Journal of Constructional Steel Research 145 (2018) 49–57

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Numerical simulation of a welding process using a prescribed

temperature approach
Karlo Seleš a, Mato Perić b, Zdenko Tonković a,⁎
a
University of Zagreb, Faculty of Mechanical Engineering and Naval Architecture, I. Lučića 5, 10000 Zagreb, Croatia
b
Bestprojekt, Bureau of Energetics and Mechanical Engineering Ltd., Petrovaradinska 7, 10000 Zagreb, Croatia

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

Article history: This paper presents an efficient finite element procedure for the prediction of welding-induced residual stresses
Received 21 April 2017 and distortions in large structures. It is based on a prescribed temperature approach using some features of an
Received in revised form 24 January 2018 Abaqus extension called Abaqus Welding Interface to significantly improve the computational efficiency and
Accepted 16 February 2018
speed up the normally time-consuming and cumbersome welding analysis setup performed by the user. To val-
Available online 24 February 2018
idate the temperature and residual stress solutions obtained by the presented method, two numerical examples
Keywords:
are analyzed. Comparison is made with the experimental measurements and the results obtained by the heat
Abaqus Welding Interface generation rate approach using the element birth and death technique. The first example is a butt-welding of
Prescribed weld temperature two plates, while the second is a T-joint fillet welding of two plates. The results obtained by the proposed
Welding process procedure demonstrate a good agreement in comparison with the heat generation rate approach as well as the
Residual stress experimental measurements. Furthermore, the computational efficiency is remarkably improved compared to
Butt-weld the heat generation rate approach as the CPU time is reduced up to ~70% in both examples.
T-joint fillet weld © 2018 Elsevier Ltd. All rights reserved.

1. Introduction method in butt and angular welded plates using temperature-

dependent material properties. Since then, enormous progress has
Welding is one of the basic procedures of metal joining and is still been made in the field of Computational Welding Mechanics (CWM).
widely used in the industry due to its reliability, ease of use and low The influence of the selection of the finite element type, the heat input
cost. Due to the great local heat input during the welding process and model, the choice of base and weld material model, and the boundary
subsequent rapid cooling, permanent plastic deformations appear in conditions, etc., has been analyzed [6–10]. This effort reflects the trend
the weld and its vicinity, causing the residual stresses and dimensional in both the scientific community and the industry to shift the welding
distortions. Such distortions can be a problem during the assembly of residual stress and distortion assessment from an experimental to a
the structure, requiring additional time and financial expenses for the computational-based procedure. A more detailed overview of CWM is
correction of the deformed structure. High tensile residual stresses, in presented in [11–13], and recent CWM standards are discussed in
combination with stresses due to workload, have a significant impact [14,15]. The transient heat transfer problem with geometrical and
on the integrity and lifetime of the structure, as they can accelerate material nonlinearities is solved by employing either fully or sequen-
the initiation of stress corrosion cracking, a fatigue crack or a brittle tially coupled thermo-mechanical analysis. With the assumption that
fracture [1–4]. To minimize these consequences, it is very important to the rate of heat generation due to mechanical dissipation energy is neg-
know the magnitude and distribution of the residual stresses and ligible in the heat transfer analysis, a sequentially coupled analysis is
strains in the structure as early as possible in the design phase. commonly applied in the welding simulation process. The temperature
Therefore, the development of an accurate and efficient technique to field determined in the thermal analysis as a function of time for each
estimate and reduce residual stresses and strains has been the subject integration point is therefore used as an input for the mechanical
of numerous studies. With the rapid development of computer technol- analysis [16–18]. Furthermore, it is well established that 3D models
ogy, numerical welding simulations have become more attractive. Ueda are required for accurate prediction of post-weld deformation and re-
and Yamakawa [5] were among the first to use the finite element sidual stress distribution. These models are often accompanied with
the element birth and death technique to simulate the addition of the
weld material. With the industrial need to solve increasingly more-
⁎ Corresponding author at: Institute of Applied Mechanics, Faculty of Mechanical demanding problems in engineering practice, such complex models
Engineering and Naval Architecture, University of Zagreb, Ivana Lučića 5, 10000 Zagreb,
Croatia.
can be computationally very expensive. This is why current thermal
E-mail addresses: [email protected] (K. Seleš), [email protected] (M. Perić), elastic-plastic finite element simulations are usually limited to small
[email protected] (Z. Tonković). and simple structures.

https://doi.org/10.1016/j.jcsr.2018.02.012
0143-974X/© 2018 Elsevier Ltd. All rights reserved.
50 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57

Due to the limitations of the nonlinear and time-dependent thermal commonly known as the element birth and death technique and is
elastic plastic finite element method, alternative solutions, such as the used to simulate filler metal addition.
inherent strain method based on the elastic material behavior assump- The prescribed temperature is assigned as a boundary condition at
tion, are widely used nowadays [19–21]. Herein, the inherent strains are the interface between the current weld bead chunks and the base mate-
the residual plastic strains caused by the welding process. The main rial or already-deposited weld bead chunks. The temperature boundary
problem in using this method is knowing the distribution of the condition is set for the time duration needed for the torch to pass the
inherent strains in advance. This particularly applies to the modeling corresponding chunk. It is linearly ramped to the targeted prescribed
of different directions and speeds in multipass welding simulations, temperature. In that time, the chunk is still deactivated and is activated
which can be properly simulated by thermal elastic plastic finite only after the torch has virtually moved to the next finite element set
element methods. representing the next discrete chunk in its path. Because an activated fi-
To partially overcome the abovementioned problem of high compu- nite element set has a constant volume temperature, the thermal profile
tation time in the thermal elastic plastic finite element methods, a novel inside the weld is neglected. The main computational advantage of such
method is presented in the first phase of the authors' study [22]. The a method can be seen from the governing finite element equation for
method is based on the 3D to shell elements transition, and it signifi- transient heat transfer analysis
cantly reduces the welding simulation time. In the current study, the
welding simulation time is further reduced using a different heat cT_ þ kT ¼ f T ð1Þ
input method. In general, there are two basic approaches for modeling
the welding heat input. The first is the heat generation rate approach where, in the weld region, there is no need to use the heat flux term on
[23–25] based on the heat flux prescribed on a surface area (surface the right-hand side to calculate temperatures, as they are already pre-
heat input method) or weld element volume (volumetric heat scribed. In Eq. (1), c represents the temperature-dependent heat capac-
input method) or a combination of both. The second is the prescribed ity matrix; k is the temperature-dependent conductivity matrix; fT is
temperature approach [26,27] where a prescribed temperature load is the thermal load vector; and T and T_ are nodal temperature and its
applied over a weld volume. time derivative matrix, respectively. On the other hand, using the stan-
In this study, the prescribed temperature approach is introduced. dard heat flux input takes a significant amount of computational time
Along with reducing the high computation time, it is also important to for the calculation of temperatures in the weld region. This is due to
reduce the user time needed for numerical model preparation, as it the high temperature gradients, which result from the temperature dif-
consists of very time-consuming and repetitive tasks. In that sense, an ference between high peak weld temperatures and its surroundings,
extension to the commercial software package Abaqus [28], the Abaqus entailing finer temporal incrementation.
Welding Interface (AWI), has recently been developed for 2D and 3D When using the prescribed temperature approach, welding parame-
models, and has been tested in [29–31]. In [29,30] the material melting ters such as the welding voltage and welding current are not explicitly
point is used as a targeted prescribed temperature, while in [31] a included as they are with the heat flux input. Therefore, the magnitude
modified parabolic thermal profile is used. The aim of this study is to of the prescribed temperature as a main input parameter cannot be eas-
determine the equivalent weld temperature which is used as the ily determined. In this study, the prescribed temperature magnitude de-
input parameter in the prescribed temperature approach in order to termination is proposed by comparison with the heat generation rate
significantly reduce the welding simulation time while concurrently approach on a smaller scale reference model where the heat flux mag-
maintaining the accuracy. Moreover, the aim is to prove the applicabil- nitude results from the known welding parameters. It can be stated
ity of the proposed method for the numerical analysis of the large that such a prescribed temperature incorporates not only the welding
welded structures used in today's industry. Certain features of AWI are parameters but also the material model and weld geometry.
used in the proposed method, exclusively in thermal analysis, in order
to speed up the numerical model preparation. For verification and 3. Analysis methodology
validation purposes, examples of the butt welding of two plates and
T-joint fillet welding are taken from the literature and authors' previous A short overview of the heat generation rate method and the pre-
study. All the computations have been performed within the FE scribed temperature method used in this study is given in this section.
software Abaqus/Standard. In this study, both methods use the same material properties for both
The paper is organized as follows. The numerical model is described base and weld filler metal, and constant radiation and convective heat
in Section 2. Section 3 provides a brief review of the used analysis pro- transfer coefficients. Welding speed is included with the size of the
cedure. In Sections 4 and 5, the efficiency and accuracy of the prescribed finite element set in the weld direction and the time duration of the
temperature approach are demonstrated on the examples of the butt heating step. As mentioned in the Introduction, the major distinction
welding of two plates and T-joint fillet welding. The results and discus- between the two methods is the heat input definition. The first is a sim-
sions are given in the Section 6. Finally, some concluding remarks are ple, commonly used method explained in [22,32] that incorporates the
given in the last section. sequential deposition of finite elements (i.e., element birth and death
technique), with constant heat flux over the given set of finite elements
2. Numerical model describing the weld bead chunk. It will be called the CHF (Constant Heat
Flux) method. The second method uses the PWT (Prescribed Weld
AWI is an Abaqus/CAE Plug-In based on the model-tree approach. It Temperature) approach as previously explained. The mechanical
provides a graphical user interface to set up the welding simulation analysis is performed with two options: M1, which is very similar to
from within Abaqus/CAE, i.e., to set up the generation of heat transfer the mechanical analysis of the CHF method incorporating the element
properties included in the welding process. In this study, AWI is used birth and death technique, and M2, which is a simple nonlinear
in thermal analysis to build the thermal interactions, steps and thermal mechanical analysis performed by simultaneous deposition of finite el-
boundary conditions as it automates these repetitive tasks usually ements, i.e., without the “model change” option. In both mechanical
performed by user. A sequentially coupled thermo-mechanical analysis analysis options, the initial temperature condition of the base material
is utilized. The prescribed temperature approach is adopted, is set to room temperature, while the element sets describing the
representing the heat input coming from the welding torch. The weld weld beads are set to the prescribed weld temperature used in the
beads are deposited in discrete chunks along the welding path. The thermal analysis. Table 1 presents the list of used methods.
chunks are represented by the finite element sets and are activated Since the CHF method has already been validated on several exam-
through the Abaqus feature called “model change”. The activation is ples in engineering practice, it will be taken as a reference solution for
 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57 51

Table 1
The list of used methods.

Method Thermal analysis heat input Mechanical analysis FE weld

name (by sequential deposition) bead deposition

CHF Constant heat flux Sequential deposition

PWT+M1 Prescribed weld temperature Sequential deposition
PWT+M2 Prescribed weld temperature Simultaneous deposition

obtaining the main parameter of the second method: the prescribed

weld temperature magnitude. In other words, to convert the welding
parameters successfully used for generating the heat flux in the CHF
method, an iterative procedure is conducted to obtain a correct equiva-
lent temperature that can be used as the prescribed weld temperature. Fig. 2. Thermal properties of SM400A steel [19].
The proposed procedure consists of a comparison of the temperature,
stress and displacement response against the CHF method solutions.

4. Butt-welding of two plates

As a first example, the butt-welded joint of two SM400A steel plates

taken from [33] is considered to validate the proposed PWT method.
The geometry of the welded plates is shown in Fig. 1. Phase changes
are not taken into consideration, as their impact was proven to be neg-
ligible for low-carbon steel [34,35]. The base material temperature-
dependent thermal and mechanical properties are shown in Figs. 2
and 3, while the chemical compositions of the base and weld
filler metal are presented in Table 2. In the mechanical analysis, an
elastic-perfectly plastic material behavior model is assumed. The
welding process is modeled as a MAG single pass weld with the follow-
ing parameters: welding current I = 240 A, welding voltage U = 28 V
and welding speed v = 300 mm/min. The following values are chosen
for the other parameters: the convective heat transfer coefficient k = Fig. 3. Mechanical properties of SM400A steel [19].
10 W/m2 K, the efficiency of heat input η = 85%, and the emissivity
coefficient ε = 0.9. root of the weld. The mesh consists of 9120 elements with the weld
The finite element mesh is shown in Fig. 4. Numerical simulations bead divided into 60 element sets in the weld direction, describing
are conducted by the CHF and PWT methods using three different pre- discrete weld bead chunks. To correctly implement the welding speed,
scribed weld temperatures. Three-dimensional, 8-node solid elements the length of each set is 5 mm in the weld direction, and the step refer-
are used: DC3D8 for the thermal analysis and C3D8 for the mechanical ring to its heating lasts 1 s.
analysis. The smallest elements are situated at the root of the welded With the welding parameters and weld chunks as stated above, the
joint. The size of the elements increases with the distance from the heat flux used in the CHF method is 5.0 × 1010 J/m3 s. A comparison of
the temperature profiles and stress distribution between the CHF
method and the three different prescribed temperature magnitudes in
PWT is given in Section 6. Furthermore, the residual stress distributions
are compared with the experimental measurements presented in [33].

5. T-joint welding of two plates

The next step is to check the prescribed temperature approach when

the weld is introduced into a model with a more complex geometry.
Thus, the T-joint fillet weld analyzed by the authors of this contribution
in [22,38] is considered. Fig. 5 shows the geometry of two T-joint fillet
welded plates with the welding sequence, restraint conditions and ex-
perimental measurement locations.
The material of the plates is nonalloyed structural steel S355JR, with
temperature-dependent thermal and mechanical material properties
adopted from Pilipenko [39] shown in Fig. 6 and Fig. 7, respectively.
The chemical compositions of the base and weld filler metal are
shown in Table 3.

Table 2
Chemical composition of SM400A base metal [36] and JIS Z3312 YGW11 wire metal [37].

Chemical composition (mass%)

C Si Mn P S Ti + Zr

Base metal 0.23 – 0.56 b0.035 b0.035 –

Weld wire 0.07 0.89 1.60 0.020 0.025 0.2
Fig. 1. Geometry of welded plates.
52 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57

Fig. 4. Finite element mesh of the welded plates.

The plates are welded with the MAG process using the wire with the Fig. 6. Thermal properties of S355JR steel [39].
diameter of 1.2 mm classified as EN ISO 14341-A. The filler material is G
42 4 M/C 3Si1 with the yield stress of 420 MPa [41]. The cooling period
after the first pass is 215 s. The material is modeled as elastic–perfectly
plastic. As described in the authors' former work [38], the tack welding
in the experimental procedure is performed for the purpose of fixing
plates together before the beginning of the welding process. Although
the tack welding is conducted, there is still a small gap between the
horizontal and vertical plates. In the numerical models, the influence
of the initial gap and the arrangement of the tack welds are not taken

Fig. 7. Mechanical properties of S355JR steel [39].

into account. Here, it is assumed that these effects have the same influ-
ence on all examined numerical models. As presented in [42], the initial
gap in the T-joint weld has a small effect on the longitudinal bending,
but it may have an effect on the angular distortion of plates. Further-
more, when the initial gap is small, the surfaces remain in contact dur-
ing welding. Thus, the use of contact models is required for post-weld
deformation and residual stress distribution [43]. However, such a
complex model can be computationally expensive or can even exceed
available resources. Therefore, a numerical T-joint welded model with-
out the initial gap is considered that gives acceptable residual stresses
and distortions compared with experimental measurements [44]. The
following thermal material properties are assumed: convective heat
transfer coefficient k = 10 W/m2 K, efficiency of the heat input η =
80%, and emissivity coefficient ε = 0.9. The finite element mesh,
obtained after the mesh sensitivity analysis using the submodeling
technique conducted in the authors' previous work [38], is shown in
Fig. 8. The presented mesh density provided satisfactory results within
the reasonable computational effort. To discretize the geometry, again,
the DC3D8 and C3D8 elements are used. The mesh consists of 22,176 fi-
nite elements with the weld bead divided into 150 element sets, de-
scribing discrete weld bead chunks. To correctly implement the
welding speed, the length of each set is 6.667 mm in the weld direction,
and the heating step lasts 1 s. As in the previous example, the size of fi-
nite elements becomes larger as the distance from the root of the weld

Table 3
Chemical composition of S355JR base metal [40] and G 42 4M/C 3Si1 wire metal [41].

Chemical composition (mass%)

C Mn Si Cr Ni Al P S

Base metal 0.19 1.05 0.2 0.008 0.11 0.006 0.0028 0.02
Weld wire 0.08 1.5 0.09 – – – 0.015 0.012
Fig. 5. Geometry of two T-joint welded plates and experimental measurement locations.
 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57 53

and 45 mm, respectively, from the middle surface of the vertical plate.
The plate deflections during welding are measured using the noncon-
tact 3D digital image correlation system ARAMIS 4 M (from GOM
mbH) on the upper surface of the horizontal plate.
For the residual stress measurements, another T-joint fillet weld is
made to completely eliminate the influence of thermocouple holes on
the residual stress state. The residual stresses are measured on the
horizontal plate's bottom surface by applying the hole-drilling strain
gauge method of stress relaxation according to the ASTM E837-95 stan-
dard [45] using the Vishay RS200 device (from Vishay Precision Group).
The strain gauge rosettes HOTTINGER type 1,5/120 RY 61 were used.
Herein, the incremental hole-drilling using a 1.6 mm hole diameter
was carried out at three locations. The measurement locations denoted
as ML-1, ML-2 and ML-3 are offset 5.5 mm, 25 mm and 70 mm from the
midplane of the vertical plate. The thermocouples and residual stress
Fig. 8. Finite element mesh of T-joint welded plates. measurement locations are given in Fig. 5.

6. Results and discussion

Table 4
Welding parameters.
This section contains detailed quantitative comparisons of the ex-
Welding current Arc voltage Welding speed Fillet weld leg
perimental and numerical results from the literature with the numerical
I = 270 A U = 29 V v = 400 mm/min 7 mm results obtained in this study. To verify the temperature and residual
stress calculations performed by the prescribed temperature approach,
the results are first compared with those obtained from the heat gener-
increases. A uniform heat flux distribution per weld chunk volume, with ation rate approach. Herein, to define the equivalent weld temperature
the magnitude rounded to 4.0 × 1010 J/m3 s, is used in the CHF method. to be used as the input parameter in the prescribed temperature ap-
The welding parameters are given in Table 4. proach, the heat generation rate approach is used as a reference solu-
The numerical simulation procedure is the same as in the previous tion. The measured temperatures, stresses and displacements are then
example. First, the CHF method is performed. It provides the reference compared with the results obtained using both the CHF and PWT
solutions to obtain the equivalent weld temperature to be used as the methods.
input parameter in the PWT method. A comparison of the temperature
profiles, plate deflections and stress distribution between the CHF and 6.1. Butt-welded plates
PWT methods using three different prescribed temperature magnitudes
is given in Section 6. Fig. 9 shows temperature profiles in butt-welded plates for both the
The measured temperature field and plate deflections are presented CHF and PWT methods 30 s, 35 s and 60 s after the start of welding
in [22], while the measured residual stresses are shown in [38]. Here, along line A–A (Fig. 1). As may be seen from Fig. 9a, the obtained max-
only the most important information is reiterated. For the surface imum temperature in the weld using the CHF method is 3030 °C, and
temperature measurements during T-joint welding, an SC2000 infrared the temperature profile has a typical parabolic shape in the weld bead.
(IR) camera (from FLIR Systems AB) is used. In addition, to evaluate the On the other hand, the temperature profile obtained using the PWT
obtained thermograms, temperature measurements are simultaneously method with three different prescribed temperatures ranging between
conducted by thermocouples (Type K). Five thermocouples are placed 1550 °C and 1900 °C shows that the weld bead temperature is constant.
inside the material 6 mm from the bottom edge of the horizontal plate The temperature profiles in the cooling process 5 s and 30 s after the
(Fig. 5). Thermocouples labeled TC-102 and TC-104 are positioned on electrode passes line A–A are presented in Fig. 9b. Here, it can be seen
lines A–A and C–C seen in Figs. 5, 18 mm from the middle surface that the PWT temperature profiles have typical parabolic profiles in
of the vertical plate. Furthermore, thermocouples on line B–B, the weld bead. The PWT profiles with the prescribed temperature
labeled TC-103, TC-105 and TC-106, are positioned 18 mm, 30 mm magnitude of 1725 °C are nearly identical to the temperature profiles

Fig. 9. Comparison of temperature profiles for the CHF and PWT methods along line A–A (Fig. 1): a) 30 s and b) 35 s and 60 s after the beginning of the welding process.
54 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57

Fig. 12. Comparison of CPU time of methods used in butt-welded plates.

by 54.1% compared with the CHF method. Moreover, the computa-

tional time is reduced by 72.7% when using the PWT + M2
Fig. 10. Residual stress profiles in the welding direction for the CHF method and for the
method.
PWT + M1 and PWT + M2 methods using the chosen equivalent temperature of 1725 °C.
6.2. T-joint weld of two plates

of the CHF method. For this reason, this temperature is taken as an Fig. 13 shows temperature profiles along line B-B seen in Figs. 5, 290
equivalent for the residual stress calculations of the butt-welded s and 365 s after the start of welding, i.e., at the start and end of welding
sample. the second pass. It can be seen that the solutions obtained by the PWT of
Fig. 10 shows the residual stress profile in the weld direction along 1325 °C match those obtained by the CHF method. A comparison of the
line B–B (Fig. 1) for the CHF, PWT + M1 and PWT + M2 methods to- temperature history on the thermocouple measurement locations for
gether with the residual stress measurements presented by Kim et al. the CHF and PWT methods (Fig. 14) shows that their temperature his-
[33]. It can be seen that the residual stresses obtained by the PWT + tory is nearly identical. The same conclusion can be drawn from
M1 and PWT + M2 methods are almost equal to the residual stresses Fig. 13a and b. For this reason, the temperature of 1325 °C is considered
obtained by the CHF method. The residual stress measured in the as the equivalent PWT for the plate deflections and residual stress calcu-
center of the weld significantly deviates from the numerically calculated lations. Experimental measurements using the thermocouples and IR
values. The reason is that the plates are welded with the wire JIS Z3312 camera are shown in Fig. 15. Although there seems to be a discrepancy
YGW11 which has a significantly higher yield stress value (527 MPa) between the experimental and numerical results up to ~10%, it can be
[37] than the base material (i.e. widely used overmatching procedure), concluded that the numerical calculations qualitatively match well the
as stated in the other paper of Kim et al. [46], but this fact is neglected in experimental measurements. This discrepancy can be attributed to the
the present analysis. With this simplification in mind, it can be stated simplification of the heat input (heat source is presumed constant over
that the results in the weld vicinity and further away match well with the weld volume in CHF) and constant convective and radiation coeffi-
the experimental measurements. cients. Although the temperature measurement using the IR camera is
Fig. 11 presents the residual stress fields obtained by the CHF, PWT + carried out on the top surface of the horizontal plate while the thermo-
M1 and PWT + M2 methods on the upper surface of the plates. It can be couples are in its depth, the results are comparable because the thermo-
seen that the residual stress fields for all the used numerical methods are couples are far enough from the weld, i.e., the temperature profile
nearly the same. through the plate thickness at the measurement points is approximately
Because the aim is to develop a computationally inexpensive constant.
procedure that still provides correct results, the CPU time con- The deflection profile of the horizontal plate at the middle surface is
sumed during the computations is presented in Fig. 12. Using the shown in Fig. 16. It can be concluded that the plate deflections of CHF
PWT + M1 method (Table 1), the computational time is reduced and both PWT options (Table 1) match very well. The experimentally

Fig. 11. Residual stress fields in the welding direction: a) CHF method, b) PWT + M1 method, and c) PWT + M2 method.
 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57 55

Fig. 13. Comparison of temperature profiles for the CHF and PWT methods along line B–B (Fig. 5): a) 290 s and b) 365 s after the beginning of the welding process.

Fig. 14. Comparison of temperature time histories at the thermocouple positions of the PWT method using 1325 °C temperature and the CHF method.

obtained plate deflections are very close to those obtained using the nu- Fig. 18 shows the CPU times spent on the numerical calculations.
merical calculations. As mentioned in Section 5, plate deflections are Again, a great reduction of 53.8% and 69.2% in CPU time is obtained
measured using the noncontact 3D digital image correlation system using the PWT + M1 and PWT + M2 methods compared to the CHF
ARAMIS 4 M on the upper surface of the horizontal plate. The results method, respectively. In addition, the user time spent in prepara-
of the deflection in the middle surface and on the upper surface of the tion of numerical models with the PWT + M1 or PWT + M2
horizontal plate are comparable, while the change in its thickness is methods is minimal compared to the CHF method, as AWI auto-
negligible. mates most of the common repetitive tasks in building the ther-
The residual stress profile on the lower plate surface (Fig. 17) along line mal models.
B–B (Fig. 5) is plotted together with the experimental measurement data. The PWT method is conceived as an efficient engineering
Fig. 17 shows two peaks corresponding to the locations of the weld beads, method to be used on large-scale welded structures where the
i.e. concentrated heat input. Similar results were obtained in [47,48] for point of interest lies not in the weld but its vicinity, neglecting
two pass fillet welding. Almost the same plate deflection is obtained the fine solution distribution inside the weld bead area. The tran-
using the CHF, PWT + M1 and PWT + M2 methods, and all the numerical sient heat transfer is governed by the difference between the base
residual stresses are very close to the experimentally obtained ones. material temperature and the prescribed temperature. The

Fig. 15. Comparison between the temperature histories obtained using the thermocouples and the IR camera 290 s and 403 s after the beginning of the welding process.
56 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57

7. Conclusions

In the presented research, the FE procedure is developed for the

prediction of welding-induced residual stresses and distortions for
application in large and complex structures. It incorporates the widely
used element birth and death technique and follows the prescribed
temperature approach for heat input modeling. The efficiency and
accuracy of the proposed method are demonstrated on examples of
the butt-welding of two plates and T-joint fillet welding. The conclu-
sions are as follows:

• The PWT method leads to great time savings in the numerical simula-
tion of the welding process compared to the commonly used CHF
method.
• The choice of prescribed temperature has a substantial effect on the
computed residual stresses and plate deflections, as the temperature
Fig. 16. Deflection profile of the horizontal plate along line B–B (Fig. 5). time history is used as a load in mechanical analysis.
• The prescribed weld temperature should be seen as an equivalent
temperature for the used material model, welding parameters and
weld geometry.
• Simultaneous deposition can be successfully employed instead of se-
quential filler deposition in the mechanical analysis of the welding
simulation. This can further accelerate the simulation process while
maintaining the computational accuracy.

As a general conclusion, it can be noted that the proposed method

can serve as an effective tool for predicting distortions and residual
stresses during the welding of large and complex structures. In future
research, further application of the proposed method will be extended
to combined shell/3D models.

Acknowledgment

The investigations described herein are part of the project “Centre of

Fig. 17. Residual stress profile of the horizontal plate in the welding direction along line B–B Excellence for Structural Health” (CEEStructHealth) supported by the
(Fig. 5).
EU under contract IPA2007/HR/16IPO/001-040513.

References
[1] S.H. Oh, T.Y. Ryu, S.H. Park, M.G. Won, S.J. Kang, K.S. Lee, S.H. Lee, M.K. Kim, J.B. Choi,
prescribed weld temperature should therefore be seen as an Evaluation of J-groove weld residual stress and crack growth rate of PWSCC in
equivalent temperature for the used material model, welding pa- reactor pressure vessel closure head, Int. J. Mech. Sci. 29 (3) (2015) 1225–1230.
[2] X.B. Ren, Z.L. Zhang, B. Nyhus, Effect of residual stresses on ductile crack growth
rameters and weld geometry. The proposed procedure would
resistance, Eng. Fract. Mech. 77 (2010) 1325–1337.
then be the determination of the equivalent temperature on a ref- [3] J. Predan, R. Pippan, N. Gubeljak, Fatigue crack propagation in threshold regime
erent smaller scale model of a large structure before numerically under residual stresses, Int. J. Fatigue 32 (2010) 1050–1056.
[4] D. Schnubel, N. Huber, The influence of crack face contact on the prediction of fa-
tackling the large structure.
tigue crack propagation in residual stress fields, Eng. Fract. Mech. 84 (2012) 15–24.
All numerical simulations are performed on an Intel® Core™ i7-4820 [5] Y. Ueda, T. Yamakawa, Analysis of thermal elastic-plastic stress and strain during
CPU @ 3.70 GHz with 64 GB RAM memory. welding by finite element method, Trans. JWS 2 (2) (1971) 90–100.
[6] G. Fu, I.L. Marcelo, M. Duan, F.E. Segen, Effect of boundary conditions on residual
stress and distortion in T-joint welds, J. Constr. Steel Res. 102 (2014) 121–135.
[7] Y. Ye, J. Cai, X. Jiang, D. Dai, D. Deng, Influence of groove type on welding-induced
residual stress, deformation and width of sensitization region in a SUS304 steel
butt welded joint, Adv. Eng. Softw. 86 (2015) 39–48.
[8] J. Ju, F. Xue, J. Zhou, J. Bai, L. Sun, Interface and bond strength of brazing cemented
carbide K20 to alloy steel AISI 4140 by high-frequency induction, Mater. Manuf.
Process. 31 (8) (2016) 1052–1060.
[9] Y. Rong, G. Zhang, Y. Huang, Study of welding distortion and residual stress consid-
ering nonlinear yield stress curves and multi-constraint equations, J. Mater. Eng.
Perform. 25 (10) (2016) 4484–4494.
[10] M.H. Nozari, S. Ziaee, K. Jafarpur, Experimental study of produced thermal stresses
in thin arc welded plates: application of constructal theory in welding paths,
Weld. World 60 (1) (2016) 51–59.
[11] D. Radaj, Welding residual stresses and distortion calculation and measurement,
DVS-Verlag GmbH, Düsseldorf, Germany, 2003 (ISBN 3871557919).
[12] J. Goldak, M. Akhlaghi, Computational Welding Mechanics, Springer-Verlag, New
York Inc., USA, 2005 (ISBN 9780387232874).
[13] L.E. Lindgren, Computational Welding Mechanics, Woodhead Publishing, UK, 2007
(ISBN 1845692217).
[14] Guide for Verification and Validation in Computation Weld Mechanics. , American
Fig. 18. CPU time comparison of different methods used on T-joint welded plates. Welding Society, USA, 2013.
 K. Seleš et al. / Journal of Constructional Steel Research 145 (2018) 49–57 57

[15] R6, Revision 4. Assessment of the Integrity of Structures Containing Defects. , EDF [32] M. Perić, D. Stamenković, V. Milković, Comparison of residual stresses in butt-
Energy, UK, 2010. welded plates using software packages Abaqus and Ansys, ESA Sci. Tech. Rev.
[16] Y. Luo, W. Jiang, W. Zhang, Z. Wei, Q. Zhang, Y. Wan, Y.J. Wang, Experimental and (ISSN: 18200206) 60 (3–4) (2010) 22–26.
numerical study on the reduction of residual stress in the fillet weld by overlay [33] Y.C. Kim, J.Y. Lee, M. Sawada, K. Inose, Verification of validity and generality of
welding and cutting method, J. Press. Vessel. Technol. 138 (6) (2016), 060401-7. dominant factors in high accurate predicting of welding deformation, Q. J. Jpn.
[17] D. Venkatkumar, D. Ravindran, 3D finite element simulation of temperature Weld. Soc. 25 (2007) 450–454 (in Japanese).
distribution, residual stress and distortion on 304 stainless steel plates using GTA [34] D. Deng, FEM prediction of welding residual stress and distortion in carbon steel
welding, J. Mech. Sci. Technol. 30 (1) (2016) 67–76. considering phase transformation effects, Mater. Des. 30 (2009) 359–366.
[18] J. Shen, Z. Chen, Welding simulation of fillet-welded joint using shell elements with [35] P. Knoedel, S. Gkatzogiannis, T. Ummenhofer, Practical aspects of welding residual
section integration, J. Mater. Process. Technol. 214 (11) (2014) 2529–2536. stress simulation, J. Constr. Steel Res. 132 (2017) 83–96.
[19] D. Deng, H. Murakawa, W. Liang, Numerical simulation of welding distortion in large [36] L. Gannon, Y. Liu, N. Pegg, M. Smith, Effect of welding sequence on residual stress
structures, Comput. Methods Appl. Mech. Eng. 196 (45–48) (2007) 4613–4627. and distortion in flat-bar stiffened plates, Mar. Struct. 23 (2010) 385–404.
[20] T.S. Jun, A.M. Korsunsky, Evaluation of residual stresses and strains using the [37] http://www.goodweld.com.tw/upload/product/th-128.pdf (last visit 3.1.2018).
eigenstrain reconstruction method, Int. J. Solids Struct. 47 (13) (2010) 1678–1686. [38] M. Perić, Z. Tonković, I. Garašić, T. Vuherer, An engineering approach for a T-joint
[21] J. Wang, S. Rashed, H. Murakawa, Y. Luo, Numerical prediction and mitigation of fillet welding simulation using simplified material properties, Ocean Eng. 128
out-of-plane welding distortion in ship panel structure by elastic FE analysis, Mar. (2016) 13–21.
Struct. 34 (2013) 135–155. [39] A. Pilipenko, Computer Simulation of Residual Stress and Distortion of Thick
[22] M. Perić, Z. Tonković, A. Rodić, M. Surjak, I. Garašić, I. Boras, S. Švaić, Numerical Plates in Multi-electrode Submerged Arc Welding. Their Mitigation Techniques
analysis and experimental investigation in a T-joint fillet weld, Mater. Des. 53 (Doctoral Thesis, Trondheim, Norway), 2001.
(2014) 1052–1063. [40] W. Piekarska, M. Kubiak, A. Bokota, Numerical simulation of thermal phenomena
[23] M. Starý, F. Novotný, M. Horák, M. Stará, Z. Vít, Experimental optimization of tab and and phase transformations in laser-arc hybrid welded joints, Arch. Metall. Mater.
slot plug welding method suitable for unique lightweight frame structures, J. Manuf. 56 (2) (2011) 409–421.
Process. 31 (2018) 453–467. [41] http://www.wutecschweisstechnik.de/fileadmin/Bilder/download/WM079500_
[24] M. Perić, Z. Tonković, I. Karšaj, D. Stamenković, A simplified engineering method for SZW_Handbuch.pdf (last visit 3.1.2018).
a T-joint welding simulation, Thermal Science, VINCA Institute of Nuclear Sciences, [42] C. Wang, Y.R. Kim, Y.W. Kim, Comparison of FE models to predict the welding distor-
2018, https://doi.org/10.2298/TSCI171108020P (ISSN 03549836), (in press). tion in T-joint gas metal arc welding process, Int. J. Precis. Eng. Manuf. 15 (8) (2014)
[25] R.L. Lostado, M.C. Bobadilla, M.A. Martínez Calvo, P.M. Villanueva Roldán, Residual 1631–1637.
stresses with time-independent cyclic plasticity in finite element analysis of welded [43] Z. Barsoum, A. Lundbäck, Simplified FE welding simulation of fillet welds - 3D effects
joints, Metals (Basel, Switz.) 7 (4) (2017) 1–25. on the formation residual stresses, Eng. Fail. Anal. 16 (2009) 2281–2289.
[26] T. Krausche, B. Launert, Pasternak, A study on the prediction of welding effects in [44] A.A. Bhatti, Z. Barsoum, H. Murakawa, I. Barsoum, Influence of thermomechanical
steel box girders, Proceedings of EUROSTEEL 2017, Copenhagen, Denmark, 2017. material properties of different steel grades on welding residual stresses and
[27] K. Nguyen, R. Nasouri, C. Bennett, A. Matamoros, J. Li, A. Montoya, Sensitivity of pre- angular distortion, Mater. Des. 65 (2015) 878–889.
dicted temperature in a fillet weld T-joint to parameters used in welding simulation [45] ASTM Standard E837-95, Standard Test Method for Determining Residual Stresses
with prescribed temperature approach, Proceedings: Science in the Age of by Hole-drilling Strain Gage Method, ASTM Philadelphia, USA, 1995.
Experience 2017, Chicago, USA, 2017. [46] Y.C. Kim, J.Y. Lee, K. Inose, Dominant factors for high accurate prediction of distor-
[28] Abaqus/Standard 6.14.1, Dassault Systemes. Simulia, 2014. tion and residual stress generated by fillet welding, Int. J. Steel Struct. 7 (2007)
[29] M. Shubert, M. Pandheeradi, An Abaqus extension for 3-D welding simulations, 93–100.
Mater. Sci. Forum (2014) 768–769. [47] G. Fu, S.F. Estefen, T. Gurova, M.I. Lourenco, Effect of material model on residual
[30] C. Parmar, C.M. Gill, P.R. Hurrell, B.M.E. Pellereau, Simulation of a multi-pass groove stress and distortion in T-joint welding, Ships Offshore Struc. 13 (1) (2018) 56–64.
weld and clad plate using Abaqus 2D weld GUI and comparison with [48] H. Pasternak, B. Launert, T. Krausche, Welding of girders with thick plates - fabrica-
measurementsSIMULIA UK Regional User Meeting, 2014. tion, measurement and simulation, J. Constr. Steel Res. 115 (2015) 407–416.
[31] A.S. Elmesalamy, H. Abdolvand, J.N. Walsh, J.A. Francis, W. Suder, S. Williams, L. Li, Mea-
surement and modelling of the residual stresses in autogenous and narrow gap laser
welded AISI grade 316L stainless steel plates, Int. J. Press. Vessel. Pip. 147 (2016) 64–78.