Investigating the influence of external restraint on welding distortion in

thin-plate bead-on joint by means of numerical simulation and

experiment
Dean Deng1, 2)*, Xiaozhan Liu 1), Jing He 1), Wei Liang 3)

1) College of Materials Science and Engineering, Chongqing University,

No.174, Shazhengjie, Shapingba, Chongqing, 400044, China
2) State Key Laboratory of Advanced Welding and Joining, Harbin Institute of Technology, Harbin
150001, China
3) College of Mechatronics and Automotive Engineering, Chongqing Jiaotong University, No. 66
Xuefu Road, Nan-an, Chongqing 400074, China

Abstract: The main objective of this study is to clarify the influence of external restraint on

welding deformation in thin-plate bead-on joint by means of both experiment and numerical

simulation. In the current study, an advanced computational approach based on ABAQUS

code with considering moving heat source, material nonlinearity and geometric nonlinearity

was developed to simulate welding-induced distortion in thin-plate joints. Meanwhile,

experiments were carried out to measure the out-of-plane deformations in the corresponding

mock-ups. Based on the experimental and numerical results the mechanisms of the

out-of-plane deformation in the thin-plate joints were discussed. In addition, the influence of

external restraint on welding distortion in thin-plate bead-on joints were clarified by means of

both numerical simulation and experiment. Both numerical models and experiments show that

external restraint can mitigate the magnitude of the final deformation to some extent, however,

it is difficult to satisfactorily eliminate welding distortion in the thin-plate joints only by

external restraint. The results obtained from this work will be meaningful in understanding the

features of welding distortion in thin-plate weldments, and also will be helpful in controlling

welding distortion.

Keywords: Welding distortion; external restraint; finite element; numerical

simulation
*Corresponding author: Dean Deng Tel: 86-23-65102079 Fax: 86-23-65102079

E-mail: [email protected]

1
1 Introduction
To improve the fuel economy and to increase the capacity of modern transport
tools, thin steel plates have been widely used to fabricate large-scale and complex
welded structures used in automobile and shipbuilding industries [1]. Meanwhile,
because fusion welding processes such gas metal-arc welding and CO2 arc welding
are high productive and practical joining methods, so they have been extensively
adopted to assemble automobiles and ships. However, because the stiffness of
thin-plate structure is relatively small, it is apt to produce welding deformation during
the assembling process. Worse, buckling distortion will occur in case that the heat
input is too large or the design of welded structure is unreasonable. Welding-induced
distortion not only negatively affects the dimensional accuracy as well as structural
performance, but also heavily delays the delivery time and increases cost owing to the
correction work [2]. Therefore, it is vital to predict and control welding distortion in
thin-plate welded structures [3, 4].
Fusion welding is a complicated thermal processing technology, so there are a lot
of factors having contribution to the final deformation. Generally, these influential
factors can be classified into three types. The first type can be taken as intrinsic factor
which includes the thermal physical and mechanical properties of both base metal and
weld metal. The second type contains welding method, heat input, preheating
temperature, post weld heat treatment, welding sequence, external restraint, gap
correction, etc., and they belong to manufacturing-related factors [5-7]. The third type
includes size of weld joint, groove shape, thickness of plate, and so on, and they can
be regarded as design-related factors [8, 9].
Fig. 1 shows six typical deformations produced in a thin-plate welded joint [2].
In this figure, longitudinal shrinkage, transverse shrinkage, and rotational deformation
belong to in-plane deformation, while longitudinal bending, angular distortion and
buckling distortion are out-of-plane deformation.
As mentioned above, welding distortion is affected by various factors and the
type of welding deformation also presents diversity, so it is really difficult to control
the dimensional accuracy in a welded structure if only relying on practical
2
experiences. If the welding distortion can be predicted beforehand, the predictions
will be helpful to the mitigation of welding distortion.
In the past decades, a number of researches were conducted to investigate the
distortion induced by fusion welding process by means of both experimental and
numerical methods. Masubuchi [10] investigated the buckling distortion occurred in a
thin-plate welded structure using experimental method, and he pointed out that
buckling is an unstable phenomenon resulting from the compressive residual stress.
Watanabe and Satoh [11] carried out a series of experiments to examine the features
of welding deformation in low carbon steel thin-plate joints. They found out that there
are only two types of deformation mode in the thin-plate bead-on joints. One is the
concave-convex mode, and the other is the convex-concave mode. With the
developments of computer technology and nonlinear finite element method, a number
of numerical models have been proposed to study welding deformation in thin-plate
welded joints or structures. Deng and Murakawa [9] investigated the characteristics of
welding deformation in a thin-plate butt-welded joint using both
thermo-plastic-plastic finite element method (FEM) and elastic FEM based on
inherent strain theory. In addition, they also examined the influences of
manufacturing-related factors and design-related factors on buckling propensities in a
large thin-plate welded structure [6]. Schenk [12] investigated the welding distortion
in a DP600 steel joint induced by gas metal arc welding, and they also examined the
influence of mesh density on the accuracy of simulation result. Mollicone et.al [13]
pointed out that even a minor variation in fabrication procedures can significantly
affect the buckling distortion in a large thin-plate welded structure. They investigated
the buckling behaviors with considering the initial geometry errors and supporting
conditions. The thickness of plate used in their research is between 3 mm and 8 mm.
Recently, Wang and his co-workers [14] studied the mechanism of buckling distortion
in thin-plate bead-on joints by mean of both numerical simulation and experiment.
As mentioned above, a number of fundamental researches have been carried out
to understand the features of welding deformation in the thin-plate welded joints.
However, the researches on how to control and mitigate welding distortion are very
3
limited. External restraint is used as a common method of on-line control to mitigate
welding deformation. Therefore, we have made some attempts to clarify the influence
of external restraint on welding distortion of thin-plate welded joint in this study. A
thermo-elastic-plastic FEM with considering moving heat source, material
nonlinearity and geometric nonlinearity was developed to simulate welding
deformation in the thin-plate joints. Meanwhile, experimental mock-ups were
manufactured to measure the out-of-plane deformations. In addition, the influence of
external restraint on the final distortion was clarified by means of both experimental
method and numerical simulation technology. The conclusions drawn from the present
study will be meaningful in understanding the features of welding distortion of
thin-plate joints, and also will be helpful in controlling the final deformation.

2 Experimental procedure
In order to examine the influence of external restraint on welding deformation in
thin-plate bead-on joints, six thin-plate joints were fabricated to measure the
deflection (out-of-plane displacement) distributions. In the experiments, three joints
were welded without external restrain, while another three ones were performed with
external restraint during welding process.
The base metal used in this study is low alloy high strength steel named Q345.
This steel has been widely used in architecture, automobile, bridge, and ship building
industries because of its excellent mechanical properties and good weld-ability [15].
Table 1 shows the chemical compositions of the base metal. The filler wire is
ER-50-6 wire with 1.2 mm diameter, and the weld metal has the same strength as that
of Q345 steel. Table 2 shows the chemical composition of the filler wire. The size of
each plate to be welded is 300 mm×100 mm×2 mm.
In the current study, CO2 arc welding process was used to perform the bead-on
welding. All the joints were welded by TAWERS TA-1400 arc welding robot system
as shown in Fig. 2(a). The length of weld bead is 290 mm in each plate. Because the
full length of thin plate is 300 mm, a 5 mm-length region was not welded at both the
start side and the end side as shown in Fig. 2(b). Before welding, the area to be
4
welded was polished by sandpapers and then cleaned by industrial alcohol to
removes contamination, oil and so on.
The welding conditions used each joint and their corresponding simulation cases
are listed in the Table 3. In all joints, the flow of CO2 gas was 15 L/min. Under the
condition without external restraint, the specimen was only placed on the welding
platform without any mechanical constraint. Under the condition with external
restraint, the welded specimen was fixed by six clamps as shown in Fig. 2(b). The
external restraints were removed when the specimen cooled to the room temperature.
After welding, a three-dimensional coordinate measuring system named ROMER
INFINITE 2.0 was used to measure welding deformation. The measuring accuracy of
the system is 0.03 mm. Three-dimensional coordinates at 48 points of each welded
specimen were measured to obtain the contour of welding deformation. In the present
study, the out-of-deformation distribution of each joint was plotted through using
MATLAB software. These profiles of welding deformation are used to verify the
simulation results.

3 Finite element modeling

During steel welding process, the thermo-metallurgical-mechanical coupling
behavior as shown in Fig. 3 is very complicated. Theoretically, when a numerical
model is employed to simulate welding process, the calculation accuracy will be
improved if all factors having influence on welding process can be taken into account.
However, the numerical model will become very complex if all details related to
welding process are considered. Sometimes, it will be very difficult or even
impossible to solve the complicated numerical model. In many cases, when a
computational approach is developed to solve problems in engineering application,
priority must be given to the main or key factors and some insignificant factors are
often neglect. For low carbon steel or low alloy high strength steels, the solid-state
phase transformation has no significant influence on welding residual stress and
deformation [16], so the factors related to microstructure are not considered in the
developed numerical model. Additionally, the strain-hardening, annealing effect and
5
creep are ignored in the mechanical analysis.
Based on the commercial ABAQUS software, a sequentially coupled
thermo-elastic-plastic finite element method with considering material nonlinearity,
geometrical nonlinearity and moving heat source was developed to simulate welding
temperature field, residual stress distribution and deformation in thin-plate joints
induced by CO2 arc welding process. The computational procedure consists of two
steps. At the first step, the temperature history of each node is computed through
using a moving heat source. In the second step, the temperature histories of all nodes
are introduced into the mechanical model as thermal load to calculate residual stress
and deformation.

3.1 Finite element model and boundary condition

The dimensions of each finite element model are as the same as the
corresponding experimental specimen. In each finite element model, the bead shape
was also carefully considered based on the experimental result. As mentioned earlier,
the average value of the reinforcement and bead breadth were used to approximately
define the size of each bead. Based on the observation of experiment, it was found
that that the restraint conditions had very limited effect on bead shape. Hence, the
bead shapes of Case 1, Case 3, and Case 5 were assumed to be the same as those of
Case 2, Case 4 and Case 6, respectively. As shown in Fig. 4, the mesh size at the weld
bead and its vicinity is finer, while it gradually increases with the distance from the
centerline of weld zone. This kind of mesh design can balance the calculation
accuracy and the computing time. The length of each element is 2.5 mm in the
longitudinal direction. The total number of node is 26,403, and that of element is
20,544.
The restraint conditions of Case1, Case 3, and Case 5 are applied just to prevent
rigid body motion. As showed in Fig. 4, the arrows towards points A, B, and C show
the boundary conditions for the model without external restraint (free model). In the
same figure, the external restraints used in fixed models are showed by solid-line at
six locations. The z-displacements of these nodes are fixed during the welding process
6
to model the clamps used in the corresponding mock-up. After cooling to room
temperature, the external restraints are removed.

3.2 Heat source and thermal analysis

The conservation of energy is the fundamental principle in thermal analysis of
welding. Therefore, during thermal analysis, stress, strain and displacement are
ignored and only the heat energy is considered. In this step, temperature history of
each node in the whole finite element model is calculated by the transient heat
conduction equation given as follows.
T
c  x , y , z , t   -  q  x , y , z , t   Q  x , y , z , t  (1)
t
where, ρ is the density of the material (g/mm3), c is the specific heat capacity
(J/(g oC)), q is the heat flux vector (W/mm2), Q is the internal heat generation rate

(W/mm3), T is the current temperature (oC), and ∇ represents the spatial gradient

operator. In addition, x, y and z are the coordinates in the reference system (mm).
The heat flux vector is defined by Fourier’s law for isotropic material as follows

q  -k  T (2)

where, k is the temperature-dependent thermal conductivity (J/(mm s oC)), and ∇T is

the temperature gradient.

The moving heat source is a very important aspect for the thermal analysis, and it
has attracted a lot of interests in the past decades. The most representative models
include the double-ellipsoidal volumetric heat source [17] which is often used to
model the heat input due to arc welding, and the cone-shaped volumetric heat source
with Gaussian distribution [18] which is employed to model the heat input induced by
high energy density welding process. Recently, Deng and his co-workers proposed a
half ellipse volumetric heat source with uniform density [5] to model the heat input of
CO2 arc welding process, and they used this heat source to have successfully
simulated the welding temperature field in a bead-on plate joint.
In order to accurately simulate welding temperature filed due to CO2 arc welding,

7
a newly combined heat source model was proposed to model the heat input in the
current study. In the combined heat source model, the whole heat input is divided into
two portions. One is the heat from the molten droplet (Part I), and the other is the heat
of the welding arc (Part II). Both of them are simulated by a half ellipse volumetric
heat source with uniform density. It is assumed that the heat from molten droplet (Part
I) is 70% of the total heat, and the heat from arc (Part II) is 30%. The combined heat
source model is schematically shown in Fig. 5. The regions of these two heat sources
are defined by Eq. (3) and Eq. (4), respectively.
( x  0 t ) 2 y 2 ( z  z0 ) 2
 2 1 (z≥0, Part I) (3)
a12 b1 c12

( x  0 t ) 2 y 2 ( z  z0 ) 2
 2 1 (z<0, Part II) (4)
a2 2 b2 c2 2
where, the parameters ai, bi and ci (i=1,2) are the three half axes of ellipsoid in x, y
and z direction, respectively. v0 is the welding speed, z0 is the initial coordinate of heat
source center in z-direction. In the current model, the value of z0 is equal to the
thickness of the plate (2.0 mm).

The parameters are determined mainly according to the cross-section of weld bead
measured by experiment. The procedure to determine the parameters of heat source is
described as follows. I) Observing and measuring the dimensions of fusion zone in
three typical cross-sections of the weld specimen. II) Based on the dimensions of
fusion zone and welding conditions, an empirical value for each parameter of heat
source model was recommended. III) Using the initial parameters recommended by
step II, the profile of fusion zone can be obtained through thermal analysis. By
comparing the calculated result with the experimental result, we can constantly adjust
the parameters of heat source until the former is very close to the latter. Once the
fusion zone obtained by finite element model matches the experimental result, the
parameters of heat source model are fixed.
In summary, determining the parameters of heat source model mainly relies on
the experimental result of fusion zone and partially depends on the trial and error.
The parameters of the combined heat source used in each case are summarized in

8
Table 4. In this table, a1, b1 and c1 are the parameters of Part I, while a2, b2 and c2 are
the parameters of Part II. The welding conditions of each case are as the same as those
of the corresponding experiment as shown in Table 3. The arc efficiency was assumed
to be 0.85 for CO2 gas metal arc welding [19, 20].
Besides considering the moving heat source, heat losses due to convection and
radiation were also taken into account in the finite element model.
The heat loss (qc) due to convection was considered for all the surfaces using
Newton’s law.
qc   hc Ts  T0  (5)

where, hc is the heat transfer coefficient, Ts is the surface temperature of the plate , and
T0 is the ambient temperature.
The heat loss due to radiation was modeled using Stefan-Boltzman law.

qr   T  273

s
4
 T0  273 
4
 (6)

where, is the emissivity factor, and  is the Stefan-Boltzman constant.

It should be pointed out that the contact heat transfer between the welded plate
and the clamps was ignored in the current model. The modeling of heat flow was also
neglected because the effect of the fluid flow on the stress and deformation field is
very limited [21].

The temperature-dependent thermal physical properties such as thermal

conductivity, specific heat and density were used in the thermal simulations. In Ref.
[22], the thermal physical and mechanical properties of SM490 steel were carefully
measured by experiment. Because the chemical composition, thermal physical, and
mechanical properties of Q345 are much similar to those of SM490 steel, the material
properties of SM490 steel can be taken as those of Q345 steel. Table 5 shows the
thermal physical properties of SM490 steel [22]. Meanwhile, the thermal effects due
to solidification of the weld pool were modeled by considering the latent heat. The
value of the latent heat of low alloy steel was set to be 270 J/g [4]. The liquidus
temperature and the solidus temperature were assumed to be 1480oC and 1430oC,
respectively.
Based ABAQUS code, a series of user subroutines were developed to model the

9
moving heat source and thermal boundary conditions. The type of the element used in
thermal analysis is DC3D8.

3.3 Mechanical analysis

In mechanical analysis, the basic equations are the equilibrium equations,
constitutive stress-strain relations and geometric compatibility equations [23, 24].
Temperature history of each node obtained from thermal analysis was introduced as
thermal load into the structural model to compute the stress-strain field. Ignoring the
strain components due to phase transformation and creep, the total increment of strain
can be decomposed into three components as follows

d   d e  d p  d th  (7)

where, d de, dp and dth represent the total, elastic, plastic and thermal strain
increment, respectively.
The isotropic Hooke’s law was applied to model the elastic strain-stress
relationship, and the regular elastic-plastic material model with von Mises yield
criterion was used in the mechanical analysis. The thermal strain was considered
through using thermal expansion coefficient. The temperature-dependent mechanical
properties such as yield strength, Young’s modulus, Possion’s ratio, and thermal
expansion coefficient as shown in Table 6 [22] were employed in the mechanical
analysis. The element type employed in mechanical analysis is C3D8I.
Because the thickness of plate is only 2 mm, it can be inferred that geometric
nonlinear phenomenon is apt to occur during welding process. Generally, the
relationship between strain and displacement can be defined in two ways. One is
based on the small deformation theory, and the other is based on the large deformation
theory. Deng and his co-workers [9] calculated welding distortions and residual
stresses in a thin-plate bead-on joint using both the large deformation theory and the
small deformation theory, and their results show that the large deformation theory can
capture the features of welding distortion and residual stress in thin plate joint, while

10
the small deformation theory largely underestimate the results. Thus, the large
deformation theory was adopted in this study. Eq. (8.1) to Eq. (8.6) shows the
nonlinear relationship between Green-Lagrange strains and the displacements. The
first order terms express linear response, and the second order terms are essential to
express non-linear behavior [1].

u 1  u      w  
2 2 2

x           (8.1)
x 2  x   x   x  

 1  u      w  
2 2 2

y           (8.2)
y 2  y   y   y  

 1  u      w  
2 2 2

z           (8.3)
z 2  z   z   z  

u   u   u         w   w  
 xy              (8.4)
y x  x   y   x   y   x   y  

u w  u   u         w   w  
 yz              (8.5)
z y  z   y   z   y   x   y  

u w  u   u         w   w  
 zx              (8.6)
z x  z   y   x   z   x   z  

4. Results and Discussions

4.1 Comparison of fusion zone

The research conducted by Wang and his co-workers [25] shows that both
transverse shrinkage and angular distortion in a thin-plate joint are sensitive to the
shape of fusion zone. Theoretically, the shape of fusion zone is determined by the type
of heat source model and its parameters. Especially, when numerical methods based
on structural finite element codes such as ABAQUS and MSC. Marc are used to
simulate welding temperature, residual stress and deformation, it is very vital to select
a reasonable heat source model and to determine the corresponding parameters.
As an example, the fusion zone calculated by Case 1 is selected to compare with

11
the experimental result. Fig. 6 (a) shows the appearance of weld bead in Mock-UpⅠ.
From this figure, it can be seen that the bead has an excellent quality and the average
value of bead width is approximately 3.3 mm. Fig. 6 (b) compares the fusion zone
simulated by Case 1 and its corresponding experimental result. The left part is the
experiment result, while the right one is the predicted result. In the right part the ashen
zone shows the area whose peak temperature is higher than or equal to 1450oC
(melting point). In Fig. 6 (b), there are some local discrepancies between the
numerical model and the experiment, but they match each other on the whole. The
reason resulted in local differences perhaps is that the effect of fluid flow was not
considered in the thermal analysis.
This example demonstrates that the fusion zone can be reproduced with a
relatively high accuracy using the combined heat source model developed by the
current study. Here, it should be stressed that accurate prediction of fusion zone is the
prerequisites to obtain good results of welding deformation and residual stress.

4.2 Comparison of welding deformation

As mentioned in Introduction, welding deformation is a serious problem in the
thin-plate welded structures. In this section, the welding deformations predicted by
FEM and experimental measurements are compared.

4.2.1 Comparison of out-of-plane deformation

Fig. 7 (a) and (b) show the contour of deflection distribution predicted by Case 1
and the profile of out-of-plane deformation of Mock-up I, respectively. This figure
indicates that the predictions generally match the measured data both in magnitude
and in distribution. Fig. 8 (a) and (b) show the contour of deflection distribution
computed by Case 2 and that of Mock-up II, respectively. From this figure, it can be
also seen that the numerical results are generally in a good agreement with the
experimental results.
Comparing Fig. 7 and Fig. 8, we can find that the direction of out-of-plane

12
deformation in the joint without external restraint is opposite to that in the joint with
external restraint. Note that Case 1 (or Mock-up I) has a convex (in longitudinal
direction)-concave (in transverse direction) shape after welding, while Case 2 (or
Mock-up II) has a concave-convex shape. The reason resulting in two converse
deformation modes can be explained as follows. The heat input used in Case 1 (or
Mock-up I) is relatively small, so angular distortion is apt to generate in the plate
because of a relatively large temperature gradient through thickness. On cooling, the
neutral plane shifted down due to angular distortion. Combined with the longitudinal
shrinkage force, the plate deformed with a convex shape in longitudinal direction and
a concave one in transverse direction. Fig. 9 (a) shows the mechanism to produce a
convex-concave deformation like Case 1. In Case 2 (or Mock-up II), even through the
heat input is identical to that used in Case 1(or Mock-up I), the angular distortion was
hard to produce during welding because of the external restraints as shown in Fig. 4.
In such situation, the final deformation shape is not controlled by angular distortion,
but is mainly governed by the longitudinal plastic strain distribution through thickness
[4]. Fig. 10 shows the longitudinal plastic strain distributions on the middle
cross-section of Case 1 and Case 2. This figure clearly tells us that the total
longitudinal plastic strain above the neutral plane is larger than that below the neutral
plane in Case 2. Correspondingly, the longitudinal shrinkage force above the neutral
plane is larger than that below the neutral plane [4]. Thus, it is easy to understand that
the final deformation mode in Case 2 (or Mock-up II) is a concave-convex shape. The
mechanisms of deformation are schematically shown in Fig. 9 (b), and it has been
discussed in Ref. [4].
Through comparing Fig. 7 and Fig. 8, we can know that external restraint can
significantly influence the final distortion in thin-plate joints. External restraint not
only can affect the magnitude of deformation, but can change the deformation mode.
The maximum values of z-displacement in Mock-up I and Case 1 are 8.8 mm and 9.0
mm, respectively, while those in Mock-up II and Case2 are 6.0 mm and 6.7 mm,
respectively. Both the numerical results and the measured data suggest that external
restraint can only reduce the out-of-plane deformation to some extent but cannot
13
completely eliminate the deformation.
Based on the above comparisons, it can be concluded that the numerical
simulation results generally are in good agreement with the experimental ones, no
matter whether the external restraint conditions are taken into account or not. This
information indicates that the current computational approach developed based on
ABAQUS code can accurately capture the features of welding deformation in
thin-plate joints.
Fig. 11(a) and (b) show the contours of deflection distribution predicted by Case
3 and the corresponding measurements in Mock-up III, respectively. From Table 3, it
can be known that the heat input used in Case 3 (or Case 4) is about 30 % larger than
that used in Case 1 (or Case 2). The difference of peak temperature through thickness
is small due to the larger heat input, so the plate could not produce a sufficient angular
distortion during welding. In such situation, the final deformation is mainly controlled
by the longitudinal plastic strain distribution through thickness [4]. Therefore, the
deformation mode of Case 3 has a concave-convex shape. The corresponding
mock-up (Mock-up III) also has the same deformation mode.
Fig. 12(a) and (b) show the contour of deflection distribution simulated by Case
4 and the profile of out-of-plane deformation in Mock-up IV, respectively. It is clear
that both Case 4 and Mock-up IV have a concave-convex deformation mode. The
maximum deflections of Case 3 and Case 4 are 12.5 mm and 8.2 mm, respectively,
while those of Mock-up III and Mock-up IV are 12.4 mm and 9.0 mm, respectively.
On the whole, the predictions match the measurements well. Comparing Fig. 12 and
Fig. 13, it can be found that external restraint can reduce welding deformation to
some extent.
Similarly, Fig. 13 shows the contour of deflection distribution predicted by Case
5 and the corresponding measured data of Mock-up V, while Fig. 14 shows the profile
of deflection distribution computed by Case 6 and the experimental results of
Mock-up VI. Similar to Fig. 11 and Fig. 12, these two figure show that the computed
results generally are in good agreement with the experimental measurements. Also,
both simulated results and experimental values suggest that external restraint can
14
reduce the final deformation to a certain extent.

4.2.2 Quantitative comparison of deflection distribution

Fig. 15 shows the defection distributions along Line AB in Case 1, Case 2,
Mock-up I, and Mock-up II. Line AB is defined in Fig. 4. From this figure, it can be
seen that Case 1 matches Mock-up I very well, while there are some differences
between Case 2 and Mock-up II. The reason perhaps is that in the external restraint
conditions of Mock-up II were not accurately modeled by the numerical model (Case
2). From the overall point of view, the numerical results are in a good agreement with
the measured data. In addition, Fig. 15 also clearly shows that the external restraint
conditions not only can reduce the magnitude of the final deformation to some extent,
but also can alter the deformation mode.
Fig. 16 shows the deflection distributions along Line AB predicted by Case 3,
Case 4. The measured data of Mock-up I and Mock-up II are also plotted in the same
figure. Similar to Fig. 15, this figure also indicates that the predictions of Case 3 are
very close to the measured data of Mock-up III, while there are some differences
between Case 4 and Mock-up IV. This information implies that to improve prediction
accuracy the external restraint conditions should be more carefully considered in the
finite element model. How to accurately model the external restraint conditions will
be our important task in the next step.
The maximum deflections of Line AB in Case 3 and Mock-up III are 13.0 mm
and 12.5 mm, respectively, and those in Case 4 and Mock-up IV are 8.0 mm and 9.2
mm. Totally, the numerical models developed in the present work basically reflects
the features of welding distortion in thin-plate joints even under the condition with a
relatively large heat input.
Fig. 17 shows the similar tendency as Fig. 16 does. The difference between these
two figures only is that the magnitudes of deformation in Case 5 (Mock-up V) and
Case 6 (Mock-up VI) are larger because the heat input used in these two cases is 25%
larger than that used in Case 3 and Case 4.
15
 From Fig. 15, Fig. 16 and Fig. 17, it can be known that external restraint can
partially mitigate the deflection, but the residual out-of-plane deformation is relatively
large. Table 7 summarizes the influence of external restraint on the maximum value
of deflection. This table clearly shows no matter whether the external restraint is
considered or not, the maximum deflection increases along with the heat input.
Meanwhile, this table tells us that the efficiency of mitigation of maximum deflection
also grows with the heat input.

4.2.3 Longitudinal shrinkage

In the present study, the longitudinal shrinkage was not measured by experiment.
Here, we only examine the influence of external restraint on the longitudinal
shrinkage based on the numerical results. To quantitatively investigate longitudinal
shrinkage distribution in thin-plate joints, we define several typical lines in the neutral
plane of welded specimen as shown in Fig. 18. Based on the results obtained from the
finite element model, the longitudinal shrinkage distribution along the width direction
can be defined as the difference distribution of X-displacement between the two edges
(Line AC and Line BD).
Fig. 19 shows the longitudinal shrinkage distribution along the width direction of
plate in Case 1, Case 2 and Case 3. It is clear that the longitudinal shrinkage
distributions along the breadth of plate in all three cases are uneven. The magnitudes
of longitudinal shrinkage at the fusion zone and its vicinity are larger, while the value
deceases with the distance from weld centerline increasing. Fig. 19 also indicates that
the magnitude of longitudinal shrinkage increases with the heat input.
Fig. 20 compares the longitudinal shrinkage distributions among Case 2, Case 4
and Case 6. Similar to Fig. 19, the magnitude of longitudinal shrinkage increases
with the heat input. Comparing Fig. 20 with Fig. 19, we can find that the external
restraint conditions have a significant influence on longitudinal shrinkage. For
example, the maximum value and the average value of longitudinal shrinkage in
Case 5 are 2.75 mm and 2.50 mm, respectively; while those in Case 6 are 0.96 mm
and 0.82 mm, respectively. Based the above comparison, it can be concluded that
16
 external restraint can significantly mitigate the longitudinal shrinkage in thin-plate
joints.

4.2.4 Transverse shrinkage

Transverse shrinkage distribution along the length direction of plate is regarded
as the difference distribution of Y-displacement between Line AB and Line CD. Line
AB and Line CD are defined in Fig. 18.
Fig. 21 compares the transverse shrinkage distributions among Case 1, Case 3
and Case 5. Due to the moving heat source and the geometrical end effect, the
transverse shrinkage distribution along welding line is uneven. Similar to the
longitudinal shrinkage, the transverse shrinkage increases with the heat input. Fig. 22
compares the transverse shrinkage distributions among Case 2, Case 4 and Case 6.
From this figure, it can be observed that the final transverse shrinkage increases with
the heat input. Through comparing Fig. 22 with Fig. 21, we have known that the
external restraint conditions can reduce transverse shrinkage. Especially, the
transverse shrinkages around the clamps were mitigated to a large extent.

4.3 Discussions
In the current study, the influence of external restraint on welding deformation
was investigated by means of both experiment and finite element model. Both
experiments and numerical models show that external restraint can mitigate the
magnitude of out-of-plane deformation to some extent, but it seems that the efficiency
of mitigation is not sufficient enough. The experimental and numerical results indicate
that the spring-back deformation is large after the clamps are removed because the
stiffness of thin-plate joint is relatively small. It is not strange that there is still a
relatively large deformation in the thin-plate joints after removing the external
constraints.
Theoretically, the location of clamp can affect the magnitude and distribution of
residual plastic strains, and hence it will also affect the final welding deformation. In
the future work, the influence of the location of clamp (e.g. the distance of clamp
17
from the centerline of weld) on welding deformation will be further clarified.
The results obtained from the present work suggest that it is not easy to effectively
mitigate welding distortion in thin plate welded joint only by using the external
constraints. We have to find other ways to control the welding deformation in a
reasonable range. Therefore, in addition to the accurate prediction of welding
deformation, how to effectively control the welding-induced distortion in thin-plate
welded structure is also an important research subject.

5．Conclusions
In this study, a nonlinear thermo-elastic-plastic FEM based on ABAQUS code was
developed to simulate welding temperature field and welding deformation in
thin-plate joints. In the developed computational approach, the moving heat source,
material nonlinearity, and geometrical nonlinearity were taken into account. The
features of welding deformation were investigated by means of both numerical
simulation and experiment. In addition, the influences of external restraint on welding
deformation in thin-plate joints were clarified. The following conclusions can be
drawn.
(1) The comparison between FEM and experiment shows that the computational
approach developed in the current study can accurately capture the features of
welding deformation in thin-plate joints.
(2) Both the simulated results and the measured data indicate that external restraint
has a significant influence on the final deformation. The present study shows that
external restraint not only can reduce both in-plane deformation and out-of-plane
deformation, but also can alter the mode of out-of-plane deformation. It is a fact
that external restraint can mitigate the final deformation to some extent in the
thin-plate joints, but it seems difficult to satisfactorily eliminate welding distortion
if only using external constraint.
(3) Based on the mechanism of out-of-plane deformation in thin-plate joints, we
found that the mode of out-of-plane deformation seems to be sensitive to heat
source model and its corresponding parameters, bead shape, and reinforcement.
18
 Therefore, when a thermo-elastic-plastic finite element model is used to predict
welding distortion in a thin-plate joint, it is essential to select a reasonable heat
source model, to determine appropriate parameters, and to carefully consider bead
shape as well as reinforcement.

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). The authors also would like to thank Mr. Y. Zhou,
Mr. J. Cai, Mr. J. Sun and Ms J. Liao of Chongqing University for their helps
provided in experiment.

References
[1] Wang JC, Investigation of buckling distortion of ship structure due to welding
assembly using inherent deformation theory, Doctoral thesis, Osaka University;
2012.
[2] Wang JC, Rashed S, Murakawa H, Mechanism investigation of welding induced
buckling using inherent deformation method. Thin-Walled Structures, 80 (2014)
103–119.
[3] Keivani R, Jahazi M, Pham T, Khodabandeh A.R, Afshar M.R, Predicting
residual stresses and distortion during multi-sequence welding of large size
structures using FEM, Int J Adv Manuf Technol, 73(2014) 409–419.
[4] Yupiter H. P, Manurung, Mohd Shahar Sulaiman, Sunhaji Kiyai Abas, Ghalib
Tham, Esa Haruman, Investigation on welding distortion of combined butt and
T-joints with 9-mm thickness using FEM and experiment, Int J Adv Manuf
Technol 77(2015) 775–782.
[5] Deng D, Murakawa H, Liang W, Numerical simulation of welding distortion in
large structures. Computer methods in Applied Mechanics and Engineering,
196(2007) 4613–4627.
19
[6] Deng D, Zhou YJ, Bi T, Liu XZ, Experimental and numerical investigations of
welding distortion induced by CO2 gas arc welding in thin-plate bead-on joints.
Material & Design, 52 (2013) 720–729.
[7] Sun JM, Liu XZ, Tong Y, Deng D, A comparative study on welding temperature
fields, residual stress distribution and deformation induced by laser beam welding
and CO2 gas arc welding. Material & Design, 63 (2014) 519–530.
[8] Deng D, Liang W, Murakawa H, Determination of welding deformation in
fillet-welded joint by means of numerical simulation and comparison with
experimental measurements. Journal of Materials Processing Technology, 183
(2007) 219-225.
[9] Deng D, Murakawa H, Prediction of welding distortion and residual stress in a
thin plate butt-welded joint. Computational Materials Science, 43(2008) 353–365.
[10] Masubuchi K, Analysis of welded structures: residual stress, distortion and their
consequences. Pergamon Press, England Oxford,1980
[11] Watanabe M, Satoh K, On the type of distortion in various welded joints:
shrinkage distortion in welded joint (Report 5), Journal of Japan Welding Society
26(6) (1957) 399-405.
[12] Schenk T, Richardson IM, Kraska M, Ohnimus S, Modeling buckling
distortion of DP600 overlap joints due to gas metal arc welding and the influence
of the mesh density, Computational Materials Science, 46 (2009), 977-986.
[13] Mollicone P, Camilleri D, Gray T, Procedural influences on non-linear
distortions in welded thin-plate fabrication, Thin-Walled Structures, 46 (2008)
1021–1034.
[14] Wang JC, Yin X, Murakawa H, Experimental and computational analysis of
residual buckling distortion of bead-on-plate welded joint, Journal of Materials
Processing Technology, 213(2013) 1447–1458.
[15] Du Z, Chen B, Welding Metallurgy, Beijing, China Machine Press, 1993.
[16] Deng D, FEM prediction of welding residual stress and distortion in carbon steel
considering phase transformation effects, Materials and Design, 30 (2009)
359–366.
20
[17] Goldak J, Chakravarti A, Bibby M, A new finite element model for welding heat
sources, Metallurgical Transactions B, 15B (1984) 299-305.
[18] Lundback A, Runnemalm H, Validation of three-dimensional finite element
model for electron beam welding of Inconel 718, Science and Technology of
Welding and Joining, 10(6) (2005) 717–724.
[19] Radaj D, Heat effects of welding: temperature field. Residual Stress, Distortion:
Springer-Verlag, Berlin, 1992.
[20] Deng D, Murakawa H, Shibahara M, Investigations on welding distortion in an
asymmetrical curved block by means of numerical simulation technology and
experimental method, Computational Materials Science, 48 (2010) 187–194.
[21] Goldak J, Akhlaghi M, Computational Welding Mechanics, Springer, New York,
2005.
[22] Ogawa K, Nakada S, Nishigawa S, Horii Y, Murakawa H, Influences of material
properties and FEM elements on welding residual stress analysis, Symposium of
Welded Structure, Osaka, Japan, Vol. 4, 2006, pp.705-712.
[23] Lindgren L-E, Computational Welding Mechanics, Woodhead Publishing,
Cambrige, 2007.
[24] Ueda Y, Murakawa H, Ma N, Welding Deformation and Residual Stress
Prevention, Elsevier, 2012.
[25] Wang N, Tong Y, Deng D, Effects of welding heat source parameters on residual
stress and distortion in thin plate joint. Transactions of the China Welding
Institution, 33(2012) 97-100.

21
Captions List
Fig. 1 Typical welding–induced deformations in a welded joint
(a) Longitudinal shrinkage (b) Transverse shrinkage (c) Longitudinal bending
(d) Angular distortion (e) Rotational deformation (f) Buckling distortion.
Fig.2 Welding robot system and the external restraint condition
(a) Welding robot system
(b) External restraint condition
Fig. 3 Couplings among temperature, stress and microstructure
Fig. 4 FEM model and boundary condition of Case 3
Fig. 5 Schematic representation of combine source model
Fig. 6 Weld bead and fusion zone of Case 1
(a) Top view of weld bead of Case 1
(b) External restraint condition
Fig. 7 Contours of deflection distribution in Case 1 and Mock-up I
Fig. 8 Contours of deflection distribution of Case 2 and Mock-up Ⅱ
Fig. 9 Mechanisms of out-of-plane deformation
(a) Convex-concave mode
(b) Concave-convex mode
Fig. 10 Longitudinal plastic strain distributions on the middle section
Fig. 11 Contours of deflection distributions in Case 3 and Mock-up Ⅲ
Fig. 12 Contours of deflection distributions in Case 4 and Mock-up IV
Fig. 13 Contours of deflection distributions in Case 5 and Mock-up V
Fig. 14 Contours of deflection distributions in Case 6 and Mock-up VI
Fig. 15 Deflection distributions along line AB of Cases 1, 2 and Mock-ups I, II
Fig. 16 Deflection distributions along line AB of Cases 3, 4 and Mock-ups III, IV
Fig. 17 Deflection distribution along line AB of Cases 5, 6 and Mock-ups V, VI
Fig. 18 Definition of lines in the neutral plane of finite element model
Fig. 10 Longitudinal shrinkage distributions along the width of plate (Cases 1, 3, 5)
Fig. 20 Longitudinal shrinkage distributions along the width of plate (Cases 2, 4, 6)

22
Fig. 21 Transverse shrinkage distributions along the length of plate (Cases1, 3, 5)
Fig. 22 Transverse shrinkage distributions of the length of plate (Cases 2, 4, 6)

23
Figures:

(a) Longitudinal shrinkage (b) Transverse shrinkage (c) Longitudinal bending

(d) Angular distortion (e) Rotational deformation (f) Buckling distortion.
Fig. 1 Typical welding–induced deformations in a welded joint

24
 (a) The welding robot system

(b) External restraint condition

Fig. 2 Welding robot system and the external restraint condition

25
Fig. 3 The coupling of temperature, stress and microstructure

Fig. 4 FEM model and boundary conditions of Case 3

26
 c1

Z
a

c2

Y X

Fig. 5 Schematic representation of combine source model

(a) Top view of weld bead of Case 1

(b) Comparison of fusion zone between experiment and FEM

Fig. 6 the weld bead and fusion zone of Case 1

27
 (a) (b)

Fig. 7 Contours of deflection distribution in Case 1 and Mock-up I

(a) (b)

Fig. 8 Contours of deflection distribution of Case 2 and Mock-up Ⅱ

Neutral plane

Tendon force

(a) Convex-concave mode (b) Concave-convex mode

Fig. 9 Mechanisms of out-of-plane deformation

28
 Neutral plane

(a) Case 1

Neutral plane

(b) Case 2

Fig. 10 Longitudinal plastic strain distributions on the middle section

(a) (b)

Fig. 11 Contours of deflection distributions in Case 3 and Mock-up Ⅲ

(a) (b)

Fig. 12 Contours of deflection distributions in Case 4 and Mock-up IV

29
 (a) (b)

Fig. 13 Contours of deflection distributions in Case 5 and Mock-up V

(a) (b)

Fig. 14 Contours of deflection distributions in Case 6 and Mock-up VI

Fig. 15 Deflection distributions along line AB of Cases 1,2 and Mock-ups I, II

30
Fig. 16 Deflection distributions along line AB of Cases 3, 4 and Mock-ups III, IV

Fig. 17 Deflection distribution along line AB of Cases 5,6 and Mock-ups V, VI

31
 300
C D

100
X
E F

A B

Fig. 18 Definition of compared lines in the neutral plane of finite element model

Fig. 19 Longitudinal shrinkage distributions along the width of plate (Cases 1, 3, 5)

32
Fig. 20 Longitudinal shrinkage distributions along the width of plate (Cases 2, 4, 6)

Fig. 21 Transverse shrinkage distributions along the length of plate (Cases1, 3, 5)

33
Fig. 22 Transverse shrinkage distributions of the length of plate (Cases 2, 4, 6)

34
 Tables:
Table 1 Chemical composition of the Q345 steel (in wt %)

C Si Mn P S Cr Ni Cu Mo

≤ 0.2 ≤ 0.5 ≤ 1.7 0.035 0.035 0.3 0.5 0.3 0.1

Table 2 Chemical composition of filler wire (in wt %)

C Mn Si S P Cu

0.06~0.15 1.4~1.85 0.8~1.15 ≤0.035 ≤0.025 ≤0.5

Table 3 Welding parameters and restraint condition used in each case

Heat input Mock-Up Numerical Welding Arc Welding External

(J/mm) Case current(A) voltage(V) speed(mm/min) restraint

95.3 Ⅰ Case 1 62.0 16.0 500 free

Ⅱ Case 2 62.0 16.0 500 fixed

123.6 Ⅲ Case 3 78.0 16.5 500 free

Ⅳ Case 4 78.0 16.5 500 fixed

154.7 Ⅴ Case 5 91.0 17.7 500 free

Ⅵ Case 6 91.0 17.7 500 fixed

Table 4 Parameters of heat source used in each case

Case a1 b1 c1 a2 b2 c2

Case 1 3.0 2.0 1.2 3.0 2.5 0.3

Case 2 3.0 2.0 1.2 3.0 2.5 0.3

Case 3 3.5 2.4 1.3 3.5 3.0 0.4

Case 4 3.5 2.4 1.3 3.5 3.0 0.4

Case 5 4.0 3.0 1.5 4.0 3.5 0.5

Case 6 4.0 3.0 1.5 4.0 3.5 0.5

35
Table 5 Temperature dependent thermal physics properties of Q345 [22]

Temperature Specific heat Thermal conductivity Density

（℃） (J／g·℃) （J/s/mm/℃） (g/mm3)
0 0.451 0.0457 7.850e-3
100 0.500 0.0450 7.850e-3
200 0.545 0.0440 7.850e-3
300 0.570 0.0425 7.850e-3
400 0.600 0.0400 7.850e-3
500 0.632 0.0367 7.850e-3
600 0.678 0.0345 7.850e-3
700 0.780 0.0256 7.850e-3
800 0.850 0.0200 7.850e-3
900 0.552 - 7.850e-3
1000 - - 7.850e-3
1200 - - 7.850e-3
1400 0.571 0.0261 7.250e-3

Table 6 Temperature dependent mechanical properties of Q345 [22]

Temperature Yield stress Young’s modulus Possion’s ratio Thermal expansion

(℃) (Mpa) (Gpa) (1／℃)

20 345 210 0.3 1.40e-5

100 320 202 0.3 1.42e-5
200 350 197 0.3 1.45e-5
300 324 185 0.3 1.49e-5
400 298 175 0.3 1.53e-5
500 175 150 0.3 1.57e-5
600 150 125 0.3 1.60e-5
700 - 90 0.3 1.64e-5
800 63 65 0.3 1.66e-5
900 - - 0.3 1.70e-5
1000 30 50 0.3 1.75e-5
1200 10 10 0.3 1.80e-5
1400 10 10 0.3 1.80e-5

36
 Table 7 Influence of external restraint on reduction of maximum deflection

Heat input Restraint Maximum Reduction of Efficiency of mitigation

(J/mm) Conditions deflection (mm) maximum (%)

deflection (mm)
95.3 Free 9.0 25.5%
2.3

Fixed 6.7

123.6 Free 12.5 4.3 34.4%

Fixed 8.2

154.7 Free 16.1 7.1 44.1%

Fixed 9.0

37