Molten pool behaviors for second pass V-groove GMAW

International Journal of Heat and Mass Transfer 88 (2015) 945–956

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt

Molten pool behaviors for second pass V-groove GMAW Dae-Won Cho a, Suck-Joo Na b,⇑ a b

Thermal Hydraulic Safety Research Division, Korea Atomic Energy Research Institute (KAERI), Daejeon, Republic of Korea Department of Mechanical Engineering, KAIST, Daejeon, Republic of Korea

a r t i c l e

i n f o

Article history: Received 15 October 2014 Received in revised form 6 May 2015 Accepted 6 May 2015 Available online 27 May 2015 Keywords: Second pass welding Gas metal arc welding Computational fluid dynamics Abel inversion Volume of fluid Positional welding Convex bead Concave bead

a b s t r a c t This study conducted three-dimensional transient numerical simulations for second pass gas metal arc welding on V-groove in various welding positions. To obtain arc models such as arc heat flux, electromagnetic force and, arc pressure, this study adopted the Abel inversion method and the resultant asymmetric arc models can be formed. Due to the different gravity effect, it is possible to obtain the different molten pool flow patterns, solidification times, temperature distributions and bead shapes along the welding positions by numerical simulations. The formation process of convex and concave weld bead has been analyzed in detail for different welding positions. The simulation results of the fusion zone were compared with the experimental ones; therefore, the various models used in this paper can be validated. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction In the welding process, the materials melt and solidify within a very short time. Moreover, it is very difficult to expect the weld bead shapes and the weld pool behavior because of the complex physical phenomena of the welding process. Many researches, therefore, expect the weld bead shapes to be formed by the welding parameters using statistical methods such as regression models [1,2] and neural networks [3–5]. However, developing numerical simulation methods with physical and practical welding models makes it possible to predict weld pool behaviors by computational fluid dynamics (CFD) simulations. Especially, the transient simulations with a volume of fluid (VOF) method described the more practical and meaningful weld pool can flow patterns [6–8]. Cao et al. showed the droplet impingement on the weld pool surface of the gas metal arc welding (GMAW) process [6]. Moreover, Cho and Na conducted laser-GMA hybrid welding, which combined the characteristics of the laser and GMAW without considering the mutual interaction effect between two welding processes [7]. Recently, the VOF transient simulation was applied to predict the alloying element distribution and pore formation in laser-arc hybrid welding [8]. However, the arc models used above did not seriously consider

⇑ Corresponding author. E-mail address: [email protected] (S.-J. Na). http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.05.021 0017-9310/Ó 2015 Elsevier Ltd. All rights reserved.

the size of the arc, so it is necessary to apply more realistic and physical arc models. Some studies have tried to obtain a distribution of the arc plasma [9,10]. Tsai and Eagar calculated the distribution of heat and current flux on an anode plate in gas tungsten arc welding (GTAW) process by using the Abel inversion method [9]. However, it takes a long to get the distribution parameters of the arc heat flux model. Cho and Na [10] proposed a robust and simple way to get the distribution of the arc plasma by using a matrix-calculated Abel inversion and a CCD camera. Moreover, this method can calculate the irradiance of axisymmetric and elliptically symmetric arc distributions, creating advanced development in the arc welding CFD simulations. Cho et al. [11] made various arc models with an Abel inversion method and physical relations among the arc irradiance, arc temperature and arc current density. Further, they applied arc models such as arc heat flux, arc pressure, and electromagnetic models to CFD simulations for gas hollow tungsten arc welding (GHTAW) on bead-on plate. Similarly, many researchers calculated arc models by using a CCD camera and Abel inversion for various CFD welding cases. The Abel inversion method with a high speed camera brings many advantages to describe alternating current (AC) welding process where the size of the arc and signals (current and voltage) are different along time [12,13]. Cho et al. [12] captured various arc images which escaped from the flux within a very short time, and then made the transient arc models for the single electrode submerged arc welding (SAW) process. Kiran et al. [13] captured the arc images and derived arc interaction models for the

946

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

two-wire tandem SAW process with a physical relation between two arc plasmas. Moreover, Cho et al. [14] applied these advanced arc interaction models to a VOF transient CFD and then described the dynamic molten pool behaviors of the two-wire tandem SAW. The gravity effect plays an important role in the molten pool flow because it is a body force throughout the entire domain. Therefore, the welding conditions should change along the various welding positions for pipe welding; otherwise, weld defects such as humping, melt-through, and lack of fusion can be formed. Cho et al. described these dynamic molten pool flow patterns on root pass GMA welding for various welding positions in VOF transient simulations [15]. This study calculated the arc models for 2nd pass GMAW with an Abel inversion method and described the various dynamic molten pool behaviors for different welding positions using commercial software, Flow-3D.

the simulations were conducted by transient numerical analysis. Welding started at 1.5 cm of the x-direction. 2.2. Governing equations The commercial software Flow-3D solved momentum, energy, mass conservation and VOF equations. The material properties used in this paper are listed in Table 2. - Momentum equation –

_s ~ @~ V ~ rp l 2 ~ m V¼ VÞ þ f b þ V  r~ þ r Vþ ðV  ~ @t q q q s

ð1Þ

- Mass conservation equation –

V¼ r~

_s m

q

ð2Þ

- Energy equation –

2. Theoretical formulation 2.1. Material shape and mesh size Cho et al.[15] found that a high welding speed (20 mm/s) with a high current (spray mode of metal transfer) can bring a stable bead shape in a vertical downward position. With similar welding conditions in Table 1, it was found that bead shapes for the root pass welding in flat and overhead positions were stable and similar to each other, as shown in Fig. 1. Thus, this study used a shape for the material in 2nd pass GMAW, as shown in Fig. 2 and the experiment conditions for the 2nd pass GMAW, as written in Table 1. The mesh density used in the simulation was 0.25 mm/mesh to determine the molten pool behaviors because previous studies used at least four meshes for the droplet diameter [12,14–16]. All

@h ~ 1 þ V  rh ¼ r  ðkrTÞ þ h_ s @t q

ð3Þ

8 qs C s T ðT 6 T s Þ > > < T  Ts h ¼ hðT s Þ þ hsl ðT s < T 6 T l Þ > Tl  Ts > : hðT l Þ þ ql C l ðT  T l Þ ðT l < TÞ

ð4Þ

where

- VOF equation –

@F VFÞ ¼ F_ s þ r  ð~ @t

ð5Þ

2.3. Boundary conditions Table 1 Welding conditions for root pass welding. Wire feed rate Voltage Electrode Current Welding speed CTWD Torch angle Shielding gas

8.2 m/min 25 V YGW15, U = 1.2 mm 260 A 20 mm/s (root pass) 10 mm/s (2nd pass) 20 mm 90 degree 80% Ar-20% CO2 20 l/min

(a) Flat position

The energy boundary condition on the top surface is considered by the arc heat flux and heat dissipation from convection, radiation and evaporation. The energy balance model on the top surface is expressed in Eq. (6).

k

@T !

@n

¼ qA  qconv  qrad  qev ap

ð6Þ

The pressure boundary on the free surface is applied as shown in Eq. (7).

p ¼ pA þ

c Rc

(b) Overhead position

Fig. 1. Weld bead shapes of root pass welding with a high welding speed (20 mm/s).

ð7Þ

947

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 2. Material shape used in the simulations.

Table 2 Properties and constants used in simulations. Symbol

Nomenclature

Symbol

Nomenclature

q

Density (solid:7.8, liquid:6.9, g/cm3) Velocity vector

V I

Welding voltage Welding current

Body force Viscosity, 0.0059 kg/(sm) Mass source of droplet Enthalpy Enthalpy source of droplet

rx ry1 ry2 fb rw

Effective radius in x-direction Left-hand side effective radius in y-direction Right-hand side effective radius in y-direction Droplet frequency (Hz) Wire radius

~ Vs Ts Tl Cs Cl F F_ s

Velocity vector for droplet source, 0.9 m/s

rd

Droplet radius

Solidus temperature, 1768 K Liquidus temperature, 1798 K Specific heat of solid, 7.26  106 erg/g s K Specific heat of liquid, 7.32  106 erg/g s K Fraction of fluid Fraction source of droplet

qd Td To

Heat input from droplet Droplet temperature, 2400 K Room temperature, 298 K Droplet heat efficiency Latent heat of fusion, 2.77  109 erg/g s Permeability of vacuum, 1.26  106 H/m

k

Thermal conductivity Normal vector to free surface

~ V fb

l _s m h h_ s

!

n qA qconv qrad qev ap pA

c Rc

gA

Heat input from arc plasma Heat dissipation by convection Heat dissipation by radiation Heat dissipation by evaporation Arc pressure Surface tension Radius of the surface curvature Arc heat efficiency

2.3.1. Arc heat source model If the seam position is on the groove center in multi-pass V-groove welding, arc forces (arc pressure, electromagnetic force) and droplet impingement may not melt the inclined side of the V-groove surface and induce the lack of fusion weld defect, as shown in Fig. 3(a), therefore, it is necessary to shift the seam position to the left or right, as shown in Fig. 3(b). The resultant arc plasma distribution on the material surface should also change because it is heavily affected by geometric material shapes [15]. Fig. 4 shows the way to apply the Abel inversion method in the 2nd pass V-groove GMA welding from two different CCD camera positions (C1 and C2). To make the arc heat flux model, it is also necessary to calculate the irradiance of the arc plasma in a yellow-dotted line, as shown in Fig. 4 [10]. However, the distribution of the gray level on the dotted line is not axisymmetric, so this study used another Abel inversion method that can be reasonably apply to 2nd pass V-groove welding with the following assumptions:

gd hsl

l0 lm Jz

Material permeability, 1.26  106 H/m Vertical component of the current density

Jr Bh c z J0 J1 da

Radial component of the current density Angular component of the magnetic field Thickness of material Vertical distance from top surface First kind of Bessel function of zero order First kind of Bessel function of first order Coefficient, 0.5

(2) P1 and P3 are symmetrical to each other. (3) P2 and P4 are symmetrical to each other. (4) P1, P2, P3 and P4 are the quarters of the ellipse. This study used an elliptically symmetric Abel inversion method for each quarter (P1, P2, P3 and P4) to get the irradiance of the arc plasma [10]. Then the calculated irradiance was applied to the Fowler-Milne method to obtain the arc heat distribution [12,17]. Finally, the asymmetric arc heat flux model, which includes several effective radii (rx ; ry1 and ry2 ) was used in the simulations, as shown in Fig. 5 and Eq. (8).

if y < arc center of y-direction   2 2 qA ðx; yÞ ¼ gA 2prxVIry av e exp  2xr2  2xr2 x

x

(1) P1, P2, P3 and P4 are the boundary area between the arc plasma and materials.

y1

if y > arc center of y-direction   2 2 qA ðx; yÞ ¼ gA 2prxVIry av e exp  2xr2  2xr2 ; where

ry av e ¼ ðry1 þ2 ry2 Þ.

y2

ð8Þ

948

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 3. Expected weld bead shape along the position of welding seam.

Fig. 4. Process of Abel inversion for 2nd pass GMA welding.

949

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 5. Process of arc heat flux model for 2nd pass GMA welding.

Previous studies [7,15,16] used a total arc heat efficiency of 0.8 which includes the heat transfer by the droplet and arc. The heat efficiency [15] of the droplet and arc can be calculated from Eqs. (9)–(12).

fd ¼

3r 2w WFR ; 4r 3d

ð9Þ

2.3.2. Arc pressure model Previous studies used the same effective radius for the arc heat flux and arc pressure because of the physical relationship [11]; therefore, the effective radii of the arc pressure model in Eq. (13) are the same as that of the arc heat flux model.

if y < arc center of y-direction   2 2 Parc ðx; yÞ ¼ 4p2 rlx0rIy av e exp  2xr2  2rx 2 x

qd ¼

4 3 pr q½C s ðT s  T o Þ þ C l ðT d  T s Þ þ hsl f d ; 3 d

ð10Þ

x

gd ¼

qd ; VI

gA GMAW ¼ 0:8  gd

y1

if y > arc center of y-direction   2 2 Parc ðx; yÞ ¼ 4p2 rlx0rIy av e exp  2xr2  2rx 2

ð13Þ

y2

ð11Þ

2.4. Electromagnetic force

ð12Þ

This study also used the same effective radius as the arc pressure and arc heat flux models due to the physical relationship

950

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

[11]. However, it is necessary to modify the effective radii as shown in Eqs. (14) and (15) because the arc distribution is not axisymmetric. Eqs. (16)–(18) show the current density and self-induced magnetic field distributions, so the electromagnetic force (EMF) model can be calculated in Eqs. (19)–(21)

if y < arc center of y-direction k1 ¼

ry1 rx ;

2

x þ

y2 k21

¼

r 2a ð14Þ

if y > arc center of y-direction k2 ¼

rr ¼

Jz ¼

Jr ¼

ry2 rx ;

x2 þ

y2 k22

I 2p

1

0

Z 0

1

Z

1

0

J 1 ðkr a Þ expðk2 r2a =4dÞ

sinh½kðc  zÞ dk sinhðkcÞ

ð18Þ

F x ¼ J z Bh

x ra

ð19Þ

F y ¼ J z Bh

y ra

ð20Þ

F z ¼ J z Bh

ð21Þ

2.5. Other welding models

ð15Þ

2 Z

lm I 2p

¼ r 2a

rx þ ry av e

I 2p

Bh ¼

kJ 0 ðkr a Þ expðk2 r2r =4da Þ

sinh½kðc  zÞ dk sinhðkcÞ

ð16Þ

kJ 1 ðkr a Þ expðk2 r2r =4da Þ

cosh½kðc  zÞ dk sinhðkcÞ

ð17Þ

The drag force surface tension and, buoyancy force were not affected by the arc plasma distribution; thus, we used the same model as previous studies [11,14–16]. 3. Results and discussion 3.1. Flat position (Case 1) In the longitudinal section (y = 2 mm), as shown in Fig. 6(a) and (b), the molten pool under the arc center started to circulate clockwise due to arc forces (arc pressure, EMF) and

Fig. 7. Temperature profiles and flow patterns on the longitudinal cross section for root pass [15].

Fig. 6. Temperature profiles and streamlines on the longitudinal cross section (y = 2.0 mm) in case 1 (flat position).

Fig. 8. Velocity profiles on the transverse cross section (x = 25 mm) at t = 1.00 s in case 1.

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 9. Temperature profiles and streamlines on the transverse cross section (y = 2.0 mm) in case 1 (flat position).

951

952

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 10. Comparison of simulation result with experiment for case 1.

droplet impingement under the arc center. The molten pool circulated only within 5 mm in the x-direction and it could not deliver to the molten pool tail. However, the molten pool length (19 mm) in the longitudinal section of the 2nd pass GMAW was larger than that of the root pass GMAW molten pool [15], as shown in Fig. 7 This was because the free top surface area of the 2nd pass GMAW molten pool in a transverse section was large to prevent fast conduction cooling through y-direction. In the transverse section, the molten pool started to circulate counterclockwise, and the velocity profiles are described in Fig. 8. Under the arc center (y = 2 mm), the arc forces (arc pressure and EMF) and droplet impingement accelerated the molten pool circulation; therefore, the velocity of molten pool was high (more than 30 mm/s) and these fast molten pool flow patterns could bring dynamic convective heat transfer. Consequently, enough of the V-groove surface melted in Part A. On the other hand, the velocity of molten pool decreased during the counterclockwise circulation so the convective heat transfer in Part B is relatively smaller than Part A; therefore, weld defects such as lack of penetration and lack of fusion can be induced in Part B. Nevertheless, this was not considered seriously because the additional weld bead by 3rd pass should be deposited and the molten pool can melt and penetrate the V-groove surface sufficiently in this region. Fig. 9(a)–(d) shows the temperature profiles and molten pool flow patterns in the transverse section where the molten pool circulated and maintained counterclockwise flow during welding. The molten pool started to solidify from the right hand side V-groove surface because it was farther from the welding seam (y = 2 mm) than left hand side V-groove surface. Additionally, the volume of material (liquid and solid) in a given transverse section decreased as the molten pool solidified because the density of liquid is smaller than that of the solid. After solidification, a convex weld bead can be formed and it is possible to compare the simulation result with the experimental one, as shown in Fig. 10.

3.2. Overhead position (Case 2) Fig. 11(a)–(c) shows the temperature profiles and molten pool flow patterns in the longitudinal section (y = 2 mm) of the overhead position. Similar to case 1, the molten pool under the arc center circulated clockwise, but the molten pool length (26 mm) was longer than case 1 because the gravity effect extracted the molten pool in the z-direction, which can delay the molten pool solidification. From Fig. 12(a)–(d), it is possible to figure out why the solidification of the molten pool should be delayed in the overhead

Fig. 11. Temperature profiles and streamlines on the longitudinal cross section (y = 2.0 mm) in case 2 (overhead position).

position. The molten pool flow patterns near the arc center in case 2 were similar to case 1, as shown in Fig. 12(b) where the molten pool circulated counterclockwise. As the given transverse (y–z) section deviated from the arc forces, and it is far from the arc center, as shown in Fig. 12(c) and (d), the molten pool flowed in the

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 12. Temperature profiles and streamlines on the transverse cross section (y = 2.0 mm) in case 2 (overhead position).

953

954

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 13. Comparison of simulation result with experiment for case 2.

z-direction. Consequently, the solidification was delayed and the reinforcement height can be increased. Fig. 13 compares the simulation result with the experimental one and it was found that a more convex weld bead can be formed in case 2. 3.3. Vertical downward position (Case 3) Fig. 14(a)–(b) shows the temperature profiles and molten pool flow patterns in a longitudinal section (y = 2 mm) of the vertical downward position. The gravity effect pushes the molten pool forward; therefore, the molten pool solidified faster and the molten pool length (15 mm) in a longitudinal direction was shorter than cases 1 and 2. Even though the arc forces suppressed the molten pool under the arc center, the accumulated molten pool volume prevented the droplet from penetrating the weld bead. Instead, the accumulated volume absorbed the momentum of droplet impingement as a cushion; therefore, it is hard to form a deep penetrated weld bead in a vertical downward position. Fig. 15(a)–(d) shows the temperature profiles and molten pool flow patterns in a transverse section (x = 25 mm) of the vertical downward position. Compared to case 1 (Fig. 9(b)) and case 2 (Fig. 12(b)), the volume of the molten pool near the arc center, as shown in Fig. 15(b) was large because the gravity effect pushed the molten pool in the x-direction. As the given transverse (y–z) section was far from the arc center, the volume of molten pool decreased and solidified very quickly, as shown in Fig. 15(c) and (d). Furthermore, it is possible to find that the shape of the top surface weld bead in case 3 was concave while those in the other cases were convex. The process of forming a concave weld bead in case 3 is summarized as follows:

Fig. 14. Temperature profiles and streamlines on the longitudinal cross section (y = 2.0 mm) in case 3 (vertical down position).

z-direction while it circulated counterclockwise. This is because the arc forces cannot suppress the molten pool in this region, but the gravity effect can still extract the molten pool in the

(1) The molten pool near the V-groove surface solidifies earlier than the welding seam (y = 2 mm). (2) While the molten pool is solidifying, the remaining molten pool near the welding seam is affected by the gravity and it flows in the x-direction. (3) As the final remaining molten pool near the welding seam flows in the x-direction, the surface height from the bottom in this region becomes short. (4) A concave top surface weld bead is formed. The mechanism of forming a concave weld bead in vertical a downward position was not indicated before; however, this study could describe this detailed process. Fig. 16 compares the simulation result with the experiment one and the models in the simulation can be validated by these comparisons.

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 15. Temperature profiles and streamlines on the transverse cross section (y = 2.0 mm) in case 3 (vertical down position).

955

956

D.-W. Cho, S.-J. Na / International Journal of Heat and Mass Transfer 88 (2015) 945–956

Fig. 16. Comparison of simulation result with experiment for case 3.

4. Conclusions

References

This paper described the CFD numerical models and the molten pool flow patterns in 2nd pass V-groove GMAW for various welding positions. The results of this work can be summarized as follows:

[1] D.W. Cho, S.J. Na, M.Y. Lee, Expectation of bead shape using non-linear multiple regression and piecewise cubic Hermite interpolation in FCA fillet pipe welding, J. KWJS 27 (2009) 42–48. [2] I.S. Kim, J.S. Son, I.G. Kim, J.Y. Kim, O.S. Kim, A study on relationship between process variables and bead penetration for robotic CO2 arc welding, J. Mater. Process. Technol. 136 (2003) 139–145. [3] K. Lee, H. Sim, J. Kwon, B. Yoon, M. Jeong, M. Park, B. Lee, A prediction of the penetration depth on CO2 arc welding of steel sheet lap joint with filet for car body using multiple regression analysis technique, J. KWJS 30 (2012) 59–64. [4] H.S. Moon, S.J. Na, Optimum design based on mathematical model and neural network to predict weld parameters for fillet joints, J. Manuf. Syst. 16 (1997) 13–23. [5] P. Dutta, D.K. Pratihar, Modeling of TIG welding process using conventional regression analysis and neural network-based approaches, J. Mater. Process. Technol. 184 (2007) 56–68. [6] Z. Cao, Z. Yang, X.L. Chen, Three-dimensional simulation of transient GMA weld pool with free surface, Weld. J. 83 (2004) 169s–176s. [7] J.H. Cho, S.J. Na, Three-dimensional analysis of molten pool in GMA-laser hybrid welding, Weld. J. 88 (2009) 35–43. [8] W.I. Cho, S.J. Na, M.H. Cho, J.S. Lee, Numerical study of alloying element distribution in CO2 laser–GMA hybrid welding, Comput. Mater. Sci. 49 (2010) 792–800. [9] N.S. Tsai, T.W. Eagar, Distribution of the heat and current fluxes in gas tungsten arcs, Metall. Trans. B 16 (1985) 841–846. [10] Y.T. Cho, S.J. Na, Application of Abel inversion in real-time calculations for circularly and elliptically symmetric radiation sources, Meas. Sci. Technol. 16 (2005) 878–884. [11] D.W. Cho, S.H. Lee, S.J. Na, Characterization of welding arc and weld pool formation in vacuum gas hollow tungsten arc welding, J. Mater. Process. Technol. 213 (2013) 143–152. [12] 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. [13] D.V. Kiran, D.W. Cho, W.H. Song, S.J. Na, Arc behavior in two wire tandem submerged arc welding, J. Mater. Process. Technol. 214 (2014) 1546–1556. [14] D.W. Cho, D.V. Kiran, W.H. Song, S.J. Na, Molten pool behavior in the tandem submerged arc welding process, J. Mater. Process. Technol. 214 (2014) 2233– 2247. [15] D.W. Cho, W.H. Song, M.H. Cho, S.J. Na, A study on V-groove GMAW for various welding positions, J. Mater. Process. Technol. 213 (2013) 1640–1652. [16] D.W. Cho, S.J. Na, M.H. Cho, J.S. Lee, Simulations of weld pool dynamics in Vgroove GTA and GMA welding, Weld. World 57 (2013) 223–233. [17] G.N. Haddad, A.J.D. Farmer, Temperature determinations in a free-burning arc. I. Experimental techniques and results in argon, J. Phys. D: Appl. Phys. 17 (1984) 1189–1196.

(a) With a CCD camera and Abel inversion method, it is possible to observe that the arc not maintaining axisymmetric distributions for the 2nd pass V-groove GMAW because of the geometrical limitations. Thus, this study used partially symmetric Gaussian arc models. (b) In transverse sections, the counterclockwise molten pool circulations brought a dynamic convective heat transfer on the left-hand side, while the weld defects such as lack of fusion, and lack of penetration could be induced on the right hand side. (c) The molten pool on the welding seam (y = 2 mm) solidified later than other regions; therefore, the molten pool is heavily affected by the gravity effect. Moreover, the molten pool lengths for various welding positions are different each other. (d) The formation process of concave and convex weld beads has been analyzed in detail. Convex weld beads are found in flat and overhead positions while a concave weld bead can be formed in vertical downward position.

Conflicts of interest None declared.

Acknowledgements The authors gratefully acknowledge the support of the Brain Korea 21 plus project, Korean Ministry of Knowledge Economy (No. 2013-10040108) and Mid-career 363 Researcher Program through NRF (2013-015605).