International Journal of Mechanical Sciences 93 (2015) 111–119

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

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

Computational analysis of the effect of welding parameters on energy

consumption in GTA welding process
Alireza Bahrami n, Daniel T. Valentine, Daryush K. Aidun
MAE Department, Clarkson University, 8 Clarkson Avenue, P.O. Box 5725, Potsdam, NY 13699, USA

art ic l e i nf o a b s t r a c t

Article history: The electrical energy needed in welding processes is a major factor of energy consumption in
Received 23 May 2014 manufacturing processes that use welds to fabricate structures. In the arc welding process an intense
Received in revised form electrical arc is used as the heat source to melt locally metallic materials to join them once they solidify.
11 September 2014
Although the arc is focused at the joint, a large portion of energy is dissipated as heat through the base
Accepted 9 January 2015
Available online 21 January 2015
metals to the surroundings. In this paper it is shown that selecting the welding parameters
appropriately, viz., the power supplied to the arc and the translation speed of the arc, can reduce the
Keywords: energy consumed per unit length of weld. The range of parameters examined was selected such that the
Arc welding vaporization of the molten metal was negligible. It is shown that increasing the arc power in conjunction
Moving heat source
with increasing the travel speed of the arc leads to reduced energy consumption per unit length of weld;
Energy consumption
this reduction in energy consumption is for equivalent welds, i.e., welds with identical geometric
Melting efficiency
Heat transfer features.
Fluid flow & 2015 Elsevier Ltd. All rights reserved.

1. Introduction decrease the energy consumed during arc welding. Of course, this
is not new. The subject of thermal energy transport in welding
In the arc welding process, which is widely used in manufac- processes has been the focus of several research studies beginning
turing industry, an electric arc is established between an electrode with the work published by Rosenthal [2] in the early 1940s; he
and the base metals. The arc provides an intense localized heat developed an analytical model for a moving heat source under the
source to melt the base metals and the filler wire to form a molten assumption that in the moving coordinates the temperature field
weld pool at the interface of the base metals to be welded. A joint was in steady state; this assumption neglected the end effects
is formed between the base metals after solidification of the weld associated with starting and stopping an automated welding
pool. The electrical energy consumed by the arc is one of the major process. Although a point heat source and the effects of convec-
portions of the total energy consumption during a manufacturing tion, phase change and heat removal from the work-piece are
process that uses welds to fabricate a product. This energy neglected in Rosenthal's model, the predicted temperature field is
consumed can be reduced by appropriate management of the acceptable outside the region where the weld pool is located.
energy flow in the work-piece [1]. This can be performed by Pavelec et al. [3] suggested the application of a Gaussian
raising the maximum temperature below the heat source and distribution of heat flux that acts on the surface of the work-
reducing the dissipation of the heat through the base metal to the piece to simulate the presence of an arc of finite dimensions. The
surroundings. heat source suggested by Palevec et al. was applied in several
To determine the energy consumption in the welding process numerical models of heat transfer and fluid flow in welding
the thermal-energy transport in the process should be investi- processes [4–8]. The latter paper reported an examination of the
gated comprehensively. Such an investigation should include the validity of the finite-element method applied in the present paper
investigation of heat transfer, fluid flow and electromagnetic field and the influence of gravitational convection on a welding process.
in the work-piece. With such a model, a set of calculations that Goldak et al. [9] suggested a Gaussian distribution of power
vary the arc power and the arc speed should be performed to density in space with a double ellipsoidal geometry. The model
determine the optimum combination of power and speed to had the flexibility to be applied in asymmetric cases like strip
electrodes and dissimilar welding.
Gery et al. [10] investigated the effect of welding parameters,
n
Corresponding author. Tel.: þ 1 315 262 7160; fax: þ1 315 268 6695.
viz., travel speed and energy input, on the temperature field in the
E-mail addresses: [email protected] (A. Bahrami), work-piece. Their results showed that the relationship between the
[email protected] (D.T. Valentine), [email protected] (D.K. Aidun). weld peak temperature and the energy input is almost linear. They

http://dx.doi.org/10.1016/j.ijmecsci.2015.01.008
0020-7403/& 2015 Elsevier Ltd. All rights reserved.
112 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119

also demonstrated that the welding parameters influence signifi-

cantly the shape of the fusion zone and the heat affected zone. Their
model does not include the fluid convection in the molten weld
pool. The fluid flow in the weld pool can affect the weld geometry
and temperature distribution in the work-piece [11].
The energy consumed to weld a unit length of the material is
directly proportional to the arc power and inversely proportional to
the travel speed of the torch [12]; therefore, the energy consumption
of welding processes with constant ratio of the arc power to the
travel speed is identical. In the present study, the geometric aspects
and the melting efficiency of the welding processes with identical
energy consumption are modeled computationally using the finite-
element software COMSOL Multiphysicss. The results of a para-
metric study are compared to find the optimum welding parameters Fig. 1. Schematic of the model geometry and the moving framework.
that lead to the least energy required per unit length of weld. To
predict accurately the shape and size of the fusion zone and the heat momentum conservation equation (Eq. (2)) are body forces that
transport in the welding process was modeled; the model included will be explained later.
the effects of electromagnetic field created by the arc, the velocity The velocity field obtained from Eqs. (1) and (2) is the velocity
field induced in the weld pool and heat transfer modes through the of the fluid relative to the moving coordinate system (xyz). This
whole work-piece. The linear GTA welding of low carbon steel (1018) coordinate system is attached to the electrode and moves at a
was examined. The power and the speed of the arc were varied constant speed equal to the travel speed, U T , along the x axis
systematically. The range of speed of the arc examined was 1.5 mm/ towards negative infinity. To simplify the computational analysis
soUT o4.5 mm/s. The range of power of the arc examined was we subtracted the translational speed from the velocity vector by
!
500 WrPa r1100 W. At the travel speeds and power inputs larger defining u ' as follows:
than the upper limits of these ranges, the maximum temperature is
! ! !
around or higher than the vaporization of alloying elements; there- u ' ¼ u  u T; ð3Þ
fore in order to avoid complexities incorporated with inclusion of !
vaporization in the model we limit the simulations to these ranges. where u T ¼  U T ^i.
We show that selecting the highest speed and power leads to the Since the travel speed is constant
lowest energy consumption per unit length of weld for the same

! ! ! ! !
u U∇ u ¼ u ' þ u T U ∇ u ';
depth of penetration.
! !
∇2 u ¼ ∇2 u ':
2. Mathematical model
hence, Eqs. (1) and (2) can be rewritten as follows:

The linear autogenous gas tungsten arc (GTA) welding (bead on !

∇ U u ' ¼ 0; ð4Þ
plate) of low carbon steel is modeled in this study. The following
assumptions are made in building the mathematical model:
!' !' !' ! ! ! !'
ρ u :∇ u ¼  ∇P þ μ∇2 u þ F b þ F d  ρ u T :∇ u : ð5Þ
 The flow of the melt in the weld pool is laminar and
incompressible; The last term at the right hand side of Eq. (5) can be interpreted as
 All material properties of the liquid metal in the weld pool, another source term.
!
other than the surface tension are not temperature dependent; The body force F b is a combination of the electromagnetic and
 The Boussinesq approximation is used to include the effect of buoyancy forces, viz.;
the changes in the mass density on the buoyancy induced
! ! ! !
convection; F b ¼ J  B þ ρβðT  T s Þ g ; ð6Þ
 Gaussian distributions are applied for the heat flux and normal ! !
current density inputs; where J and B denote the current density and the magnetic flux
 The base metal is initially at ambient temperature (298 K); density, respectively, β is the volumetric expansion factor, T and T s
 The top surface of the weld pool is flat. are the local and the alloy solidus temperatures, respectively, and
!
g ¼  g k^ is the gravity field.
The fluid flow and heat transfer in a linear welding process are The solid–liquid interface was captured by using Carman–
!
assumed to be in steady-state as modeled in a moving coordinate Kozeny theory [13]. The source term F d in the momentum
system attached to the electrode (Fig. 1). Such a model is not valid equation (Eq. (2)) is an imposed force that nulls the velocity in
for the initial and final stages of the process, but assuming the the solid state. This force is zero in the liquid phase and, hence,
weld length is large enough, these end effects on the weld are does not impact the fluid flow in the weld pool. This source term is
assumed negligible. defined as
The conservation equations of mass and momentum for a  2
! 1f !
steady incompressible flow are F d ¼ 3 l Amush u '; ð7Þ
f l þ εm
!
∇: u ¼ 0; ð1Þ
in which f l represents the liquid phase volume fraction and Amush
! ! ! ! ! and εm are constants. To generate the proper magnitude of this
ρ u :∇ u ¼  ∇P þ μ∇2 u þ F b þ F d : ð2Þ
force, Amush should be a large number and ε should be a small
!
In these equations ρ, μ u and P denote the mass density, viscosity, number. In the present investigation they are Amush ¼ 2  105 and
! !
velocity vector and fluid pressure, respectively. F b and F d in the εm ¼ 5  10  4 . The liquid phase volume fraction is determined
 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119 113

using the following equation: where j is the normal current density and b is the effective radius
8 of the current source. The bottom surface of the domain is
> 0 ðT o T s Þ
< grounded and all the other boundaries are electrically and mag-
T  Ts
f l ¼ T l  T s ðT s o T oT l Þ ; ð8Þ netically insulated.
>
:1 ðT l oT Þ At the top surface of the weld pool, the Marangoni effect is
applied to the velocity field as follows [15]:
where T l is the liquidus temperature.
Maxwell's equations are applied to calculate the electromag- 0 1
!'
netic effects. The current continuity equation is written in terms of ∂ u A ¼ ∂γ !
μ@ ∇ S T; ð18Þ
the electric potential, V, viz., ∂n ∂T
S
σ∇2 V ¼ 0; ð9Þ  
!'
where σ is the electrical conductivity. where ∂ u =∂n is the normal component of the gradient of the
! ! S
The electrical field E and the current density vector j are
velocity vector on the surface, n^ is the unit vector perpendicular to
formulated as follows: !
! the gas liquid interface, γ is the surface tension and ∇ S is
E ¼  ∇V ð10Þ derivative tangential to the interface. The surface tension, γ, is a
function of temperature and sulfur content of the material. The
! !
j ¼σ E : ð11Þ method applied to calculate the surface tension in molten steel is
explained in detail in the literature [16,17]. A no-slip boundary
Then the current density is used to calculate the magnetic flux !
density field B- as follows: condition ( u ' ¼ 0) is applied to the other surfaces.
A Gaussian distribution of heat flux is applied to the top surface
! !
∇  B ¼ μr μ0 j ; ð12Þ of the domain to simulate the presence of the arc:

where μr and μ0 are the relative magnetic permeability and the  

3P a  3ðx2 þ y2 Þ
magnetic permeability of the vacuum, respectively. qðx; yÞ ¼ exp : ð19Þ
πa2 a 2
To calculate the temperature field the thermal energy equation
is solved, viz.: In this equation q is the arc heat flux, P a is the power input and a is


! ! the effective radius of the heat source.
ρC p' u ' þ u T U∇T ¼ ∇ U ðk∇T Þ; ð13Þ

where C p' is the apparent heat capacity, which includes the heat of
fusion and k is the thermal conductivity coefficient. The apparent
heat capacity is defined as
C p' ¼ C p þ δðT ÞΔH f ; ð14Þ

where C p and ΔH f denote the actual heat capacity and the latent
heat of fusion, respectively, and δðT Þ is a Dirac delta function which
should satisfy
Z Tl
ΔH f ¼ δðT ÞΔH f dT: ð15Þ
Ts

δðT Þ is approximated as [14]

 2 
exp  ðT  T m Þ=ΔT
δðT Þ ¼ pffiffiffi ; ð16Þ Fig. 2. Boundary conditions applied to the model.
ΔT π
where T m ¼ ðT l þ T s Þ=2 is the average temperature and
ΔT ¼ ðT l  T s Þ=2 is the temperature half range in the mushy zone Table 1
(this is the material region between the solid and the melt with Thermo-physical properties of 1018 steel and other parameters applied for
numerical simulations.
temperatures between the solidus and liquidus temperatures).
Symbol Property name Value

3. Boundary conditions and numerical method T0 Ambient temperature 298 K

Tl Liquidus temperature 1802 K
Due to symmetry with respect to the xz plane, all the fluid flow Ts Solidus temperature 1770 K
dHf Heat of fusion 240 kJ/kg
and heat transfer are resolved in half of the work-piece. Moreover,
εs Surface emissivity to air exposure 0.8
melting and solidification happen in the weld pool, which is Cp,s Solid specific heat 750 J/kg K
located underneath the arc. The rest of the domain remains solid Cp,l Liquid specific heat 840 J/kg K
during the welding process. Therefore, in order to save computa- ks Solid thermal conductivity 39.4 W/m K
kl Liquid thermal conductivity 36.5 W/m K
tional resources, the fluid flow is resolved just in a semi-cylinder
ρs Solid density 7530 kg/m3
located under the heat source. This semi-cylinder is just large ρl Liquid density 7150 kg/m3
enough to accommodate the weld pool (see Fig. 2). β Volume thermal expansion of liquid 1.2  10  4/K
A Gaussian distribution of normal current density is applied at μ Dynamic viscosity 6.3  10  3 kg/(ms)
the top surface of the work-piece to simulate the current density σ Electrical conductivity 7  106 1/Ω m
h Convective heat transfer coefficient 20 W/m2 K
input from the arc, i.e.,
ε Surface emissivity 0.75
  4  10–3 m
3I  3ðx2 þ y2 Þ a Effective radius of heat source
jðx; yÞ ¼ 2 exp 2
; ð17Þ b Effective radius of current source 4.5  10  3 m
πb b
114 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119

Convective and radiative heat fluxes are applied to all lateral At the travel speed of 2.0 mm/s (Fig. 3A), the fluid flow is driven
surfaces of the domain (see Table 1 for the coefficients). At the by Marangoni convection. The fluid at the surface of the weld pool
front side, temperature is constant and it is assumed that the flows from the center towards the boundaries. As illustrated in
model is long enough to have zero heat flux at the end side (see Fig. 3B and C, at the higher travel speeds of 2.5 mm/s and 3.0 mm/
Fig. 2). s, the flow pattern undergoes a transition to a more complicated
The total heat input per unit length is a measure of the energy multiple vortex pattern. The depth of penetration is enhanced in
consumption in the welding process. This parameter is calculated these cases.
as follows: Fig. 4 presents the velocity field in a plane parallel to the yz
plane where maximum penetration occurs. This plane is at
Pa
qin' ¼ : ð20Þ x¼ 0.9 mm for Fig. 4A, x ¼1.2 mm for Fig. 4B and x ¼1.4 mm for
UT
Fig. 4C (x coordinates of the deepest point in the weld could be
To analyze the energy consumption in the GTA welding process, seen in Fig. 3). It is found that at the low travel speed of 2.0 mm/s
we run a series of computational simulations for the cases with the flow pattern is a toroid-shaped vortex, the structure of which
constant heat inputs per unit length but different travel speeds is illustrated in Figs. 3 and 4A in the xz and yz planes, respectively;
and power inputs. The power input P a is the portion of the total this structure is induced by Marangoni convection. This convection
heat generated by the welding arc which enters the base metal. In is the dominant convective motion that is adjacent to the top
order to calculate the Lorentz force in the weld pool the arc surface of the weld pool. As the travel speed increases the weld
current is calculated from the power input as follows: pool depth increases; in this case the Lorentz force induces a flow
Pa towards the edge of the weld pool along the bottom; it induces a
I¼ ; ð21Þ vortical structure that enhances the convection of heat towards
ηV
the bottom of the weld pool. This illustrates the important role the
where I is the arc current, η is the arc efficiency, and V is the arc Lorentz force plays in deepening the weld pool. Thus, between the
electric potential. travel speeds of 2.0 mm/s and 2.5 mm/s for the same thermal
In this paper we assume that a GTA welding process with a DC energy or heat input of 3.0 J/mm there is a transition from the fluid
current and a negative electrode polarity is applied and the motion of a single toroidal vortex to a system of vortical structures
shielding gas is pure argon. A constant electric potential of 12 V such that the transport of heat enhances the depth of penetration.
and an arc efficiency of 75% are applied to the simulations. It is worth noting that the velocity fields shown in Figs. 3 and 4
The system of transport equations described above that are are two dimensional views of a relatively complex three-
applied to model a linear autogenous GTA welding process was dimensional flow pattern. We illustrated this fact in Fig. 5 by
solved numerically by using COMSOL Multiphysicss (v.4.4). COM- plotting the streamlines for the weld pool for the third case; the
SOL is capable of solving model equations that include coupling of lines on the graph are streamlines in the weld pool as calculated
different physical phenomena. A tetrahedral finite-element by implementing a utility in COMSOL. The colors scale assigned to
method is applied in this tool. The discretization options for the the streamlines shows the velocity magnitude. The circulations
fluid flow equations were set to second order for velocity and first adjacent to the top surface are created by the Marangoni effect. As
order for pressure. Quadratic discretization was applied for all seen in the figure, the Marangoni effect accelerates the fluid
other partial differential equations. A very fine unstructured particles when they travel at the top surface of the weld pool.
tetrahedral mesh with the maximum element size of 0.08 mm The flow in the bottom of the pool is driven predominately by the
was applied to the fluid domain where the weld pool is located Lorentz force (this is below the toroid induced by Marangoni
(the semi-cylinder in Fig. 2). The mesh was coarser in the convection).
remaining domains where only the temperature and electromag- Let us next take a closer look at the temperature distribution on
netic fields are solved with maximum mesh size of 0.25 mm. the surface of the work-piece; this will provide additional help in
MATLABs codes were developed for post-processing the data. understanding the reason for having larger weld pools at higher
travel speeds and power inputs while the total heat input and
material properties are constant. The distributions of temperature
4. Results and discussion at the top surface of the work-piece for the combination of travel
speed and power input of 2.0 mm/s and 600 W and the 3.0 mm/s
The linear autogenous GTA welding of AISI 1018 steel is and 900 W are illustrated in Fig. 6. The value of the heat input per
modeled for different combinations of power inputs and travel unit length for both cases is 300 J/mm but the maximum tem-
speeds. The domain is 200 mm long (along the x axis), 50 mm perature in the latter case is almost 24% higher than that in the
wide (along the y axis) and 3 mm thick (along the z axis). Out of first case. However, the mean temperatures in both cases are
200 mm length of the work-piece, 30 mm is in front of the torch very close.
and 170 mm is behind the torch. The location of the torch is More than 99% of the volume of the work-piece is in solid phase
!
considered as the origin of the framework. The material properties where u ' ¼ 0. The thermal energy equation (Eq. (13)) for the solid
of 1018 steel and other parameters applied in the simulations are section of work-piece is simplified and non-dimensionalized as
presented in Table 1.
1 2
Fig. 3 shows the xz-view of the velocity field in the weld pool ∇T ¼ ∇ T; ð21Þ
Pe
for three different combinations of the power input and the travel
speed that all yield a total heat input per unit length of 300 J/mm. where Pe is a Peclet number that we defined as follows Pe ¼ V T L=α.
Note that the effective radius of the heat source (i.e., the arc) is In this dimensionless parameter, L is the effective radius of the heat
4.0 mm for the simulations performed. Therefore the entire weld source and α is the thermal diffusivity of the solid material.
pool is located underneath the heat source. The relative motion of The Peclet number can be interpreted as a ratio of the heat that
the work-piece with respect to the arc makes the longitudinal is removed through translational motion of the work-piece to the
weld pool profile asymmetric. The temperature distribution at the heat that is transferred into the material by conduction. Since the
surface of the weld pool is from the melting temperature at the material properties and the arc geometry are kept constant in this
edge of the pool to the maximum temperature located above the investigation, the Peclet number varies only with variation of the
maximum penetration depth. travel speed. In the two cases shown in Fig. 6, Peclet numbers are
 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119 115

Fig. 3. Velocity field in weld pool in the xz plane for constant heat input per unit length of 300 J/mm, (A) UT ¼2.0 mm/s and P¼ 600 W, (B) UT ¼2.5 mm/s and P¼ 750 W,
(C) UT ¼ 3.0 mm/s and P ¼900 W.

1.16 and 1.74 at the travel speeds of 2 mm/s and 3 mm/s, (WP) is increased as the combinations with higher power inputs
respectively. As illustrated in Fig. 6 the maximum temperature and travel speeds are applied. The depth of penetration is almost
increases at higher Peclet numbers which implies higher rates of doubled as the travel speed and the power input increased by a
heat removal by translational motion between the arc and the factor of 1.5 (from WP#1 to WP# 6). Fig. 7B shows that the profile
work piece as compared to the heat conducted to the material. of the top view of the WP is wider (along the y axis) and longer
However, the heat per unit distance supplied by the arc is kept (along the x axis) for higher travel speeds and power inputs.
constant; this means the power supplied to the arc increases and, However, the rate of elongation is also higher than the rate of
hence, the temperature below the arc increases. The enhancement widening.
to the mixing in the weld pool associated with the Lorentz force Fig. 8 shows the depth of penetration with respect to travel
enhances the temperature within the weld pool and, hence, more speed at three selected constant heat inputs per unit length of
metal melts at the higher travel speeds. It is the higher surface 350 J/mm, 300 J/mm, and 250 J/mm. Note that in order to main-
temperatures and the enhanced convection in the weld pool that tain the heat input per unit length constant, the arc power should
leads to greater penetration of the weld. be increased proportionally with the travel speed. For example,
The cross sectional (yz) and top views (xy) of the weld pool for when the heat input per unit length is constant at 300 J/mm and
six different travel speeds are shown in Fig. 7. The power inputs in the travel speed increases from 2 mm/s to 3 mm/s, the arc power
these cases are set such that the total heat input per unit length for also should be changed from 600 W to 900 W. It is observed that
all cases is constant at 300 J/mm. Although the same amount of for a constant heat input the combinations with higher travel
energy is put to the unit length of the weld in all six cases, the speed and power input lead in deeper penetrations.
cross sectional and top view geometries corresponding to these As illustrated in Fig. 8, to achieve a certain depth of penetration,
cases are significantly different. Fig. 7A shows the cross sectional different combinations of heat inputs per unit length and travel
profiles of the weld pools at the locations where the maximum speeds can be selected. For example, if the objective is to obtain
depth of penetration for each case occurs. As shown in Fig. 7A, the the depth of penetration of 1.25 mm as demonstrated in the figure,
depth of penetration and the cross sectional area of the weld pool the combinations of the heat inputs and travel speeds of
116 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119

Fig. 4. Velocity field in weld pool in the yz plane for constant heat input per unit length of 300 J/mm, (A) UT ¼ 2.0 mm/s and P ¼600 W, (B) UT ¼2.5 mm/s and P¼ 750 W,
(C) UT ¼ 3.0 mm/s and P ¼900 W.

can be applied. The power input for these cases are 784 W, 918 W
and 1100 W respectively. The cross sectional and top view profiles
of the weld pool for cases a–c are illustrated in Fig. 9. This shows
that by using higher power inputs and travel speeds in the welding
process, not only the production speed increases but also less
energy per unit length of weld is consumed.
The depth to width ratio of the weld pool with respect to the
travel speed for the selected heat inputs per unit length is shown in
Fig. 10. At a constant heat input per unit length, melting occurs at the
surface of the base metal at lower travel speeds and the weld pool is
shallow; accordingly, the depth to width ratio has its smallest values.
As the travel speed increases the rate of changes in the depth are
larger than the rate of changes in the width, therefore the depth to
width ratio always increases. By comparing this figure with Fig. 8, it
Fig. 5. 3D view of the flow field in the weld pool for UT ¼ 3.0 mm/s and P¼ 900 W. is seen that the trends of changes in both depth (Fig. 8) and depth to
width ratio (Fig. 10) are generally similar; however the slope in the
Case a. 350 J/mm and 2.24 mm/s, depth to width ratio curve is larger at lower travel speeds. As
Case b. 300 J/mm and 3.06 mm/s, and indicated in Fig. 10, the depth to width ratio for cases a, b and c are
Case c. 250 J/mm and 4.40 mm/s. 0.240, 0.243 and 0.246 respectively.
 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119 117

Fig. 6. Temperature distributions at the top surface of the work piece at heat input per unit length of 300 J/mm, (A) UT ¼ 2.0 mm/s and P¼ 600 W, (B) UT ¼3.0 mm/s and
P¼ 900 W.

Fig. 8. Depth of penetration with respect to travel speed for heat inputs per unit
length of 350 J/mm, 300 J/mm, and 250 J/mm.

speed and Peclet number increase. The changes in the maximum

and mean temperatures are almost linear. As shown in the figure,
the maximum temperature increases significantly as the travel
speed increases while the mean temperature only slightly
increases. Increasing the maximum temperature leads to improve-
ment in the geometric aspects of the weld such as the depth of
penetration and the depth to width ratio. On the other hand, since
the mean temperature is almost constant, the cooling rate will be
higher at higher travel speeds and power inputs. The cooling rate
is a determining factor in the solidification and intermetallic phase
formation [11]. The maximum temperature is also directly propor-
tional to metal vaporization and development of residual stresses.
It is worth mentioning that in order to avoid complexities of taking
vaporization into account the simulations were restricted to the
cases in which the maximum temperature is less than the boiling
temperature of iron.
Fig. 7. Weld pool geometry for six combinations of travel speed and arc power that
result in heat input per unit length of 300 J/mm. For the three cases with the same depth of penetration of
1.25 mm (Cases a–c), it is seen that the maximum temperatures
are significantly different. The maximum temperatures are 2547 K,
Fig. 11 illustrates the changes in the maximum and the mean 2660 K and 2836 K for cases a, b and c, respectively as shown in
temperatures of the work-piece with respect to the travel speed Fig. 11.
for different total heat inputs per unit length. This figure confirms To determine how much of the heat applied to the work-piece
the fact that the maximum temperature increases as the travel is used for the main purpose of welding, which is the melting of
118 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119

Fig. 11. Maximum and mean temperature with respect to travel speed for heat
inputs per unit length of 350 J/mm, 300 J/mm, and 250 J/mm.

Fig. 9. Cross sectional and top view of weld profile for the three cases indicated in
Fig. 8.

Fig. 12. Melting efficiency with respect to travel speed for heat inputs per unit
length of 350 J/mm, 300 J/mm, and 250 J/mm.

Fig. 10. Depth to width ratio with respect to travel speed for heat inputs per unit
length of 350 J/mm, 300 J/mm, and 250 J/mm.

the base metal, the following equation was used:

' ¼ ρAc C p ðT s  T i Þ þdH f :

qm ð22Þ

In this equation qm ' is the heat applied for heating up to the

melting point and melting of the unit length of the weld, and Ac is
the weld cross sectional area. The melting efficiency is defined as Fig. 13. Comparison of weld profiles for travel speed of 4.4 mm/s and power input
the ratio of the heat applied for melting to the total heat input, viz. of 1100 W.
ηm ¼ qm' =qin' [1]. The melting efficiencies for different heat input
values with respect to the travel speed are presented in Fig. 12. In the solid part of the base metal. As illustrated in the figure, the
this particular study, the melting efficiency varied between 1.8% melting efficiency is significantly improved by selecting the proper
and 17.3%. This shows that a small fraction of the heat is used for combinations of the travel speed and power input. As indicated in
melting the base metal and most of the heat is removed through the figure, the melting efficiency for the three cases denoted a, b
 A. Bahrami et al. / International Journal of Mechanical Sciences 93 (2015) 111–119 119

and c are 12.9%, 14.8% and 17.3%, respectively. Although the total temperature, and the melting efficiency increase significantly
heat input per unit length in case a is 40% larger than that in case when higher travel speeds are applied. In other words, more
c, it is the melting efficiency of case c that is 38% larger than that of favorable welding features are obtained by using the same
case a. This is the reason why the geometric aspects of both cases amount of energy when the welding parameters are properly
are similar. selected. Although the deeper penetrations and higher depth to
In order to verify the predictions of the weld profiles obtained width ratios and melting efficiencies are always advantageous,
from the numerical simulations the results of one case are higher temperatures could result in residual stress development,
compared with experimental data. The case examined is for the vaporization and formation of unwanted metallic phases.
travel speed of 4.4 mm/s and power input of 1100 W (case c). The  By comparing the results for different heat inputs it was found
weld profiles are compared in Fig. 13. In order to provide a power that for a certain depth of penetration (or desired weld) can be
input around 1100 W in the experiment the arc current was set to achieved by consuming less energy if the travel speed and the
125 A. The arc voltage was 12 V and the arc efficiency was arc power are properly tuned. For example, to obtain the depth
assumed to be 75%. The experimental and numerical results show of penetration of 1.25 mm the combination of travel speed and
reasonable agreement. The experimental procedure applied to arc power of 2.24 mm/s and 784 W or the combination of
obtain the experimental weld profile is described in [8]. In this 4.40 mm/s and 1100 W could be applied. This leads to a 12%
reference further validation studies of the finite element method higher maximum temperature as compared with the 2.24 mm/
in COMSOL were reported. This was in addition to examining the s case for the same depth of weld penetration; the energy
effects of high gravity on the GTA welding process. consumption in the latter case is 28% less than that in the
first case.

5. Conclusions

The autogenously gas tungsten arc (GTA) welding process on References

low carbon steel was modeled using finite element software
COMSOL Multiphysicss. The model included the electromagnetic [1] Van Elsen M, Baelmans M, Mercelis P, Kruth JP. Solutions for modelling
moving heat sources in a semi-infinite medium and applications to laser
field created by the welding arc, velocity field in the weld pool and
material processing. Int J Heat Mass Transf 2007;50:4872–82.
heat transfer in the whole work-piece. The combinations of travel [2] Rosenthal D. Mathematical theory of heat distribution during welding and
speed and arc power examined showed that for the same heat cutting. Weld J Res Suppl 1941;20:220–34.
[3] Pavelec V, Tanbakuchi R, Uyehara O, Myers P. Experimental and computed
input per unit length of weld the energy consumption could be
temperature histories in gas tungsten arc welding of thin plates. Weld J Res
reduced by selecting an appropriate combination of arc travel Suppl 1969;48:295–305.
speed and arc power. The primary conclusions of this study are as [4] Zacharia T, Eraslan A, Aidun D, David S. Three-dimensional transient model for
follows: arc welding process. Metall Mater Trans B 1989;645–659:20.
[5] Mishra S, Lienert T, Johnson M, DebRoy T. An experimental and theoretical
study of gas tungsten arc welding of stainless steel plates with different sulfur
 At a constant energy input (heat input per unit length), the concentrations. Acta Mater 2008;56:2133–46.
velocity field contains centrally upward flow of a toroid-shaped [6] Wu C, Dorn L. Computer simulation of fluid dynamics and heat transfer in full-
penetrated TIG weld pools with surface depression. Comput Mater Sci
vortex induced by Marangoni convection; it dominates the 1994;341–349:2.
convection at low travel speeds. At higher travel speeds, the [7] Salcudean M, Choi M, Greif R. A study of heat transfer during arc welding. Int J
Lorentz force induced vortical structure below the Marangoni Heat Mass Transf 1986;215–225:29.
[8] Bahrami A, Aidun DK, Valentine DT. Interaction of gravity forces in spot GTA
toroid enhance convection and, hence, deepening the weld
weld pool. Weld J Res Suppl 2014;93:139s–44s.
pool. The maximum flow velocity also increases as the combi- [9] Goldak J, Chakravarti A, Bibby M. A new finite element model for welding heat
nations with higher travel speeds and power inputs are used. sources. Metall Trans B 1984;15:299–305.
 The study of temperature distribution in the work-piece reveals [10] Gery D, Long H, Maropoulos P. Effects of welding speed, energy input and heat
source distribution on temperature variations in butt joint welding. J Mater
that the weld pool maximum temperature is higher when the Process Technol 2005;167(2–3):393–401.
combinations of higher travel speed and power inputs are [11] Kou S. Welding metallurgy. New Jersey: Wiley; 2003.
applied even though the total heat input and material proper- [12] Joshi S, Hildebrand J, Aloraier AS, Rabczuk T. Characterization of material
properties and heat source parameters in welding simulation of two over-
ties remain constant. By comparing an appropriately defined lapping beads on a substrate plate. Comput Mater Sci 2013;69:559–65.
Peclet number it is shown that the role of convective heat [13] Voller V, Prakash C. A fixed grid numerical modelling methodology for
transport (primarily associated with the relative motion convection–diffusion mushy region phase-change problems. Int J Heat Mass
Transf 1987;30:1709–19.
between the arc and the work piece) is more important than [14] Traidia A, Roger F, Guyot E, Schroeder J, Lubineau G. Hybrid 2D-3D modelling
that of conductive heat transfer at the higher travel speeds; this of GTA welding with filler wire addition. Int J Heat Mass Transf
is a result of the fact that the temperature distribution becomes 2012;55:3946–63.
[15] Lappa M. Thermal convection: patterns, evolution and stability. Chichester,
more concentrated in the vicinity of the heat source, i.e., the
West Sussex, United Kingdom: John Wiley & Sons; 2009.
temperature below the heat source is higher, and, hence, the [16] Sahoo P, DebRoy T, McNallan M. Surface tension of binary metal–surface
size of the weld pool increases (and, hence, the melting active solute systems under conditions relevant to welding metallurgy. Metall
efficiency increases). Mater Trans B 1988;19:483–91.
 The results show that at a constant heat input per unit length,
[17] McNallan M, Debroy T. Effect of temperature and composition on surface
tension in Fe–Ni–Cr alloys containing sulfur. Metall Mater Trans B
the depth of penetration, the depth to width ratio, the maximum 1991;22:557–60.