Research Article

Advances in Mechanical Engineering

2018, Vol. 10(5) 1–12
Ó The Author(s) 2018
Research on welding deformation for DOI: 10.1177/1687814018775885
journals.sagepub.com/home/ade
box girder of bridge crane based on
thermal elasto-plastic theory

Yifei Tong1 , Zhihao Ge1, Xingcheng Zhuo1, Guomin Shen1, Dongbo Li1
and Xiangdong Li2

Abstract
Cranes are mechanical devices widely used to hoist materials in engineering. The box girder of a bridge crane is formed
by welding. During welding, residual stress and welding deformation will be inevitably generated due to the instantaneous
hot temperature, which will affect not only the camber of the crane but also the operational performance of the travel-
ing trolleys. Therefore, understanding the laws governing the welding deformation can lead to better control of camber
cutting of the crane web to avoid unnecessary corrections, reasonably control the geometry shape of the girder and
ensure the good operational performances of the trolleys. To achieve this goal, thermal elasto-plastic analysis method
together with the Abaqus software is used to analyze the welding from the upper and lower covers to the web, from
the upper and lower steel angles to the web, and from the web to the ribs. For the welding simulations, Gaussian mova-
ble heat sources, with the advantages of low computational cost, fast calculation speed, easy convergence, and a reason-
able number of segments, are selected. The simulation results are compared with the experimental results to verify the
reliability, accuracy, and efficiency of the numerical welding model. Finally, the simulation results are discussed, and design
guidance for the web is proposed.

Keywords
Numerical simulation, welding deformation, thermal elasto-plastic analysis, box girder, bridge crane, Abaqus

Date received: 4 November 2016; accepted: 4 April 2018

Handling Editor: Seung-Bok Choi

Introduction and prediction of stress and strain fields for welding

are very important for contributing to an accurate
Cranes are mechanical devices that are widely used to
hoist materials in engineering and construction. The
box girder of a bridge crane is formed by welding. 1
School of Mechanical Engineering, Nanjing University of Science and
Compared to other connection methods, welding is Technology, Nanjing, People’s Republic of China
2
ideal for complex structures, saving metal raw materi- Jiangsu Province Special Equipment Safety Supervision Inspection
Institute, Nanjing, People’s Republic of China
als, simplifying production, and shortening the manu-
facturing cycle. However, welding is a complex and Corresponding authors:
special process characterized by high-temperature, Yifei Tong, School of Mechanical Engineering, Nanjing University of
instantaneous, and dynamic heat transfer. Thus, weld- Science and Technology, 401, Nanjing 210094, People’s Republic of China.
ing residual deformation is inevitably generated during Email: [email protected]
the non-equilibrium heating and cooling and can badly Xiangdong Li, Jiangsu Province Special Equipment Safety Supervision
affect the welded structures design, manufacturing pro- Inspection Institute, Nanjing 210000, People’s Republic of China.
cesses, and performance.1 Therefore, the simulation Email: [email protected]

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering

understanding of the welding stress concentration and welding stress and strain. The indirect method is less
deformation for better control of welding and its pro- accurate than the direct method but has advantages of
cess optimization. short calculation time, flexible application, and suit-
Methodologies of welding deformation prediction ability for large model systems.8
mainly include empirical, analytical, and numerical Following two decades of development, thermal
simulation methods. Empirical methods estimate the elasto-plastic theory has become increasingly mature
welding deformation for simple structures mainly using and can now simulate the changes of the temperature
data profiles and formulas based on previous experi- field and stress–strain field throughout the welding pro-
ments and statistics; these methods face significant lim- cess using iterative simulations.9 However, the compu-
itations in application to complex components. The tational cost of thermal elasto-plastic analysis is quite
analytical approach focuses on the welding residual high, especially for large and complex structures.
stress and deformation based on the classical elasticity However, computational costs can be greatly reduced
theory and several assumptions.2 Numerical simula- by simplifying the model, heat source, and welding pro-
tions have developed gradually with the developments cess, enabling the elasto-plastic analysis of welding
in welding practice, the finite element method, and deformation for large and complex structures.
computer power.3 Usually, numerical simulations are This research seeks to reveal the rules governing the
used to predict the temperature field, residual stress, welding deformation process and obtain an accurate
and welding deformation, which is helpful for the eluci- estimate of the welding deformation for a box girder of
dation of the essential welding processes; the use of a bridge crane. To achieve this goal, thermal elasto-
simulations reduces the test workload and yields accu- plastic analysis is utilized with the Abaqus (Dassault
rate post-welding and stress data.4 Because the simula- Systèmes, Vélizy-Villacoublay, France) software to
tions of welding large and complex components are analyze the welding from upper and lower covers to the
highly nonlinear, these simulations incur a large com- web, upper and lower steel angles to the web, and from
putational cost. the web to the ribs. Gaussian movable heat sources,
Numerical simulations of welding have been carried with the advantages of small computational cost, fast
out domestically and abroad. Many theoretical calculation speed, easy convergence, and a reasonable
approaches have been used to analyze welding, includ- number of segments, are selected for the welding simu-
ing the inherent strain method,5 thermal elasto-plastic lation. The simulation results are compared to the
analysis,6 and visco-elasto-plastic analysis, which con- obtained experimental results to verify the reliability,
siders the coupling effect of phase changes and thermal accuracy, and efficiency of the numerical welding
stress. For the thermal elasto-plastic analysis, finite ele- model. Finally, the simulation results are discussed,
ment software is used to simulate the welding process and design guidance for the web is proposed.
in order to determine the occurrence of welding stress
and strain.7 The thermal elasto-plastic analysis mainly
focuses on heat transfer and thermal stress. While the Finite element analysis theory for welding
temperature field can directly affect the thermal stress
Uneven welding heat can generate a temperature field
on the object whereas the thermal stress exerts a
with a very large gradient around the weldment and
smaller effect on the temperature field, the heat transfer
give rise to the heat flow inside the weldment or
and the thermal stress are coupled and interact with
between the weldment and the ambient medium. The
each other. Therefore, the thermal elasto-plastic analy-
heat flow generally appears in the form of heat conduc-
sis can be carried out in a direct or indirect manner. In
tion, radiation, and convection. The radiation and the
the direct manner, the heat and the structure are
convection normally occur in the welding process and
coupled together for the analysis, and the thermal-
after the weldment is heated, and heat conduction is
mechanism coupling cells are adopted while consider-
the dominant process.
ing the mutual effect of heat and stress to directly
The analysis of the welding temperature field can be
obtain the stress and strain results. However, the direct
considered as a typical nonlinear transient heat transfer
method has the disadvantages of difficulties in the con-
problem. According to Fourier’s heat transfer law and
vergence and long calculation times and is therefore
energy conservation law, a governing equation can be
unsuitable for simulations of large model systems. The
established for the analysis of the heat transfer prob-
indirect method considers the heat transfer as a unidir-
lem, which means that the transient temperature field
ectional coupling process and ignores the effect of
of object T (x, y, z, t) obeys equation (1)
stress and strain on the temperature, which means only
the effect of the temperature field on the stress and      
∂ ∂T ∂ ∂T ∂ ∂T ∂T
strain is considered. The indirect method divides heat kx + ky + kz + Q = rc
∂x ∂x ∂y ∂y ∂z ∂z ∂t
transfer into two processes in order to first obtain the
temperature field and then used to calculate the ð1Þ
Tong et al. 3

 
where r is the material density, Kg m3 ; c is the specific
_
e d e dT1 dT2 dTn T
heat of material, J=(Kg K); kx , ky , and kz are the coeffi- qT = qT = ,  ð9Þ
dt dt dt dt

ð "  T    T    T  # ð
∂N ∂N ∂N ∂N ∂N ∂N
KTe = kx + ky + kz dO + hc N T NdA ð10Þ
∂x ∂x ∂y ∂y ∂z ∂z
Oe se3
ð ð ð
cients of heat transfer in x, y, and z directions, respec- peT = T
rQN dO + T
qs N dA + qs T‘ N T dO ð11Þ
tively, W=(m K); and Q is the heat intensity inside
Oe se2 se3
object, W=kg.
There are three boundary conditions: where KT is the heat transfer matrix, N is a shape func-
tion, PT is the temperature load matrix, and CT is the
1. S1: given temperature of the boundary specific heat matrix.
The solution of the equation for the welding stress
Ts = Ts (x, y, z, t) ð2Þ
and the strain field is considered as a nonlinear transi-
2. S2: given heat flux distribution of the boundary ent problem of the material. Therefore, four fundamen-
tal relations are obeyed when thermal elasto-plastic
∂T ∂T ∂T analysis is performed:10
kx nx + ky ny + kz nz = qs (x, y, z, t) ð3Þ
∂x ∂y ∂z
1. The relation between strain and displacement
3. S3: given heat convection of the boundary
(compatibility condition);
∂T ∂T ∂T 2. The relation between stress and strain (constitu-
kx n x + ky ny + kz nz = hc (T‘  Ts ) ð4Þ tive relation);
∂x ∂y ∂z
3. Equilibrium condition;
where nx , ny , and nz are the direction cosines of outer 4. Response to boundary conditions.
normal on boundary S1; Ts is the given temperature on
boundary S1; qs given heat flux on boundary S2, For thermal elasto-plastic analysis, the following
W=m2 ; hc is the coefficient of heat convection between assumptions are given: the yield strength of material
the object and ambient medium, W=(m2 K); and T‘ is obeys the (von Mises) yield criterion; behavior in the
the ambient temperature. plastic zone obeys the plastic flow rule and the harden-
When using finite elements to calculate the tempera- ing rule; the elastic strain and the plastic strain cannot
ture field of welding, the object O is dispersed into ele- be separated; and temperature-related mechanical prop-
ments Oe to create the temperature field of the elements erties, stress, and strain vary linearly within a small time
and then the variational method is used to solve the increment.
generic function for the minimum. Since the analysis of Thermal stress is caused by thermal expansion due
the welding temperature field can be considered as a to the temperature difference in the object, so that the
transient heat transfer problem, the temperature field relevant physical equations are the following
of the elements changes with time as expressed by
8
>
> 1
T e (x, y, z, t) = N (x, y, z, t)  qeT (t) ð5Þ >
> exx = ½sxx  m(syy + szz ) + aT DT 
>
> E
>
> 1
where T e (x, y, z, t) denotes the temperature field of the >
>
< eyy = ½syy  m(sxx + szz ) + aT DT 
elements; qeT (t) denotes the junction temperature which E ð12Þ
changes with time >
> 1
>
> ezz = ½szz  m(sxx + syy ) + aT DT 
>
> E
>
>
>
> 1 1 1
qeT (t) = ½T1 (t), T2 (t), Tn (t)T ð6Þ : g xy = t xy , g yz = t yz , g zx = tzx
G G G
The following equation is obtained after solving
qeT (t) for the variational extremum After converting to the index form, we obtain

CTe q_eT + KTe qeT = PeT ð7Þ eij = D1 0

ijkl skl + eij
ð13Þ
sij = Dijkl (ekl  e0kl )
where
ð
where e0ij = ½aT DT , aT DT , aT DT , 0, 0, 0T .
CTe = rcT N T NdO ð8Þ Using finite elements for the analysis and solution,
Oe the object is dispersed into the elements to obtain the
4 Advances in Mechanical Engineering

following functional relationship between the mechani- Heat source model for welding
cal parameters and joint displacement The heat source model simulating the distribution of
heat flow during welding is a fundamental condition
u = Nqe ð14Þ
for carrying out a simulation of the welding process. As
e = Bqe ð15Þ a local concentrated heat input, the heat source results
in an uneven temperature field with a very large gradi-
s = D(e  e0 ) = DBqe  De0 = Sqe ent on the structural members, leading to high welding
ð16Þ
 D  aT DT ½1, 1, 1, 0, 0, 0T stress and deformation. Different welding methods can
generate different heat flow distributions, so that the
Where N, B, D, and S represent the shape function, the heat source input model shall be selected on the case-
geometric matrix, the elastic coefficient matrix, and the by-case basis. During the development of the welding
stress matrix of element, respectively, and qe is the joint analysis methodology, many new models have emerged,
displacement matrix. such as the Gaussian, hemisphere, ellipsoid, and double
Virtual displacement and virtual strain can be ellipsoid heat source models.11 The Gaussian model is
obtained by solving the variation applicable to the normal welding method, and even if
 the material performance is nonlinear, the accuracy of
du = N  dqe analysis can be improved. However, for welding with
ð17Þ
de = B  dqe high arc impact, the hemisphere, ellipsoid, or double
ellipsoid model shall be used in order to improve the
The following equations can be obtained using the accuracy.
principle of virtual work The welding process can be considered as the heat
source movement on the structural member. For the
K e qe = Pe + Pe0 ð18Þ simulation, the zone around the weld joint shall be
meshed very well. To increase the computational effi-
where ciency and reduce the dependence on the computer
ð hardware, segmented Gaussian, Gaussian string and
Ke = BT DBdO segmented ellipsoid heat source models have been
Oe developed. When the heat source is segmented for the
ð ð analysis, the computational cost can be reduced by 10
Pe = N T bdO + N T pdA ð19Þ times or more without compromising the accuracy. A
Oe sep segmented heat source refers to a long weld joint that is
ð meshed into many sections. Each section is heated in
Pe0 = BT De0 dO the welding sequence to convert the movement of the
Oe point heat source into that of a segmented heat source
and to reduce the computational cost. In this case, the
and Pe represents the normal load and Pe0 represents the computational accuracy can be increased by increasing
equivalent load of the temperature. the number of the segments. The string heat source is
based on the segmented heat source and loads the point
heat source onto the node of a discrete unit in each sec-
Welding deformation tion to improve the flexibility of this method.
Welding is a local heating process. Due to uneven heat- Generally, point and segmented heat source models of
ing and plastic deformation after the cooling of the heat movement must be realized using subroutines,
weldment, welding stress and deformation occur. The whereas the string heat source can be realized by select-
welding deformation includes the transient deformation ing nodes and increasing the analysis steps.
during welding and the residual deformation at room Literature11,12 reported numerical simulations using a
temperature (the residual deformation is also known as Gaussian heat source, a Gaussian segmented heat
the post-welding distortion). Every structural member source, and a Gaussian segmented string heat source in
can deform due to the uneven heating in each direction the heat movement and compared the obtained results.
and generate longitudinal shrinkage, lateral deforma- Due to the highest computational efficiency, the seg-
tion, bending, angular deformation, buckling distor- mented heat source is preferred for modeling, but the
tion, and so on. These distortions always co-exist and zone around weld joint must be meshed well in order
affect each other. Moreover, the assembly sequence of for this method to converge. The string heat source has
the welding parts, their rigidity and dimensions, the dis- high application flexibility, so it can be used to simulate
tribution of weld joint, and the welding process can the irregular weld joint and does not require fine
affect the welding deformation. meshes. The Gaussian heat source in the movement has
Tong et al. 5

the highest accuracy, but it requires the use of fine convergence and efficiency of the computation. The
meshes and incurs a high computational cost. density, thermal conductivity, and specific heat of the
material must be defined with boundary conditions of
thermal radiation and thermal convection.
Numerical simulation and analysis for In Abaqus, the heat source can be loaded either by
welding based on Abaqus in-element heat generation or using a subroutine. The
Welding simulation and analysis process in-element heat generation loads the heat source onto
the node of the weld joint, and heat is input for the dif-
The indirect method of thermal elasto-plastic analysis ferent nodes in each analysis step to enable those nodes
is used in this article for welding simulation of a crane to transfer heat to other nodes and obtain a variational
girder. First, heat conduction is analyzed to obtain the temperature field. In the subroutine approach, a user
temperature field of the welding process. Then, the subroutine dflux is created that determines the center
obtained temperature field is used to analyze the ther- of the heat source by analyzing the time parameter
mal stress and obtain the welding stress and strain. The
TIME(2). The heat flux of the heat source is calculated
analysis algorithm is shown in Figure 1.
according to the flux function FLUX(1), and the sub-
When heat conduction is used to analyze the tem-
routine is written in Fortran and can be called by sav-
perature field, heat elements must be adopted. For a
ing for. The string heat source is used in this article, so
three-dimensional (3D) model, the 3D solid element
that in-element heat generation is adopted.
DC3D8 or plate elements DS3 and DS4 can be used.
For analysis of thermal stress, 3D solid elements and
When a discrete model is used, since the temperature
plate elements S3 or S4 shall be used for a 3D model.
around weld joint varies greatly, the zone must be
The density, elasticity modulus, Poisson’s ratio, plastic
meshed well (small mesh size). Due to the small tem-
stress and strain, and coefficient of thermal expansion
perature variation, it is not necessary to finely mesh the
are defined as material properties. The temperature
zone far from the weld joint in order to ensure the

Heat transfer analysis Thermal stress analysis

Start

geometry Transfer into

modeling mechanical unit
& Define mechanical
Define material properties of material
properties
Define
Define load steps, boundary conditions
heat source load and temperature
boundary conditions field input temperature
field results
Modify Define heat elements
and meshing

temperature Are welding stress and

field solution welding deformation N
accurate?
temperature Y
field results
Post processing
Y
N accurate
End

Figure 1. Algorithm for numerical analysis of welding.

6 Advances in Mechanical Engineering

 Effective length of the girder: 31.5 m.

 Gauge of trolley: 3580 mm.
 H, the spacing between top and bottom plate:
1700 mm.
 B1, width of the plate: 650 mm.
 B2, the spacing between webs: 590 mm.
 T11/T12, the thickness of top/bottom plate:
24 mm.
 T21/T22, the thickness of left/right web: 6 mm.
 Spacing between large stiffening ribs: 1200–
2750 mm.
 Spacing between small stiffening ribs: 400–
500 mm.

Modeling. The crane girder is a large structural compo-

nent, and a large number of computations are required
for the numerical simulation. To increase the computa-
tional efficiency, the following assumptions are made
to simplify the model: upper and lower covers, webs,
and ribs are flat and straight without internal stress; no
external load or constraint is applied during the weld-
ing; and the solid-state phase change during the weld-
Figure 2. Model of a 50/10-t overhead crane: (a) illustration of ing is ignored.
girder structure and (b) illustration of section attributes. Moreover, since the girder structure is symmetrical
1—top plate, 2—large stiffening ribs, 3—small stiffening ribs, 4—
horizontal angle iron, 5—bottom plate, and 6—webs. in both longitudinal and transverse directions and the
welding is carried out simultaneously in four symmetri-
cal parts, only a quarter of the girder structure was
field analysis result file .odb is input as a predefined analyzed. The steel angles have less influence on the
temperature field for thermal stress analysis so that the deflection of girder and therefore are ignored. Large
stress and strain of the structure in this temperature variation of the welding temperature and the corre-
field can be analyzed. sponding change of the thermophysical properties of
the material can influence the calculation result.
Therefore, a table of listed thermophysical properties is
Numerical simulation for welding deformation of created. The girder is made of the Q235 mild steel, and
its mechanical properties are listed in Table 1.
girder cover and web
3D plate elements are used to mesh the girder, and a
The Abaqus software is used to perform the numerical plurality of element points are set with flexible tempera-
simulations for girder welding on a 50/10-t overhead ture along the thickness of each plate node so that the
crane (shown in Figure 2) in order to obtain the welding temperature can vary not only with the reference plane
deformation of the girder.13–15 Therefore, the blanking of the shell but also with the thickness. The calculation
curve of the girder web can be better controlled. The efficiency of the plate elements is much higher than that
welding between the girder web and the upper and of the solid elements, so the computational cost of the
lower covers can cause a deflection over the girder span, use of 3D plate elements for analysis is lower. Elements
the welding between the web and ribs can cause a lateral DS4 and DS3 are used for heat transfer analysis, and
deformation of the girder, and the welding between the elements S3 and S4 are used for thermal stress analysis.
covers and ribs can cause a surface deflection of the Since a string heat source was used to simulate the
covers. To reduce the welding deformation, first the welding and the welding deformation is the focused of
welding of the web with steel angles and of the covers the present study, a coarse mesh with multiple segments
with the ribs is carried out, followed by the welding of can meet the necessary accuracy. If the meshes around
the covers with the web to form a p-shaped girder. The the weld joint are dense and loose far away from the
welding of the lower cover is performed last. A clamp- weld joint, higher accuracy can be obtained. However,
ing fixture is used to ensure that the web plate and the this irregular meshing can easily cause the temperature
top/bottom plate are tightly fit for positioning welding. at some nodes to be lower than absolute zero during
In this article, the welding deformation of long weld the calculation of the temperature field, leading to the
joints on the girder covers and web was simulated first. lack of convergence. This problem can be avoided by
The parameters of the research object are as follows: adopting the coarse mesh without compromising the
Tong et al. 7

Table 1. Mechanical properties of Q235.

Temperature (°C) 25 250 500 750 1000 1500

Thermal conductivity (W/m/°C) 44 39 33 30 26 20
Total heat transfer coefficient (W/m2/°C) 100 350 520 1000 1500 3000
Density (kg/m3) 7800 7800 7800 7800 7800 7800
Specific heat (J/kg/°C) 460 480 530 675 670 700
E (GPa) 206 187 150 120 70 10
m 0.3 0.3 0.3 0.3 0.3 0.3
ss (MPa) 235 210 150 40 25 2
Thermal expansion coefficient (1/°C) 1.3e–5 1.32e–5 1.39e–5 1.48e–5 1.34e–5 1.33e-5

Selection and loading of heat source. The covers and

ribs of the girder were welded together via T-joints for
which the heat flux cannot be reasonably described by
the Gaussian function, so the Gaussian string heat
source was adopted. The Gaussian heat source is a con-
stant point heat source, the heat flux of which can be
expressed by

q(r) = qm exp (  kr2 ) ð20Þ

3UIh
qm = ð21Þ
pR2
Figure 3. Finite element model of welding. where q(r) denotes the heat flux, J=m2 ; qm denotes the
maximum specific heat flow at the center of the heat
source, J=m2 ; k denotes the energy concentration factor
(1=m2 ) and k = 3=R2 ; R denotes the effective radius of
the heat source, m; r denotes the distance from the cen-
the upper π-shaped
girder ter of the heat source, m; U denotes the welding voltage,
cover
and ribs the upper the lower V; I denotes the welding current, A; and h denotes the
cover and the end heat input efficiency.
web cover
the web and
The integrated heat input per unit time from the
steel angles
Gaussian heat source is ppm =K. If the welding speed is
v, then the heat input within a joint (with length a) is
Figure 4. Welding sequence.
(a=v)(pqm =k). However, if Gaussian heat source is
applied to each node within p this
ffiffiffiffiffiffiffiffijoint,
ffi the integrated
accuracy. A quarter of the girder is taken for meshing heat input per unit time is qm a p=k . Thus, the required
and analysis, and the number of elements is 13,812, as action time for each segmented heat source is the
shown in Figure 3. The x and z directions represent the following
span and the width of girder, respectively. pffiffiffiffiffiffiffiffiffi 1
t= p=k ð22Þ
v
Temperature field calculation. Carbon dioxide arc welding
is adopted with a current of 200 A, voltage of 20 V, This shows that the required action time for each
welding speed of 6 mm/s, and heat input efficiency of segmented heat source simply depends on the energy
0.8.11 The welding sequence for the crane girder can concentration factor and the welding speed. If there are
affect the welding deformation. The commonly used N + 1 nodes on joint a, this means that joint a is
procedure shown in Figure 4 is to first weld the upper meshed into N segments, and thus, the heat input per
cover, followed by the web, ribs, and steel angles to time unit on this joint is as follows
form a p-shaped girder and then to weld the lower qm pffiffiffiffiffiffiffiffiffi
cover. For the welding of a long joint, the cover and Qt = a p=k ð23Þ
n
ribs are welded first; next, a fillet joint is formed from
the center to both ends; and then, the second fillet joint This heat input can be applied on each node except
is formed from both ends to the center. Finally, the for the nodes at both ends, the heat input of which is
lower cover is welded from the center to both ends. equal to Qt =2. After plugging the data into the
8 Advances in Mechanical Engineering

equation, we obtain t = 1:7 s and Qt = 1:5e6 N mm. convection. It is difficult to obtain the convection coef-
Because the heat input of the string heat source is dif- ficient and the radiation coefficient that vary with the
ferent from the heat source in the movement, a certain temperature. However, the overall heat transfer coeffi-
computational accuracy must be ensured. cient is known, and therefore, the overall heat transfer
Each of the joints between the upper cover and the coefficient is set as the convection coefficient. The ini-
web of the girder is divided into five segments; each of tial temperature is set as 20°C, and the latent heat of
the joints between the web and the lower cover is phase change is not considered.
divided into four segments; and each miter joint is con-
sidered as one segment. A total of 16 load steps are cre-
Calculation result of temperature field. An automatic
ated. Each of the first 15 steps is used for analyzing a
time step was used for calculation, which means a small
joint segment in order to complete the welding from
step corresponds to a large temperature variation dur-
the upper and lower covers to the web. Cooling is per-
ing the welding and a large step corresponds to a small
formed in the final step. Each load step is allocated
temperature variation during cooling. This setting can
with a node in a joint segment with the heat source
save calculation time and ensure convergence.
applied, and the time for the load is defined as 1.7 s.
In this work, only the fillet joints from the upper
The selection of the nodes and the execution sequence
and lower covers to the ribs of the girder are analyzed
of the load steps can enable the segmented heat source
to obtain the flexural deflection of the girder after the
to move.
welding. The calculation for the temperature field of
the 3D plate elements of the girder is shown in
Setting of boundary conditions. The boundary condi- Figures 5–7.
tions for the welding process are radiation and

Figure 5. Temperature field of first weld joint.

Figure 6. Temperature field of third weld joint.

Figure 7. Temperature field of 12th weld joint.

Tong et al. 9

It can be seen that the welding temperature of the The obtained strain diagram plots in y direction after
girder reaches 799°C with the heat flux of 8.4 3 103, analysis are shown in Figures 9 and 10.
while the temperature changes with the movement of As shown in the figures, the maximum deformation
the heat source. occurs at the extension part of the girder because this
part is not constrained by the web (free deformation
occurs after welding due to the absence of boundary
Calculation of strain field. Once the stress field is obtained, conditions). The extension part of the upper cover tilts
the indirect method is used to analyze the stress and upward while the extension part of the lower cover tilts
strain fields of the girder. The elements are changed to downward. For flexural deflection of the girder weld-
structural elements S3 and S4, and nonlinear analysis ing, only the deformation on the joint between the
steps are created with the automatic step size. The ini- upper cover and the web in the y direction is consid-
tial temperature in the field is set as 20°C, the tempera- ered. It can be seen that the girder deflects downward,
ture field result file is read and the initial and final and the maximum downward deflection of the upper
analysis steps are set for reading. During the calcula- edge of the web is 1.686 mm, occurring in the middle of
tion, the temperature field result of each analysis step is the span.
obtained to calculate the stress and the strain. To obtain the deflection curve of the girder after
The mechanical properties of the material are welding, we select some discrete points along the joint
defined based on the temperature as well as on the sym- between the upper cover and the web and take the
metry plane constraint in the middle of the span and rightmost point as the origin, the distance from this
the width of the girder. A fully constrained node is also point as the x-axis, and the deformation in the y direc-
defined to avoid the rigid movement of the girder tion as the y-axis. The maximum deflection is
(Figure 8). 1.686 mm, as shown in Figure 11. Because two main
joints are arranged in the upper and lower parts of the
girder, respectively, (the upper joints are double-sided
type and the lower joints are single-sided type) and they
cancel each other out, the welding deformation is
small.

Numerical simulation for welding of web and steel

angle
The welding of the web and steel angle could result in a
Figure 8. Welding constraints. flexural deflection of the web, and it therefore must be

Figure 9. Strain plot in y direction of the girder.

Figure 10. Strain plot along the web.

10 Advances in Mechanical Engineering

Figure 14. Temperature field after cooling of welding of web

and steel angle.

Figure 11. Deformation curve for upper web after girder

welding.

Figure 15. Web deformation during welding of steel angle.

web is laid on a platform for welding, a constraint in

the z direction for the web is set. To avoid rigid move-
ment of the web, a constraint in the y direction at mid-
span is set. The web deformation during welding is
mainly analyzed, as shown in Figure 15. The web
Figure 12. Finite element model of welding for web and steel deflects downward during the welding with the maxi-
angle. mum deflection of 3.71 mm. Because the joint on the
upper steel angle is far from the centroid of the cross
section but the joint on the lower steel angle is close to
the centroid, the welding deformation on the upper
steel angle is larger, resulting in the web deforming
downward.

Numerical simulation for welding deformation of ribs

For the welding of ribs, the upper cover is welded with
the ribs separately, and the centroid is close to the cover
(the joint is close to the centroid), so the small deflec-
Figure 13. Temperature field during welding of web and steel tion of the cover during welding can be ignored. After
angle. the welding between the cover and the web is finished,
the welding of the web with the ribs is started. The
deformation caused by the welding between the web
finished before assembly. The finite element model is and the ribs is mainly analyzed via the model shown in
shown in Figure 12. Provided two weld joints are Figure 2. To simplify the analysis, it is assumed that all
formed simultaneously, each joint is divided into five ribs are welded simultaneously and each weld joint is
segments and the segmented string heat source in the heated up evenly. The temperature field analysis results
movement is adopted. are shown in Figures 16 and 17.
The obtained temperature field analysis results for The same boundary conditions as those described in
the welding are shown in Figures 13 and 14. section ‘‘Numerical simulation for welding deformation
Half of the web is used for analysis, and therefore, a of girder cover and web’’ are adopted, and the con-
symmetry constraint is set at the midspan. Since the straint of the web in the z direction is removed. The
Tong et al. 11

Table 2. Comparison of numerical simulation results with experimentally obtained welding deformations.

Welding Web and steel angles Covers and web Web and ribs

Numerical simulation result (mm) 23.71 21.686 23.33

Experimentally obtained result (mm) 24.61 21.34 24.08

Figure 18. Overall strain plot.

Figure 16. Temperature field during welding ribs.

Figure 19. Strain plot for web.

Figure 17. Temperature field after cooling of welding ribs.

deformation plots for the welding of the web and ribs roughly the same as the experimental result obtained in
are shown in Figures 18 and 19, and only the flexural our previous research.16 Therefore, the thermal elasto-
deflection of the web is considered here. It can be seen plastic finite element analysis is feasible and reliable for
that the web deflects downward after the welding and use in modeling with 3D plate elements. Although sim-
the deflection at midspan is 3.33 mm. For the welding plifications such as the simplifications of model, heat
of the ribs, small ribs are arranged close to the upper source, material properties, and welding process are
part of the web so that the upper part of the web dis- adopted for analysis, this kind of analysis has advan-
plays a larger shrinkage deformation, resulting in tages of fast calculation (short calculation time) and
downward deflection of the web. high accuracy.

2. If the deformation caused by the interaction force

Results and discussion between the parts is not considered during the weld-
ing process, the final welding deformation obtained
1. Based on the welding simulation of the upper as theP sum of all individual deformations is
and lower covers and the web, the girder deflects f= fi =  1:686  3:71  3:3 =  8:7 mm.
downward by 1.686 mm (Table 2). Based on the 3. The camber of the crane girder is considered to
welding simulation of the web and steel angles, compensate for the operation downwarping and thus
the downward deflection is 3.71 mm. Based on can decrease the energy consumption of the trolley
the welding simulation of the web and ribs, the with improved operation performance.
downward deflection is 3.33 mm.
In some studies, it was reported that there is no need
According to the above analysis, the numerical simula- to design the camber because the deformation can be
tion result for the welding deformation of girder is decreased by improving the crane’s stiffness. However,
12 Advances in Mechanical Engineering

currently, the structure of crane tends to be stable with 3. Jesudoss H, Rajesh N, Nagaraj P, et al. Numerical simu-
less room for improvement and limited material selec- lation on joining of ceramics with metal by friction weld-
tions, and the changes of the crane structure are con- ing technique. Int J Modern Phys: Conf Ser 2013; 22:
strained by design and manufacturing costs. However, 190–195.
it is undesirable to enhance the stiffness solely by 4. Bendaoud I, Matteı̈ S, Cicala E, et al. The numerical
increasing the weight of the bridge because of the simulation of heat transfer during a hybrid laser–MIG
welding using equivalent heat source approach. Opt
increased costs. Therefore, it is necessary and feasible
Laser Technol 2014; 56: 334–342.
to prefabricate the camber of the girder. To achieve this 5. Murakawa H, Deng D, Ma N, et al. Applications of
goal, thermal elasto-plastic analysis is used together inherent strain and interface element to simulation of
with the Abaqus software to analyze the welding. welding deformation in thin plate structures. Comp
Gaussian movable heat sources model with a reason- Mater Sci 2012; 51: 43–52.
able number of segments is proposed for welding simu- 6. Le Louëdec G, Pierron F, Sutton MA, et al. Identifica-
lation. The simulation results show that the camber tion of the local elasto-plastic behavior of FSW welds
should be increased by 8.7 mm during the web cutting using the virtual fields method. Exp Mech 2013; 53:
to offset the welding deformation of the girder. 849–859.
Therefore, elucidation of the laws governing the weld- 7. Ueda UY and Murakawa H. Applications of computer
ing deformation can better guide web cutting; obtain and numerical analysis techniques in welding research.
better camber curve; and ensure smooth, safe, and Trans Join Weld Res Inst 1984; 13: 165–174.
8. Chen J, Yu L and Zhe L. Analysis on welding distortion
energy-saving operation of the traveling trolley.
of large complicated structure by thermal elastic-plastic
finite element method. Trans China Weld Inst 2008; 4:
Acknowledgements 72–75+119.
The authors acknowledge the Editor. 9. Schanack F, Ramos OR, Reyes JP, et al. Pantaleón, rela-
tive displacement method for track-structure interaction.
Sci World J 2014; 2014: 397515.
Declaration of conflicting interests 10. Tremolieres R, Lions JL and Glowinski R. Numerical
The author(s) declared no potential conflicts of interest with analysis of variational inequalities. New York: Elsevier,
respect to the research, authorship, and/or publication of this 2011.
article. 11. Cai Z, Zhao H, Lu A, et al. Establishment of line-gauss
heat source model and application in welding numerical
simulation. China Mech Eng 2002; 13: 208–210.
Funding 12. Caoas J, Picon R, Paris F, Blazquez A, et al. A simplified
The author(s) disclosed receipt of the following financial sup- numerical analysis of residual stresses in aluminum
port for the research, authorship, and/or publication of this welded plates. Comput Struct 1996; 58: 59–69.
article: This work was supported by the National Key 13. Dong L. Control technology of welding deformation of
Research and Development Program of China main girder for bridge crane. Spec Equip Safe Technol
(2017YFF0209800 and 2017YFF0209804). The support was 2015; 37: 50–52.
gratefully acknowledged. 14. Cai Z, Lu A, Zhao H, et al. Model of string heat source
and numerical analysis of welding process for the side
plate in 1200T bridge crane. China Mech Eng 2002; 13:
ORCID iD 802–805.
Yifei Tong https://orcid.org/0000-0002-4054-0664 15. Wang W. Optimal design and welding numerical simulation
of box girder of bridge crane. Hebei University of Science
and Technology, Hebei, China, 2010.
References 16. Tong Y, Lu L, Shen G, et al. Camber deformation esti-
1. Huh SC, Lee JS, Kim GN, et al. A study on evaluation mation of bridge crane and energy-consumption analysis.
of residual stress by multilayer welding. International J Proc IMechE, Part B: J Engineering Manufacture 2014;
Modern Phys B 2011; 25: 4285–4288. 2: 313–318.
2. Bonifaz EA and Richards NL. Stress–strain evolution in
cast in-738 superalloy single fusion welds. Int J Appl
Mech 2010; 2: 807–826.