See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305449800

Impact of Friction Stir Welding (FSW) Process Parameters on Thermal

Modeling and Heat Generation of Aluminum Alloy Joints

Article  in  Acta Metallurgica Sinica (English Letters) · July 2016

DOI: 10.1007/s40195-016-0466-2

CITATIONS READS

14 211

6 authors, including:

Saad Aziz Mohammad Dewan

Louisiana State University Dhaka University of Engineering & Technology
7 PUBLICATIONS   57 CITATIONS    27 PUBLICATIONS   178 CITATIONS   

SEE PROFILE SEE PROFILE

M. A. Wahab Ayman M Okeil

Louisiana State University Louisiana State University
108 PUBLICATIONS   943 CITATIONS    85 PUBLICATIONS   445 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Friction Stir Welding modeling & Fatigue analysis of weld joints View project

All content following this page was uploaded by Saad Aziz on 20 September 2016.

The user has requested enhancement of the downloaded file.

Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883
DOI 10.1007/s40195-016-0466-2

Impact of Friction Stir Welding (FSW) Process Parameters

on Thermal Modeling and Heat Generation of Aluminum Alloy
Joints
Saad B. Aziz1 • Mohammad W. Dewan1 • Daniel J. Huggett1 • Muhammad A. Wahab1 • Ayman M. Okeil2 •

T. Warren Liao1

Received: 1 April 2016 / Revised: 21 April 2016 / Published online: 19 July 2016
Ó The Chinese Society for Metals and Springer-Verlag Berlin Heidelberg 2016

Abstract Friction stir welding (FSW) is a solid-state joining process, where joint properties largely depend on the amount
of heat generation during the welding process. The objective of this paper was to develop a numerical thermomechanical
model for FSW of aluminum–copper alloy AA2219 and analyze heat generation during the welding process. The ther-
momechanical model has been developed utilizing ANSYSÒ APDL. The model was verified by comparing simulated
temperature profile of three different weld schedules (i.e., different combinations of weld parameters in real weld situa-
tions) from simulation with experimental results. Furthermore, the verified model was used to analyze the effect of
different weld parameters on heat generation. Among all the weld parameters, the effect of rotational speed on heat
generation is the highest.

KEY WORDS: Friction stir welding; Frictional dissipation energy; Temperature distribution; Friction
modeling; Aluminum alloy

1 Introduction factors and is considered to be environmentally friendly.

During the FSW process, a pin tool plunges while spinning
Friction stir welding (FSW) is comparatively a new into the joint between two parts that form the workpiece
welding process invented by The Welding Institute in 1991 until the shoulder comes in contact with the workpiece as
[1]. Since no melting or fusion occurs during the welding shown in Fig. 1. A backing plate is used to clamp the
process, FSW is free of high heat input and solidification workpiece to prevent its movement while welding. The
defects. Moreover, in the absence of filler material and heat is generated due to friction between pin tool and
fumes produced in the traditional fusion arc welding, the workpiece and the plastic deformation of the workpiece
FSW process is not susceptible to defects caused by these material. After the local temperature of work material
approaches its hot working temperature (i.e., 70% to 90%
of melting temperature) and is soft enough to be stirred and
Available online at http://link.springer.com/journal/40195 displaced, the rotating tool is moved longitudinally along
the welding line. This traverse motion of the pin tool
& Muhammad A. Wahab causes the plasticized soft material at the leading edge of
wahab@me.lsu.edu
the rotating tool to be squeezed and sheared through a
Saad B. Aziz small slit formed by the displaced soft material at the side
saziz1@lsu.edu
or lateral face of the tool, preferably in the direction of tool
1
Department of Mechanical and Industrial Engineering, rotation. The displaced soft material is then deposited in
Louisiana State University, Baton Rouge, LA 70803, USA the gap that forms at the trailing edge, left by rotating pin
2
Department of Civil and Environmental Engineering, tool or probe.
Louisiana State University, Baton Rouge, LA 70803, USA

123
870 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

A transient analytical thermal model of FSW has been

proposed by Zhang et al. [16] considering all steps in FSW.
The heat generation rate was modeled using a temperature-
dependent friction coefficient, which was modeled by the
inverse solution method (ISM). An extension work of the
model is proposed by Zhang et al. [17] to study the effect
of various weld conditions on heat generation and tem-
perature. A continuum-based thermomechanical model was
developed by Buffa et al. [18], in which the workpiece
material was modeled as rigid visco-plastic and rate-de-
pendent material. Temperatures obtained from the model
were found to be in good agreement with experimental
work. However, the authors have considered material
property for thermal conductivity and thermal capacity to
Fig. 1 Schematic of friction stir welding (FSW) process be constant, which is known to vary with the temperature.
Furthermore, the workpiece was considered as a ‘‘single
block’’ to avoid contact instabilities.
Previous work on modeling FSW process can be divided
Zhang and Zhang [19] developed a rate-independent
into three main categories: pure analytical thermal models;
model based on arbitrary Lagrangian–Eulerian (ALE) for-
finite element (FE)-based solid thermal and thermome-
mulation to study the effect of plunge force on material flow
chanical models, and computational fluid dynamics (CFD)
during FSW. The effect of stick (material has the same local
models. Since the development of FSW process, numerous
velocity as the tool) and slip (the velocity maybe lower)
research articles have been published on thermal modeling
during FSW has been modeled assuming slip rate of
of FSW [2–11]. Typically, the procedure involves applying
0.5% ðSlip rate ¼ AngularAngular
rotation speed of the contact matrix layer
rotation speed of the tool Þ. The
a surface heat flux as the sole source heat. The heat source
is then moved to simulate the advancing pin tool. Most authors concluded that with the increase in plunge force, both
models require ‘‘calibration’’ parameters. To overcome this friction and plastic energy increased. However in their
problem, Schmidt and Hattel [12] have proposed a thermal research, the authors did not include any analysis of tem-
model, where heat generation is considered as surface heat perature distribution during FSW. Moreover, the friction
flux from the tool shoulder, which is dependent on the coefficient was considered to remain constant, whereas in
tool’s radius and temperature-dependent yield stress. This real life, the friction coefficient is temperature dependent.
type of model, which is often named as ‘‘thermal pseudo- Using the same material model, Zhang and Zhang [20]
mechanical model’’ as the temperature generation, is studied the effect of travel speed on material flow during
expressed as temperature-dependent yield stress by taking FSW. Nevertheless, heat generated during FSW was not
mechanical effects into account. Schmidt et al. [13] have considered in this work as well. In another study, Zhang et al.
developed another model, which employs a linear [21] developed a rate-independent material model using
weighting of the contribution from sticking and sliding in ALE formulation. The stick and slip effect during FSW has
terms of contact state variable, d. been modeled by using modified Coulomb’s law. The
Song and Kovacevic [14] developed a thermomechani- authors studied the effect of material flow on weld parame-
cal model considering heat generated between tool and ters, i.e., plunge force, rotational speed, and travel speed.
workpiece. However, the authors considered heat genera- However, in their research, heat generation was not modeled
tion of the pin tool as a moving heat source, not as a heat as a process by itself; rather they used experimental tem-
generated through process itself. In another study, Chen perature curve as an input in the model.
and Kovacevic [15] developed a Lagrangian thermome- The objective of this work is to develop a model to
chanical model, which incorporated temperature and mul- estimate heat generation due to friction during FSW. One
tilinear strain-hardening effect. The temperature effect in major difference comparing the present model with previ-
the model has been developed by considering tool thermal ous models published in the literature is that a temperature-
effect as a moving heat source. Also the effect of the dependent friction coefficient is employed that takes into
moving heat source on the workpiece material and the account both sticking and sliding friction conditions along
effect of various weld parameters on residual stress have with a rate-independent plasticity model. Second, heat
been studied. However, no contact surface between the tool generation as a process in itself is modeled by accounting
and the workpiece, and in-between the plates has been for frictional heat between tool/boundary conditions. To
considered. demonstrate the validity of the model, the model is applied

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 871

to three different weld schedules of aluminum AA2219 Table 1 Chemical composition of the workpiece (AA2219)
alloys. Finally, a parametric study was conducted on crit- Si Fe Cu Mn Mg V Zn Ti Zr
ical weld parameters including plunge force, rotational
speed, and travel speed. The plunge force was varied from 0.20 0.30 6.8 0.40 0.02 0.15 0.10 0.10 0.25
12.45 kN to 23.35 kN to cover a wide range of weld sce-
narios. Similarly, the rotational speed was varied from
200 rpm to 450 rpm, and the travel speed ranging from
1.693 mm/s to 3.386 mm/s was considered. These weld earlier, there was no heat source input in this work; rather,
schedules have been selected from the experiment. heat generation in the model was a result of friction work
between the tool and the workpiece interface. The meshed
model has 7014 nodes and 6921 elements as shown in
2 Model Description Fig. 3.

Joining aluminum alloy by FSW has been a great interest 2.1 Material and Associated Flow Model
for research nowadays [22–25]. The finite element model
presented in this paper was used to simulate an FSW As stated earlier, the FSW model presented in this paper
process of workpiece that the authors tested [26]. The used a rate-independent plasticity material, where three
welds were made with a fixed pin tool on I-STIR PDS FSW distinct criteria have been used to determine rate-inde-
machines. The experimental setup for this workpiece is pendent plasticity model and these are: (a) flow rule,
shown in Fig. 2. The workpiece material is AA2219 alu- (b) hardening rule, and (c) yield criterion.
minum alloy, whose chemical composition is listed in Flow rule determines the increment in plastic strain from
Table 1. A taper threaded pin along with a tool made of the increment in load. In the current analysis, associative
H13 steel is used for FSW. The radius of the tool shoulder flow rule is used, which is represented by Eq. (1):
is 15.27 mm, and the height of the shoulder is 38.1 mm.  
 pl  oG
The tapered angle of the tool is 10°. Two chill bars have de ¼ dk : ð1Þ
or
been placed on top of the workpiece to help clamp it. In the
model, the tool is considered as a rigid solid and the where depl = change in plastic strain, dk = magnitude of
workpiece is considered as a ductile material, whose con- the plastic strain increment, G = plastic potential (which
stitutive model is capable of simulating elastic, plastic, determines the direction of plastic straining), and
large strain, large deformation, and isotropic hardening q r = change in stress.
effect. A 3D 20-node coupled-field solid element was The von Mises yield criterion has been applied in the
selected to model both the workpiece and the tool in current analysis as a yield criterion. The von Mises yield
ANSYSÒ. By studying the speed of the FSW schedules criterion is represented by Eq. (2) [27]:
covered in this research, the plate’s elastic and plastic
behavior was assumed to be rate independent. As stated

Fig. 2 Experimental setup and process parameter of FSW (Fz = Fig. 3 Meshes and thermal boundary conditions of the finite element
plunge force, x = rotational speed, V = travel speed) model

123
872 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

  r
f r; ry ¼ re  ry ¼ 0: ð2Þ A value of smax ¼ pyffiffi3 ¼ 0:58y (distortion energy crite-
rion) is applied to determine stick/slip condition in the
where
current analysis.
re ¼ von Mises effective stress In the current analysis, according to the modified Cou-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

3 1 lomb’s model, when the contact shear stress, scontact , is less

¼ r : r  trðrÞ2 ; ð3Þ than the maximum frictional stress, smax , a sticking con-
2 3
dition is modeled. Conversely, when the contact shear
ry = yield strength and tr = Tresca criterion stress, scontact , exceeds smax , the contact and the target
The total amount of plastic work is the sum of the plastic surface will slide relative to each other, (i.e., sliding con-
work done over the history of loading as expressed by dition is modeled). The conditions of contact shear stress vs
Eq. (4): contact pressure for sticking and sliding are described in
  Fig. 4.
v ¼ r frgT ½M  depl : ð4Þ
scontact  smax ! ðStickingÞ; scontact  smax ! ðSlidingÞ:
where v = plastic work, [M] = mass matrix, and ð9Þ
r = Cauchy stress tensor.
r
The amount of frictional work has been calculated by pyffiffi
A value of smax ¼ ¼ 0:58ry (distortion energy criterion)
3
Eq. (5) [27]; is used in the current analysis, where ry ¼ yield strength of
R ¼ s  c: ð5Þ the material. Since the material yield strength is highly
temperature dependent, temperature-dependent yield
where R ¼ frictional work, s = equivalent frictional strength value of AA2219 has been used in the current
stress, and c ¼ sliding rate. analysis.

2.2 Contact Condition 2.3 Thermal Boundary Condition

The critical part in numerical modeling of FSW is simu- The initial boundary condition used for the calculation in
lating the contact condition between various parts, i.e., the model can be expressed as Eq. (10) follows:
workpiece, pin tool, and shoulder [28]. In this research, T ðx; y; z; tÞ ¼ T0 : ð10Þ
modified Coulomb’s law is applied to describe the friction
force between the tool and the workpiece. where T(x, y, z, t) represents the transient temperature field
During sticking condition, the matrix close to the tool T, which is a function of time and the spatial coordinates
surface sticks to it. Shearing is considered to address the (x,y,z), and T0 represents the initial temperature.
velocity difference between the layer of the stationary The governing equation describing transient heat trans-
material points and the material moving with the tool. The fer process during FSW process can be described by the
shear yield stress, syield , is taken as Eq. (11)
ry
syield ¼ pffiffiffi : ð6Þ
3
where ry = yield strength of the material.
In the presented model, the contact shear stress was
taken equal to the shear yield stress, which depends on the
temperature:
ry
scontact ¼ syield ¼ pffiffiffi : ð7Þ
3
During sliding condition, the tool surface and the
workpiece material slide with respect to each other.
Using Coulomb’s friction law, the shear stress necessary
for sliding is:
scontact ¼ sfriction ¼ lp ¼ lr: ð8Þ
where p is the contact normal pressure, l is the friction
coefficient, and r is the contact stress.
Fig. 4 Modified Coulomb’s law depicting sliding and sticking
conditions

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 873

2
2.4 Mechanical Boundary condition
oT o T o2 T o2 T
qcp ¼k þ þ þ Q: ð11Þ
ot ox2 oy2 oz2
Displacement boundary conditions were introduced to the
where Q is the heat generation, cp is the specific mass heat model to match the actual welding conditions. The
capacity, q is the density of the material, k is the thermal boundary condition was specified as complete displace-
conductivity, and T is the absolute temperature. ment restraint, where the workpiece was clamped:
In finite element formulation, Eq. (11) can be repre-
U ¼ 0: ð16Þ
sented by Eq. (12):
Other parts of the workpiece, where the workpiece was
C ðtÞT_ þ K ðtÞT ¼ QðtÞ: ð12Þ
supported on the backing plate, were assumed to be
where C(t) is the time-dependent capacitance matrix, T is restrained in the normal direction:
the nodal temperature vector, T_ is the temperature UZ ¼ 0: ð17Þ
derivative with respect to time (i.e., dT
dt Þ; K ðtÞ is the time-
dependent conductivity matrix, and Q(t) is the time-de- The mechanical boundary conditions used in the current
pendent heat vector. analysis are shown in Fig. 5.
It is assumed that convection from the free surfaces, as
can be seen in Fig. 3, is the main reason for heat loss in the
workpiece. The heat loss from both the side and the top 3 FSW Calibration Experiments
surfaces is calculated using Eq. (13):
Experimental results from two welding AA2219 aluminum
ql ¼ hcon ðT  To Þ: ð13Þ alloy plates have been used for model calibration. The
where T represents the absolute temperature of the workpiece length was 609.6 mm, its width 152.4 mm and
workpiece, To ambient temperature and hcon convection its thickness 8.128 mm. The pin tool used in this study is
coefficient. The experimental setup that is being modeled made of H13 tool steel. The radius of the tool shoulder is
here has a chill bar present at the top surface of the weld 15.27 mm, and the height of the shoulder (tool shank
plate, which helps to clamp the workpiece, and also the height) is 38.1 mm. The pin tool is made of MP159 nickel–
chill bar acts as a heat sink. This will cause a high heat cobalt-based multiphase alloy. The pin radius at the top is
transfer coefficient from the top surface, which has been 4.78 mm and the height of the pin tool is 7.9 mm. The
given a value of 100 W/m2. From the side surface, heat tapered angle of the tool is 10°. Temperatures were mea-
transfer of 30 W/m2 has been used for aluminum to air sured during welding from the surface of the workpiece by
convection. At the bottom, a backing plate is placed to both K-type thermocouple and FLIR thermovision A40
oppose the downward plunge force. This backing plate also thermographer. The layouts of the thermocouples are
acts as a high heat sink absorbing heat rapidly during shown in Fig. 6.
welding; consequently, a high heat transfer coefficient is
used to model the heat transfer from backing plate. The
heat loss from backing plate is modeled by Eq. (14):
qb ¼ hb ðT  To Þ: ð14Þ
where hb represents the convection heat coefficient from
backing plate. Due to the complexity associated with
determining contact conditions between the workpiece and
the backing plate, the value of hb was calibrated to match
experimental data, which was found to be 300 W/m2. Heat
loss from tool surface was calculated using Eq. (15):
qw ¼ hw ðT  To Þ: ð15Þ
where hw represents the convection heat coefficient from
the pin tool. A value of 30 W/m2 has been used as heat
transfer coefficient from tool surface in the present model,
which is calibrated to best fit the experimental data. All
other thermal boundary conditions of current analysis are
shown in Fig. 3. Fig. 5 Mechanical boundary conditions of the plate

123
874 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

Fig. 6 Layouts of the thermocouples (embedded in the surface) and

thermographer (all dimensions are in millimeter) Fig. 7 Solid226 elements [27]

4 FEA Modeling was selected because of its plasticity, stress stiffening,

large deflection and large strain capabilities [27].
4.1 Workpiece and Tool Modeling Two rectangular plates were created in the model sim-
ulating the two welded parts of the workpiece. In order to
Finite element analysis software, ANSYSÒ, has been used reduce simulation time, the length and width of the plate
to carry out the numerical simulation. The FSW modeling have been reduced, but the actual thickness was main-
is divided into three stages, namely (1) plunge, (2) dwell, tained. Thus, the simulation captured the behavior of the
and (3) traverse stages. During the plunge stage, the pin steady-state portion of the FSW process effectively without
tool first moves down vertically and then starts rotating the need to simulate the entire steady-state region. The
during the dwell stage followed by moving along the weld plate width was reduced in such a way that the regions
seam with rotation during the traverse stage. In order to away from the weld line are not affected by the welding
avoid complexity during the initial plunge stage, heat process. The dimension of the modeled plates is 152.4 mm
generation was only considered during the dwell and the (L) 9 47.625 mm (W) 9 8.128 mm (T). To improve the
traverse stages. Details of the modeling steps, i.e., duration fidelity of the results around the weld seam, the centerline
and the boundary conditions, are listed in Table 2. of the plate was modeled with a finer mesh as shown in
It should be noted that the traverse step, which is the Fig. 3.
longest step at 30 s, was used for thermal verification by The tool shank had a height of 38.1 mm and shoulder
comparing thermocouple data with FEA model. radius of 15.27 mm, which is identical to the dimensions
In the current simulation, a Lagrangian model has been used during FSW modeling. During the FSW process, heat
adopted to incorporate temperature and multilinear iso- is mainly generated from friction between the tool and the
tropic strain hardening with large strain capability and workpiece. For this purpose, a surface to surface contact
material deformation behavior. A 3-D 20-node coupled- pair was used between the tool and workpiece as shown in
field SOLID226 element, as shown in Fig. 7, was used to Fig. 11. The rate of frictional dissipation is calculated by
model both the plate and the tool. The SOLID226 element Eq. (18)
q ¼ FHTG  s  c: ð18Þ
where FHTG ¼ fraction of frictional energy converted to
Table 2 Simulation details for three steps heat.
In the model, 100% of frictional energy is considered
Stage Time step (s) Pin tool boundary condition
converted into the heat energy.
Plunging 1 Displacement along z-axis The amount of frictional dissipation on contact and
Dwelling 10 Rotation along z-axis target surface is expressed by Eqs. (19) and (20).
Traversing 16 Rotation along z-axis qc ¼ FWGT  FHTG  s  c: ð19Þ
Movement along y-axis
qT ¼ ð1  FWGTÞ  FHTG  s  c: ð20Þ

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 875

where qc ¼ contact side, qT = target side, and FWGT ¼

weight factor of the distribution of heat between the contact
and target surfaces.
Also in the current simulation, 95% of the generated
heat was distributed in the workpiece and 5% of the gen-
erated heat was distributed in the pin tool following rec-
ommendations by previous research work [29]. Heat
transfer from the tool to the workpiece was minimized by
assigning a low conductance value of 10 W/(m2 °C)
between the tool and workpiece.
The friction coefficient plays a great role in generating
heat during FSW. However, the friction coefficient in FSW
is dependent upon many factors, such as temperature,
contact geometry, relative motion between tool and
workpiece and applied force. Zhang et al. [16] have done Fig. 8 Flowchart for choice of friction coefficient
an extensive study on the above-mentioned parameters and
found out that the friction coefficient is mainly temperature
dependent. Therefore, a temperature-dependent friction
coefficient has been used in the current model varying
between 0.4 and 0.25 [30] and has been listed in Table 3.
From Table 3, we can see that as temperature rises, friction
coefficient remains constant up to 200 °C; after 200 °C,
friction coefficient starts decreasing. The choice of this
friction coefficient can be described as explained by the
work of Zhang et al. [16] as shown in Fig. 8.
Two contact element types CONTA174 and TARGE170
were used to model the contact between two plates as
shown in Fig. 9. Between the two workpieces, a high
thermal contact conductance 2 9 106 W/(m2 °C) was
introduced to develop an almost perfect thermal contact. In
general, the maximum temperature generated during
welding is about 0.7–0.9 of the melting temperature of the
material [31]. When the temperature rises over 0.7 of the Fig. 9 Contact pair between tool and workpiece and between two
melting temperature (melting temperature of AA2219 is plates
543 °C), both plates will be joined and remain joined even
after the temperature is decreased. In this current simula- work converted to heat is considered to be 80%, which has
tion, 400 °C is set as a temperature for joining. A master been found by the previous research work [29]. It should be
node/pilot node is created at the top of the tool to control noted that some researchers estimated the heat generated
the rotating and traverse speed of the pin tool as shown in from plastic work to be minimal (less than 5%) compared
Fig. 9. Also in the current model, the amount of plastic to that generated by friction [32]. From Eq. (4), total
plastic work converted to heat can be expressed by
Table 3 Temperature-dependent friction coefficient used in the Eq. (21).
model
t  
Temperature (°C) Friction coefficient qp ¼ 0:8  v ¼ 0:8  r frgT ½M  depl : ð21Þ
0
25 0.4
To account for large strain and large deformation,
100 0.4
ANSYSÒ command NLGEOM,on is used in the current
200 0.4
analysis [33]. During FSW, material properties are
300 0.35
considered to be temperature dependent since the
400 0.25
temperature gradient is large. The solution time step size
420 0.25
was set at a very small increment (in the order of 10-12 s)
543 0.01
during dwell and traverse stages to increase the accuracy.

123
876 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

4.2 Heat Generation From Pin Tool Nib

In the current simulation, the pin tool nib was not modeled
in order to avoid mesh locking due to incompressible
plastic deformation. According to Schmidt et al. [13], the
ratio between heat generated from pin nib and heat gen-
erated from tool shoulder is 16%. Therefore, the heat
generated from friction to the workpiece and plastic
deformation in the model was multiplied by 1.16 as a
compensation for the heat flow from pin tool nib.

4.3 Material Properties

The material properties of AA2219 are shown in Figs. 10,

11, and 12. Modulus of elasticity, thermal conductivity,
and specific heat capacity are temperature-dependent Fig. 11 Thermal conductivity of AA2219 as a function of temper-
ature [35]
properties and vary significantly with temperature. Con-
versely, workpiece density along with pin tool density,
thermal conductivity, and specific heat capacity have been
considered as temperature-independent properties and are
listed in Table 4.

4.4 Stress–strain Diagram

Multilinear isotropic hardening with large strain and

deformation capability has been used in the current anal-
ysis together with a strain rate-independent plasticity
model. The true stress vs plastic strain at a strain rate of
e_ = 1 s-1 for aluminum is shown in Fig. 13. The adopted
temperature-dependent yield stresses were assumed to drop
from a value of 350 MPa at ambient temperature to
25 MPa at a temperature of 370 °C according to the rela-
tionship, which can be seen in Fig. 14.
Fig. 12 Specific heat capacity of AA2219 as a function of temper-
ature [35]

4.5 Computational Time

The thermomechanical analysis performed in ANSYSÒ

used 20 Intel Ivy bridge 2.8 GHz cores processor and
64 GB of RAM memory. The CPU time was about 30 h for
27 s of simulation. This simulation was done on a Super-
computer (SuperMIC, owned by Louisiana State Univer-
sity), which has a peak performance of 557 TeraFlop (TF).

5 Thermal Verification

Rather than comparing temperature history with a single

weld schedule, three different weld schedules have been
analyzed to show the robustness of the model. The three
Fig. 10 Young’s modulus of aluminum as a function of temperature weld schedules chosen for that purpose have the same
[34]

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 877

Table 4 Material properties used in the model [36]

Density of workpiece, Density of tool, Thermal conductivity of the tool, Specific capacity of tool, ct Melting temperature of
q(kg/ m3 Þ qt (kg/ m3 Þ kt (W/m2 °C) (J/kg/°C) workpiece (°C)

2840 7800 24.4 460 543

location of y = 42.36 mm, z = 26 mm along the weld

direction for the three different weld schedules. The com-
parison shows that FEA numerical results of temperature
closely match with the experimental data.
Figures 18 and 19 represent temperature field and tem-
perature profile, respectively, along the lateral direction
from a simulation for a weld schedule with travel speed, V,
equal to 1.27 mm/s, rotational speed, x, equal to 350 rpm,
and a plunge force, Fz, equal to 15.568 kN. Figures 20, 21,
and 22 represent comparison of the results obtained from
FEA and from the experiment at transverse direction. From
these figures, it can be seen that the temperature obtained
from experiment is in close agreement with the simulated
temperature. Error analyses between experimental tem-
Fig. 13 True stress–strain diagram of the aluminum [30, 37] perature and FEA temperature have been shown in this
section. Also, from Figs. 20, 21, 22, the temperature
around the shoulder is higher than the surrounding area,
which is contributed by friction mainly and by plastic
deformation to a lesser extent [32]. For this localized
heating up to the tool shoulder radius, temperature is the
highest, and it decreases as the distance from center
increases. Maximum temperatures obtained from the sim-
ulations or experiments in all three cases are 422, 431, and
462 °C. In all cases, the maximum temperature is less than
the melting temperature of AA2219 (543 °C) and ranges
between 77.7% and 84.5% of the melting temperature,
which is typical for FSW.
The mean relative error is calculated between the
experimental and the FEA analysis value as shown in
Tables 6, 7, and 8 at different distances perpendicular to
the weld.
For all schedules, the highest absolute relative error is
Fig. 14 Temperature-dependent yield stress of aluminum at
e_ = 1 s-1 [30, 37] below 6%, and the average error for all cases is below
3.60%.

rotational speed, the same travel speed, and the same

temperature-dependent friction coefficient but different 6 Energy Generation During FSW Process
plunge force values. A summary of the weld schedules is
listed in Table 5. Temperature generated during FSW FSW causes heat generation to join workpieces together.
experiment was measured using two different methods, During the FSW process, heat is generated through two
namely attached thermocouple and thermographic device. possible ways, namely heat generation due to friction
The measured temperature results were compared with between tool/workpiece and heat generation due to plastic
simulation results. deformation. The aforementioned expressions in Eqs. (4)
Figures 15, 16, and 17 show variations of temperature and Eq. (5) have been used to calculate plastic dissipation
on the top surface of the workpiece at the thermocouple energy and friction energy dissipation converted to heat.

123
878 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

Table 5 Different weld schedule for temperature verification

Rotational speed, x (rpm) Travel speed, V (mm/s) Plunge force,FZ (kN)

350 1.27 12.455

350 1.27 15.568
350 1.27 21.351

Fig. 15 Comparison between temperature histories of thermocouples Fig. 17 Comparison between temperature histories of thermocouples
and FEA results at y = 42.36 mm, z = 26 mm location and FEA results at y = 42.36 mm, z = 26 mm location
(V = 1.27 mm/s; x = 350 rpm; Fz = 12.455 kN) (V = 1.27 mm/s; x = 350 rpm; Fz = 21.351 kN)

Fig. 18 Temperature field from simulation (V = 1.27 mm/s;

x = 350 rpm; Fz = 15.568 N)

Fig. 16 Comparison between temperature histories of thermocouples

such as Bastier et al. [32]. Bastier et al. [32] reported that
and FEA results at y = 42.36 mm, z = 26 mm location
(V = 1.27 mm/s; x ¼ 350 rpm; Fz = 15.568 kN) plastic heat generation contributed only 4.4% of total heat
generation of FSW aluminum alloy, with the remaining
95.6% heat being generated by friction. The fact that the
From Table 9, plastic energy from our model was only presented model cannot capture plastic heat generation
responsible for 0.09% for a weld schedule with a plunge accurately is mainly attributed to the fact that it only
force of 21.351 kN, rotation speed of 350 rpm and traverse considers plastic deformations occurring at the top surface
speed of 1.27 mm/s. This percentage is low compared to of the workpiece rather than around the weld nugget. This
values reported in the literature by previous researchers simplification in the current model implies an assumption

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 879

Fig. 19 Temperature variation from simulation along transverse Fig. 21 Comparison of temperature variations between experimental
direction (V = 1.27 mm/s; x ¼ 350 rpm; Fz = 15.568 kN) and simulation data along transverse direction (V = 1.27 mm/s;
x = 350 rpm; Fz = 15.568 kN)

Fig. 20 Comparison of temperature variations between experimental

and simulation data along transverse direction (V = 1.27 mm/s; x ¼ Fig. 22 Comparison of temperature variation between experimental
350 rpm; Fz = 12.455 kN) and simulation data along transverse direction (V = 1.27 mm/s; x ¼
350 rpm; Fz = 21.351 kN)
that all heat is practically generated by friction, which
should result in temperatures lower than the actual tem- effects of varying plunge force, rotational speed, and travel
perature by a few percentage points according to Bastier speed.
et al. [32]. While this is true for the results shown in
Fig. 22 and most of the observed locations in Fig. 21, it is 6.1 Effect of Plunge Force on Welding
not the case for the weld schedule, whose results are pre-
sented in Fig. 20. This may be attributed to experimental A parametric study was conducted to investigate the effects
reading errors that can surpass such a small difference of a of plunge force. Three different plunge forces of 12.455,
few percentage points. The authors are presently working 15.568, and 21.351 kN were considered. During these
on developing an FSW model capable of accurately cap- analyses, travel speed and rotational speed were kept
turing the plastic deformation around the weld nugget, constant.
which requires modeling material flow in that area. Figure 23 shows the frictional dissipation energy for
In the following sections, we will discuss the change 27 s of simulation for all three plunge force cases. It can be
friction energy, i.e., the dominant energy source, as a result seen that the energy increases with the increase in plunge
of varying welds parameters. In the current work, heat force. The total frictional dissipation energy increased
generation due to friction is investigated to study the 22.96% when plunge force is increased from 12.455 to

123
880 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

Table 6 Error analysis for Fz = 12.455 kN, x = 350 rpm, V = 1.27 mm/s weld schedule
Distance (mm) Temperature from FEA analysis (°C) Temperature from experiment (°C) Absolute error (%)

0 422 418 0.96

15 354 342 3.51
26 262 248 5.64
32 237 225 5.30
39 220 212 3.77
47 213 208 2.40
Average error 3.60

Table 7 Error analysis for Fz = 15.568 kN, x = 350 rpm, V = 1.27 mm/s weld schedule
Distance (mm) Temperature from FEA analysis (°C) Temperature from experiment (°C) Absolute error (%)

0 431.0 440 2.04

15 362.3 360 0.64
26 288.9 280 3.18
32 255.3 252 1.31
39 220.8 230 4.00
47 214.0 216 0.93
Average error 2.02

Table 8 Error analysis for Fz = 21.351 kN, x = 350 rpm, V = 1.27 mm/s weld schedule
Distance (mm) Temperature from FEA analysis (°C) Temperature from experiment (°C) Absolute error (%)

0 458.66 462 0.72

15 398.56 403 1.10
26 298.56 304 1.79
32 260.4 265 1.73
39 244.65 252 2.91
47 232.04 238 2.50
Average error 1.79

Table 9 Friction and plastic dissipation energy for weld schedule plunge force 21.351 kN, rotation rate 350 rpm, and traverse speed 1.27 mm/s
Rotational speed, x(rpm) Traverse speed, V (mm/s) Plunge force, FZ (kN) Frictional energy (J) Plastic energy (J) Total energy (J)

350 1.27 21.351 1.35 9 106 1.25 9 103 1.35125 9 106

21.351 kN. Similarly, frictional dissipation energy 6.2 Effect of Spindle Rotational Speed
increased 21.48% when plunge force is increased from
15.568 to 21.351 kN. A higher plunge force causes more Three different welding tool rotational speeds of 200, 300,
material to penetrate and spin by rotation and thus produces and 450 rpm have been considered to study the effect of
more energy. Table 10 summarizes the plunge force effect the tool’s rotational speed. A constant ravel speed,
on the frictional energy. This result is consistent with the V = 2.539 mm/s, and a constant plunge force, Fz = 26.68
experimental result reported by Tang et al. [38]. kN, have been used in the analysis.

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 881

Fig. 23 Frictional dissipation energy variation with plunge force Fig. 24 Frictional dissipation energy variation with rotational speed
(x = 350 rpm, v = 1.27 mm/s) (v = 1.27 mm/s, Fz = 26.68 kN)

Figure 24 represents frictional dissipation energy at 200, total frictional dissipation energy increased about 5.40%
300, and 450 rpm, respectively. The higher rotational when travel speed is decreased from 3.386 to 1.693 mm/s.
speed produced higher dissipation energy. The total fric- Moreover, total frictional dissipation energy increased
tional dissipation energy increased about 80.06% when about 4.50% when the travel speed is decreased from
rotational speed is increased from 200 to 450 rpm. More- 2.539 to 1.693 mm/s. The lower travel speed of the tool
over, when the rotational speed is increased from 300 to results in more time to rotate on material, and thus, the
450 rpm, total frictional energy increased about 32.25%. rate by which heat is produced locally increases. Table 12
This higher energy is produced by higher relative velocity summarizes the effect of travel speed on frictional dissi-
of the materials due to high rotational speed. Table 11 pation energy.
summarizes the effect of rotational speed on frictional
dissipation energy. Similar results have been reported by
the experimental work of Tang et al. [38]. 7 Conclusions

6.3 Effect of Welding Speed A fully coupled thermomechanical 3D model has been
developed to analyze thermal heat generation and distri-
The effect of tool travel speed on frictional dissipation bution during FSW. The goal of this research effort is to
energies was also investigated by considering three dif- advance FSW modeling a degree closer to actual weld
ferent weld speeds 3.386, 2.539, and 1.693 mm/s. A con- conditions by introducing sticking and sliding friction
stant rotational speed, x = 300 rpm, and a constant plunge along with temperature-dependent friction coefficient to
force, Fz = 26.68 kN, have been used in these analyses. study the heat generation during FSW process. Though the
From Fig. 25, it can be seen that as the welding speed developed model cannot capture plastic deformation
decreases, frictional dissipation energy increases. The accurately, it is an improvement over thermal model as it

Table 10 Summary of friction dissipation energies for various plunge forces

Rotational speed, x (rpm) Traverse speed, V (mm/s) Plunge force, FZ , (kN) Frictional energy (J) Frictional energy percentage increasea

350 1.27 12.455 1.04 9 106 22.96%

350 1.27 15.568 1.06 9 106 21.48%
350 1.27 21.351 1.35 9 106 Base1
a
With respect to Base1 weld schedule

123
882 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883

Table 11 Summary of friction dissipation energies for various rotational speeds

Rotational speed, x, Traverse speed, V (mm/ Plunge force, FZ , Total frictional energy Frictional energy percentage
(rpm) s) (kN) (J) increasea

200 2.539 26.68 3.09 9 105 80.06%

6
300 2.539 26.68 1.05 9 10 32.25%
450 2.539 26.68 1.55 9 106 Base2
a
With respect to Base2 weld schedule

measured temperatures. This implies that heat pro-

duced from the frictional work of the tool and
workpiece produces most of the energy, which is
consistent with similar findings in the literature [19].
(3) A parametric study was conducted to analyze the
effect of different weld parameters—plunge force,
rotational speed, and travel speed. Findings from this
study revealed that:
(a) The higher the plunge force, the higher the
friction dissipation energy generated. Total fric-
tional dissipation energy is increased by 23 and
21% when plunge force is increased from
12.455 to 21.351 kN.
(b) The higher the rotational speed, the higher the
Fig. 25 Frictional dissipation energy variation with welding speed
(Fz = 26.68 kN, x = 300 rpm) total amount of frictional dissipation energy.
When rotational speed is increased from 200 to
450 rpm, total frictional energy is increased to
Table 12 Summary of total friction and plastic dissipation energies 80.06% and 32.25%, respectively.
for various travel speed
(c) Lower travel speed causes more total frictional
Rotational Travel Plunge Total Frictional and plastic dissipation energy. When travel
speed, x, speed, force, FZ , frictional energy speed changes from 3.386 to 1.693 mm/s, the
(rpm) V (mm/s) (kN) energy (J) percentage
increasea total frictional is changed from 5.40% and
4.50%, respectively.
300 3.386 26.68 1.05 9 106 5.40%
300 2.539 26.68 1.06 9 106 4.50% (4) Among the three major FSW process parameters, the
300 1.693 26.68 1.11 9 106 Base3
effect of rotational speed on generating frictional
a
energy is found to be the most important parameter.
With respect to Base3 weld schedule

Acknowledgments The authors gratefully acknowledge the financial

captures heat generation due to friction. The following
support received from the Louisiana Economic Development Assis-
conclusions can be drawn from this research: tantship (EDA) program. Authors are grateful to Mr. G. Verma, senior
technology specialist at ANSYSÒ for his technical support. Authors
(1) The temperature profile obtained from simulation is are thankful to LSU’s High Performance Computing Services for their
consistent with the temperature profile obtained from support and assistance with the modeling part of this study.
experiments. Temperature profiles from three differ-
ent weld schedules have been used to compare the
result with the simulation results. In all cases, the References
highest relative error is below 6% with a mean value
of 2.47%. [1] W.M. Thomas, E.D. Nicholas, J.C. Needham, M.G. Murch, P.
Templesmith, C.J. Dawes, GB Patent Application No.
(2) Even though the current model does not capture all of 9,125,978.8, Dec 1991
the plastic energy produced by FSW in the workpiece, [2] Z. Feng, X.L. Wang, S.A. David, P.S. Sklad, Sci. Technol.
its results are in good agreement with experimentally Weld. Join. 12, 348 (2007)

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.), 2016, 29(9), 869–883 883

[3] X.K. Zhu, Y.J. Chao, J. Mater. Process. Technol. 146, 263 [23] A.H. Feng, D.L. Chen, Z.Y. Ma, W.Y. Ma, R.J. Song, Acta
(2004) Metall. Sin. (Engl. Lett.) 27, 723 (2014)
[4] M.Z.H. Khandkar, J.A. Khan, A.P. Reynolds, Sci. Technol. [24] W.F. Xu, J.H. Liu, D.L. Chen, J. Mater. Sci. Technol. 31, 953
Weld. Join. 8, 165 (2003) (2015)
[5] P. Prasanna, B.S. Rao, G.K.M. Rao, Int. J. Adv. Manuf. Technol. [25] R.Z. Xu, D.R. Ni, Q. Yang, C.Z. Liu, Z.Y. Ma, J. Mater. Sci.
51, 925 (2010) Technol. 32, 76 (2016)
[6] R. Nandan, G.G. Roy, T. Debroy, Metall. Mater. Trans. A 37, [26] M.W. Dewan, D.J. Hugget, T.W. Liao, M.A. Wahab, A.M.
1247 (2006) Okeil, Mater. Des. 92, 288 (2016)
[7] P.A. Colegrove, H.R. Shercliff, Sci. Technol. Weld. Join. 11, [27] Ansys Inc., ANSYS Mechanical APDL Theory Reference, 2013,
429 (2006) http://148.204.81.206/Ansys/150/ANSYSMechanicalTheoryRef
[8] P. Ulysse, Int. J. Mach. Tools Manuf. 42, 1549 (2002) erence.pdf. Accessed on 28 Aug 2015
[9] E. Ranjbarnodeh, S. Hanke, S. Weiss, A. Fischer, Int. J. Miner. [28] A.P. Reynolds, T. Seidel, S. Xu, in Proceedings Joining of
Metall. Mater. 19, 923 (2012) Advanced and Specialty Materials, ASM International, St.
[10] Y.H. Yau, A. Hussain, R.K. Lalwani, H.K. Chan, N. Hakimi, Int. Louis, Missouri, 9–12 Oct, 2000
J. Miner. Metall. Mater. 20, 779 (2013) [29] Y.J. Chao, X. Qi, W. Tang, Trans. ASME 125, 138 (2003)
[11] S.D. Jai, Y.Y. Jin, Y.M. Yue, S.S. Gao, Y.X. Huang, J. Mater. [30] M. Awang, Dissertation, West Virginia University, 2007
Sci. Technol. 29, 955 (2013) [31] R.S. Mishra, Z.Y. Ma, Mater. Sci. Eng. R 50, 1 (2005)
[12] H.B. Schmidt, J.H. Hattel, Scr. Mater. 58, 332 (2008) [32] A. Bastier, M.H. Maitournam, K. Dang, F. Roger, Sci. Technol.
[13] H. Schmidt, J. Hattel, J. Wert, Model. Simul. Mater. Sci. Eng. Weld. Join. 11, 278 (2006)
12, 143 (2004) [33] Ansys Inc., Theory Reference for the Mechanical APDL and
[14] M. Song, R. Kovacevic, Int. J. Mach. Tools Manuf. 43, 605 Mechanical Applications, 2009, http://orange.engr.ucdavis.edu/
(2003) Documentation12.0/120/ans_thry.pdf. Accessed on 2 Aug 2015
[15] C. Chen, R. Kovacevic, Proc. Inst. Mech. Eng. Part C J. Eng. [34] R.B. McLellan, T. Ishikawa, J. Phys. Chem. Solids 48, 603
Mech. Eng. Sci. 218, 509 (2004) (1987)
[16] X.X. Zhang, B.L. Xiao, Z.Y. Ma, Metall. Mater. Trans. A 42, [35] H.J. Zhang, H.J. Liu, L. Yu, Trans. Nonferrous Met. Soc. China
3218 (2011) 23, 1114 (2013)
[17] X.X. Zhang, B.L. Xiao, Z.Y. Ma, Metall. Mater. Trans. A 42, [36] ASM International Handbook Committee, Properties and
3229 (2011) Selection: Nonferrous Alloys and Special-Purpose Materials
[18] G. Buffa, J. Hua, R. Shivpuri, L. Fratini, Mater. Sci. Eng. A 419, (ASM Handbook, ASM International, Materials park, OH 1990),
389 (2006) p. 65
[19] Z. Zhang, H.W. Zhang, Sci. Technol. Weld. Join. 12, 226 (2007) [37] E. Semb, Dissertation, Behavior of Aluminum at Elevated Strain
[20] Z. Zhang, H.W. Zhang, Mater. Des. 30, 900 (2009) Rates and temperatures, Norwegian University of Science and
[21] H.W. Zhang, Z. Zhang, J.T. Chen, J. Mater. Process. Technol. Technology, 2013
183, 62 (2007) [38] X.G.W. Tang, J.C. Mc Clure, L.E. Murr, A. Nunes, J. Mater.
[22] D. Wang, B.L. Xiao, D.R. Ni, Z.Y. Ma, Acta Metall. Sin. (Engl. Process. Manuf. Sci. 163, 7 (1998)
Lett.) 27, 816 (2014)

123

View publication stats