Materials Science and Engineering A 419 (2006) 389396

A continuum based fem model for friction stir

weldingmodel development
G. Buffa
a,b,
, J. Hua
a
, R. Shivpuri
a
, L. Fratini
b
a
The Ohio State University, Department of Industrial, Welding and Systems Engineering,
1971 Neil Avenue, 210 Baker Systems, Columbus, OH 43210, USA
b
Dipartimento di Tecnologia Meccanica, Produzione e Ingegneria Gestionale,
Universit ` a di Palermo, Viale delle Scienze, 90128 Palermo, Italy
Accepted 10 September 2005
Abstract
Although friction stir welding (FSW) has been successfully used to join materials that are difcult-to-weld or unweldeable by fusion welding
methods, it is still in its early development stage and, therefore, a scientic knowledge based predictive model is of signicant help for thorough
understanding of FSW process. In this paper, a continuum based FEM model for friction stir welding process is proposed, that is 3D Lagrangian
implicit, coupled, rigid-viscoplastic. This model is calibrated by comparing with experimental results of force and temperature distribution, then is
used to investigate the distribution of temperature and strain in heat affect zone and the weld nugget. The model correctly predicts the non-symmetric
nature of FSW process, and the relationships between the tool forces and the variation in the process parameters. It is found that the effective strain
distribution is non-symmetric about the weld line while the temperature prole is almost symmetric in the weld zone.
2005 Elsevier B.V. All rights reserved.
Keywords: Friction stir welding; FEM; Material ow
1. Introduction
In FSW, two pieces of sheet or thin plate are joined by insert-
ing a specially designed rotating pin into the adjoining edges of
the sheets to be welded and then moving it all along the seam
[13]. At rst, the sheets or plates are abutted along edge to be
welded and the rotating pin is sunken into the sheets/plates until
the tool shoulder is in full contact with the sheets or plates sur-
face (Fig. 1). Once the pin is completely inserted, it is moved
with a small nuting angle in the welding direction. Due to the
advancing and rotating effect of the pin and shoulder of the
tool along the seam, an advancing side and a retreating side are
formed and the softened and heated material ows around the
pin to its backside where the material is consolidated to create
a high-quality, solid-state weld.
Although there have been many interesting papers focused on
the experimental aspects of FSWsuch as microstructural issues,
process parameters and visualization of material ow [16],

Corresponding author. Tel.: +39 091 6657051; fax: +39 091 6657039.
E-mail addresses: [email protected] (G. Buffa), [email protected] (J.
Hua), [email protected] (R. Shivpuri), [email protected] (L. Fratini).
very few publications report on the numerical modeling and
simulation of the process due to the complex owand difculties
in modeling the boundary conditions [7]. Important contribu-
tions in 3D modeling include that of Xu and Deng [8,9] who
developed a 3D nite element procedure to simulate the FSW
process with focus on velocity eld, material owcharacteristics
and the equivalent plastic strain distribution. The commer-
cial FEM code ABAQUS was employed and an Arbitrary
EulerianLagrangian nite element formulation with adaptive
meshing was utilized that considered large elastic-plastic
deformation and temperature-dependent material properties.
The authors compared their predicted results to experimental
data available and observed reasonable correlation between the
equivalent plastic strain distribution and the distribution of the
microstructure zones in the weld. However, although possible
with the commercial nite element code ABAQUS [10],
their FEM analysis was not a thermo-mechanically coupled
procedure. The temperature data obtained from experiment
was superimposed as a prescribed temperature eld for the
deformation analysis. This severely affected the welding
force and stress prediction as the material properties (ow
stress model) used by the author were actually temperature-
dependent.
0921-5093/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2005.09.040
390 G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396
Fig. 1. The FSW process for butt joint.
Ulysse [11] presenteda 3DFEMvisco-plastic model for FSW
of thick aluminumplates using a commercial FEMcode FIDAP.
The author investigated the effect of tool speeds on the process
parameters. It was found that the higher translational speed leads
tohigher weldingforce, while increasingthe rotational speedhas
the opposite effect, that of force reduction. Reasonable agree-
ment between the predicted and the measured temperature was
obtained and the discrepancies were explained as an inadequate
representation of the constitutive behavior of the material for
the wide ranges of strain-rate, temperatures and strains typically
found during FSW.
Chen and Kovacevic [12] developed a 3D FEM model to
study the thermal history and thermo-mechanical phenomena in
the butt-welding of aluminumalloy 6061-T6 using a commercial
FEM code ANSYS. Their model incorporated the mechanical
reaction between the tool and the weld material. Experiments
were conducted and X-ray diffraction technique used to measure
the residual stress in the welded plate. The welding tool (i.e.
the shoulder and pin) in the FEM model was modeled as heat
source, with the nodes moved forward at each computational
time step. This simple model severely limited the accuracy of
the mechanical stress and force predictions.
Colegrove and Shercliff used a CFD commercial software
for a 2D and 3D numerical investigation on the inuence of pin
geometry during FSW [13,14], comparing different pin shapes
in terms of material ow and welding forces on the basis of
both a stick and a slip boundary condition at the toolworkpiece
interface. Inspite of the goodobtainedresults, the accuracyof the
analysis is limited by the assumption of isothermal conditions.
Despite signicant recent advances in numerical modeling of
the FSWprocess, the previous models have severe limitations in
either the representation of geometry, or the material behavior,
or the boundary conditions. The objective of this research is to
develop a numerical model that can be use for optimal design of
FSW process. In this paper, a fully 3D FEM model for the FSW
process is proposed, that is thermo-mechanically coupled and
with rigid-viscoplastic material behavior. A unique feature of
this model is the representation of sheet seam(abutting edges) as
a continuum. This continuum hypothesis avoids the numerical
instabilities that result from the discontinuities present at the
edge of the two sheets. Predicted results are compared with the
experimental data to validate this model.
2. Governing equations
The commercial FEA software DEFORM-3D
TM
, Lagran-
gian implicit code designed for metal forming processes, is used
to model the FSW process. The workpiece is modeled as a rigid
visco-plastic material, and the welding tool is assumed rigid.
This assumption is reasonable as the yield strength of the sheet
(conventionally aluminum alloy) is signicantly lower than the
yield strength of the tool (tool steel or carbide). The following
is a summary of equations [15] governing the material behavior.
2.1. Constitutive equations
A rigid-visco-plastic material model with Von Mises yield
criterion and associated ow rule is used. In the deformation
zone,

ij
=
3
2

ij
(1)
with =

3
2
{
ij

ij
}
1/2
and

=

3
2
{
ij

ij
}
1/2
.
In Eq. (1), the effective stress depends on the strain-rate-
dependent function, which is to be determined by the properties
of the material being analyzed.
= ( ,

, T) (2)
2.2. Finite element formulation
The rigid-viscoplastic nite element formulation is based on
the variational approach. According this approach, the actual
velocities (i.e. the actual solution) among all admissible veloc-
ities u
i
that satisfy the conditions of compatibility and incom-
pressibility, as well as the velocity boundary conditions, gives
the following functional a stationary value
=

V
E(
ij
) dV

S
v
F
i
u
i
dS (3)
where E denotes the work function, which gives:

ij
= E/
ij
.
The incompressibility constraint on admissible velocity elds
is removed by introducing a penalized form of the incompress-
ibility in the variation of the functional. Therefore, the actual
velocity eld is determined fromthe stationary value of the vari-
ation as follows,
=

dV + K

V

v

v
dV

S
F
F
i
u
i
dS = 0 (4a)
where = ( ,

), and
v
=
ii
is the volumetric strain-rate.
The penalty constant K should be very large positive constant
for incompressibility so that
lim
K

V

v

v
dV= lim
K
1
K

S
F
F
i
u
i
dS

dV

= 0
(4b)
2.3. Thermal effects
The temperature generated in FSW process can be quite high
and have a considerable inuence on the mechanical response.
G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396 391
This heat generation and transfer is expressed in the form of
energy balance as follows,
k
1
T
,ii
+ r c

T = 0 (5)
where k
1
T
,ii
presents the heat transfer rate, r the heat generation
rate and c

T the internal energy-rate. It is assumed that the heat

generation in the deformation zone is due only to plastic defor-
mation and frictional conditions at the toolwork piece interface.
Insides the work piece,
r =
ij

ij
(6)
The fraction of mechanical energy transformed into heat and
is usually assumed to be 0.9. The fraction of the rest part of the
plastic deformation energy (1 ), stored in the body, causes
changes in dislocation density, grain boundaries, and phases
and so on. This energy is usually recoverable by annealing. The
energy balance is written in the variation form

V
k
1
T
,i
T dV +

V
c

TT

ij

ij
T dV

S
q
q
n
T dS = 0 (7)
where q
n
is the heat ux across the boundary surface S
q
, and
q
n
= k
q
T
,n
(8)
To solve problems of this nature, it is required that the temper-
ature eld satises the prescribed boundary conditions and Eq.
(8) for arbitrary perturbation T. The nite element formulation
for temperature analysis can be expressed as
[C]{

T} + [K
c
]{T} = {Q} (9)
The theory necessary to integrate Eq. (9) can be found in vari-
ous numerical analysis textbooks. The new temperature is often
found by the nite difference approximation
T
t+t
= T
t
+ t[(1 )

T
t
+

T
t+t
] (10)
The convergence of the time marching scheme depends on the
choice of the parameter . It is usually considered that should
be larger than 0.5 to ensure an unconditional stability and a value
of 0.75 is usually selected.
3. FEM model for FSW
The FSW modeling is divided into two stages: (1) sinking
stage and; (2) welding (advancing) stage. In other words, FSW
is modeled in this research from its initial state to steady state.
During the sinking stage the tool rst moves down vertically at
0.1 mm/s witha rotatingspeedof 1000 rpm, then, duringwelding
or advancing stage, the rotating tool moves along the weld line
(seam) joining the two workpieces. Before these two stages, the
tool is tilted 2

about the vertical axis, namely the nuting angle.

The sinking stage is modeled to obtain high enough temperature
for the subsequent welding process and the advancing stage is
modeled to investigate the thermo-mechanical phenomena in the
formation of weld nugget.
3.1. Material model
For the thermal characteristics of the AA7075 aluminum
alloy taken into account, the following values were utilized:
thermal conductivity k
1
=180 N/(s

C) and thermal capacity

c =2.4 N/(mm
2
C) taken from literature for aluminum alloys.
No variation of k
1
and c with temperature was taken into account.
this assumption linearizes the thermal Eq. (9) and results in bet-
ter convergence. Arigid-viscoplastic temperature and strain rate
dependent material model is employed.
= KT
A
(

)
B
( )
C
(11)
where K=2.69E10, A=3.3155, B=0.1324 and C=0.0192,
are material constants determined by a numerical regression
based on experimental data. As it can be seen, an increase in
temperature leads to a decrease in owstress (A<0). On the con-
trary, an increase in both strain and strain rate leads to an increase
in ow stress (B>0, C>0). A constant interface heat exchange
coefcient of 11 N/(mms

C), based on previous works [16,17],

was utilized for the tool sheet contact surface. The value of the
heat transfer coefcient is not very important in FSW due to
the fact that the tool quickly reaches a steady-state temperature
very close to the sheet temperature, resulting in a poor thermal
exchange between the tool and the workpiece. A preliminary
sensitivity analysis for different interface heat exchange coef-
cient value conrmed that there was no signicant variations of
temperature as interface heat exchange coefcient changes.
3.2. Workpiece and tool models
In order to obtain the desired vertical material ow, different
pin shapes can be adopted such as cylindrical or conical, either
smooth or threaded. Nevertheless, it is noticed that tool threads
wear during FSW quickly lead to a self-optimized smooth
pin surface [1318]. For this reason and to avoid possible con-
tact instabilities due to the presence of the threads, a cylindrical
smooth pin shape was taken into account in this paper. Based on
previous experimental campaign [19], the following tool design
was utilized (Fig. 2): a pin height of 2.8 mm, a pin diameter
equal to 3 mm, a shoulder diameter equal to 10 and a 0.5 mm
llet radius at the pin-shoulder interface. The tool was mod-
eled as rigid body and meshed, for the thermal analysis, with
about 3000 tetrahedral elements. Fig. 3 schematically shows the
sketch of the model and the assumptions made on the thermal
and mechanical phenomena.
During the modeling of workpiece, a single block con-
tinuum model (sheet blank without a gap) is used in order to
avoid contact instabilities due to the intermittent contact at the
sheetsheet and sheettool interfaces. The rotating tool moves
forward and welds a crack left behind the pin as it advances
along the welding line. The sheet blank, 3 mm in thickness,
was meshed with about 10,000 tetrahedral elements with sin-
gle edges of about 0.75 mm; in this way, about four elements
were placed along the sheet thickness. Anon-uniformmesh with
adaptive re-meshing was adopted with smaller elements close
to the tool and a re-meshing referring volume was identied
392 G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396
Fig. 2. FSW tool mesh utilized.
Fig. 3. Model sketch and frictional heat generation at the interface toolsheet.
all along the tool feed movement (Fig. 4). Experience in previ-
ous FEMsimulation shows that a coarser mesh leads to incorrect
results and a ner mesh results in unaffordable computation time
without signicant improvement of simulation results. A con-
stant shear friction factor of 0.46 is used for toolsheet interface
on the basis of a previous experimental thermal characterization
and a numerical sensitivity analysis for the shear friction fac-
tor m [20]. All the simulations are performed using the same
procedure and process parameters; in particular, tool sinking of
0.1 mm, tool rotating speed of 1000 rpm, and nuting angle of
Fig. 5. Vertical force vs. time compared with experimental values
for R=1500 rpm and v
f
=200 mm/min (FSW3), and R=1500 rpm and
v
f
=100 mm/min (FSW4).
2

and a varying advancing velocity were chosen. Advancing

velocity of the tool was varied to investigate its effect on the
weld zone.
4. FSW experiment
The proposed model has been calibrated with several exper-
imental tests. In particular, the shear friction factor for the
sheettool interface was selected in order to best reproduce
the welding force and temperature distribution during the FSW
process. Acomparison between numerical results and the exper-
imental results from reference [21] for welding forces is shown
in Fig. 5; it is seen that there is a very good agreement between
the predicted force and experimental data for maximum mea-
sured value.
The temperature predicted by the proposed model was also
compared to the experimental results (Fig. 6). It is observed
that the predicted temperature matches the experimental results
very well (see [20] for experimental details). It has to be noticed
that although the comparisons were performed using an AA
6082-T6 aluminum alloy, it can be reasonably accepted that the
validity of the thermal model is not strongly affected by the use
of a different aluminum alloy. This is because that there is no
signicant difference in thermal properties for 6000 and 7000
series aluminum alloys. Furthermore, it is reasonable to assume
Fig. 4. The FE continuum model at the beginning of the simulation.
G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396 393
Fig. 6. Temperature histories at 8 mm from the welding line for R=1040 rpm
and v
f
=100 mm/min.
Fig. 7. Temperature prole in a xz section for v
f
=100 mm/min.
that the same friction factor can be used for the contact interface
between the tool material and the sheet when using different
aluminum alloys.
5. Results and discussion
5.1. Temperature prole
It is known that during the FSW process, joining is achieved
through frictional heating between the tool and the sheet, plas-
ticizing, mixing, and extrusion action of a rotating pin-shoulder
tool that moves between two parts being joined. The ultimate
requirement for a FSW process is to create a certain amount of
friction heat, which can keep the welding material in a well-
plasticized state with a suitable temperature, and to generate a
high hydrostatic pressure along the joint line so that a sound
weld can be generated. The heat generation in FSW is in direct
proportion to deformation and frictional energy created during
the stirring process. The latter depends on the friction factor and
friction area between the tool shoulder and workpiece surface, as
well on the rotation speed of the welding head pin and the pres-
sure applied to the welding tool head shoulder. Fig. 7 shows the
temperature proles around weld zone for the advancing speed
Fig. 9. Effective strain prole along a xz section for v
f
=100 mm/min.
of 100 mm/min. The typical v-shape of temperature distribution
around the pin, due to the high thermal contribution of the fric-
tion forces at the toolworkpiece interface, is very similar to that
observed in FSW experiments.
Especially at lower tool advancing speed, both an expansion
of the heat affected zone (HAZ) and a higher value of maximum
temperature is seen with decreasing advancing speed. Fig. 8
illustrates temperature distribution in the FSWjoint along trans-
verse section of the weld zone for different pin advancing speeds.
The prole of temperature distribution is directly related to the
distortion of the welded workpiece, and can be object of further
investigations. The more uniformtemperature distribution along
the thickness direction of the workpiece, the less distortion in
the workpiece will take place after welding. Especially for the
thin sheet, this post-weld distortion can be critical.
5.2. Strain prole
As it is known that the thermal history and the plastic defor-
mation of the weld region are strongly relevant to the microstruc-
ture of weld itself, better understand of the effects of process
parameters on the strain distribution in the joint during the FSW
is very helpful for optimal design and process control. Fig. 9
illustrates the straindistributioninthe weldzone inthe transverse
section. It is seen fromthis gure that the strain distributions are
not symmetric about the weld line and maximum strains are on
the material advancing side where a positive combination of
the tool feed rate and of the peripheral tool velocity is observed.
Fig. 10 presents the effective strain generated in FSWjoint along
transverse section for different advancing speeds. Again, the pic-
ture indicates that the maximum strain is not in correspondence
with the welding line, but it is shifted towards the advancing
side. A further consideration can also be made: not only the
maximum strain but also the average strain increases with the
Fig. 8. Temperature distributions along xz section in FSW joint as a function of advancing speed.
394 G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396
Fig. 10. Effective straindistributions alongxz sectioninFSWjoint as a function
of advancing speed.
decrease of the advancing speed, especially in the centre of the
weld where the stirring action of the pin sensibly extended the
plastically deformed area due to the increased time to heat up
and weld the same length.
5.3. Non-symmetry nature of the process
During welding, as long as the tool sinking is kept constant,
the stresses that the tool is subjected to vary with the process
parameters. In addition, material ow pattern and ow rate are
also strongly dependent on such parameters. This FSWmodel is
able to predict the material ow in the weld zone. The material
ow pattern is shown in Fig. 11(a,b) where the arrows represent
the material velocity vectors. It can be seen from this gure
that no vertical material ow is observed with a cylindrical pin.
It is observed that a different pin shape is likely to cause an
additional down ow creating a helical movement produced by
the circular movement component in the horizontal (xy) plane,
observed with a cylindrical pin (Fig. 11a), and another circular
movement component in vertical (yz) plane. This will produce
more uniformplastic deformation in the weld zone and provides
better nugget integrity as dened in reference [22].
Further consideration can be made based on the results
obtained in terms of temperature and strain. As seen fromFig. 8,
temperature distribution along a xz section is nearly symmet-
ric, while strain distribution presents high asymmetry (Fig. 10).
Fig. 12. Schematic sketch of the advancing speed and peripheral velocity com-
position at pin surface.
This can be explained considering actual rotating speed and
advancing speed values used in FSW industrial applications
and adopted for the simulations in this paper: with a rotat-
ing speed of 1000 rpm and a tool pin diameter of 3 mm, the
peripheral pin velocity due to rotation is about 9420 mm/min
at pin surface, much higher then the advancing speed (rang-
ing from 50 to 200 mm/min in this research); this difference
will of course be even greater at a generic point of the shoulder
surface, where most of the frictional heat is generated. Further-
more, frictional heat that supplies most of the heat conferred
to the welding, is directly proportional to the relative velocity
between tool and workpiece. Since the relative velocity due to
rotating speed is much higher than the advancing speed, the
heat generation is dominated by the pin rotation resulting in
an almost symmetric temperature distribution about weld line.
Fig. 12 shows a schematic that highlight the difference between
pin peripheral velocity and advancing speed. Strain, on the con-
trary, is non-symmetric about the welding line, which indicates
that the material ow, and/or deformation, is highly inuenced
by the combination of rotating and advancing speeds of the tool.
Fig. 13(ac) shows the deformation pattern of two particles, P1
Fig. 11. Material ow pattern in weld zone (a) along the welding line; and (b) in a 3D view.
G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396 395
Fig. 13. Particles ow pattern during the welding: (a, b) circular movement in
the horizontal (xy) plane; and (c) nal position in the advancing side.
and P2, initially located, respectively, on the advancing and on
the retreating side of the welding at a distance of 1 mm from the
welding line and at z =1.5 mm (Fig. 13a). After being involved
in the circular movement around the pin (Fig. 13b), their nal
position in the xy plane is characterized by a negative value of
x, indicating that the particles are both in the advancing side of
the welding, and by a different value of y (Fig. 13c), accordingly
to the different strain values observed in Figs. 9 and 10. A dif-
ference in temperature histories between the two particles can
be seen in Fig. 14: after sharing the same temperature history
during the rst few seconds, as the tool gets closer, particle P1
is more involved in the circular movement around the pin thus
experiencing higher temperature, with a maximum difference
of about 40

C, and deformation. Then, once completed their

deformation pattern, temperature will tend to the same value
during the cool down of the sheet.
Finally, it should be observed that no signicant movement
along the vertical (z) axis is observed, due to the cylindrical
pin shape utilized. Once again, a vertical velocity component is
likely to be obtained once a different pin shape, i.e. conical, is
chosen.
Fig. 14. Temperature history for particle P1 and P2.
6. Conclusions and future work
A 3D numerical model for FSW process is proposed, that is
thermo-mechanically coupled, uses rigid-viscoplastic material
description and a continuum assumption for the weld seam. The
proposed model is capable of predicting the effect of process
parameters on process thermo-mechanics, such as the temper-
ature, strain, strain rate as well as material ow and forces.
Predicted results are compared with experimental data and
good agreement is obtained. Base on the preliminary simula-
tion results, it is found:
an expansion of the heat affected zone and an increase of
the maximum temperature and maximum strain in the nugget
with the decrease in advancing speed,
temperature distribution about the weld line is nearly sym-
metric because the heat generation during FSW is dominated
by rotating speed of the tool which is much higher than the
advancing speed,
material ow (deformation) in the weld zone is non-
symmetrically distributed about the weld line because the
material ow during FSW is mainly controlled by both
advancing and rotating speeds.
This continuum model has been used to investigate the effect
of welding tool geometry and advancing speed on the process
variables and grain size distribution in the welding zones for pro-
cess optimization of FSW. The results of this study are included
in reference [22].
Acknowledgements
This work was supported by the Dean of College of Engineer-
ing, Prof. Bud Baeslack III, and the Manufacturing Research
Group (Professor Shivpuri, Director). Gianluca Buffa was addi-
tionally supported by the Italian Ministry of Education for his
thesis work at the Ohio State University. The authors wish to
express their grateful thanks to Prof. Fabrizio Micari, Univer-
sity of Palermo, Italy, for his constant help and encouragement
during this research. Scientic Forming Technologies Corpora-
tion provided help with the DEFORM software.
References
[1] H.J. Liu, H. Fujii, M. Maeda, K. Nogi, J. Mater. Proc. Tech. 142 (2003)
692696.
[2] C.G. Rhodes, M.W. Mahoney, W.H. Bingel, R.A. Spurling, C.C. Bamp-
ton, Scripta Mater. 36/1 (1987) 6975.
[3] M. Guerra, C. Schmidt, L.C. McClure, L.E. Murr, A.C. Nunes, Mater.
Charact. 49 (2003) 95101.
[4] Y. Li, L.E. Murr, J.C. McClure, Scripta Mater. 40 (19991) 10411046.
[5] K. Colligan, Weld. J. (July 1999) 229237.
[6] A.P. Reynolds, Sci. Technol. Weld. Joi. 5/2 (2000) 120124.
[7] V. Soundararajan, S. Zekovic, R. Kovacevic, Int. J. Mach. Tool. Manu.
45 (2005) 15771587.
[8] X. Deng, S. Xu, Trans. NAMRI/SME SME 29 (2001) 631638.
[9] S. Xu, X. Deng, Proceeding of the 21st Southeastern Conference on
Theoretical and Applied Mechanics (SECTAM XXI), Orlando, Florida,
May 1921, 2002, pp. 699704.
396 G. Buffa et al. / Materials Science and Engineering A 419 (2006) 389396
[10] H. Schmidt, J. Hattel, Model. Simul. Mater. Sci. Eng. 13 (2005) 7793.
[11] P. Ulysse, Mach. Tool. Manu. 42 (2002) 15491557.
[12] C.M. Chen, R. Kovacevic, Mach. Tool. Manu. 43 (2003) 13191326.
[13] P.A. Colegrove, H.R. Shercliff, Sci. Technol. Weld. Joi. 9/6 (2004)
483492.
[14] P.A. Colegrove, H.R. Shercliff, Proceedings of the 4th International Sym-
posium on FSW, 2003.
[15] S. Kobayashi, S.K. Oh, T. Altan, Metal Forming and The Finite-Element
Method, Oxford University Press Inc., 1989.
[16] G. Buffa, L. Fratini, Proceedings of 8th NUMIFORM Conference,
Columbus, Ohio, 2004, pp. 12241229.
[17] L. Fratini, G. Buffa, Int. J. Mach. Tool. Manu. 45 (2005) 11881194.
[18] R.A. Prado, L.E. Murr, K.F. Soto, J.C. Mc Clure, Mater. Sci. Eng. A
349 (2002) 156165.
[19] A. Barcellona, G. Buffa, L. Fratini, Keynote Paper of the VII ESAFORM
Conference, Trondhaim, Norway, April 2830, (2004) 371374.
[20] L. Fratini, S. Beccari, G. Buffa, Trans. NAMRI/SME 33 (2005) 259266.
[21] G. Buffa, L. Fratini, F. Micari, CIMNE International Conference for
Coupled Problems in Science and Engineering, Santorini Island, Greece,
May 2528, 2005.
[22] G. Buffa, J. Hua, R. Shivpuri, L. Fratini, Mater. Sci. Eng. A 419 (2006)
381388.