Journal of Materials Processing Technology 214 (2014) 1710–1720

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

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

Numerical simulation of full penetration laser welding of thick steel

plate with high power high brightness laser
L.J. Zhang a,b , J.X. Zhang b , A. Gumenyuk c , M. Rethmeier c , S.J. Na a,∗
a
Department of Mechanical Engineering, KAIST, Daehak-ro 291, Yuseong-gu, Daejeon 305-701, Republic of Korea
b
State Key Laboratory of Mechanical Behavior for Materials, Xi’an Jiaotong University, Xi’an 710049, China
c
Division 9.3, BAM, Unter den Eichen 87, 12205 Berlin, Germany

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

Article history: Full penetration laser welding was carried out on a 10 mm steel plate using a 16 kW maximum power
Received 3 November 2013 continuous wave thin disk laser. Upper surface and lower surface of molten pool were observed syn-
Received in revised form 22 February 2014 chronously with two high speed CCD cameras during the welding process. The lower surface was much
Accepted 8 March 2014
longer and more unstable than the upper one. A three dimensional laser deep penetration welding model
Available online 15 March 2014
in which volume of fluid (VOF) method was combined with a ray-tracing algorithm was used to simulate
the dynamic coupling between keyhole and molten pool in laser full penetration welding. The calculated
Keywords:
weld cross-section morphology and molten pool length on both upper side and lower side agree well
Full penetration laser welding
Molten pool
with experimental results. Evolution of molten pool in lower side during full penetration laser welding
Keyhole was analyzed, periodical features of energy coupling, molten pool behavior and keyhole dynamics in laser
Thick plate full penetration welding were identified and discussed.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction
In recent years, a new generation of high-power high-brightness
solid state laser marked by fiber laser and thin disk laser has
entered the market. Therefore, thick metal welding technology
using high brightness solid state laser has obtained a great devel-
opment potential overcoming some previous limitations. Kawahito
et al. (2007) obtained a bead with a depth-to-width ratio of about 11
at 130 ␮m spot diameter, which can compete with the bead aspect
ratio produced by electron beam welding. Zhang et al. (2011a) suc-
cessfully carried out single pass full penetration welding of up to
50 mm thick metal plate using high brightness laser. However, laser
full penetration welding of thick plate would have a higher demand
in process control than laser partial penetration welding. Avilov
et al. (2012) observed a complete drop through of the liquid metal
during fiber laser welding of 20 mm thick AlMg3 plate. Bachmann
et al. (2012) pointed out that thermo capillary convection would
be very strong on both upper and lower sides of the specimen dur-
ing full penetration laser welding and the weld pool surfaces would
become destabilized by the large convection velocities. Considering
the increased plate thickness involved in laser welding, it becomes
much more difficult to stably maintain a keyhole formed in laser
∗ Corresponding author. Tel.: +82 42 350 3216; fax: +82 42 350 3210.
E-mail address: sjoona@kaist.ac.kr (S.J. Na).
full penetration welding process, and thus weld defects such as
blowhole, porosity and hot crack become more likely to happen
accordingly.
Experimental study of the coupling dynamics of molten pool
and keyhole with satisfactory details is an extremely difficult
mission. Fabbro (2010) experimentally studied the keyhole dynam-
ics of full penetration welding process and found that different
effects depending on the weld pool dynamics and plume inter-
action strongly disturbed the keyhole stability when the welding
speed was reduced. It is a pity that only thin metallic sheet was
considered in Fabbro’s study. Kawahito et al. (2011) performed
a X-ray transmission in situ observation in 10 kW fiber laser
welding process. They found that with the increasing of weld-
ing speed the keyhole appeared to be expanded backward by
strong ejection of laser-induced plume generated from the front
wall. X-ray transmission in situ observation method is effective
way of studying molten pool and keyhole behavior. However,
the application of X-ray transmission in situ observation is lim-
ited because of the high cost and the possible harm from X-ray
radiation. In order to improve the understanding of laser full
penetration welding process, many research efforts have been
devoted to numerical simulation. Ye and Chen (2002) carried out
a 3D modeling of laser full penetration welding and found that
Marangoni convection played critical role in determining the tem-
perature distribution in the workpiece. Du et al. (2004) simulated
full penetration laser beam welding of titanium alloy and found

http://dx.doi.org/10.1016/j.jmatprotec.2014.03.016
0924-0136/© 2014 Elsevier B.V. All rights reserved.
 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
that melt flow driven by surface tension was responsible for the
formation of “hourglass” cross-section profile. Rai et al. (2008)
created a laser welding model taking into account the enhanced
heat and mass transfer due to turbulence in the weld pool. They
found that convective heat transfer was the main mechanism of
heat transfer in the weld pool and affected the weld pool geom-
etry and the solidification characteristics of welds. Zhang et al.
(2011b) numerically found that the interaction between metallic
vapour and molten melt could be well controlled by using the side
gas flow during laser full-penetration welding. These researches
improved our knowledge of full penetration laser welding. At the
same time, it should be noted that dynamic effects of recoil pres-
sure on molten pool during laser full penetration welding were
neglected or underestimated in the numerical researches intro-
duced above.
Recent years, some numerical models including free surface
tracking or capturing technique have been successfully used to
study the laser partial penetration welding process. Most notice-
able are the VOF method-based laser welding model and the Level
Set Method-based laser welding model. Up to now, VOF method-
based laser welding model has already been employed to study the
temperature field (Wu et al., 2009), keyhole dynamics (Zhao et al.,
2011), melt pool movements (Amara and Fabbro, 2008), porosity
formation (Zhao and DebRoy, 2003), humps formation (Amara and
Fabbro, 2010) and transport phenomena (Zhou et al., 2006) dur-
ing laser welding process. Otto et al. (2011) simulated the laser full
penetration welding of 1 mm stainless steel by using VOF method.
They found that the waves running down the keyhole front was the
reason for keyhole oscillations. They also reported that in the simu-
lations with high welding speed the formation of pores in the weld
seam can be observed. Most recently, Tan et al. (2013) created a
three-dimensional transient model employing the sharp interface
method. They found that the interplay of the multiple reflections
and the plume attenuation played important roles in partial pene-
tration laser welding.
A series of progress in three-dimensional transient modeling
on laser deep welding based on VOF method was made by Na’s
group. Cho and Na (2006) established a three-dimensional laser
welding model in which VOF method was combined with a ray-
tracing algorithm to calculate laser energy deposition on keyhole
wall. Furthermore, Cho and Na (2009) extended their model from
laser welding simulation to laser-arc hybrid welding simulation.
Recently, Cho et al. (2012) improved the ray-tracing algorithm in
their model.
Although better understanding of the underlying scientific prin-
ciples of laser deep penetration welding have already be obtained
based on these numerical studies, there remains plenty of unex-
plored topics. For example, numerical studies of full penetration
laser welding of thick plates which is most widely used in industry
field are really limited. In present study, the VOF-based laser deep
welding model developed by Na’s group is extended from partial
penetration laser welding to full penetration laser welding of thick
plate. The general features of laser full penetration welding process
are simulated and analyzed.
2. Mathematical model
2.1. Assumptions
Assumptions employed in simulation are listed below:
(1) The molten metal is assumed to be laminar, incompressible and
Newtonian fluid.
(2) Power density of laser beam is assumed to be with a Gaussian
distribution.
1711
(3) Evaporation induced recoil pressure was decided by
Clausius–Clapeyron equation.
(4) Inside the blind keyhole, the vapor velocity is assumed to
increase linearly along the depth from zero velocity at the bot-
tom of the keyhole to its calculated velocity at the entrance of
the keyhole.
(5) Inside the upper half of through keyhole, the vapor flows
upward and the vapor velocity is assumed to increase linearly
from zero velocity at half thickness of the plate to its calculated
velocity at the upper entrance. Inside the lower half of through
keyhole, the vapor flows downward and the vapor velocity is
assumed to increase linearly from zero velocity at half thickness
of the plate to its calculated velocity at the lower entrance.
(6) Radiation heat transfer from plasma/metallic vapor to work-
piece is considered at both upper and lower surface of
workpiece.
(7) The buoyancy force is modeled with the Boussinesq approxi-
mation.
2.2. Governing equations and laser welding model
The governing equations describing the melt flow and heat
transfer in the molten pool and the theory of VOF method can be
easily found in literatures (Flow Science, 2008), so they are not
repeated here. A detailed introduction of the laser beam profile,
sub-model of multiple reflections in keyhole, sub-model of Fresnel
absorption, sub-model of recoil pressure, the calculation method
of vapor velocity, the calculation method of shear stress and the
calculation method of the impact of the vapor on opposite keyhole
walls can be found in the previously published papers of Na’s group
(Cho and Na, 2006; Cho et al., 2010, 2012).
2.3. Material properties and boundary conditions
Low carbon steel plate with a thickness of 10 mm was con-
sidered in this study. The physical properties of the material and
coefficients used in the simulation are shown in Table 1. Fig. 1
shows the coordinate system and computational domain used in
the study. The analytic domain was set as 56 mm in length, 26 mm
in width, and 20 mm in height including the void regions at both
top and bottom side for free surface tracking. Totally, 871,396 cells
have been used in the computation, and the size of the minimum
cell is 0.2 mm × 0.2 mm × 0.2 mm, as shown in Fig. 2.
The governing equations are solved using the computational
fluid dynamics code FLOW3D. The energy on both top and bot-
tom free surface is balanced between the laser heat flux, the heat
Table 1
Material properties used in simulation.
Physical properties Value
Density of liquid metal  (kg m−3 ) 6900
Density of solid metal  (kg m−3 ) 7800
Thermal conductivity of liquid kL (W m−1 K−1 ) 26.9
Thermal conductivity of solid kS (W m−1 K−1 ) 32.3
Viscosity  (kg m−1 s−1 ) 0.0059
Surface tension  (N m−1 ) 1.87
Surface tension gradient d/dT (N m−1 K−1 ) −4.3 × 10−4
Specific heat of solid CS (J kg−1 K−1 ) 726
Specific heat of liquid CL (J kg−1 K−1 ) 732
Latent heat of fusion hSL (J kg−1 ) 2.77 × 105
Latent heat of vaporization hV (J kg−1 ) 7.34 × 106
Coefficient of thermal expansion ˇ 1 × 10−5
Liquidus temperature TL (K) 1798
Solidus temperature TS (K) 1768
Boiling temperature TV (K) 2900
Convection heat transfer coefficient h (W m2 K−1 ) 10
Emissivity ε 0.4
1712 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720

Table 2
Disk laser parameters.

Parameters Parameters

Wavelength (␮m) 1.03

Polarization Circular
Beam quality, M2 29.6
Beam quality, BPP (mm mrad) 9.7
Focal length (mm) 300
Beam diameter on optics (mm) 37.3
Focal radius (␮m) 181
Rayleigh length (mm) ±2.2
Focal plane z0 (mm) 0.0

Fig. 1. Schematic of computation domain.

dissipation by convection and radiation, and the heat loss due to

surface evaporation. Cho et al. (2010) assumed that the change of
energy and mass by evaporation could be calculated by consid-
ering that roughly 18% of all evaporating particles return to the
surface and good agreement between simulation and experiment
was observed. In this study, the same assumption is adopted to
evaluate the change of energy and mass induced by evaporation.
The top boundary and bottom boundary (i.e. ADHE and BCGF
in Fig. 1) of computation domain are set as pressure outlet bound-
ary, while the other four boundaries of computation domain (i.e.
ABCD, EFGH, ABFE and DCGH in Fig. 1) are set as continuative
boundary. On vapor/liquid interface, recoil pressure is calculated
and applied on vapor/liquid interface through an user defined
subroutine, while static pressure, dynamic pressure, surface ten-
Fig. 3. Photograph of the experimental facilities.
sion pressure and ambient atmospheric pressure are automatically
treated by FLOW3D itself. Details of computation procedure can be
found in the previously published papers of Na’s group (Cho et al., 4. Results and discussion
2012).
4.1. Comparison between simulation and experiment
3. Experiment
Fig. 4a–f shows the calculated temperature fields and velocity
Laser welds were made autogenously on 10 mm steel plate using fields from a cross-sectional side view. The initial stage of laser full
a 16 kW maximum power disk laser. Laser beam is focused on the penetration welding is similar to that of laser partial penetration
workpiece surface using a 300 mm focal length lens. The focal spot welding. After 50 ms, molten pool depth has already reached about
is approximately 0.36 mm in diameter. The characteristic parame- three fourths of the plate thickness and the keyhole has deeply
ters of the disk laser are listed in Table 2. penetrated into the sheet. After about 200 ms, the steel sheet is
Fig. 3 shows the experimental facilities. Welds with a length of almost completely penetrated.
100 mm were produced with a welding speed of 1 m/min and a laser Fig. 4g is the calculated temperature field at 1.8 s from a cross-
power of 8 kW. Two high speed CCD cameras were used to take the sectional side view, of which the temperature range is from 1768 K
images of molten pool surface on both the upper and lower side. to 2900 K. It can be seen from Fig. 4g that the temperature of
The capturing speed is 1000 shoots a second. some parts of keyhole surface where the laser beam irradiates,

Fig. 2. The mesh used in the simulation.

 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720 1713

Fig. 4. Calculated temperature fields and velocity fields in cross-sectional side views.

reaches the evaporation temperature and accordingly experiences As a whole, the simulation results and experimental results agree
the recoil pressure. well with each other.
In Fig. 4, it is noted that the length of molten pool surface in
lower side is much larger than that in upper side. Fig. 5 shows the 4.2. Extreme spreading in bottom side
variation of molten pool length in the whole welding process. It
can be seen from Fig. 5 that the length of upper surface of molten It can be seen from Fig. 9a–e that volume of the melt in lower
pool becomes relatively stable after about 0.9 s, while the length part continues to increase in the initial stage of laser full penetra-
of lower surface of molten pool continues to increase until about tion welding. Theoretically, this could be related to the influence of
1.4 s. In the quasi steady stage, the calculated length of molten pool gravity, the action of recoil pressure and the difference of surface
is about 7.8 mm in upper side and about 15 mm in lower side.
Fig. 6 shows a comparison between simulation and experiment.
Fig. 6a shows the computed weld pool geometry at t = 1.8 s, while
the experimentally captured weld pool surface in upper side and
lower side are shown in Fig. 6b and c respectively. According to
Fig. 6b and c, the observed molten pool length is about 7 mm in
upper side and about 13 mm in lower side, respectively. So the
simulation results are reasonable. It can also be seen from Fig. 6b
and c that during full penetration laser welding the lower surface
of molten pool is more unstable than the upper surface of molten
pool.
Calculated temperature fields at 1.8 s are displayed in Fig. 7 in
the front views. Fig. 8 shows the comparison of cross-sectional
fusion zone between simulation and experiment. In Fig. 8a and b,
the dashed lines are the boundary of experimentally decided fusion
zone. According to Fig. 8, the simulated top width of fusion zone
agrees well with experimental result, while the simulated bottom
width of fusion zone is a little larger than the experimental result.

20
Length of melt pool surface / mm

unsteady quasi-steady
upper surface
16 lower surface

12

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Time / sec

Fig. 5. Simulation results of molten pool length evolution. Fig. 6. Comparison of molten pool length between simulation and experiment.
1714 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
Fig. 7. Calculated temperature fields in cross-sectional front views at 1.8 s.
tension between top surface and lower surface. But in this subsec-
tion the role of flow pattern of the lower part of molten pool is
mainly discussed.
It is no doubt that the melt tends to dropdown under the effect of
gravity and the effect of recoil pressure on the bottom of a blind key-
hole. Difference of surface tension between top surface and lower
surface may also contribute to the continuous volume increase in
the lower part. Fig. 9b–e reveals that temperature of the bottom
surface is higher than that of the top surface of molten pool, which
indicates that surface tension is less at lower surface than at the
upper surface. This may partially explain why the lower surface of
molten pool is more unstable than the upper surface.
Fig. 10 shows the dynamic fluctuation of keyhole radius and
keyhole depth during the initial stage of laser full penetration
welding. Fig. 10 reveals that through keyhole can be formed inter-
mittently in full penetration welding process. When the molten
pool is fully penetrated, the melt near lower exit is driven by recoil
pressure and flows toward the rear-bottom part of molten pool,
as shown in Fig. 9d. At the rear-bottom part of molten pool, the
overheated melt moving backward will be blocked by solid/liquid
interface and then flows upward, as shown in Fig. 9e–g. Finally, a
clockwise vortex is formed in the lower part of molten pool and this
flow pattern, i.e. clockwise vortex, helps transport the over-heated
melt to the rear-bottom part of the molten pool intermittently but
endlessly. Hence, the molten pool is stretched and gets longer and
longer in the lower part. Fig. 9 shows that the upward flowing of
melt is gradually weakened with the spreading of molten pool in
the lower side, as indicated by the arrowed curves in Fig. 9e–g. So
the molten pool length at half thickness decreases gradually and
the final shape of molten pool is as shown in Fig. 9h.
4.3. Periodic fluctuation of molten pool
Fig. 11 shows the transient evolution of keyhole and molten
pool from 1.702 s to 1.733 s. It seems that the transient behavior
of molten pool and keyhole in quasi steady stage of laser full pen-
etration welding has an obviously periodic feature. A fluctuation
cycle of molten pool and keyhole may consist of four typical stages
and the corresponding characteristic signs are as follows: (a) devel-
opment of keyhole into molten pool (i.e. Fig. 11a, e, i and m); (b)
formation of through keyhole (i.e. Fig. 11b, f, j and n); (c) melt flow-
ing toward keyhole (i.e. Fig. 11c, g, k and o); (d) keyhole collapse
(i.e. Fig. 11d, h, l and p). Fig. 12a–d shows the variation of absorbed
laser energy along keyhole depth at the moment of 1.702 s, 1.705 s,
1.706 s and 1.708 s respectively. The instantaneous change of key-
hole size and laser energy coupling in quasi steady stage of laser
full penetration welding are given in Figs. 13 and 14, respectively.
From 1.702 s to 1.733 s, keyhole experienced about four cycles of
fluctuation and a cycle of fluctuation takes about 8 ms.
As shown in Fig. 12a, the energy absorbed by the lower part
of a blind keyhole is significantly greater than that absorbed by
the other part. Therefore, the melt below keyhole bottom will flow
 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720 1715

Fig. 8. Comparison of weld cross-section between (a) experiment and (b) simulation.

downward under the action of recoil pressure, as shown in Fig. 11a,
e, i and m, and resultantly, the keyhole will develop downward into
the molten pool rapidly.
When the through keyhole is just formed, the melt around
keyhole still flows away from keyhole and keyhole size keeps
increasing, as shown in Fig. 11b, f, j and n. A comparison between
Fig. 12a and b shows that the energy absorbed by the lower part
of keyhole wall decreased when a through keyhole is formed. In
addition, the laser power irradiated on workpiece decreases and
become less than the output power (i.e. 8 kW) when the workpiece
is fully penetrated, as shown in Fig. 14. This can be understood as
that a part of laser beam will directly pass through the through
keyhole, without any interaction with keyhole wall, and escape
from the lower exit of keyhole. It can also be seen from Fig. 14 that
decrease in the laser energy irradiated on workpiece will result in
the decrease of absorbed energy. As the absorbed energy decreases,
the recoil pressure on keyhole wall will decrease. At the same time,
the keyhole expansion leaded to the humping of molten pool sur-
face at both upper and lower side. Accordingly the effect of surface
tension which is closely related to configuration of molten pool sur-
face and the diameter of keyhole is gradually strengthened. Finally
the recoil pressure cannot continue to support the expansion of
keyhole, and the melt will flow toward the keyhole, as shown in
Fig. 11c, g, k and o. Decrease of diameter of the middle part of

Fig. 9. Simulation results of the flow pattern of lower part in cross-sectional side views.
1716 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
Fig. 10. Fluctuation of keyhole radius and keyhole depth in initial stage.
keyhole will lead to the increase the energy coupling there, as
shown in Fig. 12b and c.
Then, under the effect of surface tension, melt flows toward
the keyhole axis and eventually lead to the collapse of keyhole,
as shown in Fig. 11d, h, l and p. After that, the liquid metal located
in the beam path will be heated by laser beam and the molten pool
Fig. 11. Calculated velocity fields and temperature profiles in a cross-sectional side views.
surface will be depressed by recoil pressure. Soon a blind keyhole
will be formed in the molten pool and energy deposition at key-
hole bottom again increases greatly as shown in Fig. 12d, similar
to those presented in Fig. 12a. During the period from 1.702 s to
1.733 s, the fluctuation process discussed above totally repeated
about four times.
In Fig. 15g, point A, B, C and D represent four typical locations
in molten pool. Fig. 15a–f shows the simulation results of veloc-
ity of these four points during the period from 1.702 s to 1.733 s.
It is shown in Fig. 15 that the transient behavior of molten pool
has an obvious cyclical characteristic. According to Figs. 11 and 15,
the molten pool experienced about four cycles of fluctuation from
1.702 s to 1.733 s. It can also bee seen from Fig. 15 that not only the
magnitude of fluid velocity but also the direction of fluid flow may
change in full penetration laser welding.
4.4. Flowing patterns in one cycle
This section will focus on the flow patterns of quasi steady
molten pool in a typical cycle. Fig. 16 shows the transient evolu-
tion of molten pool and keyhole during a typical cycle of molten
pool fluctuation. From 1.7154 s to 1.7240 s, molten pool and key-
hole experienced a complete cyclical fluctuations: the formation of
blind keyhole, as shown in Fig. 16a; development of blind keyhole
in molten pool, as shown in Fig. 16b; formation of through keyhole,
as shown in Fig. 16c; backflow of melt toward keyhole, as shown in
 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
Fig. 12. Distribution of laser energy on keyhole wall.
Fig. 13. Fluctuation of keyhole radius and keyhole depth during quasi steady stage.
Fig. 14. Dynamics of energy coupling during the quasi steady stage.
1717
Fig. 16d; collapse of keyhole, as shown in Fig. 16e; once again the
formation of blind keyhole, as shown in Fig. 16f.
The calculated streamlines during the period from 1.7154 s to
1.7240 s is shown in Fig. 17. The flow patterns of molten pool in
one cycle are identified based on the streamlines shown in Fig. 17
and demonstrated in Fig. 16 by the arrowed curves. In Fig. 16, “S”
represents that the fluid flow is caused by surface tension, while
“R” represents that the fluid flow is dominated by recoil pressure.
From the simulation results shown in Fig. 16, it can be found that
very little melt exists in front of keyhole. Accordingly the flow
behavior of melt behind the keyhole is mainly discussed here. Dur-
ing laser full penetration welding process, factors influencing the
dynamic behavior of molten pool mainly include surface tension
which would close keyhole and recoil pressure which helps main-
tain keyhole.
Pressure difference across air/liquid interface is proportional to
the surface tension coefficient, , and inversely proportional to the
effective radius of the interface. When the rear part of melt pool
surface heaves up, which happens periodically in welding process,
the effective radius there greatly decreases and the pressure differ-
ence across melt pool surface sharply increases. Under this surface
tension induced pressure, the liquid metal in the rear part of melt
pool will flow toward the keyhole. Under the action of recoil pres-
sure, keyhole will expand and the melt around keyhole will flow
away from keyhole. In the quasi steady stage of laser full pene-
tration welding process, the laser energy coupling efficiency and
the shape of molten pool surface change periodically. Therefore,
both surface tension and recoil pressure will vary periodically.
The flow field of molten pool is finally decided by the outcome
of dynamic competition between recoil pressure and surface ten-
sion.
At 1.7154 s, the keyhole formed in the previous cycle collapsed.
As shown in Fig. 16a, a new blind keyhole is formed under the action
of recoil pressure and the melt around keyhole is flowing away from
keyhole, as indicated by the arrow curves R1, R2 and R3 in Fig. 16a.
It is shown in Fig. 16a that in the region far away from the newly
formed blind keyhole, flow pattern of molten pool is dominated by
surface tension, which can be described by the arrowed curves S1,
S5, S6, S7, S8, S9 and S10 in Fig. 16a. In the upper part of molten pool,
1718 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720

Fig. 15. Calculated velocity evolution curves (dimension of x- and z-axes: cm).

Fig. 16. Calculated flow patterns from a cross-sectional side view in one cycle.
 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
Fig. 17. Calculated streamlines from a cross-sectional side view in one cycle.
a counterclockwise vortex (i.e. as indicated by the arrowed curves
S1 + S2 in Fig. 16a) is formed. Under the effect of surface tension, a
part of the melt from rear-bottom portion of molten pool will flow
upward and finally enter into the upper portion of molten pool, as
indicated by the arrowed curves S8 + S3 in Fig. 16a. At the same
time, some melt from the rear-bottom portion of molten pool will
pass by the front solid/liquid interface and then flow downward due
to the principle of continuity, as indicated by the arrowed curve S4
in Fig. 16a. Finally, a clockwise vortex as indicated by the arrowed
curves S9 + S4 in Fig. 16a is formed in the lower part of molten pool.
In Fig. 16b, arrowed curves S1, S5 and S6 are given to show the
driving effect of surface tension at 1.7176 s. With the expanding
of keyhole, melt around keyhole flows away from keyhole (i.e. as
indicated by the arrowed curves R1, R2, R3 and R4 in Fig. 16b).
The difference between the arrowed curve S3 in Fig. 16a and the
arrowed curve S3 in Fig. 16b is obviously caused by the expanding
of keyhole. It seems that with the expanding of keyhole, the melt
flowing from left to right is restricted, while the upward flowing of
the melt driven by surface tension is promoted (i.e. as indicated by
the arrowed curve S3 in Fig. 16b).
Fig. 16c is the calculated results at 1.7192 s. At 1.7192 s a through
keyhole is formed in molten pool. As mentioned, formation of
through keyhole will lead to decline of energy coupling efficiency.
Moreover, molten pool surface is raised due to keyhole expand-
ing, which would enhance the role of surface tension. Therefore,
melt will flow from the rear part of molten pool to the front part of
molten pool, as indicated by arrowed curves S2, S4, S5 S6 in Fig. 16d
and S2, S4, S5, S11 in Fig. 16e, under the effects of surface tension
and finally results in the keyhole collapse.
The fluid flow caused by surface tension is indicated by arrowed
curves S1, S3, S7, S8, S9 and S10 in Fig. 16d and e. Fig. 16f shows the
calculation results of flow field and temperature field at 1.7240s.
When the keyhole collapsed, laser beam will be blocked by molten
pool surface and extensive evaporation of metal will occur at the
upper surface of molten pool. The molten pool surface around the
laser beam path will be depressed rapidly under the action of recoil
pressure (as indicated by the arrowed curves R1, R2 and R3 in
Fig. 16f). After that, a blind keyhole will be formed in molten pool
and the next cycle of molten pool fluctuation will start.
1719
Typical dwell time considering 0.36 mm spot diameter and
1 m/min weld speed is about 22 ms and considering the average
keyhole diameter this will be more than 50 ms. Thus, one would
expect several cycles of partial/full penetration welding during the
dwell time as the cycle time is about 8 ms. So this cyclic process may
have substantial effects on the escaping of bubbles from molten
pool and the mixing of alloy elements in molten pool. Formation of
humping weld may also relate to the cyclic process demonstrated
in Figs. 11, 15 and 16. In addition, molten pool is stretched in both
upper and lower side due to the cyclic process, which may influence
the shape of solid/liquid interface and therefore influence direction
of crystal growth and properties of welded joint. Further study on
the influence of process parameters on cyclic behavior of molten
pool in laser full penetration welding is necessary.
5. Conclusions
(1) The agreement between experimental results and the corre-
sponding simulations shows that the three dimensional model
presented here is pertinent for a description of laser full pene-
tration welding process.
(2) In full penetration welding, lower surface of molten pool may
be longer and more unstable than upper surface.
(3) Formation of through keyhole lead to that a part of laser
beam directly passes through the keyhole and escapes from
the lower exit without any contact with metal, which is one
of the important differences between partial penetration and
full penetration laser welding.
(4) Transient evolution of molten pool and keyhole in quasi steady
stage of laser full penetration welding has an obviously periodic
feature. The transient flow field of molten pool is the outcome
of the dynamic competition between recoil pressure and sur-
face tension. Under the conditions considered in this work one
typical cycle of molten pool fluctuation takes about 8 ms.
Acknowledgements
This work was supported by BK21 program of Korea,
Mid-career Researcher Program through NRF (Grant no.
1720 L.J. Zhang et al. / Journal of Materials Processing Technology 214 (2014) 1710–1720
2013R1A2A1A01015605), National Natural Science Founda-
tion of China (Grant no. 51275391) and Alexander von Humboldt
Stiftung of Germany.
References
Otto, A., Koch, H., Leitz, K.H., Schmidt, M., 2011. Numerical simulations – a versatile
approach for better understanding dynamics in laser material processing. Phys.
Proc. 12, 11–20.
Amara, E.H., Fabbro, R., 2008. Modelling of gas jet effect on the melt pool movements
during deep penetration laser welding. J. Phys. D: Appl. Phys. 41, 055503.
Amara, E.H., Fabbro, R., 2010. Modeling of humps formation during deep-penetration
laser welding. Appl. Phys. A 101, 111–116.
Avilov, V.V., Gumenyuk, A., Lammers, M., Rethmeier, M., 2012. PA position full pene-
tration high power laser beam welding of up to 30 mm thick AlMg3 plates using
electromagnetic weld pool support. Sci. Technol. Weld. Join. 17, 128–133.
Bachmann, M., Avilov, V.V., Gumenyuk, A., Rethmeier, M., 2012. Numerical simu-
lation of full-penetration laser beam welding of thick aluminium plates with
inductive support. J. Phys. D: Appl. Phys. 45, 035201.
Cho, J.H., Na, S.J., 2006. Implementation of real-time multiple reflection and Fresnel
absorption of laser beam in keyhole. J. Phys. D: Appl. Phys. 39, 5372.
Cho, J.H., Na, S.J., 2009. Three-dimensional analysis of molten pool in GMA-laser
hybrid welding. Weld. J. 88, 35–43.
Cho, W.I., Na, S.J., Cho, M.H., Lee, J.S., 2010. Numerical study of alloying ele-
ment distribution in CO2 laser/GMA hybrid welding. Comput. Mater. Sci. 49,
792–800.
Cho, W.I., Na, S.J., Thomy, C., Vollertsen, F., 2012. Numerical simulation of molten
pool dynamics in high power disk laser welding. J. Mater. Process. Technol. 212,
262–275.
Du, H.B., Hu, L.J., Liu, J.H., Hu, X.Y., 2004. A study on the metal flow in full penetration
laser beam welding for titanium alloy. Comput. Mater. Sci. 29, 419–427.
Fabbro, R., 2010. Melt pool and keyhole behaviour analysis for deep penetration
laser welding. J. Phys. D: Appl. Phys. 43, 445501.
Flow Science, I., 2008. FLOW 3D User Manual V9.2.
Kawahito, Y., Matsumoto, N., Abe, Y., Katayama, S., 2011. Relationship of laser
absorption to keyhole behavior in high power fiber laser welding of stainless
steel and aluminum alloy. J. Mater. Process. Technol. 211, 1563–1568.
Kawahito, Y., Mizutani, M., Katayama, S., 2007. Elucidation of high-power fibre laser
welding phenomena of stainless steel and effect of factors on weld geometry. J.
Phys. D: Appl. Phys. 40, 5854.
Rai, R., Kelly, S.M., Martukanitz, R.P., DebRoy, T., 2008. A convective heat-transfer
model for partial and full penetration keyhole mode laser welding of a structural
steel. Metall. Mater. Trans. A 39, 98–112.
Tan, W., Bailey, N.S., Shin, Y.C., 2013. Investigation of keyhole plume and molten pool
based on a three-dimensional dynamic model with sharp interface formulation.
J. Phys. D: Appl. Phys. 46, 055501.
Wu, C.S., Wang, H.L., Zhang, Y.M., 2009. Numerical analysis of the temperature pro-
files and weld dimension in high power direct-diode laser welding. Comput.
Mater. Sci. 46, 49–56.
Ye, X.H., Chen, X., 2002. Three-dimensional modelling of heat transfer and fluid flow
in laser full-penetration welding. J. Phys. D: Appl. Phys. 35, 1049.
Zhang, L.J., Zhang, J.X., Zhang, G.F., Bo, W., Gong, S.L., 2011a. An investigation on
the effects of side assisting gas flow and metallic vapour jet on the stability of
keyhole and molten pool during laser full-penetration welding. J. Phys. D: Appl.
Phys. 44, 135201.
Zhang, X.D., Ashida, E., Tarasawa, S., Anma, Y., Okada, M., Katayama, S., Mizutani, M.,
2011b. Welding of thick stainless steel plates up to 50 mm with high brightness
lasers. J. Laser Appl. 23, 022002.
Zhao, H., DebRoy, T., 2003. Macroporosity free aluminum alloy weldments through
numerical simulation of keyhole mode laser welding. J. Appl. Phys. 93,
10089–10096.
Zhao, H.Y., Niu, W.C., Zhang, B., Lei, Y.P., Kodama, M., Ishide, T., 2011. Modelling
of keyhole dynamics and porosity formation considering the adaptive keyhole
shape and three-phase coupling during deep-penetration laser welding. J. Phys.
D: Appl. Phys. 44, 485302.
Zhou, J., Tsai, H.L., Lehnhoff, T.F., 2006. Investigation of transport phenomena and
defect formation in pulsed laser keyhole welding of zinc-coated steels. J. Phys.
D: Appl. Phys. 39, 5338.