Materials and Design 63 (2014) 519530

Contents lists available at ScienceDirect

Materials and Design

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

A comparative study on welding temperature elds, residual stress

distributions and deformations induced by laser beam welding and CO2
gas arc welding
Jiamin Sun, Xiaozhan Liu, Yangang Tong, Dean Deng
College of Materials Science and Engineering, Chongqing University, No. 174 Shazhengjie, Shapingba, Chongqing 400045, China

a r t i c l e

i n f o

Article history:
Received 22 April 2014
Accepted 24 June 2014
Available online 1 July 2014
Keywords:
Thin plate
Finite element method
Numerical simulation
Geometrical nonlinearity
Welding distortion

a b s t r a c t
Welding-induced distortion in thin-plate structure is a serious problem which not only hinders the
assembling process but also negatively affects the performance of product. Therefore, how to control
welding deformation is a key issue both at design stage and at manufacturing stage. During welding process, there are a number of factors which can signicantly affect manufacturing accuracy. Among these
factors, the heat input is one of the largest contributors to the nal deformation. Generally, when laser
beam welding (LBW) is used to join parts the total heat input is far less than that used in a conventional
welding method such as gas metal arc welding, so it is expected that LBW can signicantly reduce welding distortion especially for thin-plate joints. As a fundamental research, we investigated the welding
deformations in low carbon steel thin-plate joints induced by LBW and CO2 gas arc welding by means
of both numerical simulation technology and experimental method in the current study. Based on the
experimental measurements and simulation results, we quantitatively compared the welding deformation as well as residual stress induced by LBW and those due to CO2 gas arc welding. The results indicate
that the out-of-plane deformation of thin-plate joint can be largely reduced if CO2 gas arc welding
method is replaced by LBW. Moreover, the numerical results indicate that the residual stresses induced
by LBW are superior to those produced by CO2 gas arc welding both in distribution and in magnitude.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
In automobile industry, it is a current trend in design to adopt
thin-plate steel components to achieve weight reduction and fuel
savings. Meanwhile, the designers have been forced to use thinner
steel or light alloy structures to reduce topside weight for improving fuel economy, and enhance mission capability in shipbuilding
industry [1]. Fusion welding technology has been extensively
applied to assemble parts both in automobile and in shipbuilding
industries owing to its high productivity and exibility for design.
However, when a thin-plate structure is welded by traditional
fusion welding processes such as gas metal arc welding (GMAW)
and shielded metal arc welding (SMAW), it can be foreseen that
welding-induced deformation will be a serious problem because
of the relatively small stiffness. If the heat input cannot be reasonably controlled, signicant deformation even buckling distortion is
apt to occur especially in the welded structure whose thickness is
only a few millimeters [2]. In many cases, additional costs and
schedule delay are incurred from straightening welding distortion.
Corresponding author. Tel.: +86 23 6512 2079; fax: +86 2365127340.
E-mail address: deandeng@cqu.edu.cn (D. Deng).
http://dx.doi.org/10.1016/j.matdes.2014.06.057
0261-3069/ 2014 Elsevier Ltd. All rights reserved.

On the other hand, increasingly, the design of engineering components and structure relies on the achievement of small tolerance.
Therefore, controlling welding-induced deformation has become
of critical importance.
Welding deformation can be controlled both at design stage and
at manufacturing stage [3]. In principle, distortion in a welded
structure can be minimized through designing reasonable positions for the joints and/or by adjusting the thickness of plate at
the design stage. Meanwhile, controlling heat input and adopting
appropriate assembling sequence can reduce the welding-induced
deformation to some extent at the manufacturing stage. Recently,
the welding processes with high energy density such as laser beam
welding [4] and laser-arc hybrid welding [5] have been introduced
into manufacturing industry for both improving productivity and
reducing welding distortion.
LBW has many advantages over conventional fusion welding
methods such as low heat input per unit length, small heat affected
zone, high speed and non-contact welding, deep penetration, effective integration with industrial robots and the capability of joining
materials by single side access [3,4,6]. On the other hand, LBW
involves several complex phenomena like the formation of a keyhole, ionization and vaporization of material, circulation of molten

520

J. Sun et al. / Materials and Design 63 (2014) 519530

metal within the weld pool due to buoyancy and Marangoni forces
[7], solidication at the liquidsolid interface, and so on. Thus, it is
not easy to simulate LBW process with considering all the above
factors through using a single numerical model. However, various
simplications are admissible in numerical simulation of welding
process with minimal loss in accuracy [8]. Han et al. [9] investigated numerically the temperature eld in a thin-plate stainless
steel AISI304 joint caused by LBW. Their results suggested that
the temperature grads are very large in the fusion zone and its
vicinity. Daniel and his co-workers [10,11] studied the distortion
and residual stress distribution induced by LBW in a thin-plate
AA6065T4 joint by means of both numerical simulation technology
and experimental method. Their research indicated that the simulated results matched the experimental measurements well.
As for welding deformation of thin-plate joint induced by traditional arc welding process, Wang and his co-workers [12] studied
the characteristics of welding distortion in a bead-on-plate weld
performed by metal inert gas welding process. Deng et al. [13]
investigated the inuence of heat input on welding distortion in
a thin-plate joint induced by CO2 gas arc welding by means of both
thermo-elasticplastic FEM and experimental method. Recently,
several experimental mock-ups and numerical models have been
used to study welding residual stresses and deformations in thinplate joints, however, very limited literatures have quantitatively
compared the deformation and residual stress distribution induced
by LBW with those induced by a traditional fusion welding method
such as CO2 gas arc welding in a thin-plate joint or welded
structure, especially in a weldment whose thickness is less than
3.0 mm [14]. Theoretically, it is easy to qualitatively distinguish
the welding distortion induced by LBW and that due to CO2 gas
arc welding in a given joint. However, to help choosing a reasonable welding method with balancing manufacturing cost and
dimensional accuracy, it is necessary to quantitatively clarify the
differences between the welding distortion and residual stress
caused by LBW and those produced by CO2 gas arc welding. In
addition, the comparison can provide fundamental knowledge for
welding engineers.
As a fundamental research, the objective of the present study
was to compare the welding deformation and the residual stress
distribution induced by LBW with those due to CO2 gas arc welding
in a thin plate joint. First, experiments were carried out to measure
the transverse shrinkages and the deection (out-of-plane deformation) induced by LBW and CO2 gas arc welding. Then, based
on the commercial software MSC.Marc code [15], a Thermo-elastic-plastic nite element method with considering moving heat
source, material nonlinearity and geometrical nonlinearity was
developed to simulate welding temperature eld, residual stress
distributions and distortions for thin-plate joints. In the developed
computational approach, we made certain efforts to design moving
heat sources and to select their corresponding parameters for modeling the heat inputs of both LBW and CO2 gas arc welding. Moreover, we claried the difference between the welding deformation
predicted by large deformation theory and that computed by small
deformation theory in a low carbon steel thin-plate joint performed by LBW. Finally, we quantitatively compared the welding
deformation and the residual stress distribution induced by LBW
and those caused by CO2 gas arc welding in thin-plate joints with
2.3 mm thickness. It is hoped that the present study will provide
both designers and practicing engineers with a method to control
welding distortion.

deformations in two thin-plate joints. The base metal is low carbon

steel (Q235), and its chemical components are shown in Table 1.
The dimension of each weld specimen is 300 mm  100 mm 
2.3 mm. All of the specimens were welded without any external
restraint. In the LBW, the laser beam was used to only melt the
plate, and no ller metal was added to the plate. The laser beam
welding system used in the current study is shown in Fig. 1, and
the welding parameters used in the experiment are summarized
in Table 2. The joint type performed by CO2 gas arc welding is a
bead-on joint, and the wire is YGW16 with 1.2 mm diameter. The
CO2 gas arc welding was performed by a Panasonic TAWERS
TA-1400 arc welding robot as shown in Fig. 2, and the welding conditions are shown in Table 3.
To measure the transverse shrinkage, six holes were drilled in
each specimen, and their locations are shown in Fig. 3. Here, we
take hole a and hole b (location 1) which are a pair as an example
to explain how to measure the transverse shrinkages on both the
top surface and the bottom surface. Before welding, the distances
between hole a and hole b on the top surface and the bottom

Table 1
Chemical composition of Q235 (low carbon steel).
Material

Si

Mn

Cr

Fe

Q235

0.17

0.26

0.46

0.007

0.009

0.02

Bal.

(a) Laser welding machine

2. Experimental procedure
To compare the welding deformations induced by LBW and CO2
gas arc welding, we carried out experiments to measure the

(b) Weld specimen and torch of laser machine

Fig. 1. Laser beam welding system. (a) Laser welding machine. (b) Before welding.

521

J. Sun et al. / Materials and Design 63 (2014) 519530

Table 2
Laser beam welding conditions.
Parameter

Power (W)

Welding speed (m/min)

Efciency

Nominal heat input (kJ/mm)

Shielding gas

Shielding gas ow rate (L/mm)

Focus length (mm)

Value

2400

1.8

0.8

0.08

Ar

10

200

Fig. 3. Locations of transverse shrinkage measuring points (mm).

Fig. 2. CO2 gas arc welding robot system.

surface were measured by a vernier caliper, respectively. The

distance on the top surface was recorded as lab1, and that on the
0
bottom surface as lab1 . After welding, these two distances were
0
measured again, and their values were recorded as lab2 and lab2 ,
respectively. The transverse shrinkage on the top surface (Dlab)
0
and that on the bottom surface Dlab can be calculated by Eqs.
(1) and (2), respectively.

Dlab lab1  lab2

0

Dlab lab1  lab2

Meanwhile, the out-of-plane deformation of each specimen

after welding was measured by a three-coordinates measuring system [16]. The deection distributions along the edges of the thin
plates were obtained through using the coordinates in z direction
before and after welding.

from the weld center. As the heat input used in LBW is smaller than
that used in CO2 gas arc welding, a ner mesh is required in such
situation. The length of each element along the welding direction
is 1.5 mm in LBW, and that is 2.5 mm in CO2 gas arc welding.
The total number of node is 43,617, and the total number of element is 36,800 in the LBW model. The total number of node is
22,773, and the total number of element is 17,644 in the CO2 gas
arc welding model. These two nite element models are shown
in Figs. 4 and 5, respectively. The sizes of these nite element models are the same as those of the mock-ups. It should be stressed
that the length of welding line is 290 mm in the CO2 gas arc welding model. The element type used in the thermo-mechanical coupled analysis is type 7 (3-D brick element) [15].
3.1. Heat source and thermal analysis
In the present study, transient nonlinear heat transfer analysis
with given welding conditions were performed by using full three

3. Finite element analysis

Based on MSC.Marc code [15], a thermo-elasticplastic nite
element method was developed to simulate welding temperature
elds, residual stress distributions and deformations induced by
LBW and CO2 gas arc welding in low carbon steel thin-plate joint.
The welding temperature eld was rstly computed according to
the given welding conditions. Then, the temperature history of
each node was introduced into the mechanical analysis as thermal
load to calculate the displacement of each node, and strain and
stress of each element. In order to capture the geometrical nonlinearity behaviors in the thin-plate joints, besides the material nonlinearity, the large deformation theory was also incorporated into
the mechanical model.
In order to balance the computing time and the calculation
accuracy, ner meshes were designed at the weld zone and its
vicinity, while the mesh size gradually increased with the distance

Boundary
condition

Weld bead

n
ectio
g dir
n
i
Y
d
l
We

Z
X

Fig. 4. Finite element model and boundary conditions for LBW.

Table 3
CO2 gas arc welding conditions.
Parameter

Current (A)

Voltage (V)

Welding speed (m/min)

Efciency

Nominal heat input (kJ/mm)

Shielding gas

Shielding gas ow rate (L/mm)

Value

107

17.9

0.5

0.8

0.23

CO2

10

522

J. Sun et al. / Materials and Design 63 (2014) 519530

Boundary
condition

Weld bead

ectio

g dir

in
Weld

Z
Fig. 7. Conical heat source model.

Fig. 5. Finite element model and boundary conditions for CO2 gas arc welding.

dimensional (3-D) nite element models with moving heat source

models. The temperature cycle during the entire welding process
at each node and the transient temperature distribution within
the whole range of model are acquired by solving the following
non-linear transfer governing equation:

qc

@T
x; y; z; t r  qx; y; z; t Qx; y; z; t
@t

where q is the density of the materials (g/mm3), c is the specic

heat capacity (J/(g C)), T is the current temperature (C), q is the
heat ux vector (W/mm2), Q is the internal heat generation rate
(W/mm3), x, y and z are the coordinates in the reference system
(mm), t is the time (s), and r is the spatial gradient operator. The
non-linear isotropic Fourier heat ux constitutive equation was
employed:

q krT

where k is the temperature-dependent thermal conductivity (J/

(mm s C)).
Generally, the heat source model is considered to be an important aspect for the welding thermal analysis, especially when the
thickness of joint is small [13]. For the numerical simulation of
LBW process, the emphasis of this work was placed on how to
accurately consider the thermo-mechanical behaviors during the
welding. Thus, to simplify the thermal analysis some detailed phenomena such as formation of keyhole and heat ow in the weld
pool were neglected. In fact, these neglected factors have no significant inuence on the simulation accuracy of welding residual
stress and distortion [17]. However, the penetration shape produced by the LBW was carefully taken into account. In the present
study, we developed a combined heat source model to simulate the
welding temperature eld produced by LBW. In the combined heat
source model, one part is a half ellipsoid volumetric heat source
with uniform density as shown in Fig. 6, and the other part is a
3-D conical heat source with Gaussian distribution as shown in
Fig. 7. The combined heat source model is schematically shown
in Fig. 8. The heat ux qh(x, y, z) in the half ellipsoid volumetric

Fig. 6. Half ellipsoid volumetric heat source model.

Fig. 8. Schematic representation of combine source model. (a) Boundary of heat

source (3D image). (b) Cross-section of combined heat source (2D view).

heat source with uniform density can be expressed by the following equation:

qh Q H =V H

where QH is the power of the half ellipsoid volumetric heat source,

and VH is the volume of this heat source. The region of the heat
source is dened by Eq. (6), and VH can be calculated according to
Eq. (7).

x  t0 t2 y2 z  z0 2
2
61
a2
c2
b
2
V H pabc
3

z  0

6
7

where the parameters a, b and c are three half axes of the half ellipsoid in x, y and z direction, respectively. t0 is the LBW speed. z0 is
the initial coordinate of heat source center (Point O as shown in
Fig. 6) in z-direction. In the current model, the value of z0 is equal
to the thickness of the plate (2.3 mm).
In the combined heat source, QH was assumed to be 50% of the
total heat input, and that of the 3-D conical heat source with
Gaussian distribution (QC) was assumed to be the rest 50% of the
total heat input:

523

J. Sun et al. / Materials and Design 63 (2014) 519530

Q H gP  Q C

QH QC

where P is the power of laser beam, and g is the thermal efciency

of LBW. In the current study, g was assumed to be 0.8.
For the 3-D conical heat source, the heat ux qv(r, z) can be
expressed by the following equation [10]:

qv r; z

9Q C e3
pe3  1

1
B

exp @ 
ze  zi r 2e r e r i r 2i

1
3r 2
ri r e 

i
r i zzz
e zi

C
2 A
10

where the parameters re and ri are the larger and smaller radii of the
cone respectively. Similarly, ze and zi are the upper and lower
z-coordinates of the cone, respectively. r is the current radius as a
function in Cartesian coordinate system.
The parameters used in the combined heat source are summarized in Table 4. We determined the parameters of heat source
using the following procedures. (i) Measure the size of fusion zone
in the middle cross-section of the weld specimen. (ii) According to
the dimension and the shape of the measured fusion zone, give an
initial value of each parameter for the combined heat source
model. For example, parameter a, and parameter b can be set to
be the half width of weld bead, and c is set to be the half of thickness at rst. Similarly, parameters re, ri, ze and zi are also determined according to the size and the shape of penetration. (iii)
Using the initial values of parameters given by step (ii), the prole
of fusion zone can be obtained through a trial calculation. Through
comparing the fusion zone simulated by FEM and that measured
by experiment, adjust the value of each parameter till to obtain a
reasonable fusion zone which is similar to the experimental result
both in shape and in size. When the fusion zone computed by FEM
matches that obtained from experiment, the values of parameters
can be xed.
For the CO2 gas arc welding, a full ellipsoid volumetric heat
source with uniform density was developed to simulate the heat
input, and the image of this heat source is shown in Fig. 9. The
region of this heat source is dened by Eq. (11).

x  t1 t2 y2 z  z1 2
2
61
a21
c21
b1

Fig. 9. Full ellipsoid volumetric heat source model. (a) Image of heat source used in
CO2 gas arc welding. (b) Dimensional parameters of ellipsoid heat source.

Table 5
Parameters of heat source in CO2 gas arc
welding.
Parameter

Value (mm)

a1
b1
c1

4
1.8
1.8

11

where the parameters a1, b1 and c1 are three half axes of the full
ellipsoid in x, y and z direction, respectively. t1 is the CO2 gas arc
welding speed. Here, the location of heat source center in the z
direction (z1) is 2.3 mm.
All parameters of the full ellipsoid volumetric heat resource are
summarized in Table 5. Similar to the heat source model used in
LBW, the value of each parameter in this heat source model was
also determined mainly according to the fusion zone obtained from
experiment.

Table 4
Parameters of combined heat source in laser
beam welding.
Parameter

Value (mm)

a
b
c
ze
zi
re
ri

0.65
0.65
1.5
1.2
0
0.4
0.38

The heat ux qf(x, y, z) of the full ellipsoid volumetric heat

source with uniform density can be expressed by Eq. (12):

qf g1 UI=V F

12

where U is the arc voltage, I is the welding current, g1 is the arc efciency, and VF is the volume of this heat source. The volume VF can
be expressed by the following equation:

VF

4
pa1 b1 c1
3

13

Besides considering the moving heat source, heat losses due to

convection and radiation were also taken into account in the nite
element model. Heat loss (qc) due to convection was considered for
all the surfaces using Newtons law:

qc hc T s  T 0

14

where hc is the heat transfer coefcient; Ts is the surface temperature of the plate and T0 is the ambient temperature. In addition,
the heat loss due to radiation was modeled using StefanBoltzman
law:

n
o
qr er T s 2734  T 0 2734

15

524

J. Sun et al. / Materials and Design 63 (2014) 519530

where e is the emissivity and r is the StefanBoltzman constant.

In this study, the temperature-dependent thermal physical
properties [14] of mild steel (Q235) such as thermal conductivity,
specic heat and density are employed in the thermal simulation
as shown in Fig. 10. In addition, the thermal effects due to solidication of the weld pool are modeled by taking into account the
latent heat for fusion. The value of the latent heat of mild steel
was set to be 270 J/g [18] in the nite element model. The liquidus
temperature and the solidus temperature were assumed to be
1500 C and 1450 C, respectively.
3.2. Mechanical analysis
The mechanical analyses were conducted by using temperature
history of each node in the FE model as the input information. The
same nite element models used in the thermal analyses were
employed in mechanical analyses. The restraint conditions used
in the LBW model and the CO2 gas arc welding model are shown
in Figs. 4 and 5, respectively. Similar to thermal analysis, the temperature dependent mechanical properties such as Youngs modulus, yield strength and thermal expansion coefcient were used in
the FE model, and they are shown in Fig. 11.
For low carbon steel, because solid-state phase transformation
has an insignicant inuence on the welding residual stress and
deformation [19], the phase change was neglected in the current

mechanical model. In addition, the period with high temperature

during the entire thermal cycle was very short (only a few seconds), so the creep behavior was not taken into account. Except
for the strain increments due to phase transformation and creep,
the total strain increments can therefore be decomposed into three
components as follows:

fdetotal g fdee g fdep g fdeth g

e

16

th

where {de }, {de } and {de } are the increments of elastic strain,
plastic strain and thermal strain, respectively.
The elastic behavior was considered using the isotropic Hookes
rule with temperature-dependent Youngers modulus and Possions ratio. For the plastic strain component, a plastic model was
employed with the following features: the Von Mises yield surface
and temperature dependent mechanical properties. Because the
effect of work hardening is not signicant in low carbon steel, it
was neglected in this study.
In this paper, the thickness of the specimens is only 2.3 mm, so
it can be expected that the geometrically non-linear phenomenon
potentially occurs during welding. Therefore, besides the material
nonlinearity, the geometrical nonlinearity was also considered in
the current simulation. When the large deformation is considered,
GreenLagrange strain which is the nonlinear function of the displacement is required. Eqs. (17)(22) shows the nonlinear relationships between GreenLagrange strain and the displacements.

( 
 2  2 )
2
@u
@m
@w

@x
@x
@x
( 
 2  2 )
2
@m 1
@u
@m
@w
ey

@y
@y
@y 2
@y
( 
 2  2 )
2
@w 1
@u
@m
@w
ez

@z 2
@z
@z
@z
        
@u @ m
@u @u
@m @m
@w @w

cxy
@y @x
@x
@y
@x
@y
@x @y
        
@ v @w
@u @u
@m @m
@w @w

cyz
@y
@z
@y
@z
@z @y
@y @z
        
@u @w
@u @u
@m @m
@w @w

czx

@z @x
@z
@x
@z
@x
@z
@x

ex

@u 1

@x 2

17
18
19
20
21
22

where ex, ey and ez are GreenLagrange strain in x, y and z directions;

cxy, cyz and czx are the shear strain on the xy, yz and zx planes. u,m
Fig. 10. Temperature-dependent thermal physical properties of mild steel.

and w are the displacement in x, y and z directions, respectively. In

Eqs. (17)(22), the rst order terms express linear response, and the
second order terms are essential to express non-linear behavior
such as buckling [2].
3.3. Simulated cases
To compare the simulated results computed by large deformation theory with that computed by small deformation theory,
two simulation cases (Case A for large deformation and Case B
for small deformation) were performed for LBW. Since the heat
input for CO2 gas arc welding is much larger than that for LBW,
referring to the research conducted by Deng et al. [13], it can be
inferred that the small deformation theory will underestimate

Table 6
Simulated cases.

Fig. 11. Temperature-dependent mechanical properties of mild steel.

Case

Deformation theory

Case A
Case B
Case C

Large deformation (LBW)

Small deformation (LBW)
Large deformation (CO2)

J. Sun et al. / Materials and Design 63 (2014) 519530

the welding deformation. Therefore, we simulated welding deformation and residual stresses induced by CO2 gas arc welding only
using large deformation theory, and this case is named Case C.
Totally, three simulation cases as shown in Table 6 were computed
in this study.
4. Results and discussion
4.1. Results of welding temperature eld
Fig. 12(a) and (b) compare the fusion zones (peak temperature
above 1450 C) predicted by FEM and those measured by experiment in the LBW joint and the CO2 gas arc welding joint, respectively. In these two gures, the white broke lines show the
boundaries of fusion zone in the mock-ups, while the gray areas
represent the fusion zones simulated by FEM. Through comparing
the fusion zones predicted by FEM and those measured by experiment, we can nd that the simulated results generally match the
measurements both in the LBW joint and in the CO2 gas arc welding joint. This information suggests that through carefully designing the heat source type and reasonably selecting the parameters
for the corresponding heat source, the essential features of fusion
zones induced by either LBW or CO2 gas arc welding can be captured by the current computational approach. Theoretically, the
thermo-elasticplastic behavior at the elevated temperature above
the melting point has very limited contribution to the nal residual
stress and deformation [20] because both the elastic modulus and
the yield strength are close to zero as shown in Fig. 11. Thus, it can
be inferred that the calculation accuracy of welding residual stress
and distortion will not be signicantly sacriced if the size and
shape of fusion zone obtained by FEM are roughly similar to these
of the actual fusion zone.
From Fig. 12(a), one can observe that the width of heat affected
zone (peak temperature between 750 C and 1450 C) on the top
surface of LBW joint predicted by FEM is approximately 0.6 mm,
and the size of heat-affected zone (HAZ) is also roughly match
the experimental result. From Fig. 12(b), we can know that the size
and shape of HAZ predicted by FEM in the CO2 gas arc welding joint
is similar to those obtained from experiment. Comparing these two
gures, we can nd that both the fusion zone and HAZ due to LBW
are much smaller than those induced by CO2 gas arc welding.

(a)

Fusion line

Tmax 1450oC
Tmax

(b) Fusion line

Fig. 13(a) and (b) shows the top views of weld bead induced by
LBW and by CO2 gas arc welding, respectively. The width of bead
measured by experiment in the LBW joint is 1.2 mm, while that
in the CO2 gas arc welding joint is 4.2 mm. It is clear that the width
of bead generated by the former is much narrower than that
caused by the latter. On the contrary, the penetration of the LBW
joint is deeper than that of the CO2 gas arc welding joint as shown
in Fig. 12.
Fig. 14 compares the thermal cycles of Point A (Case A) and
Point B (Case C), and these two points are at the fusion zone of
middle cross-section as dened in the same gure. From this gure, it can be observed that the heating speed of Case A is much
faster than that of Case C, and the peak temperature of Point A
and Point B are 2667 C and 1722 C, respectively. It is clear that
the peak temperature generated by LBW is far higher than that
caused by CO2 gas arc welding. Meanwhile, the cooling rate of
the former is much faster than that of the latter. Because the total
heat input used in LBW is signicantly smaller than that used in

Fig. 13. Top view of LBW and CO2 welded section. (a) Top view of weld bead
induced by LBW. (b) Top view of weld bead induced by CO2 gas arc welding.

750 C

Tmax

525

1450oC

Tmax

750oC

Fig. 12. Comparison of fusion zone between experiment and FEM. (a) Fusion zone
and HAZ induced by LBW. (b) Fusion zone and HAZ induced by CO2 gas arc welding.

Fig. 14. Temperature histories of Points A and B.

526

J. Sun et al. / Materials and Design 63 (2014) 519530

CO2 gas arc welding. Fig. 15 compares the thermal cycles of Point C
and Point D. It should be pointed out that Point D is at the HAZ of
middle cross-section in the CO2 gas arc welding joint (Case C),
while Point C whose location is the same as point D is at the
LBW joint (Case A). From this gure, we can know that the peak
temperature of Point C and Point D are 380 C and 798 C, respectively. It is clear that Point C is not within the HAZ. This information implies that the range with high peak temperature above
the mechanical melting point (about 750 C for Q235 steel) in the
LBW joint is far narrower that in the CO2 gas arc welding joint.
4.2. Welding deformation
4.2.1. Transverse shrinkage
Figs. 16 and 17 show the transverse shrinkages on the top surface (weld face) and the bottom surface of the welded plate in Case
A and Case B, respectively. In this gure, the ordinate numbers 1,
2 and 3 in horizontal abscissa represent the locations 1, 2
and 3 as dened in Fig. 3, respectively. Both the simulated results
and the measured data are plotted in these two gures. In these
gures, both simulation results and experimental measurements
suggest that the transverse shrinkages on the top surface are
slightly larger than those on the bottom surface. Comparing Case
A and Case B, we can nd that the values of the former are slightly
larger than those of Case B. Even though there is no marked difference between Case A and Case B, it seems that the simulation

Fig. 17. Transverse shrinkages in Case B.

results computed by Case A (large deformation theory) are closer

to the measured data. The average values of transverse shrinkage
calculated according to an empirical formulation [21] are also plotted in the same gures. The formulation used for calculating transverse shrinkage is as follows.

S=h 2:85  106 

Q net
h

23

Fig. 15. Temperature histories of Points C and D.

where Qnet is the net heat input, and h is the thickness of plate.
According to the above formulation, the average value of transverse shrinkage is 0.08 mm in the LBW joint. In Fig. 16, it is clear
that the transverse shrinkages predicted by Case A are in a good
agreement with the empirical value. From Fig. 17, it can be found
that the transverse shrinkages calculated by small deformation
theory (Case B) are smaller than the value obtained by Eq. (23)
to a certain extent.
Fig. 18 shows the transverse shrinkages on the top surface and
the bottom surface of the welded plate in Case C. In this gure,
both simulation results and experimental measurements indicate
that the transverse shrinkages on the top surface are slightly smaller than that on the bottom surface. This is opposite to the result of
the LBW joint. The reason is that the out-of-plane deformation of
Case C has a convexconcave shape [13]. Fig. 18 also tells us that
the transverse shrinkages predicted by Case C matches the measured data well, and the average value (0.27 mm) of three locations

Fig. 16. Transverse shrinkages in Case A.

Fig. 18. Transverse shrinkages in Case C.

527

J. Sun et al. / Materials and Design 63 (2014) 519530

is also in a good agreement with the empirical value (0.23 mm) calculated by Eq. (23).
Comparing Figs. 16 and 18, we can nd that the average value of
transverse shrinkage due to LBW is much smaller than that
induced by CO2 gas arc welding. According to the empirical formulation, we can understand that when the thickness of plate is xed,
the average value of transverse shrinkage is proportion to heat
input. Meanwhile, the numerical results obtained in the current
study also support this conclusion.

3.628e- 009
-2.744e- 002
-4.116e- 002
-5.487e- 002
-6.859e- 002
-8.231e- 002
-9.603e- 002

4.2.2. Longitudinal shrinkage

Fig. 19 shows the longitudinal shrinkages of the neural plane in
Case A and Case C. The longitudinal shrinkage is the difference of
longitudinal displacement between line 1 and line 2. In this gure,
it can be found that the average value of longitudinal shrinkage is
only 0.08 mm in Case A, while that of Case C is 0.68 mm. It is clear
that the longitudinal shrinkage of Case A is far less than that of
Case C. Unlike transverse shrinkage, the longitudinal shrinkage
has no direct proportion relation with heat input, but it increases
more quickly than heat input in the current situation.
4.2.3. Out-of-plane deformation
Figs. 20 and 21 show the contours of deection distributions
computed by Case A and Case B, respectively. The magnitude of
deformations has been enlarged 100 times in these gures. Comparing Case A and Case B, we can observe that the model of outof-plane deformation in Case A is different from that in Case B.
The out-of-plane deformation of Case A has a concaveconvex

Line 2

Line 1

100

-1.372e- 002

Z
Y

-1.097e- 001
-1.235e- 001

-1.372e- 001

Fig. 21. Deection distribution of Case B.

shape [13], so both longitudinal bending and transverse bending

(angular distortion) can be seen in Fig. 20 even if their magnitudes
are small. In Fig. 21, the transverse bending (angular distortion)
can be seen, but the longitudinal bending is too small to be
observed. Fig. 22 compares the deection distributions along line
3 computed by Case A and Case B, and the measurements are also
plotted in the same gure. This gure tells us that the predictions
of Case A generally matches the measurements well both in shape
and in magnitude, but the simulation results of Case B are signicantly smaller than the measured data. Based on the above comparisons, we can know that even though the heat input used in
the LBW is not large enough, the out-of-plane deformation predicted by large deformation theory is different from that simulated
by small deformation theory. Therefore, to obtain an accurate prediction the large deformation theory is recommended to simulate
the thermo-mechanical behavior for thin-plate joints.
Fig. 23 shows the contour of deection distribution simulated
by Case C. In this gure, the magnitude of deformation has been
enlarged 3 times. Because the heat input used in CO2 gas arc welding is relatively large, the welded plate has a convexconcave
deformation, which is completely opposite to Case A. From this gure, it can be found that the maximum displacement in z direction
is about +1.4 mm, while the minimum one is 7.8 mm. The maximum deection is 9.2 mm, and it is four times as large as the thickness of plate (2.3 mm). Fig. 24 compares the deection distribution
along line 3 computed by FEM and that measured by experiment.
Even though the magnitudes measured by experiment are slightly
larger than simulation results, however, the predictions generally

Fig. 19. Longitudinal shrinkages in the neural plane.

100

1.198e- 001

Line 3

9.737e- 002
7.491e- 002
5.244e- 002
2.998e- 002
7.514e- 003
-1.495e- 002
-3.741e- 002
-5.988e- 002

Z
Y

-8.234e- 002
-1.048e- 001

Fig. 20. Deection distribution of Case A.

Fig. 22. Comparison of deection distributions along Line 3 in Case A and Case B.

528

J. Sun et al. / Materials and Design 63 (2014) 519530

1.418e+000
4.989e- 001
-4.203e- 001
-1.340e+000
-2.259e+000
-3.178e+000
-4.097e+000
-5.017e+000
-5.936e+000

Z
Y

-6.855e+000
-7.774e+000

Fig. 23. Deection distribution of Case C.

4.2.4. Plastic strain distribution

Fig. 26 shows the longitudinal plastic strain distributions in the
neutral plane in the middle transverse cross section of Case A and
Case C. In this gure, it can be seen that both the magnitude and
the width of longitudinal plastic strain caused by LBW are smaller
and narrower than those induced by CO2 gas arc welding. Fig. 27
compares the transverse plastic strain distribution caused by
LBW and that induced by CO2 gas arc welding. This gure shows
that the transverse plastic strain of Case A is much smaller than
that of Case C both in width and in magnitude.
Comparing Figs. 26 and 27, it can be found an interesting phenomenon that the region with transverse plastic satin is much narrower than that with longitudinal plastic strain, but the magnitude
of transverse plastic strain is larger than that of longitudinal plastic
strain. The reason is that the restraint density in the longitudinal
direction (parallel to welding direction) is stronger than that in
the transverse direction (perpendicular to welding direction), and
the generation mechanism of the former is different from that of
the latter [22]. Figs. 26 and 27 also tell us that both longitudinal
plastic strain and transverse plastic strain are mainly determined
by heat input. From the viewpoint of inherent strain, we can conclude that the LBW can effectively reduce the nal welding deformation in the thin-plate joint.

Line 3

Fig. 24. Deection distribution along Line 3 in Case C.

match the measured data. This information suggests that the computational approach developed in the current study has sufcient
computing accuracy.
Comparing Case A with Case C, we can understand that the outof-plane deformation due to LBW is far less than that induced by
CO2 gas arc welding. Fig. 25 compares the welding deformation
of the specimen performed by LBW and that conducted by CO2
gas arc welding. In this gure, it is clear that the nal deformation
in the specimen jointed by LBW is very small, while the specimen
performed by CO2 gas arc welding has a signicant bending deformation. From the viewpoint of deformation prevention, LBW is
superior to CO2 gas arc welding.

Fig. 25. Specimens were welded by (a) LBW and (b) CO2 gas arc welding.

Fig. 26. Longitudinal plastic strain distribution in the neutral plane of mid-section.

Fig. 27. Transverse plastic strain distribution in the neutral plane of mid-section.

J. Sun et al. / Materials and Design 63 (2014) 519530

529

4.3. Results of welding residual stress and discussion

Middle section

Fig. 28 shows the longitudinal residual stress distributions on

the top and bottom surface of the middle cross-section in Case A
and Case C. From this gure, we can observe that the longitudinal
residual stress distribution on the top surface is almost identical to
that on the bottom surface in Case A, whereas the longitudinal
residual stress distribution on the top surface in Case C is signicantly different from that on the bottom surface. The heat input
used in LBW is small and the penetration is relatively uniform
through thickness as shown in Fig. 12(a), so the difference of longitudinal stress distribution between the top surface and the bottom surface is very small. However, the heat input used in CO2
gas arc welding is relatively large, which causes the geometrical
nonlinear phenomenon is remarkable after welding and can result
in a large out-of-plane deformation. It is for this reason we can see
a big difference of longitudinal residual stress distribution between
the top surface and the bottom surface in the weld joint performed
by CO2 gas arc welding. The longitudinal residual stresses on the
top surface are higher than those on the bottom surface within
the regions with Y P 10 mm and Y 6 10 mm in Case C. However,
the longitudinal residual stresses in the rest region are signicantly
lower than those on the bottom surface. In theory, if the deformation due to geometrical nonlinearity is not large enough after
welding, the longitudinal residual stress on the top surface with
a convexconcave deformation mode as shown in Fig. 23 will be
smaller than that on the bottom surface because of the direction
of longitudinal bending. However, if the geometrical nonlinear
deformation is large enough or buckling distortion occurs, the
residual stresses will redistribute within the whole range of the
plate. Because the heat input used in Case C is large enough, it possibly made the plate buckled. This is the reason that the longitudinal residual stresses on the top surface are larger than those of the
bottom surface [13]. In Fig. 28, not only the magnitude of longitudinal stresses between Case A and Case C are considerably different
but also the region with high tensile stress in Case A is much narrower than that in Case C. The peak tensile stress (300 MPa) of Case
A is very close to the yield strength of base metal (292 MPa) at
room temperature, while the maximum tensile stress (340 MPa)
in Case C is slightly higher than the yield strength. Overall, it seems
that the longitudinal residual stress distribution caused by LBW is
superior to that produced by CO2 gas arc welding.
Fig. 29 shows the transverse residual stress distributions on the
top and bottom surfaces of the middle cross-section in Case A and
Case C. In Case A, there is almost no difference between the top
surface and the bottom surface on the whole range, and the
maximum absolute value is approximately 50 MPa. In Case C, the
Middle section

Fig. 28. Longitudinal residual stress in mid-section computed by Case A and Case C.

Fig. 29. Transverse residual stress in mid-section computed by Case A and Case C.

transverse residual stresses on the top surface are almost opposite

to those on the bottom surface because of the convex bending in
transverse direction. In addition, the maximum value of transverse
residual stress is 150 MPa. Fig. 29 shows that the transverse residual stresses caused by LBW are signicantly smaller than those
induced by CO2 gas arc welding.
Finally, it should be pointed that even though the welding
deformation in the LBW joint calculated by large deformation theory (Case A) is different from that computed by small deformation
theory (Case B) to some extent, the residual stress distributions
predicted by Case A and Case B have no signicant difference.
The reason is that the heat input used in LBW is relatively small
and the corresponding distortion is not large enough.
5. Conclusions
In this study, a thermo-elasticplastic FEM with considering
moving heat source was developed to simulate the welding temperature elds, residual stresses and distortions induced by LBW
and CO2 gas arc welding. Meanwhile, experiments were also carried out to measure transverse shrinkage and deection in the
weld specimens performed by both LBW and CO2 gas arc welding.
On the basis of the calculated and the experimentally determined
results, the following conclusions are made.
(1) In comparison to CO2 gas arc welding, LBW can signicantly
reduce the welding deformation. In this study, the maximum deection resulted from the thin-plate welded by
LBW is only 0.23 mm, while that resulted from CO2 gas arc
welding is 8.7 mm.
(2) The range with high longitudinal tensile stress in the joint
welded by LBW is signicantly narrower than that generated
by CO2 gas arc welding. Moreover, the maximum value of
longitudinal residual stress generated by the former is smaller than that caused by the latter.
(3) The simulation result shows the welding distortion predicted by large deformation theory is different from that
simulated by small deformation theory to some extent.
Moreover, the deections predicted by the large deformation theory are in good agreement with the measured data.
The simulation results suggest that to accurately predict
welding distortion in a thin-plate joint the large deformation
theory is recommended.
(4) An accurate description of the heat source and the appropriate choice of the relevant parameters are essential to the
numerical simulation of thin-plate welded structures.

530

J. Sun et al. / Materials and Design 63 (2014) 519530

Acknowledgements
This research is supported by National Natural Science Foundation of China (Project No. 51275544) and the Fundamental Research
Funds for Central University (Project No. CDJZR12130036).

References
[1] Deng D, Murakawa H. FEM prediction of buckling distortion induced by
welding in thin plate panel structures. Comput Mater Sci 2008;43(4):591607.
[2] Wang J. Investigation of buckling distortion of ship structure due to welding
assembly using inherent deformation theory. Doctoral thesis. Osaka
University; 2012.
[3] Pan M. Minimization of welding distortion and buckling. Woodhead
Publishing Limited; 2011.
[4] Amit K, Duck Y, Darek C. Correlation analysis of the variation of weld seam and
tensile strength in laser welding of galvanized steel. Opt Laser Eng
2013;51(10):114352.
[5] Hee S, Han S, You C, Ik H. A study on mechanical and microstructure
characteristics of the STS304L butt joints using hybrid CO2 laser-gas metal arc
welding. Mater Des 2011;32(4):232833.
[6] Chen W, Paul A, Pal M. CO2 laser welding of galvanized steel sheets using vent
holes. Mater Des 2009;30(2):24551.
[7] Jandaghi M, Parvin P, Torkamany M, Sabbaghzadeh J. Alloying elemental
change of SS-316 and Al-5754 during laser welding using real time laser
induced breakdown spectroscopy (LIBS) accompanied by EDX and PIXE
microanalysis. Phys Proc 2010;5B:10714.
[8] Lindgren LE. Numerical modeling of welding. Comput Methods Appl Mech Eng
2006;195(4849):671036.
[9] Han G, Zhao J, Li J. Dynamic simulation of the temperature eld of stainless
steel laser welding. Mater Des 2007;28(1):2405.

[10] Muhammad Z, Daniel N, Jullien J, Dominique D. Prediction of laser beam

welding-induced distortions and residual stresses by numerical simulation for
aeronautic application. J Mater Process Technol 2009;209(6):290717.
[11] Muhammad Z, Daniel N, Jullien J, Dominique D. Experimental investigation
and nite element simulation of laser beam welding induced residual stresses
and distortions in thin sheets of AA 6056-T4. Mater Sci Eng A
2010;527(12):302539.
[12] Wang J, Shibahara M, Zhang X, Murakawa H. Investigation on twisting
distortion of thin plate stiffed structure under welding. J Mater Process
Technol 2012:170515.
[13] Deng D, Zhou Y, Bi T, Liu X. Experimental and numerical investigations of
welding distortion induced by CO2 gas arc welding in thin-plate bead-on
joints. Mater Des 2013;52:7209.
[14] Deng D, Murakawa H. Prediction of welding distortion and residual stress in a
thin plate butt-welded joint. Comput Mater Sci 2008;43:35365.
[15] MARC Analysis Research Corporation. Theory and user information. Palo Alto;
2007.
[16] Yan D, Wu A, Silvanus J, Shi Q. Predicting residual distortion of aluminum alloy
stiffened sheet after friction stir welding by numerical simulation. Mater Des
2011;32(4):228491.
[17] Deng D, Kiyoshima S. Numerical simulation of residual stresses induced by
laser beam welding in a SUS316 stainless steel pipe with considering initial
residual stress inuences. Nucl Eng Des 2010;240(4):68896.
[18] Zhang W, Elmer J, DebRoy T. Modeling and real time mapping of phases during
GTA welding of 1005 steel. Mater Sci Eng 2002;333A:3205.
[19] Deng D. FEM prediction of welding residual stress and distortion in carbon
steel considering phase transformation effects. Mater Des 2009;30(2):35966.
[20] Ueda Y, Murakawa H. Applications of computer and numerical analysis
techniques in welding research. Mater Des 1985;6(3):10311.
[21] Deng D. Theoretical prediction of welding distortion in thin curved structure
during assembly considering gap and misalignment. Doctoral thesis, Osaka
University; 2002.
[22] Ueda Y, Yamakama T. Analysis of thermal elasticplastic stress and strain
during welding by nite element method. Trans Jpn Weld Soc
1971;2(2):90100.