points investigated. The temperature correspond-
ingtoeachpoint is the average value of the temperatures calculated
during the time interval corresponding to non-zero pressure val-
ues. The points for which the criterion gives a wrong prediction are
highlightedinred. Inparticular, all the points of test Aare belowthe
Fig. 10. Tensile test results for the considered case studies.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
8 G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx
Fig. 11. Calculated Wparameter and threshold value limit curve.
limit curve, correctly indicating that solidbonding was not reached.
A satisfying prediction is obtained also for test B and C. The tunnel
defect of test B corresponds to the points P1 and P2, which are
below the curve, as well as P1 for test A. The remaining observa-
tion points are above the curve indicating a sound weld in those
areas. The only exception is P6 for test B, which is well below the
curve. For all the performed tests, maximumpressure in the direc-
tion orthogonal to the sheets separation line is observed at about
mid thickness. Being P6 close to the top surface of the joint, small
values of pressure are calculated inhibiting reaching the expected
bonding parameter value. The network correctly predict test E and
F, being all the observation points above the limiting curve. Finally,
a completely wrong prediction is obtained for test D. Although the
experimental evidenceclearlyshownthat asoundweldis obtained,
all the sixpoints are well belowthe thresholdvalues. Again, the rea-
son is the insufcient pressure and time due to the large welding
speed used for the test. It should be observed that the points not
shown in Fig. 11, namely C-P4, C-P5, E-P5, F-P1, F-P1, F-P2, F-P2
and F-P6, exceeds the maximumvalue of the scale selected for the
graphic in order to focus on the wrong predictions. Finally, it can be
noted that, even using a specically dened limiting curve instead
of the one taken from literature, no advantage in the prediction
accuracy could be obtained. In fact, the six points of test Dand B-P6
have a too lowWvalue. Using a lower limiting curve would result
in wrong prediction for B-P1, B-P2, C-P1 and many points of test
A. In this way, the curve taken from literature and dened for a
different process appears as the best possible thus conrming the
robustness of proposed approach.
The reasons explaining the wrong predictions of the Wparam-
eter leaded to include the effect of nodal velocity in the bonding
criterion, as shown in Eq. (2). In Fig. 12, the values calculated for
the six observation points of each case study are reported. A few
Table 2
Experimental observation and calculated and threshold values for the Wand W

parameters.
Test Point W[s] W

[mm] Experiments
Calculated value Threshold value Calculated value Threshold value
A P1 1.556 5.372 1.039 3.644 not welded
P2 0.697 5.297 1.058 3.474 not welded
P3 1.040 5.334 2.337 3.558 not welded
P4 0.694 5.269 1.936 3.411 not welded
P5 0.715 5.304 1.743 3.489 not welded
P6 1.107 5.341 3.452 3.574 not welded
B P1 2.552 4.810 1.548 2.501 not welded
P2 3.586 4.740 2.158 2.380 not welded
P3 5.565 4.673 4.112 2.268 welded
P4 6.095 4.702 5.178 2.316 welded
P5 4.849 4.583 9.505 2.122 welded
P6 1.280 4.635 3.391 2.205 welded
C P1 3.917 4.095 0.749 1.448 not welded
P2 7.018 4.106 1.943 1.460 welded
P3 8.679 4.079 3.020 1.428 welded
P4 10.898 4.079 4.223 1.428 welded
P5 10.915 4.057 1.965 1.402 welded
P6 7.835 4.128 0.644 1.487 welded
D P1 1.797 3.928 1.412 1.256 welded
P2 1.756 3.869 1.960 1.193 welded
P3 1.760 3.847 2.954 1.170 welded
P4 2.038 3.771 4.619 1.093 welded
P5 1.069 3.752 5.832 1.075 welded
P6 1.161 3.757 5.411 1.080 welded
E P1 4.726 3.492 2.287 0.842 welded
P2 7.523 3.443 5.971 0.802 welded
P3 4.379 3.473 2.698 0.826 welded
P4 3.590 3.496 2.974 0.845 welded
P5 12.092 3.344 19.930 0.726 welded
P6 5.558 3.550 27.614 0.890 welded
F P1 10.059 3.073 2.556 0.545 welded
P2 12.034 3.045 4.594 0.528 welded
P3 16.326 3.038 4.992 0.524 welded
P4 6.419 3.026 3.870 0.517 welded
P5 5.279 2.989 17.791 0.496 welded
P6 14.774 2.977 35.671 0.489 welded
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx 9
Fig. 12. Calculated W

parameter and threshold value limit curve.

observations on the obtained values can be made. First, for all the
tests, P1 is the point characterized by the lowest W

value. This is
consistent with what experimentally observed: at the bottom of
the joints the least favorable conditions for solid bonding occurs,
in terms of temperature, pressure and velocity. That is why if a
owdefect is observed, it is located in that area. Additionally, more
consistent results are obtained for test D, showing the six points a
W

value larger than the corresponding values of test A, B and C.

Finally, focusing on test C, it is noted that P6 has an extremely low
W

value, due to too lowvelocity values calculatedby the numerical

model. This observation could lead to deduce that a sound weld is
not obtained in that area, which is not conrmed by experimental
observations. Numerical instabilities may have led to an incorrect
predictionof the eldvariables by the FEMmodel. Hence, the bond-
ing criterion cannot give a correct prediction and underestimates
the W

parameter. As for the previous Fig. 11, points E-P5, E-P6, F-

P5 and E-P5 are not shown in the graph as their W

value exceeds
the scale selected for better readability.
The better consistency of the results with respect both to the
experimental evidences and to the expected effect of the process
parameters, in terms of heat input conferred to the joints, per-
mits to state that the W

parameter is better suited to predict the

solidbondingphenomena occurringinFSWprocesses. Basedonthe
experimental evidences and the calculated values of the W

param-
eter, a regression curve can be obtained representing the threshold
values, as a function of temperature, of the pressuretimeow
criterion for the considered AA6061 aluminumalloy:
W

= 1327661974400T
4.797
(4)
Finally, Table 2 summarizes the obtained results: for each of the
36 points considered, the calculatedandthresholdWandW

values
are reportedtogether withthe experimental observationindicating
if, at a given area, material continuity can be observed.
4. Conclusions
An experimental and numerical campaign was performed
on FSW of AA6061 aluminum alloy. The applicability to FSW
of analytical binding criteria, originally developed for different
manufacturing processes, was assessed. FSW experiments were
developed with tool rotation and welding speed varying in a wide
range in order to obtain different conditions of joint integrity.
An already developed and veried numerical model for FSW was
used to calculate the main eld variables values and the occur-
ring material ow needed to implement the considered criteria.
The following main conclusions can be drawn:

In FSW the material close to the welding line ows around the
tool in the retreating side and is eventually left at the advancing
side after a few turns depending on the ratio between the tool
rotation and welding speed;

The Pivnik and Plata criterion can be used for FSWbut problems
arise when large welding speed is selected; in those cases low
pressure is obtained and values of the W parameter lower that
the threshold are calculated, indicating that no bonding occurred
even when sound joints are experimentally obtained;

The pressuretimeowcriterion is better suited for FSWas the

effect of velocityis considered. Inthis way results moreconsistent
with what expected based on the process input parameters are
calculated;

For the same criterionit was possible toidentifya thresholdcurve

and a power function of temperature was calculated through a
regression. Using this limit curve, just one of the 36 analyzed
points, i.e. P6 of test C, gives back a wrong prediction.
References
ASM Handbook: Properties and Selection: Nonferrous Alloys and Special-Purpose
Materials, 1990, 10 ed.
Ackeret, R., 1972. Properties of pressure welds inextrudedaluminiumalloy sections.
Journal of the institute of metals 100, 202L 207.
Azushima, A., Kopp, R., Korhonen, A., Yang, D.Y., Micari, F., Lahoti, G.D., Groche,
P., Yanagimoto, J., Tsuji, N., Rosochowski, A., Yanagida, A., 2008. Severe plastic
deformation (SPD) processes for metals. CIRP Annals - Manufacturing Technol-
ogy 57, 716735.
Balasubramanian, V., 2008. Relationship between base metal properties and fric-
tion stir welding process parameters. Materials Science and Engineering A 480,
397403.
Buffa, G., Campanella, D., Fratini, L., 2013. On tool stirring action in friction stir weld-
ing of work hardenable aluminium alloys. Science and Technology of Welding
and Joining 18, 161168.
Buffa, G., Donati, L., Fratini, L., Tomesani, L., 2006a. Solid state bonding in extru-
sion and FSW: process mechanics and analogies. Journal of Materials Processing
Technology 177, 344347.
Buffa, G., Hua, J., Shivpuri, R., Fratini, L., 2006b. A continuum based fem model for
friction stir welding model development. Materials Science and Engineering
A 419, 389396.
Buffa, G., Hua, J., Shivpuri, R., Fratini, L., 2006c. Design of the friction stir welding
tool using the continuumbased FEMmodel. Materials Science and Engineering
A 419, 381388.
Ceretti, E., Fratini, L., Gagliardi, F., Giardini, C., 2009a. Anewapproach to study mate-
rial bondinginextrusionportholedies. CIRPAnnals ManufacturingTechnology
58, 259262.
Ceretti, E., Mazzoni, L., Giardini, C., 2009b. Simulationof metal owandwelding pre-
diction in porthole die extrusion: the inuence of the geometrical parameters.
International Journal of Material Forming 2, 101104.
Chen, H.B., Yan, K., Lin, T., Chen, S.B., Jiang, C.Y., Zhao, Y., 2006. The investigation of
typical welding defects for 5456 aluminum alloy friction stir welds. Materials
Science and Engineering A 433, 6469.
DUrso, G., Longo, M., Ceretti, E., Giardini, C., 2012. Coupled simulative-experimental
procedure for studying the solid state bonding phenomena. 2011 International
Conference on Extrusion and Benchmark, ICEB 2011, Bologna, Italy, 3 October
20115 October 2011, pp. 181188.
Donati, L., Tomesani, L., 2004. The prediction of seam welds quality in aluminum
extrusion. Journal of Materials Processing Technology 153154, 366373.
Donati, L., Tomesani, L., 2005. The effect of die design on the production and seam
weld quality of extruded aluminum proles. Journal of Materials Processing
Technology 164165, 10251031.
Donati, L., Tomesani, L., Minak, G., 2007. Characterization of seam weld quality
in AA6082 extruded proles. Journal of Materials Processing Technology 191,
127131.
Donati, L., 2004. Evaluation of a new nite element method (FEM) criterion for
seam welds quality prediction in aluminum extruded proles. Proceedings of
the Eighth International Aluminum Extrusion Technology Seminar, vol. II, pp.
221235.
Fratini, L., Beccari, S., Buffa, G., 2005. Friction stir welding fem model improve-
ment through inverse thermal characterization. North American Manufacturing
Research Conference, NAMRC 33, NewYork, NY, United States, 2427 May 2005,
pp. 259266.
Guerra, M., Schmidt, C., McClure, J.C., Murr, L.E., Nunes, A.C., 2002. Flow patterns
during friction stir welding. Materials Characterization 49, 95101.
Please cite this article in press as: Buffa, G., et al., Analytical bonding criteria for joint integrity prediction in friction stir welding of
aluminumalloys. J. Mater. Process. Tech. (2014), http://dx.doi.org/10.1016/j.jmatprotec.2014.02.014
ARTICLE IN PRESS
G Model
PROTEC-13902; No. of Pages 10
10 G. Buffa et al. / Journal of Materials Processing Technology xxx (2014) xxxxxx
Heinz, B., Skrotzki, B., Eggeler, G., 2000. Microstructural and mechanical char-
acterization of a friction stir welded Al-alloy. Materials Science Forum 331,
II.
Jo, H.H., Jeong, C.S., Lee, S.K., Kim, B.M., 2003. Determination of welding pres-
sure in the non-steady-state porthole die extrusion of improved Al7003
hollow section tubes. Journal of Materials Processing Technology 139,
428433.
Mishra, R.S., Ma, Z.Y., 2005. Friction stir welding and processing. Materials Science
and Engineering R: Reports 50, 178P.
Plata, M., Piwnik, J., 2000. Theoretical and experimental analysis of seamweld for-
mation in hot extrusion of aluminum alloys. 7th Int. Al. Extr. Tech. Sem., pp.
205211.
Reynolds, A.P., Lockwood, W.D., Seidel, T.U., 2000. Processing-property correlation
in friction stir welds. Materials Science Forum331, II.