International Journal of Heat and Mass Transfer 103 (2016) 521–537

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Three-dimensional numerical simulation of arc and metal transport

in arc welding based additive manufacturing
Xiangman Zhou a, Haiou Zhang a,⇑, Guilan Wang b, Xingwang Bai c
a
State Key Laboratory of Digital Manufacturing Equipment and Technology, Huazhong University of Science and Technology, Wuhan 430074, PR China
b
State Key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and Technology, Wuhan 430074, PR China
c
School of Mechanical Engineering, University of South China, Hengyang 421001, PR China

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

Article history: In arc welding based additive manufacturing, the surface topographies of deposited layer are more
Received 5 April 2016 complex than conventional welding, therefore, the distribution of the electromagnetic force in molten
Received in revised form 20 June 2016 pool, arc pressure, plasma shear stress and heat flux on molten pool surface are not the same as the con-
Accepted 24 June 2016
ventional welding. A three-dimensional weak coupling modeling method of the arc and metal transport is
Available online 9 August 2016
developed to simulate the arc, molten pool dynamic and droplet impingement in arc welding based
additive manufacturing. In the arc model, the molten pool is simplified to be solid state on the basis of
Keywords:
experimentally observed results. The arc is simulated firstly, and then the electromagnetic force, arc
Arc welding based additive manufacturing
Arc
pressure, plasma shear stress and heat flux are extracted and transmitted to metal transport model.
Metal transport The volume of fluid (VOF) method is employed to track free surface of molten pool and droplet, and
Numerical simulation the continuum surface force (CSF) method is applied to transform all the surface forces on free surface
as localized body forces. This weak coupling model has better accuracy than empirical model and
decreases computational consumption. The molten pool morphology and cross-sectional profile of
simulated results accord well with experimental results in both single-bead deposition and overlapping
deposition, which indicates that this weak coupling modeling method is capable of simulating the com-
plex heat and mass transfer phenomena in arc welding based additive manufacturing.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction have focused on investigating the arc welding process by way of

Additive manufacturing (AM) is a new technique of a great
potential in reducing high cost of producing conventional compo-
nents with complex structures or made from relatively expensive
materials. Many AM techniques have been developed to fabricate
metal components, such as selective laser melting (SLM) [1], selec-
tive laser sintering (SLS) [2], electron beam freeform fabrication
(EBF3) [3], and arc welding based additive manufacturing (AWAM)
[4–7]. Compared with other AM techniques, the AWAM is an
efficient AM technology which is a promising option to traditional
subtractive manufacturing for fabricating large aeronautical and
naval components that feature the higher deposition rate and
lower cost.
In AWAM process, the gas metal arc welding (GMAW) or gas
tungsten arc welding (GTAW) are usually employed as heat source,
therefore the heat and mass transfer during the AWAM deposition
are similar to the conventional arc welding process. A lot of works
⇑ Corresponding author.
E-mail address: zholab@hust.edu.cn (H. Zhang).
experiments [8,9] or simulations [10–32]. Although the informa-
tion obtained through experimental observations is very useful, it
is not sufficient for revealing underlying mechanisms due to the
complexity of the welding process which involves high-
temperature plasma and non-transparent metal. Mathematical
modeling provides a convenient way to understand the physical
phenomena observed in the welding process. There are two main
types of mathematical models on arc welding simulation: unified
model [10–20] and separated model [21–31]. Hu et al. [10,11]
and Rao et al. [12,13] developed a two-dimensional unified com-
prehensive model to simulate the transport phenomena occurring
during the gas metal arc welding process. And a harmonic mean of
thermal conductivities was applied to process the heat conduction
between the arc and the metal. Rao et al. [12,13] investigated the
effect of the argon–helium mixtures on arc shape and metal trans-
port phenomena. It found that the increase of helium content in
the mixture shielding gas results in the change of arc shape and
pressure distribution along the substrate surface as well as leads
to the formation of larger droplets and the decrease of droplet
detachment frequency. Based on previous studies, Xu et al. [14]

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.06.084
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
522 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537

Nomenclature

A magnetic vector potential Tref reference temperature

Amush mushy zone constant Tp,c plasma temperature adjacent to cathode
af front ellipsoid radius of initial temperature field Tamb ambient temperature
ar rear ellipsoid radius of initial temperature field Tc cathode temperature
B magnetic flux density Tcenter center temperature of initial temperature field
b the ellipsoid radius of width direction of initial temper- T0(x, y, z) initial temperature field of molten pool
ature field t time
c the ellipsoid radius of depth direction of initial temper- x x-coordinate value
ature field y y-coordinate value
cp specific heat z z-coordinate value
Dw wire diameter V velocity vector
e elementary electric charge Vps velocity of plasma adjacent to metal surface
F volume of fraction v0 initial velocity of droplet
Fb buoyance vf wire feeding rate
Fms momentum source of enthalpy-porosity model
Fmag electromagnetic force Greek symbols
Fvol localized body forces b thermal expansion coefficient
fd droplet transfer frequency bl volume fraction of liquid metal
g acceleration of gravity c surface tension coefficient
Hev latent heat for liquid-vapor phase-change @ c=@T surface tension temperature gradient
h enthalpy e radiation emissivity
hconv heat convection coefficient j curvature of free surface
Iw deposition current l viscosity
J current density lps viscosity of plasma
Keff effective thermal conductivity l0 permeability
k thermal conductivity U electric potential
kb Boltzmann constant u shape function of free surface
n vector tangential to the local free surface re electric conductivity
p pressure q density
Q flow rate of shielding gas q1 density of first phases
qev melt mass evaporation rate q2 density of second phases
qnet net heat flux qw density of wire
Rn nozzle inner radius qd density of droplet
Rw wire radius sps plasma shear stress
r radius value of XY plane sMa Marangoni shear stress
Smass mass source terms sst surface tension
SM momentum source terms sSF surface force on free surface
SE energy source terms d length of the cathode sheath
SR radiation heat loss p circumference ratio
Sarc arc heat x a small number to prevent division by zero
Sdrop droplet heat gd proportion of droplet energy to deposition
SL latent heat Dt droplet generation time
T temperature

developed an integrated comprehensive three-dimensional model
to study the transport phenomena in gas metal arc welding consid-
ering the moving arc. Murphy [15,16] presented a self-consistent
three-dimensional model to investigate the heat and mass trans-
port phenomena in gas metal arc welding by taking a special treat-
ment of energy transfer between the workpiece and arc, namely
the heating effect of the excited atoms that cause electrons to be
emitted from the workpiece and accelerated by the sheath voltage
was added to the energy conservation equation. In addition, an
equilibrium surface (ES) method was used to track the free surface.
Compared with the widely used volume of fluid (VOF) method, the
ES method had higher computational efficiency. Wang et al.
[17,18] developed a unified model of coupled arc and weld pool
for double electrodes GTAW. The weld pool dynamic was described
by taking into account buoyance, electromagnetic force, surface
tension and plasma drag force. And the influence of electrode sep-
aration on the flow and temperature fields of arc and weld pool
were studied [18]. Yin et al. [19] established a full coupled model
of arc and weld pool for single electrodes GTAW with applied axial
magnetic fields. It found that the axial magnetic fields resulted in a
reverse flow appearing over the anode in arc, and the fluid flow
cycle brought about a wide and shallow weld pool. The models
of Wang and Yin took the sheath area as a special internal bound-
ary condition, and the free surface of weld pool was not considered.
Recently, Jian and Wu [20] developed a unified fluid flow and heat
transfer model for stationary plasma arc welding. The VOF method
was employed to track the keyhole boundary. And the whole evo-
lution processes of keyhole formation were numerically simulated.
The separated model also has been widely applied in arc weld-
ing simulation, in which the model of arc and metal transport are
not simulated simultaneously. Consequently, the complex energy
and momentum boundary conditions between the arc and the
metal do not need to calculate directly. Wu et al. [21] presented
a numerical simulation method for predicting the profile of the free
surface deformation of fully-penetrated GTAW weld pool. The
physical terms and the sign of Lagrange multiplier were used to
derive the both front- and back-side deformation of weld pool sur-
faces equations. Chen et al. [22] developed a three-dimensional
weld pool dynamics model for groove gas metal arc welding
processes to study the influence of groove angle on welding of
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
low carbon structural steel plates. The electromagnetic forces were
calculated by empirical analytical formula, and the heat flux of arc
was described as double elliptical distribution. Xu et al. [23]
developed a three-dimensional arc model to study the non-
axisymmetric arc caused by external magnetic field. Later, Xu
et al. [24] developed a three-dimensional mathematical model of
arc in GTAW and presented a method to calculate 3D arc pressure,
heat flux, and current density on the surface of the weld pool. Hai-
dar [25] established an arc model to study the dynamic effects of
metal vapor in gas metal arc welding. Hu et al. [26], Rao et al.
[27] developed a three-dimensional transient model for weld pool
dynamics and periodical impingement of filler droplets and inves-
tigated the effect on the formation of ripples in three-dimensional
moving gas metal arc welding. The heat flux and arc pressure on
surface of molten pool were considered to be a Gaussian distribu-
tion. Liu and Rao [28] applied this model [27] to investigate of weld
pool behaviors and ripple formation for a moving GTAW under
pulsed currents. Cho et al. [29] developed a fluid dynamics simula-
tion model for submerged arc welding process. The Abel inversion
method with CCD camera images was adopted to develop the heat
flux model of arc. Later, the modeling method of arc heat flux is
performed to establish three-dimensional transient numerical sim-
ulation for second pass gas metal arc welding on V-groove in var-
ious welding positions [30]. Recently, Zhou et al. [31] developed a
three-dimensional numerical model of electric arc based on
magnetic fluid dynamics for arc based additive forming process
with pure argon shielding gas. The influences of surface topogra-
phies of deposited layer on the shape and state, heat and mass
transfer of electric arc were investigated in details.
In view of the overlapping deposition and stacking deposition
are usually employed in AWAM process (as shown in Fig. 1), which
will result in more complex surface topographies of deposited
layer, as compared to the conventional GMAW, then the tempera-
ture and pressure of arc are no longer symmetrical distribution and
the heat and mass transfer of arc will also be affected by the sur-
face topographies of deposited layer [31]. In the arc welding based
deposition, the heat flux and forces of molten pool are transmitted
by the arc. Consequently, the heat resource model, arc pressure
model, and electromagnetic force model of AWAM simulation are
not identical with conventional welding and no longer symmetri-
cal distribution. The unified model has advantages of solving these
problems. However, the modeling and calculation processes are
complex and time consuming [14], and physical mechanism of
the sheath area are not completely understood. Especially, the con-
vergence of the temperature field near the free surface is limited
for widely used VOF method due to large temperature gradient.
Therefore, a lot of simplifications are usually introduced in unified
model [10–14,17,18]. The calculation accuracy of separated model
need to be enhanced, because the electromagnetic force and arc
pressure of weld pool are usually calculated by symmetrical ana-
lytical formula [26–29], or the parameters of heat flux model and
Fig. 1. A schematic representation of AWAM deposition: (a) single-bead deposition; (b) overlapping deposition; (c) stacking deposition.
523
arc pressure model are obtained through complex experiments
[28,29].
In this study, a three-dimensional weak coupling modeling
method of arc and metal transport has been presented to simulate
the arc, molten pool dynamic and droplet impingement in GMAW
based additive manufacturing. The arc model is firstly developed
based on magnetic fluid dynamics with pure argon shielding gas,
and the molten pool is simplified to be solid state based on exper-
imentally observed results that molten pool remains approxi-
mately constant surface shape during the AWAM deposition.
Then the calculated results of electromagnetic force, arc pressure,
plasma shear stress and heat flux are extracted and transmitted
to metal transfer model to simulate the molten pool dynamic
and droplet impingement. This modeling method can avoid the
complex calculation of sheath and guarantee simulated accuracy
at the maximum degree. It also provides a general method for solv-
ing complex heat and mass transport phenomena in arc welding
simulation. The CFD codes Fluent is used to solve these models,
and the interpolation function of user-defined scalar (UDS) is
applied to do the transmissions.
As illustrated in Fig. 2, a robot based AWAM experimental setup
with a high-speed camera system is established to carry out the
AWAM experiments and capture the molten pool and arc profile
in this study. A MOTOMAN robot is used to execute deposition tra-
jectory and a SAF GMAW welder is used as heat source. A pulsed
laser illumination system provides light for shooting. An 810 nm
narrowband filter is used to filter out arc light for capturing the
molten pool and an attenuation slide is applied to weaken the
arc light for capturing the arc profile.
2. Mathematical models
2.1. Basis s of the weak coupling modeling method
Fig. 3 is the images of molten pool and weld bead of single-bead
deposition captured by high-speed camera system at different
instants. The process parameters of present deposition are the
same as simulation and given in Table 1 of Section 2.2.3. It can
be seen from Fig. 3 that there has little change of the molten pool
surface as well as the morphologies and sizes of molten pool
remain stable at different instants under the used process
parameters. In addition, the entire finished deposited layers
(Figs. 19 and 20) remain the same appearance under the used pro-
cess parameters, which indicate that the molten pool remain stable
during the deposition. The above experimental observations also
mean that the heat flux and forces of the molten pool remain stable
substantially during the deposition. Therefore the molten pool is
simplified to be solid state (as shown in Fig. 4) based on experi-
mental observations in the arc model to simulate arc firstly, and
then the heat flux and forces of the molten pool are extracted
and transmitted to metal transport simulation.
524 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537

Fig. 2. The schematic diagram of robot based AWAM experimental setup.

Fig. 3. The images of weld bead and molten pool of single-bead deposition at different instants: (a) 0.5 s; (b) 0.51 s; (c) 0.52 s; (d) 0.53 s.

As shown in Fig. 4, in the arc models, the molten pool is simpli-
fied to be solid state on the basis of above experimentally observed
results. The distance between wire tip and substrate is 5 mm, the
bead width is 10 mm and the bead height is 3 mm under the used
process parameters. The distance between beads is 5 mm in over-
lapping deposition. The cross-sectional profile of single weld bead
can be fitted by circular arc curve under the used process parame-
ters [34]. The deposition direction is the positive direction of the x-
axis.
To improve the computational efficiency, the domains of metal
transport model are no longer the same with the arc model in pre-
sent study. The initial morphologies of weld bead and molten pool
of single-bead deposition and overlapping deposition are initial-
ized in the same way as the arc simulation in metal transport
model. The schematic representation of metal transport models
are illustrated in Fig. 5. The surface IJLK is the free surface.
2.2. Governing equations
In this study, the single-bead deposition and overlapping depo-
sition are modeled to verify the weak coupling modeling method.
The schematic representation of single-bead deposition and over-
lapping deposition are shown in Fig. 1(a) and (b), respectively. In
order to simplify the model, only substrate, deposited layers, wire
and shielding gas are modeled in arc model (as shown in Fig. 4),
and deposited layers and droplet are modeled in metal transport
model (as shown in Fig. 5). Many researchers took this simplifica-
tion in the modeling arc welding simulation investigations and
obtained the credible results [10–14,26–30].
In the present numerical models, the following simplifying
assumptions are made for the arc and metal transport model:
(1) The arc and molten metal are considered to be laminar flow.
The Reynolds number of arc and molten pool are about 670
and 2400, respectively. The transition from laminar flow to
turbulent occurs at a Reynolds number of around is about
100,000 for a free jet [24] and 4000 for a normal flow. There-
fore, the arc and molten metal can be modeled as laminar
flow.
(2) The arc is in local thermodynamics equilibrium (LTE). This
means that electron and heavy particle temperatures are
equal [10,12].
(3) The plasma arc is optically thin which indicates that the
optical thin radiation loos per unit volume can be used in
the arc model [10,12,23,33,35].
(4) The shielding gas is pure argon at 1 atm, and the effect of
metal vapor on plasma properties and presence of surround-
ing atmosphere are neglected [10,12].
These assumptions were widely used in the arc welding
simulation, and the simulated results were good agreement with
experimental results [13,14,24,33,35].
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537 525

Fig. 4. The schematic representation of arc models: (a) single-bead deposition; (b) overlapping deposition.

Fig. 5. The schematic representation of metal transport models: (a) single-bead deposition; (b) overlapping deposition.
526 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537

With the above assumptions, the governing equations of heat ð1  b Þ2

and mass transfer for arc and metal transport are given below: F ms ¼  3 l  Amush ðV  Vp Þ ð13Þ
bl þ -
Mass continuity
The electromagnetic force, Fmag, obtains from the computational
@
ðqÞ þ r  ðqVÞ ¼ Smass ð1Þ result of arc simulation.
@t
Momentum 2.2.3. Source terms of energy
@ In arc model, the source term of energy is given by
ðqVÞ þ r  ðqVVÞ ¼ rp þ r  ðlrVÞ þ qg þ SM ð2Þ

@t JJ 5kb J
SE ¼ þ  r h  SR ð14Þ
Energy
re 2e cp

 where the terms on the right-hand side of Eq. (14) are the Joule
@ k
ðqhÞ þ r  ðVqhÞ ¼ r  rh þ SE ð3Þ heating, the diffusive transport of enthalpy due to the electronic
@t cp
flux and radiation losses, respectively.
In the arc model, Maxwell equations are employed to solve the In metal transport model, the arc heat, Sarc, latent heat, SL, and
electromagnetic field. droplet heat, Sdrop, are the source terms of weld pool (expressed
Current continuity by Eq. (15)). The Sarc is determined by internal energy boundary
r  ðre rUÞ ¼ 0 ð4Þ condition, and the SL is defined by Eq. (16).

Ohm’law SE ¼ Sarc þ SL þ Sdrop ð15Þ

J ¼ re ðrUÞ ð5Þ Dh @bl

SL ¼ ð16Þ
Magnetic vector potential cp @t
The droplet heat, Sdrop is considered by the droplet temperature.
r2 A ¼ l0 J ð6Þ
The initial temperature of droplet equals to 2100 K [26,27].
Magnetic flux density The deposition process and numerical model parameters [27]
are shown in Table 1.
B¼rA ð7Þ
The physical properties of argon are temperature-dependent
In this study, the arc model and metal transport model have the [31]. The substrate and wire are mild steel, and the thermophysical
independent source terms. properties of mild steel are shown in Table 2.

2.2.1. Source terms of mass 2.3. Tracking of free surface

In arc model, Smass equals to 0. In metal transport model, the
metal transfer mode is the globular transfer under the used process In present study, the VOF method is employed to track the free
parameters [27], and the droplet is considered to generate at a cer- surface. A volume of fluid function, F(x, y, z, t), is defined to indicate
tain frequency and inject into molten pool [25–27]. Then the the topology of metal flow. This VOF function satisfies the follow-
source term of mass can be written as ing equation
qd dF @F
Smass ¼ ð8Þ ¼ þ ðV  rÞF ¼ 0 ð17Þ
Dt dt @t
The droplet is assumed to be a sphere. The initial diameter of
According to the definition, a unit value corresponds to cells full
droplet is determined by wire feeding rate, vf, wire radius, Rw,
of fluid (metal), while zero value corresponds to cells empty of
and droplet transfer frequency, fd [16]
fluid. Cells with F lies in 0 and 1 mean that the cells are located

1=3
q vf at the free surface. The gradient direction of F is the normal direc-
Dd ¼ 6R2w w ð9Þ tion of free surface.
qd f d
2.4. Boundary conditions
2.2.2. Source terms of momentum
In arc model, the source term of momentum is electromagnetic 2.4.1. Internal boundary conditions
force 1) Internal momentum boundary conditions
SM ¼ J  B ð10Þ In the arc model, the plasma shear stress acting on molten pool
surface is determined by viscosity and velocity gradient of the arc
In metal transport model, the source terms of momentum can plasma (Eq. (18)). And this surface force is added on metal surface
be written as (surface D in Fig. 4) as an internal boundary condition.

SM ¼ F b þ F ms þ F mag ð11Þ
Table 1
Fb is buoyance which can be calculated by Boussinesq Deposition process and numerical model parameters.
approximation
Parameters Value (unit)
F b ¼ qbðT  T ref Þg ð12Þ Deposition current 200 A
Deposition voltage 23 V
The enthalpy-porosity technique is applied to treat the solid- Deposition speed 570 mm/min
liquid mushy zone as a porous medium with porosity equal to Wire feeding rate 7.2 m/min
the liquid fraction. The liquid fraction is used to implicit track Shielding gas flow rate 15 L/min
the solid-liquid interface. A momentum source, Fms, is added to Ambient temperature 300 K
Wire diameter 1.6 mm
the momentum equation for representing the influence of mushy Initial velocity of droplet 0.6 m/s
zone on flow
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537 527

Table 2 Table 3
Thermophysical properties of mild steel [26]. The internal boundary conditions of arc model.

Property Value (unit) Boundary V (m/s) p (Pa) T (K) U (V) A (Wb s1)
1 1
Specific heat of solid phase 700 (J kg K ) B – – 1000 Coupled Coupled
Specific heat of liquid phase 780 (J kg1 K1) C – – 3000 Coupled Coupled
Thermal conductivity of solid phase 22 (W m1 K1) D – – Eq. (22) Coupled Coupled
Thermal conductivity of liquid phase 22 (W m1 K1)
Density of solid phase 7200 (kg m3)
Density of liquid phase 7200 (kg m3)
Dynamic viscosity 0.006 (kg m1 s1) The all internal boundary conditions of arc model are listed in
Radiation emissivity 0.4 Table 3. In Table 3 and Fig. 4, B represents the sidewall surface of
Heat convection coefficient 100 (W m2 K1)
wire. The surface B is close to the arc, so the temperature is set
Latent heat of fusion 2.47e5 (J kg1 K1)
Solidus temperature 1750 K to 1000 K. C represents bottom surface of wire where is the outlet
Liquidus temperature 1800 K of the deposition current of wire. The temperature of C is set to
Latent heat of vaporization 7.34e6 (J kg1 K1) 3000 K [20]. D represents the top surfaces of the substrate and
Thermal expansion coefficient 4.95e5 (K1)
weld metal, therefore the heat flux acts on the surface D. The ‘Cou-
Permeability 1.26e6 (H m1)
Electric conductivity 7.7e5 (m1 W1)
pled’ in Fluent means the coupling between the metal and arc
Surface tension coefficient 1.2 (N m1) which is to ensure the continuity of electric current and magnetic
Surface tension temperature gradient 1e4 (N m1 K1) potential at the interface.

2.4.2. External boundary conditions of computational domain

@Vps The external boundary conditions of arc models are listed in
sps ¼ lps ð18Þ Table 4. In Table 4 and Fig. 4, A represents the top section of wire
@n
where is the inlet of the deposition current. E is the bottom surface
In addition, the pressure on metal surface is determined by of the substrate and the outlet of deposition current, so the poten-
solving the uniform momentum conservation equations in arc tial is set to 0. F represents the inlet of shielding gas. Sensitivity
model. studies has shown the inlet temperature has an insignificant effect
In the metal transport model, the momentum boundary condi- on arc [14], so the temperature of F is set to 1000 K. G represents
tions of free surface (surface IJLK in Fig. 4) are plasma shear stress, the outlet of arc and shielding gas, and H represents the sidewall
arc pressure, surface tension and Marangoni shear stress, in which surface of substrate. Because the computational domains of arc
the plasma shear stress and arc pressure are obtained from the model are smaller than the actual situation, so the gradient of tem-
computational result of arc simulation and transmitted to metal perature of G and H are set to 0.
transport model. The convection and radiation boundary conditions are used to
The surface tension of free surface can be written as describe heat loss of external boundaries [32]
sst ¼ cj ð19Þ @T  
k ¼ hconv ðT  T amb Þ þ ekb T 4  T 4amb ð24Þ
@n
ru
j¼r ð20Þ The velocity of shielding gas inlet can be written as [12]
jruj
n   o
2 2 2 lnðr=Rn Þ
2Q Rn  r þ Rn  Rw lnðRn =Rw Þ
2
where u is the shape function of free surface defined by the volume pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fraction F. v z ðrÞ ¼  ; r ¼ x2 þ y2 ; r 2 ½Rw ; Rn 
p ðRn Rw Þ
2 2 2

Marangoni shear stress tangential to the metal free surface is R4n  R4w þ lnðR n =Rw Þ
given by
ð25Þ
@ c @T
sma ¼ ð21Þ The boundary conditions of metal transport models are listed in
@T @n
Table 5. The electromagnetic equations are no longer solved in
All the surface forces on free surface are transformed to local-
metal transport simulation. Therefore the boundary conditions
ized body forces and added to the momentum equations as source
are only velocity, temperature and pressure. The temperature
terms [10]. The detailed statements of the transforming method
and pressure boundary conditions of outlets (ABJI, AIKE, EFLK,
are discussed in Section 3.1.
and FBJL) and inlet (ABFE) are set to the same with arc models.
2) Internal energy boundary conditions
In the arc model, the net heat flux of metal surface (surface D in
Fig. 4) can be expressed by [10,12,15] 3. Numerical considerations

T p;c  T c 3.1. Coupling methods of the arc and metal transport model
qnet ¼ K eff  ekb T 4c  qeV HeV ð22Þ
d

 The molten pool is simplified to be solid state on the basis of
18; 836
logðqeV Þ ¼ 2:52 þ 6:21   0:5 logðT c Þ ð23Þ experimental results and the arc is simulated firstly. Then the
Tc
Table 4
The terms on the right-hand side of Eq. (22) are the contribution The external boundary conditions of arc model.
due to thermal conduction from the arc to metal, the radiation loss
Boundary V (m/s) p (Pa) T (K) U (V) A (Wb s1)
and the heat loss due to the evaporation of metal, respectively.
In the metal transport model, the net heat flux on molten pool A – – 1000 r @U
e @z ¼ Iw @A=@n ¼ 0
pR2w
free surface is determined by arc simulation, and transmitted to E – – Eq. (24) 0 @A=@n ¼ 0
metal transport model by UDS. And the heat flux boundary F Eq. (25) 101,325 1000 @ U=@n ¼ 0 @A=@n ¼ 0
G – 101,325 @T=@n ¼ 0 @ U=@n ¼ 0 A¼0
condition of metal transport model is applied on surface IJLK (free
H – – @T=@n ¼ 0 @ U=@n ¼ 0 A¼0
surface of molten pool) in Fig. 5.
528 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
Table 5
The external boundary conditions of metal transport model.
Boundary V (m/s) T (K)
IJCD – Eq. (24)
ICGK – Eq. (24)
KLHG – Eq. (24)
JDLH – Eq. (24)
ABJI – @T=@n ¼ 0
AIKE – @T=@n ¼ 0
EFLK – @T=@n ¼ 0
FBJL – @T=@n ¼ 0
ABFE Eq. (25) 1000
CDHG – Eq. (24)
electromagnetic force, plasma shear stress, arc pressure and net
heat flux are transmitted to metal transport model. In the Fluent,
the result variables can be written in user-defined scalars (UDSs)
and exported as files. These files can be imported in another model
by the way of interpolation. Therefore, the interpolation function of
UDS is used to do the transmissions, and this interpolation function
is capable of adapting different grids under the same coordinate
system.
Actually, the molten pool surface keeps fluctuating during the
deposition. Therefore, the obtained results of heat flux and forces
on molten pool surface based on the simplification of solid molten
pool in arc model will cause errors comparing with the actual sit-
uation. Nevertheless, as shown in Fig. 3, the topography of molten
pool change very little during the deposition, hence these errors
can be accepted. Simultaneously, to ensure the boundary condi-
tions can be applied to the molten pool surface, the heat flux and
forces of interface are projected to the free interface (the volume
of fraction of cells equals to 0.5 are considered to be the interface
of the metal and atmosphere) at every time step.
In metal transport model, the electromagnetic force can be
added to the momentum conservation equations as a source term
directly by UDS. The plasma shear stress, Marangoni shear stress,
arc pressure and surface tension act on molten pool surface. There-
fore, a continuum surface forces (CSF) method is applied to trans-
form all the surface forces to localized body forces. And then these
localized body forces are added to the momentum conservation
equations as source terms at boundary cells [10,12,14].
qrF
F v ol ¼ sSF 1
2
ð q1 þ q2 Þ
where rF is VOF gradient, q1 and q2 are the density of first and sec-
ond phases, respectively. sSF represents the surface forces on free
surface.
The net heat flux is also accounted for by adding source term to
the energy conservation equation at the interface layer cells [20].
Schematic diagram of the coupling method is shown in Fig. 6.
Fig. 6. Schematic diagram of coupling method.
3.2. Initialization
p (Pa) The moving electrode is not considered in this study. Therefore,
– the initial morphologies of weld bead and molten pool of single-
– bead deposition and overlapping deposition are initialized in the
– same way as the arc model in metal transport model (as shown
– in Fig. 5). To reproduce the actual shape of molten pool, a double
101,325
101,325
ellipsoid temperature field is used to initialize the temperature
101,325 field of molten pool, the formula of this initial temperature field
101,325 is expressed as
101,325 8    2  
3x2 3y 3z2
– >
< T center exp a2r exp b2 exp c2 ; x < 0
T 0 ðx; y; zÞ ¼
  2   ð27Þ
>
: T center exp 3x2 2 exp 3y2 exp 3z2 2 ; x P 0
a b c
f
where Tcenter is set to 2200 K in present simulation.
3.3. Solution techniques and grid treatments
In the present study, the heat and mass transport phenomena in
arc and metal transport model are calculated separately. The arc is
considered as in steady state [24]. When the calculation of arc
becomes steady after about 6000 iterations, the electromagnetic
force, plasma shear stress, arc pressure and net heat flux are
extracted and transmitted to metal transport model. The metal
transport model is in unsteady. When the calculated time equals
to 0.02 s, the droplet begins to generate and fall toward the molten
pool. The time step is set as 1  105 s.
In the arc model, the standard SIMPLE algorithm is employed to
treat the pressure-velocity coupling, and the second order upwind
method is applied to discretize the equations of pressure, momen-
tum, energy, potential and magnetic vector potential. In the metal
transport model, the improved SIMPLE algorithm, namely, the PISO
algorithm is used to treat the pressure-velocity coupling. The Body
Force Weighted method is applied to discretize the pressure equa-
tion, and the QUICK method is applied to discretize the equations
of momentum, VOF and energy.
In addition, a non-uniform grid is employed with finer grid sizes
near the high temperature gradient region and free surface region,
such as near the wire tip, substrate, droplet and molten pool
region. The minimum grid size is set as 0.2 mm.
ð26Þ 4. Results and discussions
4.1. Results of arc simulation
As shown in Fig. 7, a planar substrate model is modeled to verify
the temperature distribution of arc model. Fig. 8 represents the
Fig. 7. The planar substrate model and the temperature field of arc.
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537 529

Fig. 8. Comparison of distribution of isotherm between experiment and simulation (unit: 1000 K).

Fig. 9. Simulated results of arc temperature and molten pool electromagnetic force at YZ plane (X = 0): (a) single-bead deposition; (b) overlapping deposition.

Fig. 10. Arc profiles are captured by high-speed camera: (a) single-bead deposition; (b) overlapping deposition.

comparison of arc isotherm between the planar substrate model
and experimental measurement [36]. It shows that the simulated
result accords well with the experimental result as a whole, but
some isotherms of present model (14,000–18,000 K) are not exact
match with experimental measurements. In the present simula-
tion, the effect of metal vapor on plasma properties is not consid-
ered. Actually, the metal vapor in GMAW arc can increase the
radiation loss as well as specific heat of low temperature area of
arc [37], which will result in arc constriction, namely the isotherms
shrink toward the arc center in the experiment [38,39]. And it
causes the errors between the experimental and simulation result.
Fig. 9 shows the simulated results of arc temperature and mol-
ten pool electromagnetic force at YZ plane (X = 0). It can be seen
from Fig. 9 that the temperature field and molten pool electromag-
530 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
Fig. 11. Schematic diagram of paths on surface of deposition layer.
netic force are symmetric distribution in single-bead deposition,
but asymmetric distribution in overlapping deposition. In addition,
the maximum temperature of the arc center in single-bead
Fig. 12. Distributions on Path x and Path y of a single-bead deposition: (a) electromagnetic force; (b) net heat flux; (c) pressure; (d) plasma shear stress.
deposition and overlapping deposition are lower comparing with
planar substrate model, which because the protuberant deposited
layer surface of single-bead deposition and overlapping deposition
model result in a larger contact area and a smaller current density
as well as lower temperature in the arc center [31].
Fig. 10 displays the arc profiles captured by high-speed camera.
After the comparison of Figs. 9 and 10, it can be found that the arc
profiles of simulation and captured by high-speed camera are in
good agreement.
Fig. 11 exhibits the paths on the surface of single-bead depos-
ited layer (the same as overlapping deposition). Figs. 12 and 13
show the distributions on Path x and Path y. The curves labeled ‘-
emfx’, ‘-emfy’, ‘-hx’, ‘-hy’, ‘-px’, ‘-py’, ‘-plax’ and ‘-play’ represent
the electromagnetic force, net heat flux, pressure and plasma shear
stress on Path x and Path y, respectively. From Fig. 12, the electro-
magnetic force and net heat flux deviate to the rear of molten pool
on the Path x, but the pressure and plasma shear stress deviate to
the front of molten pool. All the distributions on Path y are sym-
metrical distribution. It can be seen that even if in the single-
bead deposition, the distributions on molten pool surface are also
affected by topography of molten pool and are not the same as
empirical analytical formula [26–29].
From Fig. 13, the distributions on Path x and Path y are more
complex. The electromagnetic force and net heat flux are asym-
metrical distribution, and deviate to the rear of molten pool on
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
Path x. Because the arc deviating to solidified bead (as shown in
Fig. 9(b)), the electromagnetic force and net heat flux on Path y
deviate to the solidified bead (left of molten pool). The net heat flux
on Path y presents the bimodal distribution due to the height dif-
ference between the molten pool and the solidified weld bead.
The centers of pressure distribution on Path x and Path y show a
small deviation. The plasma shear stress on Path x is similar to
single-bead deposition (Fig. 12(d)) substantially. But there has a
small reverse fluctuation between 0 and 0.005 m on Path y due
to saltation of the deposited layer morphology.
Figs. 12 and 13 demonstrate that because of the asymmetrical
deposited layer morphology, the electromagnetic force, net heat
flux, pressure and plasma shear stress on the surface of molten
pool are not completely symmetrical confirming that the morphol-
ogy of the deposited layers has important influence on these distri-
butions. In the conventional separated model, the effects of
topography of deposited layers on the distributions are usually
neglected. Apparently, these asymmetric distributions can’t be
neglected in AWAM deposition. Compared with the symmetrical
empirical model of electromagnetic force and pressure
[21,26,28], this modeling method is more realistic and has higher
accuracy. Furthermore, the distributions of electromagnetic force,
net heat flux and pressures on the overlapping deposited layer sur-
face are smaller than single-bead deposition due to a protuberant
Fig. 13. Distributions on Path x and Path y of overlapping deposition: (a) electromagnetic force; (b) net heat flux; (c) pressure; (d) plasma shear stress.
531
deposited layer surface resulting in a larger contact area and a
smaller current density [31].
4.2. Results of metal transport simulation
Fig. 14 displays the temperature and velocity vector distribu-
tion of single-bead molten pool at different instants. The tempera-
ture and velocity of molten pool distribute symmetrically at YZ
plane (X = 0). Before the droplet impingement, the metal in the
center of molten pool flow to the bottom due to the downward
electromagnetic force and positive surface tension temperature
gradient. The high temperature region in molten pool deviate to
rear of molten pool due to the backward biased heat input (as
shown in Fig. 12(b)) and backward flow of molten metal at the bot-
tom of the molten pool (as shown in Fig. 14(a and b)). The droplet
impingement (as shown in Fig. 14(b–f)) brings the heat input and
results in increasing temperature and oscillation of molten pool.
Additionally, this impingement also drive molten metal flow to
rear of molten pool (as shown in Fig. 14(e and f)). In a continuous
deposition process, these metal transport phenomena will form the
final morphology of weld bead and periodic arc-shaped ripples on
solidified weld bead [26].
Fig. 15 shows the schematic diagram of path 1 on the
Z = 1 mm plane. Fig. 16 represents the distributions on path 1
532 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537

Fig. 14. The temperature and velocity vector distribution of single-bead molten pool at different instants: (a) 0.020 s; (b) 0.024 s; (c) 0.028 s; (d) 0.032 s; (e) 0.036 s; (f)
0.040 s.
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
of single-bead deposition at different instants. The temperature
and velocity magnitude of the molten pool distribute symmetri-
cally on path 1. It can be seen clearly from Fig. 16(a) that the
temperature of the molten pool increases with time. With the max-
imum temperature increasing to 2450 K at 0.028 s due to the dro-
plet impingement and arc heat input. Subsequently, the maximum
temperature at different instants changes little. Fig. 16(b) also
shows that the velocity of molten pool center reaches its maximum
value at 0.028 s. At 0.032 s, the velocity of the molten pool center
drops to 0.076 m/s due to the impact energy absorbed by the mol-
ten pool oscillation. At that same time, the velocity at the molten
pool border begins to increase. Subsequently, the plurality of peaks
and valleys appearing on path1 means that oscillation in the mol-
ten pool becomes mitigation.
Fig. 17 displays the temperature and velocity vector distribu-
tion of the overlapping molten pool at different instants. The tem-
perature and velocity of molten pool distribute asymmetrically at
YZ plane (X = 0) and the molten pool deviates to solidified weld
bead (the positive direction of Y coordinate), which result from
asymmetric morphology of deposited layer caused leftward biased
heat input (as shown in Fig. 13(b)). Comparing with Fig. 14, the
high temperature region of molten pool is narrower than the
single-bead molten pool due to the smaller net heat flux of over-
Fig. 15. Schematic diagram of path 1 on Z = 1 mm plane.
Fig. 16. Distributions on path 2 of single-bead deposition at different instants: (a) temperature; (b) velocity magnitude.
533
lapping deposition. The molten pool dynamics are more complex
than the single-bead.
Fig. 18 represents the distributions on path 1 of overlapping
deposition at different instants. It can be seen that the temperature
and velocity magnitude of molten pool distribute asymmetrically
on path 1. As shown in Fig. 18(a), the position of maximum tem-
perature deviates from the center of the substrate about 1 mm in
Y direction. At 0.028 s, the maximum temperature is equal to
2300 K due to the lower arc heat comparing with the single-bead
deposition. Subsequently, the maximum temperature increases to
2370 K gradually. From Fig. 18(b), it can be seen that the velocity
distribution on path 1 does not present a particular law which
means that the molten pool dynamics are more complex than
single-bead deposition due to the completely asymmetric electro-
magnetic force and molten pool. At 0.028 s, the maximum velocity
reaches to 0.035 m/s. And the maximum velocity reaches to
0.037 m/s at 0.032 s, but the position of the maximum velocity
deviates far from the geometric center of the substrate. Subse-
quently, the velocity distribution on path 1 declines with the same
rhythm.
Figs. 19 and 20 display the comparison of molten pool morphol-
ogy between the simulated and experimental result. Fig. 19 shows
the results of single-bead deposition. A protrusion forms at the
center of cooled molten pool in both simulated and experimental
result due to the oscillation and high cooling rate of the molten
pool. In Fig. 20, because the arc heat flux and electromagnetic force
of overlapping deposition deviate to solidified weld bead, the
solidified weld bead is melted partly and forms a nick on solidified
weld bead in both simulated and experimental result. From
Figs. 19 and 20, it can be seen that the molten pool morphology
of simulated and experimental result are in good agreement in
both single-bead deposition and overlapping deposition. In addi-
tion, because only the molten pool rather than whole deposition
process was simulated, there had no periodic arc-shaped ripples
forming on weld bead surface in simulated result.
Fig. 21 is the comparison of the measured and simulated
deposition profiles at X = 0 cross-section. It can be seen that the
simulated deposition profiles are basically in agreement with the
experimentally measured ones. But there have some difference at
local areas. The main reason of the difference is that the initial
morphologies of weld bead and molten pool of arc and metal trans-
port model are not exactly the same with experimental result.
534 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537

Fig. 17. The temperature and velocity vector distribution of overlapping molten pool at different instants: (a) 0.020 s; (b) 0.024 s; (c) 0.028 s; (d) 0.032 s; (e) 0.036 s; (f)
0.040 s.
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537 535

Fig. 18. Distributions on path 2 of overlapping deposition at different instants: (a) temperature; (b) velocity magnitude.

Fig. 19. The molten pool morphology of single-bead deposition: (a) simulated result; (b) experimental result.

5. Conclusions
A three-dimensional weak coupling modeling method of arc
and metal transport has been developed to simulate the arc and
metal transport in GMAW based additive manufacturing. This
weak coupling modeling method can avoid the complex calcula-
tion of sheath and guarantee simulated accuracy at the maximum
degree. Additionally, this modeling method provides a general
method for solving complex heat and mass transport phenomena
in arc welding simulation. The following main conclusions are
gained:
(1) The evolvements of the distribution on surface of deposited
layer show that the electromagnetic force and net heat flux
are not completely symmetrical distribution in both single-
bead deposition and overlapping deposition. The surface
topography of deposited layer surface has significant influ-
ence on the value and distribution of electromagnetic force,
net heat flux, plasma shear stress and arc pressure of molten
pool. The distribution of electromagnetic force, net heat flux
and pressures on surface of overlapping deposition layer are
smaller than single-bead deposition due to the protuberant
deposited layer surface resulting in a larger contact area
and a smaller current density.
(2) In single-bead deposition, the temperature and velocity of
molten pool distribute symmetrically at YZ plane (X = 0),
and temperature distribution deviate to rear of molten pool
due to the backward biased heat input and backward flow of
metal at the bottom of the molten pool. And in overlapping
deposition, the temperature and velocity of molten pool dis-
tribute asymmetrically at YZ plane (X = 0) and the molten
pool deviates to solidified weld bead, which result from
asymmetric morphology of deposited layer caused leftward
biased heat input and electromagnetic force. The droplet
impingement brings the heat input and results in increasing
of temperature and oscillation of molten pool.
536 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
Fig. 20. The molten pool morphology of overlapping deposition: (a) simulated result; (b) experimental result.
Fig. 21. Comparison of the measured and simulated deposition profiles at X = 0 cross-section: (a) single-bead; (b) overlapping deposition.
(3) The high temperature region of molten pool of overlapping
deposition is narrower than the single-bead due to the smal-
ler net heat flux of overlapping deposition.
(4) The molten pool morphology and cross-sectional profile of
simulated results accord well with experimental results in
both single-bead deposition and overlapping deposition,
which show that this weak coupling modeling method is
capable of simulating the complex heat and mass transfer
phenomena in arc welding based additive manufacturing.
Conflict of interest
The authors declare that they have no conflict of interest.
Acknowledgments
This work was supported by the National Natural Science Foun-
dation of China (Grant nos. 51175203, 51374113 and 51505210).
The authors would like to thank Youheng Fu and Feixiang Mei
for their experimental contributions.
References
[1] S.L. Sing, W.Y. Yeong, F.E. Wiria, Selective laser melting of titanium alloy with
50 wt% tantalum: microstructure and mechanical properties, J. Alloy. Compd.
660 (2015) 461–470.
[2] J. Liu, C. Gao, P. Feng, C. Shuai, Selective laser sintering of b-TCP/nano-58S
composite scaffolds with improved mechanical properties, Mater. Des. 84
(2015) 395–401.
[3] W. Yan, Z. Yue, J. Zhang, Study on the residual stress and warping of stiffened
panel produced by electron beam freeform fabrication, Mater. Des. 89 (2016)
1205–1212.
[4] H. Zhang, X.P. Wang, G.L. Wang, Y. Zhang, Hybrid direct manufacturing method
of metallic parts using deposition and micro continuous rolling, Rapid
Prototyping J. 19 (2013) 387–394.
[5] F. Wang, S. Williams, M. Rush, Morphology investigation on direct current
pulsed gas tungsten arc welded additive layer manufactured Ti6Al4V alloy, Int.
J. Adv. Manuf. Technol. 57 (2011) 597–603.
[6] J. Xiong, G.J. Zhang, Adaptive control of deposited height in GMAW-based layer
additive manufacturing, J. Mater. Process. Technol. 214 (2014) 962–968.
[7] D. Ding, Z. Pan, D. Cuiuri, H. Li, A multi-bead overlapping model for robotic
wire and arc additive manufacturing (WAAM), Robotics Comput.-Integrated
Manuf. 31 (2015) 101–110.
[8] C.S. Wu, M.A. Chen, Y.F. Lu, Effect of current waveforms on metal transfer in
pulsed gas metal arc welding, Meas. Sci. Technol. 16 (2005) 2459–2465.
[9] Y.L. Chang, X.L. Liu, L. Lu, A.S. Babkin, B.Y. Lee, F. Gao, Impacts of external
longitudinal magnetic field on arc plasma and droplet during short-circuit
GMAW, Int. J. Adv. Manuf. Technol. 70 (2014) 1543–1553.
 X. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 521–537
[10] J. Hu, H.L. Tsai, Heat and mass transfer in gas metal arc welding, Part I: The arc,
Int. J. Heat Mass Transf. 50 (2007) 833–846.
[11] J. Hu, H.L. Tsai, Heat and mass transfer in gas metal arc welding, Part II: The
metal, Int. J. Heat Mass Transf. 50 (2007) 808–820.
[12] Z.H. Rao, J. Hu, S.M. Liao, H.L. Tsai, Modeling of the transport phenomena in
GMAW using argon–helium mixtures, Part I – The arc, Int. J. Heat Mass Transf.
53 (2010) 5707–5721.
[13] Z.H. Rao, J. Hu, S.M. Liao, H.L. Tsai, Modeling of the transport phenomena in
GMAW using argon–helium mixtures. Part II – The metal, Int. J. Heat Mass
Transf. 53 (2010) 5722–5732.
[14] G. Xu, J. Hu, H.L. Tsai, Three-dimensional modeling of arc plasma and metal
transfer in gas metal arc welding, Int. J. Heat Mass Transf. 52 (2009) 1709–
1724.
[15] A.B. Murphy, A self-consistent three-dimensional model of the arc, electrode
and weld pool in gas–metal arc welding, J. Phys. D – Appl. Phys. 44 (2011)
194009.
[16] A.B. Murphy, Influence of droplets in gas–metal arc welding: new modelling
approach, and application to welding of aluminium, Sci. Technol. Weld. Joi. 18
(2013) 32–37.
[17] X.X. Wang, D. Fan, J.K. Huang, A unified model of coupled arc plasma and weld
pool for double electrodes TIG welding, J. Phys. D – Appl. Phys. 47 (2014)
275202.
[18] X.X. Wang, D. Fan, J.K. Huang, Numerical simulation of arc plasma and weld
pool in double electrodes tungsten inert gas welding, Int. J. Heat Mass Transf.
85 (2015) 924–934.
[19] X. Yin, J. Gou, Zhang J. Sun, Numerical study of arc plasmas and weld pools for
GTAW with applied axial magnetic fields, J. Phys. D – Appl. Phys. 45 (2012)
285203–285215 (13).
[20] X. Jian, C.S. Wu, Numerical analysis of the coupled arc–weld pool-keyhole
behaviors in stationary plasma arc welding, Int. J. Heat Mass Transf. 84 (2015)
839–847.
[21] C.S. Wu, J. Chen, Y.M. Zhang, Numerical analysis of both front- and back-side
deformation of fully-penetrated GTAW weld pool surfaces, Comput. Mater. Sci.
39 (2007) 635–642.
[22] J. Chen, C. Schwenk, C.S. Wu, M. Rethmeiera, Predicting the influence of groove
angle on heat transfer and fluid flow for new gas metal arc welding processes,
Int. J. Heat Mass Transf. 55 (2011) 102–111.
[23] G. Xu, J. Hu, H.L. Tsai, Three-dimensional modeling of the plasma arc in arc
welding, J. Appl. Phys. 104 (2008) 103301.
[24] G. Xu, J. Hu, H.L. Tsai, Modeling three-dimensional plasma arc in gas tungsten
arc welding, J. Manuf. Sci. Eng. 134 (2012) 40–80.
537
[25] J. Haidar, The dynamic effects of metal vapour in gas metal arc welding, J. Phys.
D: Appl. Phys. 43 (2010) 165204.
[26] J. Hu, H. Guo, H.L. Tsai, Weld pool dynamics and the formation of ripples in 3D
gas metal arc welding, Int. J. Heat Mass Transf. 51 (2008) 2537–2552.
[27] Z.H. Rao, J. Zhou, S.M. Liao, et al., Three-dimensional modeling of transport
phenomena and their effect on the formation of ripples in gas metal arc
welding, J. Appl. Phys. 107 (5) (2010) 54905.
[28] J.W. Liu, Z.H. Rao, S.M. Liao, H.L. Tsai, Numerical investigation of weld pool
behaviors and ripple formation for a moving GTA welding under pulsed
currents, Int. J. Heat Mass Transf. 91 (2015) 990–1000.
[29] D.W. Cho, W.H. Song, M.H. Cho, S.J. Na, Analysis of submerged arc welding
process by three-dimensional computational fluid dynamics simulations, J.
Mater. Process. Technol. 213 (2013) 2278–2291.
[30] D.W. Cho, S.J. Na, Molten pool behaviors for second pass V-groove GMAW, Int.
J. Heat Mass Transf. 88 (2015) 945–956.
[31] X.M. Zhou, H.O. Zhang, G.L. Wang, X.W. Bai, Simulation of the influences of
surface topography of deposited layer on arc shape and state in arc based
additive forming, Acta Phys. Sinica 65 (2016) 038103.
[32] X.W. Bai, H.O. Zhou, G.L. Wang, Modeling of the moving induction heating
used as secondary heat source in weld-based additive manufacturing, Int. J.
Adv. Manuf. Technol. 77 (2015) 717–727.
[33] K.C. Hsu, K. Etemadi, E. Pfender, Study of the free-burning high-intensity argon
arc, J. Appl. Phys. 54 (1983) 1293–1301.
[34] J. Xiong, G. Zhang, H. Gao, L. Wu, Modeling of bead section profile and
overlapping beads with experimental validation for robotic GMAW-based
rapid manufacturing, Robotics Comput.-Integrated Manuf. 29 (2013) 417–423.
[35] J. Mckelliget, J. Szekely, Heat transfer and fluid flow in the welding arc, Metall.
Trans. A 17 (1986) 1139–1148.
[36] A.J.D. Farmer, G.N. Haddad, P. Kovitya, Temperature distributions in a free-
burning arc. IV. Results in argon at elevated pressures, J. Phys. D – Appl. Phys.
21 (1988) 432–436.
[37] M. Schnick, U. Fuessel, M. Hertel, et al., Modelling of gas-metal arc welding
taking into account metal vapour, J. Phys. D – Appl. Phys. 43 (2010) 2460–
2468.
[38] M. Tanaka, H. Terasaki, M. Ushio, J.J. Lowke, A unified numerical modeling of
stationary tungsten-inert-gas welding process, Metall. Mater. Trans. A 33
(2002) 2043–2052.
[39] D. Fan, Z.C. Huang, J.K. Huang, X.X. Wang, Y. Huang, Three-dimensional
numerical analysis of interaction between arc and pool by considering the
behavior of the metal vapor in tungsten inert gas welding, Acta Phys. Sinica 64
(2015) 108102.