Acta Metall. Sin. (Engl. Lett.

)
DOI 10.1007/s40195-017-0658-4

A Fully Coupled Thermomechanical Model of Friction Stir

Welding (FSW) and Numerical Studies on Process Parameters
of Lightweight Aluminum Alloy Joints
Saad B. Aziz1 • Mohammad W. Dewan1 • Daniel J. Huggett1 • Muhammad A. Wahab1 • Ayman M. Okeil2 •

T. Warren Liao1

Received: 24 June 2017 / Revised: 9 August 2017

Ó The Chinese Society for Metals and Springer-Verlag GmbH Germany 2017

Abstract This paper presents a new thermomechanical model of friction stir welding which is capable of simulating the
three major steps of friction stir welding (FSW) process, i.e., plunge, dwell, and travel stages. A rate-dependent Johnson–
Cook constitutive model is chosen to capture elasto-plastic work deformations during FSW. Two different weld schedules
(i.e., plunge rate, rotational speed, and weld speed) are validated by comparing simulated temperature profiles with
experimental results. Based on this model, the influences of various welding parameters on temperatures and energy
generation during the welding process are investigated. Numerical results show that maximum temperature in FSW process
increases with the decrease in plunge rate, and the frictional energy increases almost linearly with respect to time for
different rotational speeds. Furthermore, low rotational speeds cause inadequate temperature distribution due to low
frictional and plastic dissipation energy which eventually results in weld defects. When both the weld speed and rotational
speed are increased, the contribution of plastic dissipation energy increases significantly and improved weld quality can be
expected.

KEY WORDS: Aluminum alloy; Friction stir welding; Temperature distribution; Plastic energy; Frictional
energy; Rate-dependent model; Friction modeling

1 Introduction the absence of fumes that are generally produced in the

traditional fusion arc welding, the FSW process is con-
Friction stir welding (FSW) has generated widespread sidered to be environmentally green. Furthermore, FSW
attention after its invention by The Welding Institute in joints generally have higher mechanical properties and
1991 [1]. In FSW, there is absence of melting or fusion lower post-weld tensile residual stress than conventional
during the welding process itself; thus, this welding is free fusion welding. The FSW process starts by plunging a
of high heat input, and eventual melting, and solidification rotating pintool with a shoulder into the workpiece at the
defects. Moreover, due to the absence of filler material and joint between two metal pieces to be joined as shown in
Fig. 1. The rotational motion of the pintool causes fric-
tional heat and consequently large plastic deformations due
Available online at http://link.springer.com/journal/40195 to material stirring in the workpiece. The shoulder also
helps prevent the material from being expelled out of the
& Muhammad A. Wahab weld center. A backing plate placed at the bottom of the
[email protected]
workpiece prevents its movement and deformation while
1
Department of Mechanical and Industrial Engineering, welding.
Louisiana State University, Baton Rouge, LA 70803, USA In FSW, heat generation occurs in three distinct stages:
2
Department of Civil and Environmental Engineering, plunge, dwell, and travel. During the dwell stage, the
Louisiana State University, Baton Rouge, LA 70803, USA position of the pintool remains the same at the end of the

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

[4–12]. Zhang et al. [13] developed an analytical thermal

model which considered all necessary steps in FSW. In the
model, an inverse solution method (ISM) has been used in
the analysis to calculate rate of heat generation. The
extension of the proposed model by Zhang et al. [14] has
been analyzed with the consequence of different weld
parameters on temperature and heat generation.
An ALE formulation was employed by Zhang et al. [15]
to develop rate-independent material model. However, the
results did not include any discussion on plastic and fric-
tional energy dissipation. Another rate-independent model
was developed by Zhang et al. [16–18] to analyze influ-
ences of plunge force, welding speed, and rotational speed
upon material flow. During FSW, influence of stick (pin-
Fig. 1 Schematic representation of friction stir welding (FSW) tool and material have the same velocity) as well as slip
process [2] (pintool and material have different velocities) has been
modeled assuming a slip rate of 0.5% ðSlip
plunge stage without changing the rotation speed. Addi- rateðdimensionlessÞ ¼ AngularAngular
rotation speed of the contact matrix layer
Þ
rotation speed of the tool
tional heat is generated from the shoulder–workpiece
[19]. Buffa et al. [20] developed a continuum model using
interface, which raises the temperature of the workpiece
a Lagrangian-based code, DEFORM-3DTM. However, it
close to its melting temperature. The two main sources of
was assumed that the material properties of the workpiece
heat generation are: (1) the friction between pintool and
(thermal conductivity and thermal capacitance) were con-
workpiece and (2) plastic deformation of the workpiece
stants, and in real life, it varies with temperature. Besides,
material.
the workpiece plate was modeled as ‘‘single block,’’ rather
Previous work on modeling FSW process can be divided
than two parts next to each other, to avoid contact insta-
into two main categories: pure thermal models and coupled
bilities. Aziz et al. [21] developed a FSW model using a
thermomechanical models, which account for the interac-
rate-independent material model. The model used three
tion between thermal and mechanical deformation. The
different weld schedules to verify temperature profile of
pure thermal models only compute temperature fields. On
simulation with experimental results. Furthermore, the
the other hand, thermal and deformation coupled models
verified model was used to analyze the effects of different
may be classified into two types, namely computational
weld parameters on heat generation and its sources (friction
solid mechanic (CSM)-based models and computational
vs. plastic deformation). Another FSW model was devel-
fluid dynamic (CFD)-based models [3]. A Lagrangian
oped using ABAQUS by Aval et al. [22], who showed that
formulation is commonly used in CSM models, and an
the temperature in FSW process was distributed unevenly
Eulerian formulation is commonly used in CFD models.
along the weld line.
Lagrangian formulation permits mesh and material to move
Lasley et al. [23] developed a model to model the plunge
at the same time, which enables tracking surfaces. Its
stage of FSW only using commercial software called
limitations lay in the difficulty to achieve convergence
‘‘Forge3.’’ Heurtier et al. [24] developed a semi-analytical
when elements become severely distorted. On the other
model considering microstructure change during welding
hand, in the Eulerian approach materials move through the
in lightweight aluminum alloy. The kinematic model was
mesh. The shortcoming of this approach is that it is chal-
divided into two parts. One location is just below the
lenging to tracking surface and the boundary conditions.
shoulder of the tool and corresponds to the ‘‘flow arm
Nevertheless, distortion of mesh is not a problem since
zone.’’ The other zone is ‘‘nugget zone’’—located in the
mesh distortion never occurs in the Eulerian approach.
depth zone. Grujicic et al. [25] developed a 3D flow model
Another approach is known as arbitrary Lagrangian–Eu-
using a modified Johnson–Cook model which takes into
lerian (ALE) method which incorporates the benefit of both
consideration of the dynamic recrystallization effect during
Lagrangian and Eulerian approach and, hence, is suit-
FSW. Assidi et al. [26] developed a 3D FSW model using
able for large deformation problem. In the arbitrary
commercial software Forge3Ò FE software. Two different
Lagrangian and Eulerian (ALE) method, the node points
models formulated with Eulerian and ALE techniques have
can be moved arbitrarily which enables the material to
been developed and compared. From the model, several
move independent of mesh.
conclusions were drawn. First, the arbitrary Lagrangian–
Several research articles have been published so far on
Eulerian technique is in good agreement with the measured
the aspects of thermal modeling of friction stir welding

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

forces and tool temperature. Second, Coulomb’s friction 2 Model Description

model provides better result than Norton friction model.
Hamilton et al. [27] developed a thermomechanical model A finite element (FE) model is developed for simulation of
using the Johnson–Cook material model. In order to cap- FSW process of aluminum alloys. The developed model is
ture material flow and high deformation underneath the validated using experimental results taken from the FSW
pintool, authors used adaptive re-meshing technique. process of workpieces tested by the authors [29]. An
However, like Assidi et al. [26], the friction coefficient I-STIR PDS-FSW machine with a fixed pintool was used in
between the pintool and workpiece was considered to be welding the workpieces selected for model validation. The
constant in this study. In reality, the friction coefficient is experimental setup is shown in Fig. 2. Both validation
temperature dependent; as the temperature increases, the weld workpieces were made of lightweight AA2219-T87
friction coefficient decreases [28]. aluminum alloy, whose chemical composition is listed in
Most of the thermal modeling calculates source of heat Table 1.
using heat flux per unit area which is distributed over the The developed finite element (FE) model consists of two
tool surface. This heat flux is inputted by the user in the simulated parts: workpiece and pintool, as shown in Fig. 3.
model as a constant intensity or it is varied as a function of In the model, the pintool is considered as a 3D rigid body
the distance from the tool’s axis. However, this type of rather than a deformable material as they are manufactured
modeling is far from real-life welding scenario and is from a hardened material that can stir the workpiece metal
incapable of capturing real-life friction interface between and to avoid computational complexities. The pintool
tool and the workpiece because the effect of stick/slip shoulder has a diameter of 30 mm and a height of 4 mm.
during FSW cannot be captured. Furthermore, strain rate The diameter of the pin nib is 10 mm (the part that plunges
dependency cannot be captured with thermal modeling. into the workpiece) and its height is 6 mm. The pintool has
The FSW process consists of several complex physical a tapered angle of 20°.
processes. An ideal and fully thermomechanical model of In the model, the motion of the pintool is controlled by
FSW (1) should be temperature dependent, (2) should be using a reference node. The reference node has translation,
able to capture tool material interface condition (stick/slip rotation, and thermal degrees of freedom. The pintool is
condition), and (3) needs to be strain rate dependent. assumed to be isothermal. The workpiece’s elastic as well
Moreover, in all of the aforementioned models, only the
models developed by Buffa et al. [20] and Hamilton et al.
[27] simulated all three stages of FSW; however, these two
models assume a temperature-independent friction
coefficient.
The main objective of this current work is twofold:
Foremost, to build a rate-dependent thermomechanical
model which can evaluate heat generation during FSW by
considering the frictional heat and plastic deformation. A
major difference between the presented model and previ-
ous models published so far is that all steps of FSW pro-
cessing were successfully simulated including plunge,
dwell, and travel stages. And secondly, the model used a
temperature-dependent friction coefficient that considers
sticking/sliding conditions. Two separate experimentally
obtained weld schedules have been used to validate the
numerical model. Fig. 2 Dominant FSW process parameters and experimental setup
The subsequent objective of this study is to conduct a
parametric study of the critical weld parameters, namely
rotational speed, weld speed, and plunge rate, for furthering Table 1 Chemical compositions (wt%) of AA2219
understanding of FSW processing. These weld schedules Element Ti Zn Fe V Cu Mn Zr Si Mg
were selected from an earlier experimental investigation by
wt% 0.10 0.10 0.30 0.15 6.8 0.40 0.25 0.20 0.02
the Louisiana State University’s research team [29].

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Fig. 3 a Finite element model of the workpiece and pintool; b dimensions of the pintool

as plastic behavior was modeled using strain rate-depen- eight nodes with trilinear displacement. The element
dent material model, as in real life strain rate varies greatly C3D8RT has the capability of reduced integration and
during FSW [30]. As mentioned previously, no heat flux hourglass control.
was used as a heat source in the present work. Conversely,
friction between the pintool and the workpiece causes heat 2.1 Material and Associated Flow Model
generation in the present model. As such, any errors or bias
due to user’s choice of heat source input and distribution is In the current study, a temperature- as well as strain rate-
eliminated. In the FEA model, the workpiece has a length dependent material model with an elasto-plastic Johnson–
of 80 mm, width of 80 mm, and thickness of 8.13 mm. The Cook material law [31] has been used. The Johnson–Cook
meshed model has 25,600 elements and 204,800 nodes as plasticity model is widely used for high strain rate defor-
shown in Fig. 4. As will be described later in detail, the mation. The model can be represented by Eq. (1) [32],
solid element C3D8RT used to build the model consists of     
  e_pl T  Tref m
ry ¼ A þ B½epl n 1 þ cln 1 ;
e_0 Tmelt  Tref
ð1Þ
where ry is the yield stress, e_pl is the effective plastic strain
rate, epl is the effective plastic strain, and e_0 is the nor-
malizing strain rate. Here, A, B, C, n, and m represent
material/test constants, and Tref and Tmelt are reference and
melting temperatures, respectively. A summary of the
AA2219-T87 material property for the Johnson–Cook
model is listed in Table 2.

2.2 Yield Condition

The present model used von Mises yield criterion. This

yield criterion is expressed by Eqs. (2) and (3) [34],
 
f r; ry ¼ re  ry ¼ 0; ð2Þ
Fig. 4 Schematic representation of boundary conditions of thermal
analysis and meshes of the finite element model (all dimensions are in where
millimeter)

Table 2 Johnson–Cook material plastic model input [27]

A (MPa) B (MPa) n m Melting temperature (°C) [33] Reference temp (°C)

369 684 0.73 1.7 543 25

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
3 1 Tiþ1 ¼ ðDtiþ1 ÞT_i þ Ti ; ð10Þ
re ¼ von Mises effective stress ¼ r : r  trðrÞ2 ;
2 3 Thus, by substituting for T_i from Eqs. (8) into (10),
ð3Þ nodal temperature rate can be expressed as Eq. (11) [28],
ry is the yield strength, and tr is the Tresca criterion. Tiþ1 ¼ ðDtiþ1 ÞC 1 ðF  KTi Þ þ Ti ; ð11Þ
In general, convection acts as a main source for heat loss
2.3 Flow Rule in the workpiece. The top and side surface of the workpiece
has heat loss that is calculated using Eq. (12) [21];
The increments of plastic strain are determined by flow
rule. In present model, an associative flow rule is used. This ql ¼ hcon ðT  Ta Þ; ð12Þ
flow rule is expressed by Eq. (4) [32, 34]; where T represents workpiece’s absolute temperature, Ta is
  oN the temperature in the ambient, and hcon is the coefficient
e_pl ¼ dk ; ð4Þ for convection.
or
  In the experimental setup, a chill bar is used at the upper
where e_pl is the plastic strain change, dk is the plastic surface of the workpiece (see Fig. 2). This chill bar per-
strain increment magnitude, N is the plastic potential, and forms as a heat sink in addition to clamping the workpiece
or is the stress change. to resist distortion during welding. Due to the presence of
this chill bar, high heat transfer occurs from top surface.
2.4 Thermal Boundary Condition Therefore, the weld plate’s top surface was assigned a
value of 100 W/m2 for heat transfer coefficient [21]. For
Figure. 4 shows the schematic sketch of the physical aluminum to air convection from the side surface, low heat
model. The initial temperature boundary condition can be transfer coefficient 30 W/m2 is used [4]. Also a backing
represented by Eq. (5) [28]; plate is used at the bottom of the plate to counter the plunge
T ðx; y; z; tÞ ¼ Ti ; ð5Þ force. The backing plate absorbs heat during welding.
Therefore, high value of heat transfer coefficient is applied
The transient heat transfer process during FSW can be
in the model to calculate heat transfer from backing plate.
represented by Eq. (6) [28];
Equation (13) is used in the model to calculate the heat loss
oT o oT oT oT from backing plate [21];
qcp ¼ kx þ ky þ kz þ Q; ð6Þ
ot ox ox oy oz qback ¼ hback ðT  Ta Þ; ð13Þ
where, cp is the specific mass heat capacity, k is the thermal where hback represents backing plate convection heat
conductivity (kx ; ky ; kz are heat conductivity at x; y; z coefficient. For simplicity, hback was calibrated to match
directions), q is the material density, Q is the heat gener- experimental data, which was found to be 100 W/m2 for
ation, and T is the workpiece’s absolute temperature. the validation FSW test setup. Figure 4 represents all
Equation (6) can be rewritten by Eq. (7) [28]; thermal boundary condition of present analysis.
C ðtÞT_ þ K ðtÞT ¼ QðtÞ; ð7Þ
2.5 Mechanical Boundary Condition
where T is the temperature vector, T_ is the time-dependent
temperature derivative (i.e., dT
dt Þ; K ðtÞ is the conductivity Figure 5 shows the schematic sketch of the physical model.
matrix which is dependent on time, and CðtÞ is the time- The bottom of the workpiece is restrained in the normal
dependent capacitance matrix. direction as shown in Fig. 5 [21]
The temperature derivative rate from Eq. (7) results in Uy ¼ 0 at y ¼ 0; ð14Þ
Eq. (8) [28],
T_i ¼ C 1 ðQðtÞ  KTi Þ; ð8Þ 2.6 FEA Modeling

Equation (9) represents forward difference integration 2.6.1 Workpiece and Pintool Modeling
for temperature rate [28],
Tiþ1  Ti The numerical simulations presented in this study were
T_i ¼ ; ð9Þ analyzed in ABAQUS/Explicit. In the present model, the
Dtiþ1
pintool was modeled using analytic rigid shell element. As
The above expression can be rewritten as shown in stated earlier, the workpiece was modeled using C3D8RT
Eq. (10) [28], elements, which is an eight-node 3-D coupled temperature

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

displacement degree of freedom. Moreover, the element was 0.4 mm/s until a plunge depth of 6.08 mm was
has the capability of reduced integration and hourglass reached.
control. It should be noted that the workpiece was modeled In the current analysis, the frictional energy dissipation
as a single workpiece rather than a two-part workpiece. energy rate is calculated by Eq. (15) [32],
The whole workpiece was defined as an adaptive domain, R ¼ sc;
_ ð15Þ
which enables the mesh and material to be moved inde-
pendently. The workpiece surface is treated as a sliding where s is the stress due to friction and c_ is the slip rate.
type, which allows the mesh to trail the material in the The heat energy released on each surface is assumed to be
route normal to the surface but rests stationary in the other generated by Eq. (16) [32],
two orthogonal directions. The length as well as width of PA ¼ f gR and PB ¼ ð1  f ÞgR; ð16Þ
the workpiece was reduced to decrease simulation time, but
the actual thickness was maintained. It should be noted that where g is the fraction of dissipated energy, f is the
the workpiece width in the model was reduced by trun- weighting factor, PA is the heat flux into the slave surface,
cating 112 mm from the plate’s extremities, i.e., away and PB is the heat flux into the master surface.
from the weld line. The parts that are eliminated from the The heat generation due to plastic dissipation is calcu-
model are known to have far less effect on the process of lated by Eq. (17) [32],
welding. The dimension of the modeled plate is length: F pl ¼ gA rd e_pl ; ð17Þ
80 mm, width: 40 mm, and thickness: 8.13 mm.
The FSW modeling is separated in three stages—(1) where gA is the user-defined factor, rd is the deviatoric
plunge, (2) dwell, and (3) travel stage. At plunge stage, the stress, e_pl is the rate of plastic straining, and F pl is the
pintool moves down vertically while rotating; it keeps on plastic dissipation energy.
rotating on the same location during the dwell stage. Later, The present model considered 90% of the frictional and
the pintool moves along weld line with rotation during the plastic energy transformed to heat energy. The current
travel stage. Details of the steps needed for modeling, i.e., simulation also considers that the majority of the heat
time steps and boundary conditions, are mentioned in generated (95% of the total heat produced) was dispersed
Table 3. Also, during the analysis, the pintool plunge rate in the workpiece and the remaining amount of generated
heat (5% of the total heat produced) was dispersed in the
pintool as recommended by earlier research work [35].
During FSW heat generation, the choice of friction
coefficient between pintool and the workpiece plays an
important role. However, the friction coefficient is depen-
dent on several factors, e.g., temperature, tool and work-
piece relative motion, contact geometry, and applied force.
An extensive study was conducted on the factors that affect
the friction coefficient for FSW processes by Zhang et al.

Table 4 Friction coefficient (temperature dependent) used in present

model
Temperature (°C) Friction coefficient

25 0.30
300 0.25
420 0.20
Fig. 5 Mechanical boundary conditions of workpiece 543 0.01

Table 3 Simulation details for three steps (Plunge, Dwell, and Traverse)
Step Time duration of the step Boundary condition

Plunging 15.2 s Displacement in y-axis, rotation in y-axis

Dwelling 0.1 s Rotation in y-axis
Traversing 20 s Rotation in y-axis movement in x-axis

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

€ þ C w_ þ Kw ¼ F;
Mw ð18Þ
where M represents mass, C is damping, K is stiffness
€ w;
coefficient, and F is an external force. Also, w, _ and w
represent nodal acceleration, nodal velocity, and nodal
displacement, respectively.
The above equation can be rewritten in the form of
Eq. (19) [28],
€i ¼ M 1 ðF  C w_ i  Kwi Þ;
w ð19Þ
An explicit central difference formula has been used for
integration. The acceleration equation can be expressed as
Eq. (20) [28],
w_ iþ12  w_ i12
€i ¼
w ; ð20Þ
Fig. 6 Flowchart used to explain selection of friction coefficient used ðDtiþ1 þ Dti Þ=2
in this study [13]
The velocity can be expressed by Eq. (21) [28],
 
Dtiþ1 þ Dti
[13], who found that the friction coefficient depends greatly w_ iþ12 ¼ €i þ w_ i12 ;
w ð21Þ
2
on temperature. Therefore, current analysis used a tem-
perature-dependent friction coefficient analysis which By substituting for the nodal acceleration in Eq. (20) by
varies between 0.3 and 0.2 [28]. The value of friction Eq. (19) [28], we get
coefficient is listed in Table 4.  
Dtiþ1 þ Dti
From Table 4, we can see that with the increase in w_ iþ12 ¼ M 1 ðF  C w_ i  Kwi Þ þ w_ i12 ; ð22Þ
2
temperature, coefficient of friction is constant up to
300 °C; but as the temperature reaches 300 °C, coefficient 2.6.4 Computational Strategies for Arbitrary Lagrangian–
of friction starts decreasing. The selection of coefficient of Eulerian (ALE) Methods
friction can be explained by the analysis of Zhang et al.
[13] with a help of flowchart shown in Fig. 6. Computational cost of FSW process by ALE method can
become easily prohibitive. During coupled thermome-
2.6.2 Material Properties chanical analysis of FSW, both mechanical and thermal
parts have their own stable time increment. The stable time
The weld base metal is AA2219-T87. The density of increment is defined as the smaller among the two.The
AA2219-T87 is 2840 (kg/m3), specific heat capacity is mechanical stable time is specified by the condition that
1100 (J/(kg °C)), and the heat conductivity is 160 (W/ stress should not translate more than the distance of the
(m °C)), which are assumed to be independent of temper- minimum element length of the dimension. The increment
ature variation. The melting point of AA2219-T87 is for stable time is described as
543 °C [33]. Lsmall
As mentioned previously, the FSW pintool is considered Dtmax;mech ¼ ; ð23Þ
cd
to be rigid and, therefore, no mechanical properties were
defined for its material. However, as the pintool was where Lsmall is the smallest characteristic element length
obtaining a portion of the heat generated due to pintool– and cd is the dilatational wave speed of the material.
workpiece interaction, during the FSW process its thermal The dilatational wave speed defined in a linear elastic
capacity had to be identified. Heat capacitance from bottom material is defined as [32],
sffiffiffiffi
of the pintool surface is assigned a value of 350 W/m2 in
E
the model. cd ¼ ; ð24Þ
q
2.6.3 Mechanical Analysis where E is the Elastic modulus and q is the density of the
material. In case of aluminum alloys, E ¼ 70 GPa and
In the analysis, mechanical response of FSW is represented q ¼ 284 kg/m3. Therefore, the value of cd in Eq. (24) is
by the following form of the equation of motion (Eq. (18)) 4964.66 m/s. The smallest workpiece element size present
[28], in the current work is 0.001 m. Therefore, using Eq. (23),

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

the stable time increment for Dtmax;mech is determined to be problem is modeled using ALE technique. In the ALE
* 2.0 9 10-11 s. technique, the node points need not to be fixed in space;
Since the FSW process is a thermomechanical problem, thus, this formulation enables the mesh to move indepen-
the thermal stable time increment also has to be checked. dently of the material, making it possible to maintain a
According to its definition, during this time increment, the high-quality mesh during an analysis. Combining the
thermal wave should not spread a distance longer than the advantages of both the Lagrangian and the Eulerian tech-
minimal element length. Therefore, thermal stable time niques, the ALE method is well suited for dealing with
increment is determined using Eq. (25) [32]as follows: large deformation problems avoiding numerical difficulties
L2small which is attributed to excessive element distortion.
Dtmax;therm ¼ ; ð25Þ Furthermore, the adaptive meshing feature of ABAQUS/
ð2aÞ
Explicit plays a great role in modeling FSW [27]. The main
where Lsmall is the smallest characteristic element length criteria of the adaptive mesh scheme are that a fine mesh is
and a is the thermal diffusivity of the material. For alu- automatically retained even when the model is subjected to
minum alloys, the value of a is 2.44 9 10-5 (m2/s). Thus, high deformations, which empowers the mesh to travel
from Eq. (25) the thermal stable time increment, independent of its associated material at any time step.
Dtmax;therm , equals to 2.04 9 10-4 s. When adaptive meshing is applied, a new mesh is gener-
Based on these time increments, it was determined that ated over the adaptive mesh domain. In each new mesh
the computational resources available for current research generation, the nodes are relocated in the domain. In this
work would require an estimated 108 h for every simulated paper, the whole workpiece was considered as an adaptive
case. Due to the high time for computation, a mass scaling mesh domain and each time increment meshing sweep was
algorithm is applied in the model [25]. The purpose of set to 40.
mass scaling algorithms is to increase material density
artificially so it increases time for stable increment. Mass 2.6.6 Contact Condition
scaling has no effect on the amount of heat generated by
dissipation of plastic deformation and friction. Both fixed An important part for modeling of FSW process is to
mass scaling and variable mass scaling have been used in simulate the workpiece and pintool contact condition. The
the present analysis. Fixed mass scaling is used in the effect of Norton and Coulomb friction model on FSW has
analysis on the entire model at the beginning of the step. In been analyzed in the work of Assidi et al. [26]. Using
the current study, the element whose time for stable incre- Norton friction model led to unsatisfactory tool tempera-
ment is below 10-4 s is assigned for fixed mass scaling. ture and to significantly overestimate welding forces
Variable mass scaling is used for scaling elements whose beyond the experimentally observed values. On the other
stable time increment is drastically reduced due to large hand, results obtained using Coulomb friction model are
deformations. In variable mass scaling, calculations are closer to experimentally obtained welding forces. There-
done periodically during each time increment step. In the fore, a modified Coulomb’s law is used in the present work
present analysis, variable mass scaling is applied for every to model the friction contact between the pintool and the
10 increments for the elements whose stable time incre- workpiece.
ment is below 10-5 s. During FSW, sticking or sliding occurs between the
materials in contact (workpiece and pintool) depending
2.6.5 Arbitrary Lagrangian–Eulerian (ALE) Formulation upon the contact shear stress. At the time of sticking, the
material nearby the tool surface sticks to the pintool. The
During FSW, high deformations occur underneath the velocity dissimilarity among stationary material and
pintool and around the pintool periphery, which involves a materials moving along the tool causes shearing action. For
large amount of plastic deformation. In such conditions, the sticking, the shear yield stress,syield , is expressed as
pure Lagrangian technique is not suitable for capturing the Eq. (26) [15],
high plastic deformation due to mesh distortion and ele- ry
ment entanglement of highly deformed surfaces with large syield ¼ pffiffiffi ; ð26Þ
3
plastic strains. Hence, another FSW process formulation
approach can be attempted using Eulerian technique. In the current work, shear stress due to contact, scontact ,
Eulerian technique enables the material to move through calculated equals to temperature-dependent yield stress due
mesh, which is appropriate for resolving problems in fluid to shear,
dynamics. The shortcoming of this technique is that surface
and boundary conditions are challenging to track. To
subdue this problem, the current finite element analysis

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

ry Dr^medium ¼ jr^medium  r^course j and Dr^fine

scontact ¼ syield ¼ pffiffiffi ; ð27Þ
3 ¼ jr^fine  r^medium j; ð29Þ
The shear stress due to sliding is represented by If the successive stress increments reduce by at least
Coulomb’s friction law by Eq. (28) [19], 10%, the three mesh sequence is considered to have
scontact ¼ sfriction ¼ lp ¼ lr; ð28Þ converged. This can be expressed by Eq. (30),
where p is the contact normal pressure, r is the contact Dr^medium [ 1:1Dr^fine ! converging;
stress, and l is the friction coefficient. In the current Dr^fine
r e^ ¼ \es ! converged, ð30Þ
analysis, distortion energy criterion, smax ¼ pyffiffi3 ¼ 0:58ry , r^fine
is used to identify stick/slip criterion. where e^ is the error estimate and es is the error level. In the
As defined in the modified Coulomb’s model, while present analysis, es of less than 3% is considered to be
shear stress due to contact, scontact , is lower than the max- satisfactory. Therefore, the stress value from the fine mesh
imum friction stress, smax , contact due to sticking condition was considered to be acceptable.
is modeled in current analysis. On the other hand, as the
contact shear stress, scontact , surpasses smax , the contact and
the target surface cause sliding. Modified Coulomb’s law 3 Thermal Validation
for sticking and sliding conditions is illustrated in Fig. 7
[21]. 3.1 Calibration of the FSW Experiments
jjscontact jj  smax ! ðStickingÞ; jjscontact jj  smax
! ðSlidingÞ; In the current analysis, temperature readings from the
workpiece surface during welding of two weld schedules
2.6.7 Mesh Size Study have been used for calibration. The pintool used in welding
these plates has been fabricated from of H13 tool steel with
Mesh sensitivity was investigated by studying the conver- a shoulder diameter equal of 30 mm. The tool has a tapered
gence of stress results using one model that was discretized angle of 20°. During welding, the temperature was mea-
into three different mesh sizes: coarse (1 mm 9 sured at the surface of the workpiece simultaneously by
1 mm 9 4 mm), medium (1 mm 9 1 mm 9 2.66 mm), both K-type thermocouple and FLIR thermovision A40
and fine (1 mm 9 1 mm 9 2 mm). All three models used thermographer. The thermocouples layout is presented in
the same element type C3D8RT. Let the coarse, medium, Fig. 8.
and fine mesh stress of interest be designated by
r^course ; r^medium ; r^fine , respectively. Therefore, the stress 3.2 Thermal Validation During FSW Schedule
increments can be represented by Eq. (29),
While comparing temperature history analysis from a sin-
gle weld schedule has been considered adequate for model
verification by other researchers, it was deemed more
appropriate to use two different weld schedules for

Fig. 7 Modified Coulomb’s law for sticking and sliding conditions Fig. 8 Arrangement of thermocouples (inserted within the surface)
[21] and thermographer (all dimensions are in millimeter)

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Table 5 Weld schedule used in temperature validation

Weld schedule Rotational speed, N(rpm) Weld speed, v (mm/s)

Case-1 350 1.27

Case-2 350 2.54

improving the sensitivity of the model over a wide range of

welding process parameters. The two different weld
schedules selected for this task had the identical rotational
speed, identical friction coefficient which is temperature
dependent but different travel speeds. A summary of the
two weld schedules chosen for model verification is shown
in Table 5.
Figures 9 and 10 show the time histories of the tem-
perature on the top surface at thermocouple location 3
(x = 42.36 mm, z = 26 mm) along the weld direction for Fig. 10 Temperature variation histories from experiment and FEA
two different weld schedules. The comparison shows that analysis at location 3 for Case-2 weld schedule
result from FEA analysis is in good agreement with
experimental data and also shows a similar temperature
profile. From the temperature profile, it can be seen that as
the time progresses there is a swift increase in temperature
when the pintool moves close to the thermocouple. The
maximum temperature takes place at the top surface of the
workpiece underneath the pintool shoulder as shown in
Fig. 11. After the temperature reaches its peak value, it
starts decreasing as the pintool moves farther away from
the thermocouple. Based on these results, it can be said that
the simulation profiles are close to the experimental pro-
files at the starting of welding. As the weld progresses
forward, there is a small divergence of FEA profile from
the experimental profile as the temperature profile reaches
its peak temperature. This small deviation in temperature
profile continues after the experimental profile reaches its
peak temperature. This may be attributed to the fact that the

Fig. 11 Temperature field from simulation (for Case-1 weld

schedule)

simulation does not take into account the thermal proper-

ties of the pintool which was modeled as rigid element to
avoid numerical complexities. Another reason for the
observed deviation of numerical results may be that the
thermal property of the workpiece is assumed to be tem-
perature independent. Previous published work [21] has
shown that using temperature-dependent properties for
both pintool and workpiece provided better result.
Figures 12 and 13 show the variation between FEA
result and experimental results in the transverse direction
of weld line. From both figures, it is obvious that experi-
mental temperature profile is in close arrangement with the
Fig. 9 Temperature variation histories from experiment and FEA temperature from FEA analysis results as shown in
analysis at location 3 for Case-1 weld schedule

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

between the experimental and the FEA temperature values

is given in Tables 6 and 7. From these results, we can see
that the absolute error is less than 4% for both weld
schedules at the weld centerline. The results show that as
the distance from the weld center increases, the absolute
error slightly increases, but that the maximum absolute
error is less than 10%. This may be due to the fact that the
pintool was modeled as a rigid element where thermal
properties are ignored. Also thermal properties of the
workpiece were considered to be temperature independent,
which may have caused small deviation between FEA
analysis and temperature obtained from experiment. For
both the weld schedules, the maximum absolute relative
error is less than 10%, and the mean value of the error is
below 7.08%.
Fig. 12 Temperature variations along transverse direction between
experimental and simulation data (for Case-1 weld schedule) 3.4 Mass Scaling Verification

As mentioned in Sect. 2.6.4, mass scaling was used to

reduce the computational costs in the current simulation.
The ratio of kinematic energy and internal energy is ana-
lyzed to evaluate the effect of mass scaling as shown in
Figs. 14 and 15 for both simulated weld schedules. Fig-
ures 14 and 15 show that the ratio is very small in the
current simulations with a maximum ratio of less than 0.1.
This represents that the FSW process is a quasi-static
problem though both fixed and variable mass scales are
used. Based on these results, it was concluded that the used
mass scaling factor does not affect the numerical solutions
of FSW significantly.

4 Energy Generation During FSW Process

Fig. 13 Temperature variations along transverse direction between
The heat produced during FSW causes the workpieces to
experimental and simulation data (for Case-2 weld schedule)
join together. During the FSW, heat is generated through
two possible ways, namely heat generation due to both
Sect. 3.3. Also, Figs. 12 and 13 show that the temperature
friction between tool/workpiece and plastic deformation of
around the shoulder of the pintool is higher than adjoining
the workpiece material. In the current study, both friction
location due to friction and plastic deformation. The
and plastic energy are studied by analyzing energy histories
maximum simulated temperatures in both cases were
during FSW.
406.74 and 450.6 °C for simulated Case-1 and Case-2 weld
schedules, respectively. For both the weld schedule, the Table 6 Error analysis for weld schedule Case-1 along transverse
maximum temperature during welding is lower than melt- direction
ing temperature of AA2219-T87 (543 °C) [33], which is
Distance from Temperature Temperature from Absolute
quite typical for FSW. weld center (mm) from FEA (°C) experiment (°C) error (%)

3.3 Error Analysis 0 406.7 422 3.6

15 318.0 345 7.8
The error between experimentally recorded temperatures 26 227.2 248 8.3
and FEA-simulated temperatures was calculated to quan- 32 208.5 231 9.7
tify the deviation between the two temperatures at various 39 200.4 214 6.3
distances perpendicular to the weld line. The error analysis Average error 7.1

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Table 7 Error analysis for weld schedule Case-2 along transverse direction
Distance from weld center(mm) Temperature from FEA (°C) Temperature from experiment (°C) Absolute error (%)

0 450.5 458 1.6

15 335.1 364 7.9
26 257.6 280 8.0
32 228.5 251 8.9
39 208.2 228 8.6
Average error 7.0

which is about 90.8 and 88.6% for Case-1 and Case-2,

respectively. The high rotational speed and the pintool
pressure cause this high percentage of heat production due
to differential velocity (slip rate) on the workpiece surface.
As given in Table 8, the plastic energies obtained from the
model are about 9.4 and 11.4% of the total energy dissi-
pated for Case-1 and Case-2, respectively. These percent-
ages are in the similar ranges compared to the values
described by Bastier et al. [36]. Bastier et al. [36] reported
that the contribution from plastic heat energy is only 4.4%
of the total heat energy of FSW aluminum alloy, while the
residual 95.6% heat energy being produced due to friction.
In the following sections, parametric studies are con-
ducted to study the effects of rotational speed, travel speed,
and plunge rate heat generation due to friction and plastic
Fig. 14 Ratio of kinematic and internal energies with time for Case-1 energy.
weld schedule
4.1 Effect of Pintool Rotational Speed

Various rotational speeds have been analyzed in a para-

metric study to investigate its influence on energy gener-
ation. Three different rotational speeds, namely 200, 350,
and 450 rpm, were consideration. A fixed weld speed of
v = 1.27 mm/s and a steady plunge rate of 0.4 mm/s have
been used in the analysis.
Figures 16 and 17 represent frictional dissipation energy
and plastic dissipation histories, respectively. Based on
these results, it is clear that higher dissipation energy is
produced with higher rotational speeds. Figure 16 shows
that the frictional dissipation energy increases almost lin-
early for different rotational speeds. Furthermore, frictional
dissipation energy generated for rotational speed of 350
and 450 rpm is quite close. The amount of frictional heat
Fig. 15 Ratio of kinematic and internal energies with time for Case-2 energy generated for 200 rpm is low compared to 350 and
weld schedule 450 rpm. Experiments conducted by this research group
have shown that this inadequate temperature promotes the
development of defects like wormholes or internal cavities
Table 8 lists the total plastic and frictional energies
and trenching or surface cavities [29]. The increase in total
obtained from the FE simulations for the simulated weld
frictional dissipation energy is about 19.9% as spindle
schedules, i.e., Case-1 and Case-2. It can be seen that the
rotational speed is raised from 200 to 350 rpm. Moreover,
friction work between tool and workpiece contributes the
as the rotational speed is increased from 350 rpm to
majority of the energy to the welding process. The fric-
450 rpm, the overall frictional energy increased about
tional energy contributes the majority of the total energy

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Table 8 Plastic/total energy ratio of different weld schedules

Weld Rotational Weld speed, Total friction Total plastic Total Total plastic energy
Total energy  100%
schedule speed, N(rpm) v (mm/s) energy (J) energy (J) energy (J)

Case-1 350 1.27 4.62 9 104 4.76 9 103 5.09 9 104 9.4%
4 3 4
Case-2 350 2.54 4.92 9 10 6.30 9 10 5.55 9 10 11.4%

Fig. 16 Variation of frictional energy with rotational speed (plunge Fig. 17 Variation of plastic energy with rotational speed (plunge
rate = 0.4 mm/s, v = 1.27 mm/s) rate = 0.4 mm/s, v = 1.27 mm/s)

3.0%. At high rotational speed, the high relative velocity of summarizes the effect of rotational speed on frictional and
the material causes this high energy production. Figure 17 plastic dissipation energy. Table 9 shows that when the
shows that the plastic dissipation energy increases sharply rotational speed increases, plastic dissipation energy
until the completion of the plunge stage for rotational increases considerably and the frictional dissipation energy
speeds 350 and 450 rpm. After the plunge stage, there is increases only slightly. Similar results have been reported
linear increase in plastic energy dissipation. This can be by Tang et al. [37] based on experimental work. Also,
explained by the fact that during plunge stage there is large numerical simulations by Zhang et al. [18] and Awang
amount of material penetrated by the pin nib (apex of the et al. [38] have shown that higher rotational speed causes
pintool) which causes high plastic deformation underneath higher dissipation energy.
the pintool. At the end of plunge stage, plastic deformation Figure 18 represents the variation of ratio of plastic
becomes linear during dwell and travel stages as high energy to the total energy, (Total plastic energy
Total energy ) with time for
deformations due to the compounded effect of plunging different rotational speeds. As the plunge stage starts
and stirring no longer exist. However, a similarly clear (around 0.304 s when the contact between the pintool and
trend is not observed for the rotational speed of 200 rpm the workpiece first occurs), there is a sudden rise in the
plastic dissipation energy profile, for which a sharp plastic energy ratio for all rotational speed as the pin nib
increase in plastic deformation during the plunge stage was starts penetrating the workpiece followed by drop until the
not observed. Experiments by this research group have shoulder becomes in contact with the workpiece surface.
shown that this inadequate lack of plastic deformation Once the shoulder starts stirring the workpiece material, a
promotes the development of defects like incomplete second rise in the plastic energy ratio is observed, which is
penetration which is classified as cold welds [29]. The more pronounced for N ¼ 350 rpm and N ¼ 450 rpm
amount of frictional heat energy generated for 200 rpm is during the plunge and dwell stages, but not as much for
also low compared to 350 and 450 rpm weld schedules. By N ¼ 200 rpm. During the plunge and dwell stages, the pin
increasing the rotational speed from 200 to 350 rpm, the nib first perforates into the workpiece and then the pintool
total plastic heat energy dissipation increased is 68.1%, keeps on rotating in the same location, respectively, which
whereas increasing it from 350 to 450 rpm resulted in an causes large plastic deformations in the workpiece mate-
increase in plastic heat energy of about 24.2%. Table 9 rial. This high plastic deformation is what allows FSW to

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Table 9 Synopsis of energies for various rotational speeds


Plunge rate Rotational Weld speed, Total Change in Total Change Total Totalplasticenergy
 100%
Totalenergy
(mm/s) speed, v (mm/s) frictional frictional plastic in plastic energy
N (rpm) energy (J) energya energy (J) energya

0.4 450 1.27 4.76 9 104 3.0% 5.91 9 103 24.2% 5.35 9 104 11.04%
4 3
0.4 350 1.27 4.62 9 10 Base1 4.76 9 10 Base1 5.09 9 104 9.35%
0.4 200 1.27 3.70 9 104 19.9% 1.52 9 103 68.1% 3.85 9 104 3.94%
a
with respect to Base1 weld schedule

From, it can be seen that frictional dissipation energy

rises as the welding speed increases. For different weld
speeds, frictional dissipation energy is the same during the
plunge stage, which is expected as there is no variation of
weld speed during the plunge stage, i.e., all weld schedules
have the same plunge rate and rotational speed. After the
plunge stage, frictional dissipation energy for different
weld speeds shows small variation with respect to time.
Earlier work reported by Aziz et al. [21] also shows a
similar behavior. Plastic dissipation energy profile shows
similar behavior as frictional dissipation energy profile.
After the plunge stage, plastic energy increases linearly
with the increase in weld speed. The overall frictional
energy rises about 5.3% when the weld speed is increased
from 1.27 to 1.69 mm/s. Besides, the overall frictional
Fig. 18 Variation of Total plastic energy
Total energy with time for different rotational energy increased about 0.8% when the weld speed is raised
speeds (plunge rate = 0.4 mm/s, v = 1.27 mm/s) from 1.69 to 2.54 mm/s. Furthermore, when weld speed is
increased from 1.27 to 1.69 mm/s, overall plastic dissipa-
tion energy is increased about 8.8%. Also as the welding
create a suitable weld joint. For weld schedules that pro- speed increases from 1.69 to 2.54 mm/s, plastic dissipation
duce low plastic deformations, e.g., N ¼ 200 rpm, no such energy increases about 20.7%. Table 10 summarizes the
rise in energy ratio is observed during the plunge or dwell influence of weld speed on frictional and plastic energy,
stage. Consequently, the insufficient plastic deformation which shows that when the travel speed increases, the
eventually causes cold welds that typically result in weld plastic dissipation energy increases significantly and the
defects as observed by our research group’s experimental frictional dissipation energy increases marginally. The
work [29]. During the travel stage, the plastic energy ratio higher weld speed of the pintool results in faster time to
starts declining for all rotational speed. This indicates rotate the workpiece material, which causes an increase in
lower plastic deformation during the travel stage of FSW in the production of local heat. Similar result has been
comparison with the plunge and dwell stages. Finally, it is reported by Zhang et al. [17] (Figs 19 and 20).
clear from the results in Table 9 that the weld schedule Figure 21 shows variation of energy ratio with time for
with N ¼ 450 results in a higher energy ratio compared to different welding speeds. As discussed earlier, the behavior
other weld schedules due to the fact that higher rotational during the plunge and dwell stages can be summarized in
speeds cause more material to be deformed through plastic an abrupt rise in the plastic energy ratio during the initial
deformation. plunge stage (around 0.304 s) corresponding to the start of
pin nib penetration into the workpiece. A drop in the plastic
4.2 Effect of Welding Speed energy ratio after the initial rise is observed until the
shoulder becomes in contact with the workpiece followed
The influence of pintool weld speed on energy production by a continuing increase in the plastic energy ratio during
during FSW was also investigated by considering three the remainder of plunge and dwell stages. Finally, a
weld speeds 1.27, 1.69, and 2.54 mm/s. A fixed rotational gradual decline in the plastic energy ratio during the travel
speed of N = 350 rpm has been used in these analyses.

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Table 10 Synopsis of energies for various weld speeds


Plunge Rotational Weld speed, Total friction Change in Total plastic Change in Total Total plastic energy

Total energy
rate speed, N (rpm) v (mm/s) energy (J) frictional energy (J) plastic energyb energy
(mm/s) energyb 100%

0.4 350 2.54 4.92 9 104 0.8% 6.30 9 103 20.7% 5.55 9 11.35%
104
0.4 350 1.69 4.88 9 104 Base2 5.22 9 103 Base2 5.40 9 9.66%
104
0.4 350 1.27 4.62 9 104 5.3% 4.76 9 103 8.8% 5.09 9 9.35%
104
b
with respect to Base2 weld schedule

Fig. 21 Variation of Total plastic energy

Total energy with welding speed
Fig. 19 Variation of frictional dissipation energy with welding speed (N = 350 rpm, plunge rate = 0.4 mm/s)
(N = 350 rpm, plunge rate = 0.4 mm/s)

stages. The rate of decline in plastic energy ratio differs

based on the travel speed of the weld schedule. Table 10
shows that the weld schedule with v ¼ 2.54 mm/s pro-
duced a higher plastic energy ratio compared to other weld
schedules as higher weld speed results in stirring more
material along the path of the pintool, hence increasing
plastic deformations.

4.3 Effect of Plunge Rate on Welding

The last parameter investigated in this work was the plunge

rate. Various plunge rates of 0.3, 0.4, and 0.6 mm/s were
considered. The weld speed was kept fixed at 1.27 mm/s
and spindle rotational speed was also fixed at 350 rpm
during these simulations. The weld tool plunge depth was
also fixed at 6.08 mm.
Fig. 20 Variation of plastic dissipation energy with welding speed Figures 22 and 23 show the friction and plastic dissi-
(N = 350 rpm, plunge rate = 0.4 mm/s)
pation energy of simulation for all three considered plunge
rates, respectively. Based on the plots, the energy increases
stage is observed for all welding speed. This implies that
with the decrease in plunge rate. A lower plunge rate
the plate material undergoes lower plastic deformations
allows more time to rotate the workpiece material and
during travel stage of FSW compared to plunge and dwell

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

Fig. 22 Variation of frictional energy with plunge rate (welding Fig. 24 Effect of plunge rate on plastic energy ratio at the end of
speed = 1.27 mm/s, N = 350 rpm)
plunge stage (welding speed = 1.27 mm/s, N = 350 rpm)

plastic dissipation energy. The overall plastic dissipation

energy increased 28.5% when the plunge rate is decreased
from 0.6 to 0.4 mm/s and increased 37.2% when the plunge
rate is decreased from 0.4 to 0.3 mm/s.
Table 11 summarizes the plunge force influence on
frictional and plastic energies. Comparable results have
been mentioned in the work of Awang et al. [38].
Figure 24 shows the effect of various plunge rates on the
plastic energy ratio at the end of plunge stage. It can be
seen that the weld schedule with a plunge rate of 0.3 mm/s
results in a higher plastic energy ratio compared to the
other two considered cases with plunge rates equal to 0.4
and 0.6 mm/s. This demonstrates that a slower plunge rate
enables more time to rotate the workpiece material and
therefore generate more plastic energy.
Fig. 23 Variation of plastic energy with plunge rate (welding
speed = 1.27 mm/s, N = 350 rpm)

therefore generate more energy. The overall frictional 5 Conclusions

energy increased 28.0% when the plunge rate is decreased
from 0.6 to 0.4 mm/s. In the same manner, frictional A fully coupled thermomechanical 3D finite element (FE)
energy increased 24.6% when the plunge rate is decreased model has been developed to analyze thermal heat gener-
from 0.4 to 0.3 mm/s. The plastic dissipation energy also ation and distribution during friction stir welding. In this
follows the same trend observed for frictional dissipation study, both heat dissipation from friction and plastic
energy. The slower the penetration speed, the higher the

Table 11 Synopsis of energies for various plunge rates

Plunge Rotational Weld speed, Total Change in Total plastic Change in Total Totalplasticenergy
Totalenergy  100%
rate speed, v (mm/s) friction frictional energy (J) plastic energy
(mm/s) N (rpm) energy (J) energyc energyc

0.3 350 1.27 2.89 9 104 24.6% 5.53 9 103 37.2% 3.44 9 104 16.07%
4 3
0.4 350 1.27 2.32 9 10 Base3 4.03 9 10 Base3 2.72 9 104 14.81%
0.6 350 1.27 1.67 9 104 28.0% 2.88 9 103 28.5% 1.95 9 104 14.76%
c
with respect to Base3 weld schedule

123
S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

deformation are considered. The results obtained are (NCAM), NASA’s Michoud Assembly Facility at New Orleans-
summarized as follows: Louisiana, and NASA’s Marshall Space Flight Center for their help in
providing the facility for the experimental program.
1. Temperature field obtained from the FE simulations is
similar to the temperature field obtained experimen-
tally. Two separate weld schedules temperatures were References
investigated for comparing the experimental result
[1] J.C.N.W.M. Thomas, M.G. Murch, P. Templesmith, C.J. Dawes,
with the result from the numerical simulation. For all International Patent Application No. PCT/GB92102203 and
cases considered, comparison from both weld sched- Great Britain Patent Application No. 9125978.8 (1991)
ules, the maximum relative error is lower than 10%, [2] M. Pirizadeh, T. Azdast, S.R. Ahmadi, S.M. Shishavan, A.
having a mean value of 7.1%. Bagheri, Mater. Des. 54, 342 (2014)
[3] J.H. Hattel, M.R. Sonne, C.C. Tutum, Int. J. Adv. Manuf.
2. Heat produced by friction between the pintool and Technol. 76, 1793 (2015)
workpiece produces majority of the energy (about [4] X.K. Zhu, Y.J. Chao, J. Mater. Process. Technol. 146, 263
90.8% for Case-1 and 88.6% for Case-2 weld (2004)
schedule). [5] M.Z.H. Khandkar, J.A. Khan, A.P. Reynolds, Sci. Technol.
Weld. Join. 8, 165 (2003)
3. Investigation was performed to study the influence of [6] P. Prasanna, B.S. Rao, G.K.M. Rao, Int. J. Adv. Manuf. Technol.
different weld process parameters—rotational speed, 51, 925 (2010)
weld speed, and plunge rate. This study revealed that: [7] R. Nandan, G.G. Roy, T. Debroy, Metall. Mater. Trans. A-Phys.
Metall. Mater. Sci. 37A, 1247 (2006)
a. High rotational speed causes higher amount of [8] P.A. Colegrove, H.R. Shercliff, Sci. Technol. Weld. Join. 11,
frictional and plastic dissipation energies. Fric- 429 (2006)
tional dissipation energy behaves almost linearly [9] P. Ulysse, Int. J. Mach. Tools Manuf 42, 1549 (2002)
[10] E. Ranjbarnodeh, S. Hanke, S. Weiss, A. Fischer, Int. J. Miner.
with respect to time for different rotational speeds Metall. Mater. 19, 923 (2012)
investigated. Plastic energy dissipation energy [11] Y.H. Yau, A. Hussain, R.K. Lalwani, H.K. Chan, N. Hakimi, Int.
increases sharply during the plunge stage. Low J. Miner. Metall. Mater. 20, 779 (2013)
rotational speed causes inadequate temperature [12] S.D. Ji, Y.Y. Jin, Y.M. Yue, S.S. Gao, Y.X. Huang, L. Wang, J.
Mater. Sci. Technol. 29, 955 (2013)
due to the low frictional and plastic dissipation [13] X.X. Zhang, B.L. Xiao, Z.Y. Ma, Metall. Mater. Trans. A Phys.
energies which eventually causes defects such as Metall. Mater. Sci. 42A, 3218 (2011)
wormholes, surface cavities, and incomplete [14] X.X. Zhang, B.L. Xiao, Z.Y. Ma, Metall. Mater. Trans. A Phys.
penetration. Metall. Mater. Sci. 42A, 3229 (2011)
[15] H.W. Zhang, Z. Zhang, J.T. Chen, J. Mater. Process. Technol.
b. Higher weld speeds cause more total frictional and 183, 62 (2007)
plastic dissipation energies. However, it was found [16] Z. Zhang, H.W. Zhang, Sci. Technol. Weld. Join. 12, 226 (2007)
that friction dissipation energy and plastic dissi- [17] Z. Zhang, H.W. Zhang, Mater. Des. 30, 900 (2009)
pation energy remain the same during the plunge [18] Z. Zhang, H.W. Zhang, J. Mater. Process. Technol. 209, 241
(2009)
stage for different weld speeds. After the plunge [19] H. Schmidt, J. Hattel, J. Wert, Model. Simul. Mater. Sci. Eng.
stage, the frictional energy for different weld 12, 143 (2004)
speeds shows small variation with respect to time. [20] G. Buffa, J. Hua, R. Shivpuri, L. Fratini, Mater. Sci. Eng.
c. At plunge and dwell stages of FSW, energy ratio A Struct. Mater. Prop. Microstruct. Process. 419, 389 (2006)
[21] S.B. Aziz, M.W. Dewan, D.J. Huggett, M.A. Wahab, A.M.
(Total plastic energy
Total energy ) increases as the time progresses. Okeil, T.W. Liao, Acta Metall. Sin. (Engl. Lett.) 29, 869 (2016)
[22] H.J. Aval, S. Serajzadeh, A.H. Kokabi, Int. J. Adv. Manuf.
Throughout travel stage, a decline in the energy Technol. 52, 531 (2011)
ratio is observed. [23] M. J. Lasley, MS Thesis, Brigham Young University (2004)
d. During plunge stage, a lower plunge rate causes [24] P. Heurtier, M.J. Jones, C. Desrayaud, J.H. Driver, F. Mon-
higher friction and plastic dissipation energy. theillet, D. Allehaux, J. Mater. Process. Technol. 171, 348
(2006)
e. When rotational speed and weld speed are [25] M. Grujicic, T. He, G. Arakere, H.V. Yalavarthy, C.F. Yen, B.A.
increased, plastic dissipation energy increased Cheeseman, Proc. Inst. Mech. Eng. Part B-J. Eng. Manuf. 224,
significantly. The corresponding increase in the 609 (2010)
frictional dissipation energy under the similar [26] M. Assidi, L. Fourment, S. Guerdoux, T. Nelson, Int. J. Mach.
Tools Manuf 50, 143 (2010)
conditions is shown to be minimal. [27] R. Hamilton, D. MacKenzie, H.J. Li, Eng. Comput. 27, 967
(2010)
Acknowledgements The authors are grateful for the financial sup- [28] M. Awang, PhD Dissertation, West Virginia University, 57
port provided by Louisiana Economic Development Assistantship (2007)
(EDA) program. This work is also partially supported by NASA [29] M.W. Dewan, D.J. Huggett, T.W. Liao, M.A. Wahab, A.M.
through the NASA-SLS Grant # NNM13AA02G. The authors would Okeil, Mater. Des. 92, 288 (2016)
like to thank the National Center for Advanced Manufacturing [30] M.W.M. Rajiv S. Mishra, ASM Int. 309 (2007)

123
 S. B. Aziz et al.: Acta Metall. Sin. (Engl. Lett.)

[31] G.R.C. Johnson, W.H., in Proceedings of the 7th International [35] Y.J. Chao, X. Qi, W. Tang, J. Manuf. Sci. Eng. Trans. ASME
Symposium on Ballistics, 1983, p. 541 125, 138 (2003)
[32] ABAQUS Theroy Manual, Daasault Systemes Simulia Corp [36] A. Bastier, M.H. Maitournam, K.D. Van, F. Roger, Sci. Technol.
(Providence, RI, 2014) Weld. Join. 11, 278 (2006)
[33] Matweb, http://www.matweb.com/Accessed on September 14, [37] X.G.W. Tang, J.C. McClure, L.E. Murr, A. Nunes, J. Mater.
2016 Process. Manuf. Sci. 7, 163 (1998)
[34] ANSYS, Mechanical APDL Technology Demonstration Guide [38] M. Awang, V.H. Mucino, Mater. Manuf. Process. 25, 167 (2010)
14.5, 435 (2012)

123