Journal of Manufacturing Processes 23 (2016) 278286

Contents lists available at ScienceDirect

Journal of Manufacturing Processes

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

Technical Paper

A study on the variation of forces and temperature in a friction stir

welding process: A nite element approach
Rahul Jain a , Surjya K. Pal a, , Shiv B. Singh b
a
Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, 721302, India
b
Department of Metallurgical and Materials Engineering, Indian Institute of Technology Kharagpur, 721302, India

a r t i c l e i n f o a b s t r a c t

Article history: A three dimensional coupled thermo-mechanical nite element model (FEM) is proposed to simulate
Received 23 November 2015 a friction stir welding (FSW) process based on Lagrangian incremental technique. Since FSW is a large
Received in revised form 7 March 2016 deformation process, workpiece is considered as a rigid visco-plastic material. The model has been devel-
Accepted 15 April 2016
oped for predicting forces, spindle torque, temperature and plastic strain for a butt welding between two
Available online 27 May 2016
AA2024-T4 metals having thickness of 5.9 mm each. The developed model has been validated with exper-
imental results (forces, spindle torque) obtained from literature. Maximum force is obtained during the
Keywords:
plunging phase of the tool and this makes tool susceptible to failure. Forces and spindle torque reduce
Friction stir welding
Finite element simulation
with the increase in rotational speed due to increase in heat generation rate which is also reected in
Temperature distribution temperature distribution. Effect of welding speed and frictional boundary condition are studied. Conical
Pin shape pin shape produces higher material velocity as compared to cylindrical with reduced plunge force.
2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction
FSW is a joining process which welds two materials in solid state.
This technique was successfully executed in the last two decades
in TWI Cambridge [1]. Since FSW welds materials in solid state,
it has got many advantages over conventional welding techniques,
such as no consumable materials, no defects related to solidication
of the material, higher weld strength, reduced power consump-
tion, etc. [2]. A rotating, non-consumable, specically designed tool
plunges on the faying surfaces of the clamped workpiece to gen-
erate sufcient heat and to stir the material beneath the shoulder,
and subsequently, a relative motion is given to the tool in weld-
line direction to perform the weld. A schematic of different stages
of friction stir welding process is shown in Fig. 1. Heat generation
due to friction between the tool and the work-surface, and plas-
tic deformation inside the workpiece are responsible to bring the
material into a viscous state [3].
Experimental investigation reveals the generation of higher
forces in the plunging stage as compared to the welding stage [4].
Different methodologies have been implemented to experimen-
tally measure the forces such as traditional load cell to measure
the axial force[5], three component rotating type piezoelectric
dynamometer to measure forces and torque generated during the
Corresponding author. Tel.: +91 3222 282996.
E-mail address: skpal@mech.iitkgp.ernet.in (S.K. Pal).
process [6], responses of electrical motor (like power, current, etc.)
to measure the forces and torque [7,8].
FEM of a FSW process is complex due to highly nonlinear con-
tact interaction between workpiece and tool, large deformation of
workpiece leading to higher strains and strain rates. Some of the
notable FE models are discussed below.
Chao et al. [9] have simulated heat transfer during FSW as
two boundary value problems: rstly, steady state heat trans-
fer for the tool, and second a transient heat transfer problem for
the workpiece. They have considered heat generation due to fric-
tional force only neglecting the heat generation due to plastic
deformation. Based on temperature, residual stress over the weld-
ing domain was also calculated. Deng and Xu [10] carried out
a two dimensional analysis in ABAQUS to predict material ow
and material spatial velocity. They reported that predicted mate-
rial ow pattern is comparable to experimental pattern. Chen and
Kovacevic [11,12] have developed three dimensional symmetric
thermo-mechanically coupled model to simulate a FSW process by
using ANSYS to predict temperature history and stress distribu-
tion. They have also predicted forces by integrating the stress over
the area. Buffa et al. [13] have simulated three dimensional cou-
pled thermo-mechanical analysis using Lagrangian implicit code
in DEFORM-3D. They have simulated plunging and welding stages
by dening workpiece as a rigid visco-plastic material. Trimble
et al. [6] have simulated force generation during FSW process using
Johnson-Cook material model and compared force, yield strength,
% elongation between a cylindrical pin and a threaded cylindrical

http://dx.doi.org/10.1016/j.jmapro.2016.04.008
1526-6125/ 2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286
Fig. 1. Schematic of various stages of FSW process (a) plunging (b) Dwelling (c) Welding [3]. Note: Dashed line indicates abutting edges, while arrows indicates the tool
movement direction.
pin. Instead of constraining the mesh nodes, they have enclosed the
workpiece inside the box to restrict the rigid motion. Trimble et al.
[14] have studied the effect of different pin shape and rotational
speed for increased welding speed. They achieved a good weld
with triute pin at a speed of 355 mm/min. Assidi et al. [15] have
adopted Arbitrary Eulerian Lagrangian (ALE) method to predict and
validated the forces and power with the experimental values. They
have compared coulombs law of friction and Nortons frictional
model. Some other researchers have simulated the material ow
and force generation for magnesium alloy and have also studied
the effect of shoulder diameter and pin diameter on strain, tem-
perature and heat affected zone [16,17]. A few researchers [18,19]
have also studied the different constitutive material models and
their inuence on the output responses of the FSW process.
Measurement of forces and torque is quite difcult as it involves
costly equipment and extra xture. One of the alternative ways of
estimating forces is through the numerical modeling of the process.
Though modeling of FSW has been done by a lot of researchers but
only a few [6,20,15] have predicted the forces generated during
the process. FSW process is performed for a range of input process
parameters (rotational speed, welding velocity, etc.) and have var-
ious output responses viz. temperature, forces, torque, etc. Mostly,
validation of FEM model is carried out with numerically predicted
temperature-time plot for a single input process parameter. A com-
plete validation of model can only be achieved if it is validated with
different output responses and under different input conditions.
Also, modeling is only carried out for a steady state welding condi-
tion [2123] or for plunging and welding stages [20,24]. A complete
realistic modeling of a FSW process, including all the three stages
having proper duration i.e. plunging, dwelling and welding has not
yet done.
In the current research work, a fully coupled 3-dimensional
thermo-mechanical model is developed based on Lagrangian
implicit technique to simulate the three stages of a FSW process.
Workpiece is considered as a rigid visco-plastic material and stick-
ing friction boundary condition is dened at the interface. The
model is capable of predicting forces and spindle torque generated
during the process as these parameters inuence the tool design.
Distribution of temperature and plastic strain are also predicted,
which inuences the microstructure and grain distribution in the
nugget zone. Axial force, welding force and spindle torque are val-
idated with the experimental results published in the literature
[8] for different input parameters. Effects of tool rotational speed,
welding speed and pin shape (cylindrical and tapered cylindrical)
on the output responses (forces, torque, and temperature) are also
studied.
2. Simulation details
FSW is a large deformation process and for the simulation, it is
very crucial to control the mesh distortion, with the use of effective
279
re-meshing technique. Contact interaction between tool and work-
piece is nonlinear and precise denition of frictional boundary
condition is also required. Keeping the above criteria in mind,
DEFORM-3D software is selected to simulate the process. Apart
from efcient re-meshing capability the software also has a unique
point tracking method, which is helpful in simulating material ow
and to visualize the deformation pattern. The FEM model has been
validated with forces and spindle torque values obtained from the
research work of Su et al. [8]. They have monitored and recorded
the electrical signals of three phase AC induction motor in FSW
machine. Based on the electrical signals forces and spindle torque
are calculated. The experiments were performed for different input
process parameters.
FSW modeling is divided into three stages viz. plunging,
dwelling and welding stages. During assembly of workpiece and
tool; tool is titled by 2.5 towards its trailing edge. During the plung-
ing stage, tool moves in negative Z direction at a dened rotational
speed (600, 800 and 1000 rpm are used in this study) and a feed
velocity of 8 mm/min, afterwards dwell time of 10 s is dened, in
which tool rotates at the same rpm in its own position to further
increase the temperature and make material beneath the material
softer, and nally welding speed is dened to the tool to simulate
the welding stage till steady stage is achieved. Time required for
plunging is 43 s to achieve a plunge depth of 0.1 mm.
All simulations are performed with the similar procedure
explained above with a tilt angle of 2.5 degrees, plunge velocity
of 8 mm/min, plunge depth of 0.1 mm, and welding veloc-
ity of 80 mm/min. Three rotational speeds (600 rpm, 800 rpm
and 1000 rpm) are chosen to investigate their effect on output
responses. All input parameters and dimensions are same as the
experimental data of Su et al. [8].
2.1. Geometric modeling and boundary conditions
Workpiece and tool dimension are taken as per the experimen-
tal data available in literature [8]. Workpiece material is AA2024-T4
aluminum alloy with 5.9 mm thickness and has a length and width
of 120 mm and 60 mm, respectively each. Workpiece is modeled
as rigid visco-plastic material and meshed with tetrahedral ele-
ments. To improve the computational efciency of the model,
biased meshing is dened using mesh window option as shown
in Fig. 2. Here, a mesh window at the interface is created and ner
mesh with a uniform element size of 0.8 mm is dened, such that
it has approximately 7 nodes along the thickness direction and
remaining workpiece is meshed with the minimum element size
of 2 mm, with a size ratio of 3 to have 40,000 tetrahedral elements
on workpiece. Tool is modeled as a rigid body with dened thermal
degree of freedom to calculate the heat transfer between workpiece
and tool. Since yield strength of tool material is much higher as
compared to workpiece material, assumption of rigid tool is valid.
Tapered cylindrical pin tool has following dimensions: shoulder
280 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286

Fig. 2. Geometric model of tool-workpiece used in simulation.

diameter 15 mm, top and bottom pin diameter is 6 mm and 2.5 mm, Table 1
Properties of workpiece and tool material [26].
respectively, and pin height of 5.5 mm whereas cylindrical pin has
pin diameter of 6 mm and other parameters are same as the tapered Properties AA 2024 Tool steel H13
cylindrical pin. Tool material is tool steel H13. Tool is also meshed Youngs modulus (N/mm2 ) 68,900 210,000
with tetrahedral element using biased meshing, as shown in Fig. 2. Thermal conductivity (N/s C) 181 24.5
Tool shoulder and pin which comes in contact with the workpiece Heat capacity (N/mm2 C) 2.43 2.78
have a ner mesh size of 0.3 mm to have dened precise geome- Coefcient of thermal expansion (mm/mm C) 22 11.7
Poissons ratio 0.33 0.3
try and boundary of pin and shoulder. Tool is meshed with 36382
elements.
Various boundary conditions are required to constrain the rigid form [25]. Flow stress is dened for a range of strain, strain rate
body motion of workpiece and also to dene various heat transfer (0.3100 s1 ) and temperature (20500 C).
modes. Various boundary conditions, dened in the model, are as
follows:  = (,
T)
, (1)

where,  is ow stress, is strain and is strain rate in s1 . Mechan-

ical and thermal properties of workpiece and tool material are
(a) All degree of freedom of workpiece need to be constrained since mentioned in Table 1.
rotational and travel velocities have been dened on the tool. To
achieve this, workpiece bottom face is arrested in vertical (along 2.3. Thermal model
Z axis) direction. Since material ow take place in XY plane
therefore, bottom face of workpiece cannot be constrained in FSW is a coupled thermo-mechanical problem; and heat gener-
X and Y directions, therefore, side faces of the workpiece in X ation of the problem is governed by Eq. (2), which is Fourier law of
and Y directions are constrained. heat conduction.
(b) Rotational speed and welding speed are dened on the tool in
T
Z and Y directions, respectively. k 2 T + q = c (2)
(c) Heat transfer due to radiation and convective is dened t
between free surfaces of workpiece/tool and environment with where,  is mass density of the material, c is specic heat capacity,
convective heat transfer coefcient of 0.02 N/s/mm C. t is time and q is heat generation rate which comprises of frictional
(d) It has been reported that the heat transfer between workpiece and plastic deformation heat as mentioned in Eq. (3).
and tool is higher than the workpiece and environment. There-
q = q f + q p (3)
fore, heat transfer co-efcient of 11 N/s/mm C [24] has been
dened between them. where, q f is the frictional heat generation rate and q p is the heat gen-
(e) To improve the computational efciency of the simulation, eration rate due to plastic deformation of the material as mentioned
backing plate is neglected and heat transfer coefcient is taken by Eq. (4)
as 11 N/s/mm C [24].
q p = ( )
(4)

where,  is inelastic heat fraction, which is dened as the amount

2.2. Material model
In FSW process, material undergoes large deformation along
with high temperature and strain rate. Therefore, to capture the
correct behavior of the material it is important to dene the ow
stress of material as a function of temperature, strain and strain
rate. Hence, workpiece is dened as rigid visco-plastic material i.e.
material is rate dependent but elastic deformation of the material
is neglected to achieve better convergence. This is a valid assump-
tion because in FSW process effective strain is in a range of 680
[3,20,24] as compared to elastic strain which is considered as 0.002
in offset method to calculate yield point. Here, ow stress is dened
as per Eq. (1) and data is fed to the simulation in the tabular
of mechanical energy converted to heat energy and it is taken as 0.9
[16], the fraction of the remainder energy is utilized in changing dis-
location density, sub-grain boundaries, phase transformation and
evolution, etc.
2.4. Frictional model
Contact condition at the interface of tool and workpiece is quite
complex and also there is not much experimental evidence avail-
able, which suggest the exact frictional conditions. Researchers
have simulated FSW with different frictional boundary conditions.
A few have used Coulombs law of friction [15,27]; some have used
sticking condition [16,24]; and a few have considered coefcient
 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286 281

of friction as a function of pressure and slip rate [21,22]. In this problem, where the rst order variation of the functional vanishes,
simulation, sticking frictional condition is assumed as per Eq. (5). as given in Eq. (10).
Sticking condition is selected because it is best suited for most of  
the forming operations, and material state in FSW can be correlated  = dV
 Fi ui dS = 0 (10)
with such operations [28] V SF

Incompressibility in penalized form is utilized to remove the

 = mmax (5)
incompressibility constraint on admissible velocity eld (Eq. (10)).
With this, Eq. (11) forms as fundamental equation for nite element
where,  is contact pressure,  max is the shear yield strength of
formulation
material (0.577 times of the yield strength as per von-Mises yield   
criteria), and m is shear factor. To calibrate the shear factor value,
 = dV +
 v v dV Fi ui dS = 0 (11)
three different simulations are carried out at three different shear V V SF
factor values. For each case, forces and torque are compared with
experimental data and good match is observed for a shear factor where and ui are large penalty constant and arbitrary variation
value of 0.4. Table 2 shows the comparison of average welding respectively. and v are variations in strain rate derived from
forces and torque for three different shear factor for rotational ui . v = ii , is the volumetric strain rate.
speed of 800 rpm and welding velocity of 80 mm/min.
3. Results and discussions

2.5. Governing equations 3.1. Forces and torque

Strain rate dependent process along with large deforma- Force study is a critical aspect of FSW, as it inuences the tool
tion is generally simulated with rigid visco-plastic FEM method. design and machine structure. Fig. 3 shows the evolution of axial
This method is based on variational approach which gives non- force during FSW. Initial increase is observed due to plunging of
linear equations, which is solved by iterative methods like pin in the workpiece, subsequently it attains a peak value during
NewtonRaphson and direct iteration methods. Following are the plunging of shoulder. Maximum force is generated during plunging
various fundamental governing equations for rigid visco-plastic which makes pin susceptible to failure due to buckling. A gradual
material [28]. Eq. (6) is the constitutive relation for the method reduction in force is due to dwelling of the tool, where, material
while Eqs. (7) and (8) are supporting equations. attains a sufciently high temperature leading to drop in force.
40% reduction of forces is observed from the peak value during
3  this period; subsequently force attains a steady value during the
ij =  (6)
2  ij welding phase.
 Fig. 4 shows, validation of the model with average experimental
3   forces during welding and the model is successfully able to predict
 = {  } (7)
2 ij ij axial (Fz ) and welding (Fy ) forces. Percentage errors in axial force
 for 600, 800 1000 rpm are 12.5%, 6.5% and 9.1%, respectively.
2   Pin geometry and dimension inuence the welding force and
= ij ij (8) this makes study of welding force important. Though axial force
3

where, , , ij , ij are effective stress/ow stress, effective strain

rate, strain rate components and deviatoric stress components,
respectively. Also effective stress is a function of temperature,
strain and strain rate as per Eq. (1).
The variational approach is based on one of two variational
principles, such that among admissible velocities ui should satisfy
compatibility and incompressibility conditions, along with veloc-
ity boundary conditions. The actual solution gives the following
functional (function of a function) a stationary value
 
= E( ij )dV Fi ui dS (9)
V SF

where Fi , V, SF , and Eij are surface tractions, volume of workpiece,

force surface and work function, respectively. Based on second
extremum principle for rigid visco-plastic material, work func-
tion is given by ij = E/ ij . The solution to the original boundary
value problem is obtained from the solution of the dual variational Fig. 3. Evolution of axial force with time during FSW process.

Table 2
Comparison of experimental and simulated force and torque values for different shear factors.

Axial force (N) Percentage error (%) Spindle torque (N m) Percentage error (%)

Experimental 10,175 13.326

m = 0.3 11,353 11.5 10.65 20
m = 0.4 9500 6.6 11.93 10.5
m = 0.5 8186 19.6 17 27.6
282 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286
Fig. 4. Validation of experimental forces results with simulation data for
80 mm/min.
has much higher value as compared to the welding force, but pin is
subjected to shear stress due to bending and torsion, and the shear
stress is mainly driven by the welding force. It is a well-known
fact that material is more prone to failure under shearing action
rather than compression. Welding force (Fy ) is also validated with
the experimental results and a good correlation has been observed.
With increase in the rotational speed from 600 rpm to 1000 rpm,
welding force has reduced from 4270 N to 2523 N. This indicates
that welding force decreases with increase in rotational speed of the
tool, this similar behavior of welding force has also been reported
in the literature [2931].
Similar to axial force, spindle torque also increases in plung-
ing phase and attains a maximum value during impingement of
shoulder, followed by a drop during plunging phase and attaining
a steady value during welding, as shown in Fig. 5.
Spindle torque indicates the power consumption during the pro-
cess and is the main driving force for selection of motor capacity
for the FSW machines. Simulation results show a good correla-
tion with experimental values for 800 and 1000 rpm, but for lower
rpm simulation torque is under predicted to those of experimen-
tal values. Spindle torque reduces with increase in rotational speed
because at the higher rpm, ow of material becomes easier as com-
pared to the lower rpm due to higher temperature. It has been
observed that from 600 to 800 rpm, maximum temperature has
increased by 12.8%, leading to easier ow of the material and hence
Fig. 5. Evolution of spindle torque with time during FSW process.
Fig. 6. Validation of simulation results for torque with experimental values.
lowered spindle torque at 800 rpm as compared to 600 rpm. But
from 800 rpm to 1000 rpm increase in temperature is merely 4%
and therefore spindle torque has almost remain unchanged in the
simulation results (Fig. 6).
Welding speed has an inuential effect on forces and spindle
torque as shown in Fig. 7. Forces and spindle torque are predicted
for ve different welding speeds, and the same increasing trend is
observed for axial force, welding force and spindle torque. This phe-
nomenon is due to decrease in heat generation rate and hence lesser
rise in temperature. Due to reduced heat generation, workpiece
resistance increases, leading to increase in axial force, welding force
and spindle torque. Fig. 8 shows the effect of rotational speed on
forces and spindle torque. Simulation is done for ve different rota-
tional speeds for a constant welding velocity of 80 mm/min. Forces
and spindle torque decrease with the increase in rotational speed
(except at 1000 rpm) due to increase in heat generation at the inter-
face of the tool and workpiece. Simulation results show a slight
increase in axial force and spindle torque at 1000 rpm (merely by
4.2% and 4.7%, respectively).
3.2. Temperature and strain distribution
It is well known that proper mixing of material by rotating
shoulder and pin leads to an efcient FSW process which is directly
Fig. 7. Effect of welding speed on forces and Spindle torque.
 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286 283

Fig. 8. Effect of rotational speed on forces and spindle torque.

Fig. 9. Temperature distribution for rotational speed of (a) 600 rpm, (b) 800 rpm, (c) 1000 rpm along the transverse section for welding speed of 80 mm/min.
284 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286

dependent on temperature generated during the process. It is

required to have sufcient temperature beneath the shoulder to
keep material in plasticized state. Temperature is mainly gener-
ated because of plastic deformation and frictional heat. The latter
depends on shear factor, contact area between workpiece and tool
shoulder and also on rotational speed.
In Fig. 9 temperature distributions along the transverse direc-
tion of welding for three different rotational speeds are shown.
Temperature distribution forms a typical V shape structure near
the nugget zone because of high heat generation at shoulder,
which is 2.5 times bigger than the pin, which leads to higher
heat generation at the top interface as compared to the bottom.
Also, heat dissipation from the bottom surface of the workpiece
is much higher as compared to the top surface. Effect of rota-
tional speed studied on temperature distribution is also studied.
Maximum temperatures calculated for 600, 800 and 1000 rpm are
483 C, 545 C and 570 C, respectively. This indicates increase in
the temperature with increase in rotational speed due to increase
in relative velocity between workpiece and the tool. This leads
to higher softening of the material and reduction in forces and Fig. 11. Temperature distribution on tool for 800 rpm and 80 mm/min speed at
torque generated during the process, which is explained in Section frictional factor of 0.3.
3.1.
Temperature distribution on assembly of tool and workpiece
indicates that tool temperature comes in steady state with the
temperature of workpiece and both attain nearly equal temper-
ature. Maximum temperatures of tool and workpiece are 504 C
and 512 C respectively. Maximum temperature on tool is observed
on interface area between the tool and workpiece, and also based
on the thermal conductivity of the tool material a thermal gradi-
ent can be observed. Also, with frictional factor of 0.3, maximum
temperature obtained is 512 C, which is lesser than the max-
imum temperature obtained for 0.4. This indicates larger axial
forces for lower frictional factor and same is indicated in Table 2
(Figs. 10 and 11).
Temperature distribution and plastic strain play a vital role in Fig. 12. Effective strain distribution in transverse direction for rotational speed of
800 rpm and 80 mm/min velocity.
microstructure and grain size of the welded sample. Therefore,
study of plastic strain is important to have a better understanding
on the physics of the process 3.3. Comparison of pin shape
In Fig. 12 contour plot of effective strain distribution along
the transverse direction of the weld is shown. Effective strain is Pin shape plays an important role in the efcient welding. It
descending from top to the bottom, due to the fact the higher stir- is the responsibility of the pin to make sure proper mixing of the
ring takes place on top surface leading to higher deformation of the material on the faying surface and in the thickness direction [3].
material and also higher strain is observed at the center, which is A cylindrical and tapered cylindrical pin has been compared as
the stirred zone formed by pin rotation. It is also evident that strain shown in Fig. 13. Maximum axial force (during plunging phase)
is higher on advancing side (AS) as compared to the retreating side is higher for cylindrical pin as compared to tapered cylindrical
(RS) of the workpiece. Positive velocity component in AS leads to pin, which makes former more susceptible to failure during the
more shearing action of the material compared to the other. This plunging phase. Change in average axial force during welding for
differential velocity gradient on two sides of the weld makes FSW cylindrical and tapered cylindrical pins is approximately 4%, which
process asymmetrical with higher deformation in AS, compared indicates axial force is mainly driven by the size of the shoulder and
to RS. is independent of the pin shape [6]. Average welding force for the

Fig. 10. Temperature distribution on workpiece tool assembly for rotational speed of 800 rpm and 80 mm/min speed at frictional factor of 0.3.
 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286 285

Fig. 13. Comparison of cylindrical and tapered cylindrical pin.

cylindrical pin is higher than the tapered cylindrical pin because

welding force directly opposes the translational movement of the
tool and is mainly driven by the cross-sectional area of the tool
coming in contact with the workpiece. Therefore, welding force is
mainly driven by the pin design and is independent of the shoulder
dimension. It can be also observed that the spindle torque for the
cylindrical pin is higher the tapered cylindrical pin.
Spindle torque determines the ease of material ow. Lower spin-
dle torque for tapered cylindrical pin indicates the better material
ow as compared to the cylindrical pin. Improved material ow
will lead to the better mixing of the material and hence efcient
welding.Velocity of the material ow is a critical factor for the ef-
cient material ow. Fig. 14 indicates higher material velocity for
the tapered cylindrical pin leading to efcient mixing of the mate-
rial. Truncated shape of the pin allows the material to ow in plane Fig. 14. Velocity for cylindrical and tapered cylindrical pin for 800 rpm and
direction as well as in vertical direction [13]. A study on material 80 mm/min.
velocity along the thickness direction for cylindrical and tapered
cylindrical pins is conducted, for which 11 points are selected along is on the bottom face of the workpiece. Fig. 16 shows higher velocity
the centerline of the workpiece in thickness direction as shown in for tapered pin in vertical (Z) direction as compared to the cylin-
Fig. 15. Point 1 is located at 1 mm beneath the top surface and other drical pin. Point 1 has much higher velocity due to the combined
points are in an incremental distance of 0.5 mm, such that point 11 inuence of the shoulder and the pin. For cylindrical pin velocity

Fig. 15. Location of points for velocity prediction.

286 R. Jain et al. / Journal of Manufacturing Processes 23 (2016) 278286
Fig. 16. Velocity (Z direction) distribution of the points.
is nearly in a range of 23 mm/sec except for point 1, whereas for
tapered pin velocity graph has negative gradient because of the
reduction in radius from top to bottom since material velocity is
a function of tool radius. Point lying on the bottom face of the
workpiece has almost negligible velocity for both the cases.
4. Conclusions
A three dimensional thermo-mechanical model is developed
based on Lagrangian incremental technique. Model is success-
fully validated for forces and spindle torque at different rotational
speeds with experimental data obtained from literature. Proposed
FEM model can predict temperature, forces and strain distribu-
tion. Point tracking method has been used to predict the material
velocity. Two different pin shape have been critically examined.
Following are the summary of the results
(i) Lower welding force and spindle torque is predicted for the
tapered cylindrical pin as compared to the cylindrical pin.
(ii) Material has higher velocity in case of tapered cylindrical pin as
compared to the cylindrical pin
(iii) Forces are reduced by 40% from plunging to welding phase of
FSW indicating pin are more susceptible to failure in plunging
stage as compared to the welding stage.
(iv) Forces decrease with increase in rotational speed and increase
with increase in welding speed. This is also reected in the
temperature distribution which increases with the increase in
rotational speed.
(v) Plastic strain is higher on advancing side as compared to the
retreating side indicating the asymmetry nature of the FSW
process.
References
[1] Thomas WM. Patent FrictionWelding ThomasTWI.pdf; 1995.
[2] Mishra RS, Ma ZY. Friction stir welding and processing. Mater Sci Eng R: Rep
2005;50:178, http://dx.doi.org/10.1016/j.mser.2005.07.001.
[3] Jain R, Kumari K, Kesharwani RK, Kumar S, Pal SK, Singh SB, et al. Friction stir
welding: scope and recent development. Modern Manuf Eng 2015:179228,
http://dx.doi.org/10.1007/978-3-319-20152-8.
[4] Zimmer S, Langlois L, Laye J, Bigot R. Experimental investigation of the inu-
ence of the FSW plunge processing parameters on the maximum generated
force and torque. Int J Adv Manuf Technol 2009;47:20115, http://dx.doi.org/
10.1007/s00170-009-2188-3.
[5] Kumar K, Kalyan C, Kailas SV, Srivatsan TS. An investigation of friction
during friction stir welding of metallic materials. Mater Manuf Process
2009;24:43845, http://dx.doi.org/10.1080/10426910802714340.
[6] Trimble D, Monaghan J, ODonnell GE, Donnell GEO. Force generation during
friction stir welding of AA2024-T3. CIRP Ann Manuf Technol 2012;61:912,
http://dx.doi.org/10.1016/j.cirp.2012.03.024.
[7] Mehta M, Chatterjee K, DA. Monitoring torque and traverse force in friction stir
welding from input electrical signatures of driving motors, 18; 2013. p. 1917,
http://dx.doi.org/10.1179/1362171812Y.0000000084.
[8] Su H, Wu CS, Pittner A, Rethmeier M. Simultaneous measurement of tool
torque, traverse force and axial force in friction stir welding. J Manuf Process
2013;15:495500, http://dx.doi.org/10.1016/j.jmapro.2013.09.001.
[9] Chao YJ, Qi X, Tang W. Heat transfer in friction stir welding experimen-
tal and numerical studies. J Manuf Sci Eng 2003;125:138, http://dx.doi.org/
10.1115/1.1537741.
[10] Deng X, Xu S. Two-dimensional nite element simulation of material ow in
the friction stir welding process. J Manuf Process 2004;6:12533.
[11] Chen C, Kovacevic R. Thermomechanical modelling and force analysis of friction
stir welding by the nite element method. Proc Inst Mech Eng C: J Mech Eng
Sci 2004;218:50919, http://dx.doi.org/10.1243/095440604323052292.
[12] Chen CM, Kovacevic R. Finite element modeling of friction stir
weldingthermal and thermomechanical analysis. Int J Mach Tools Manuf
2003;43:131926, http://dx.doi.org/10.1016/S0890-6955(03)00158-5.
[13] Buffa G, Hua J, Shivpuri R, Fratini L. Design of the friction stir welding tool
using the continuum based FEM model. Mater Sci Eng A 2006;419:3818,
http://dx.doi.org/10.1016/j.msea.2005.09.041.
[14] Trimble D, ODonnell GE, Monaghan J. Characterisation of tool shape
and rotational speed for increased speed during friction stir welding of
AA2024-T3. J Manuf Process 2015;17:14150, http://dx.doi.org/10.1016/
j.jmapro.2014.08.007.
[15] Assidi M, Fourment L, Guerdoux S, Nelson T. Friction model for fric-
tion stir welding process simulation: calibrations from welding experi-
ments. Int J Mach Tools Manuf 2010;50:14355, http://dx.doi.org/10.1016/
j.ijmachtools.2009.11.008.
[16] Asadi P, Mahdavinejad RA, Tutunchilar S. Simulation and experimental inves-
tigation of FSP of AZ91 magnesium alloy. Mater Sci Eng A 2011;528:646977,
http://dx.doi.org/10.1016/j.msea.2011.05.035.
[17] Shojaeefard MH, Khalkhali A, Akbari M, Asadi P. Investigation of friction
stir welding tool parameters using FEM and neural network. Proc Inst
Mech Eng L: J Mater Des Appl 2015;229:20917, http://dx.doi.org/10.1177/
1464420713509075.
[18] Ammouri AH, Hamade RF. On the selection of constitutive equation for mod-
eling the friction stir processes of twin roll cast wrought AZ31B. Mater Des
2014;57:67388.
[19] Kuykendall K, Nelson T, Sorensen C. On the selection of constitutive laws
used in modeling friction stir welding. Int J Mach Tools Manuf 2013;74:7485,
http://dx.doi.org/10.1016/j.ijmachtools.2013.07.004.
[20] Tutunchilar S, Haghpanahi M, Besharati Givi MK, Asadi P, Bahemmat P. Simu-
lation of material ow in friction stir processing of a cast AlSi alloy. Mater Des
2012;40:41526, http://dx.doi.org/10.1016/j.matdes.2012.04.001.
[21] Arora A, Nandan R, Reynolds AP, DebRoy T. Torque, power requirement and stir
zone geometry in friction stir welding through modeling and experiments. Scr
Mater 2009;60:136, http://dx.doi.org/10.1016/j.scriptamat.2008.08.015.
[22] Nandan R, Roy GG, Lienert TJ, Debroy T. Three-dimensional heat and material
ow during friction stir welding of mild steel. Acta Mater 2007;55:88395,
http://dx.doi.org/10.1016/j.actamat.2006.09.009.
[23] Colegrove PA, Shercliff HR. 3-Dimensional CFD modelling of ow round
a threaded friction stir welding tool prole. J Mater Process Technol
2005;169:3207, http://dx.doi.org/10.1016/j.jmatprotec.2005.03.015.
[24] Buffa G, Hua J, Shivpuri R, Fratini L. A continuum based fem model for fric-
tion stir weldingmodel development. Mater Sci Eng A 2006;419:38996,
http://dx.doi.org/10.1016/j.msea.2005.09.040.
[25] Heinemann HH [Dissertation] Flow stress of different aluminum and copper
alloys for high strain rates and temperature. TH Aachen; 1961.
[26] DEFORM v 11.0 documentation; 2015.
[27] Zhang Z, Zhang HW. A fully coupled thermo-mechanical model of friction stir
welding. Int J Adv Manuf Technol 2008;37:27993, http://dx.doi.org/10.1007/
s00170-007-0971-6.
[28] Kobayashi S, OH S-I, Altan T. Metal forming and the nite element method.
Oxford Series; 1989.
[29] Ammouri AH, Hamade RF. Correlating process parameters to thrust forces and
torque in the friction stir processing of AZ31B. Proc NAMRI/SME 2014:42.
[30] Astarita a, Squillace A, Carrino L. Experimental study of the forces acting on
the tool in the friction-stir welding of AA 2024 T3 sheets. J Mater Eng Perform
2014;23:375461, http://dx.doi.org/10.1007/s11665-014-1140-3.
[31] Arora A, Mehta M, De A, DebRoy T. Load bearing capacity of tool pin during fric-
tion stir welding. Int J Adv Manuf Technol 2011;61:91120, http://dx.doi.org/
10.1007/s00170-011-3759-7.