Article

Computational Techniques in Numerical Simulations

of Arc and Laser Welding Processes
Tomasz Kik
Department of Welding Engineering, Silesian University of Technology, Konarskiego 18A,
44‐100 Gliwice, Poland; [email protected]; Tel.: +48‐32‐1681

Received: 2 December 2019; Accepted: 20 January 2020; Published: 29 January 2020

Abstract: The article presents a comparison of modern computational techniques used in numerical
analyses of welding processes. The principles of the “transient” technique calculations with a
moving heat source, the “macro‐bead” (MBD) technique, with an imposed thermal cycle on a
selected weld bead section and the “local–global” approach with shrinkage calculation technique
were described. They can be used, depending on the variant chosen, both for individual, simple
weld joints and those made of many beads or constructions containing dozens of welds and welded
elements. Differences in the obtained results and time needed to perform calculations with four
different calculation examples of single and multipass arc and laser beam welding processes were
presented. The results of calculations of displacements and stresses distributions in the welded
joints using various computational techniques were compared, as well as the calculation times with
the described techniques. The numerical analyses in the SYSWELD software package have shown
the differences between the described computational techniques, as well as an understanding of the
benefits and disadvantages of using each of them. This knowledge allows preparing an efficient and
fast optimization of the welding processes, often aimed at minimizing deformations in the first
place, as well as detection of potential defects of both simple and complex welded structures. In
general, the possibilities and flexibility of modern numerical calculation software have been
presented.

Keywords: FEM; numerical analyses; simulations; computational techniques; SYSWELD;

displacements; stresses

1. Introduction
Terms such as simulation, calculations or numerical solutions have been known to engineers for
a long time, not only in technical fields but also in medicine or natural sciences. Of course, with the
development of electronics, at this time, great progress was made both in terms of the hardware itself
and the possibilities that the modern software offers us. Many branches of technology can now show
almost comprehensive numerical solutions, from the design stage, production preparation, through
the stage of determining technological and production conditions, to the overall optimization of the
production process. It is also possible to simulate the actual or assumed conditions of use of the
product together with determining its durability. The use of modern manufacturing techniques such
as laser or electron beam welding places requirements that are increasingly difficult to meet.
Increasing requirements regarding the accuracy of element preparation and maintaining minimum
dimensional deviations during and after the welding process mean that we readily use modern
computational tools [1–7].
The development of numerical analysis software has recently focused mainly on advanced
second‐generation simulation calculations. These types of simulations support a unique way of
creating the final product using various industrial technologies. In this way, it is possible to simulate
the entire production process and its relationships with the limit states of structures and materials.

Materials 2020, 13, 608; doi:10.3390/ma13030608 www.mdpi.com/journal/materials

Materials 2020, 13, 608 2 of 18

This is a typical path, e.g., in the automotive industry, where the processes of sheet metal forming,
body welding and in the last stage of crash tests are analyzed. However, in the case of numerical
analyses of welding processes, despite the long and intensive development of computational
techniques, it is still difficult to achieve all of the mentioned operations in one calculation program
[1–5,8].
Since welding is one of the most widespread technological processes, this technology has also
received a lot of attention when developing modern simulation tools. The problem with the
simulation of this process is the fact that the most frequently moving heat source is the reason for the
uneven distribution of displacements and stresses in welded elements [5–7]. However, they are not
only dependent on the moving heat source itself. Their values and distributions are also strongly
dependent on many other technological factors such as clamping condition, mechanical and thermal
properties, type of the welding technology and welding parameters, preheating temperatures, weld
joints design, temperature of surroundings, etc. The residual stresses in the structure after welding
have a negative influence on the durability of the structure and its reliability under various operating
conditions. The degree of complexity of the process itself and the non‐linearity of phenomena that
occur in the welded element during its heating and cooling, and their dependence on the type of
material and technology used, significantly increase the degree of complexity of numerical analyses
[5–7,9,10].
Currently, there are two views on the issue of conducting and solving tasks in the field of
numerical analysis of welding processes. The first of them is associated with a detailed, often partial
solution of selected effects of the welding process, consisting of determining the shape and
dimensions of the weld molten pool, heat affected zone (HAZ) and related metallurgical and
mechanical changes in the area of the welded joint [1–5]. The result of this type of analysis is always
a detailed description of the solution, unfortunately only in relation to a specific, individually
considered data and process parameters. The second approach is to apply numerical analyses to
large, complete parts and production units. Due to the complexity of the structure, this method must
contain a number of simplifications, which to some extent reduce the resulting accuracy of
calculations. Despite this disadvantage, this allows conducting analyses that in the first approach,
even despite the currently available computing power, are practically impossible to perform,
precisely because of the number of variables that should be taken into account. Not without
significance, in this case, is the constant focus on reducing production costs and reducing to the
absolute minimum the time needed to prepare a new product [11–14].
The use of numerical analyses today allows for a better understanding of the production process,
finding relationships that connect individual process parameters to its results and increases product
quality. It also leads to a significant cost reduction in the production preparation stage by limiting
and often excluding costly prototypes. This is particularly important when making samples or
prototypes involving a lot of material, time and are energy‐consuming. Based on the results of
numerical simulations, it is possible to determine the optimal solution, even if their results are not
100% consistent (in terms of value) with the results of real tests. Thanks to the constant development
of modern simulation software, it offers us today various possibilities of approach to the issue of
numerical analyses, allowing for the mentioned analyses within both one joint and entire
constructions. In the following, based on the SYSWELD software from ESI Group, typical
computational techniques are presented. Each of them are successfully used in today’s modern
computational packages [1–5,8,11–18].

2. Description of the Problem

As it was already mentioned, depending on the selected calculation method, numerical analyses
of welding processes carried out in the SYSWELD environment can be divided into the following
groups:
Local analysis: Used to determine the distribution of temperature fields, metallurgical phases as
well as stresses and deformations within one welded joint;
Materials 2020, 13, 608 3 of 18

Global analysis: This applies to the entire structures and usually involves changing the size of
the structures (displacements) and stresses distributions around the welds [1–4].
In the case of welding, essentially the only load on the workpiece is the non‐stationary
distribution of the temperature field both on its surface and in many cases in thickness. Thus, the
temperature in the model is determined by means of a variable function in time and dimensional
space. Accurate determination of the temperature field distribution is, therefore, the basic and most
important step in determining its influence on the metallurgical properties of the material and the
stresses and deformation distributions of the welded structure. In order to refine the record of
changes in temperature function in time and space, it is necessary to define a mathematical model of
a heat source [1]. This description, however, brings with it a lot of problems due to the complexity of
the issue. There is a need to take into consideration many input parameters and apply some
simplifications in the description of this model. When creating it, the influence of convection in the
liquid metal pool, active elements, type of protective gas, wire diameter or method of transferring
metal in the arc are not taken into account. Models of heat sources can be divided into three basic
groups while defining the degree of complexity of the analyses carried out [1]:
One‐dimensional;
Two‐dimensional;
Three‐dimensional.
Therefore, depending on the dimension of the heat source model used, point, linear or flat
models of heat sources were used in the analyses. Their advantage was low “resource consumption”
in the calculations. The limitations of these heat source models resulted primarily from the available
computing power. Nowadays, when we have sufficient computing power, in most cases we almost
exclusively use three‐dimensional models [5,11–18].

2.1. Transient Technique: Moving Heat Source

In the case of local analyses, when the moving heat source model is used, the calculations are
carried out by the so‐called “transient” technique, also called the “step‐by‐step” technique.
Calculations are made in this case for each subsequent time increased by a given time step, which is
automatically adjusted depending on the mesh density of the model or set by the user. The
mathematical model of the heat source in this type of analysis moves along the line that determines
the welding trajectory. The parameters used to describe the heat source model, in this case, are the
actual welding process parameters, i.e., welding current, arc voltage, welding speed or thermal
efficiency of the method, etc. So, this is mostly data that can be found, among others, at the Welding
Procedure Specification Form (WPS) but also are part of the knowledge of every welding engineer.
One of the most popular models used in current numerical analyses is the volumetric double ellipsoid
model, also known as Goldak’s model [1,5,15–20]. It consists of two ellipsoids which, placed in two
perpendicular planes, create an image of the weld pool in both horizontal and vertical space (Figure
1). Transferred heat into a volume is described by Equations (1) and (2) [1,5,19,21]:
for the front part of the heat source model, the Equation is:

6√3𝑓 𝑄 k𝑥 l𝑦 m𝑧
𝑄 𝑥, 𝑦, 𝑧 𝑒𝑥𝑝 𝑒𝑥𝑝 𝑒𝑥𝑝 (1)
𝑎𝑏𝐶 𝜋√𝜋 𝑎 𝑏 𝑐

and for the rear part of the heat source model, the Equation is:

6√3𝑓 𝑄 k𝑥 l𝑦 m𝑧
𝑄 𝑥, 𝑦, 𝑧 𝑒𝑥𝑝 𝑒𝑥𝑝 𝑒𝑥𝑝 (2)
𝑎𝑏𝐶 𝜋√𝜋 𝑎 𝑏 𝑐

where Qf, Qr are volumetric heat flux density in front and rear part of the model [W/m3]; Q is total
power source; a b, cf, cr are width, depth and length of the front and the rear part of the molten pool;
Materials 2020, 13, 608 4 of 18

ff, fr are constants which influence energy flow intensity into the material (ff + fr = 2); and k, l, m are
coefficients enabling modification of the liquid metal pool shape.
The efficiency of the heat transfer into parent material is given by the applied welding method
[20]. For the proper heat source calibration, it is necessary to compare the results with experimentally
determined and measured values. It is mainly the overall heat, the geometrical parameters of the
molten pool and the efficiency of the heat transfer from the source into the material. After specifying
the input parameters, it remains to add the constants ff, fr affecting the distribution of energy flow to
the material and choose the coefficients k, l, m. Practice shows that we usually accept intercourse of
ff to fr as 60:40. The choice of coefficients k, l, m is more complicated. If the unmodified heat source
model is used, all three factors will be equal to three. The unmodified source is particularly suitable
for simulated welding with coated electrodes. In the case of welding simulation, e.g., with the GMAW
method, the model must be modified to obtain the correct shape of the liquid metal pool and the
values of the coefficients chosen experimentally [1,5,15–22].

Figure 1. Double ellipsoid and 3D conical heat source models [1,5,19].

While Goldak’s source successfully covers the scope of numerical analyses of arc welding
processes, in the case of laser or electron beam welding, the conical model with a Gaussian
distribution is currently most commonly used. Its conical shape allows good reproduction of welding
processes carried out using welding heat sources with a high concentration of energy (for example
laser and electron beam welding). The mathematical description of this model can be represented by
two Equations [1,5,22]:

𝑥 𝑦
𝑄 𝑥, 𝑦, 𝑧 𝑄 𝑒𝑥𝑝 (3)
𝑟 𝑧
𝑟 𝑟
𝑟 𝑧 𝑟 𝑧 𝑧 (4)
𝑧 𝑧
where Q0 is the maximum value of volumetric heat flux density; re, ri are upper and lower 3D cone
radius dimensions parameters; ze, zi are the 3D cone length parameters; and x, y, z are point
coordinates (Figure 1).
Equation (3) describes the volumetric heat flow density into the material in the dependence of
coordinate data. Equation (4) supplements Equation (3) by the definition of the radius change in the
direction of the depth.
This way of defining a heat source model and conducting numerical analyses based on a moving
heat source requires its appropriate calibration procedure (Figure 2). Only in this case will it be
possible to recreate with it the actual distribution of temperature fields that affect the metallurgical
transformations, stress and deformation distributions that occur. It is a long‐term process but can be
carried out on partially simplified models (including 2D cross‐section models). Incorrect calibration
of the heat source model leads to duplication of errors during this stage and simulation, which in
turn leads to incorrect results.
Materials 2020, 13, 608 5 of 18

Figure 2. An example of heat source model parameters calibration. Comparison of the calculated
shape and size of the molten zone in the laser welding process with a real macrograph of laser weld
bead (laser beam power 2000 W, welding speed 0.5 m/min).

The analysis carried out using the “transient” technique is divided into three parts: preparation
of the material base, calculation of thermal phenomena and related metallurgical changes as well as
mechanical phenomena (i.e., stresses and displacement distributions, etc.) [1–4]. Taking into account
the fact of performing calculations at each subsequent time, the user after completing the analysis
receives a powerful set of thermo‐metallurgical and mechanical data related to the simulated process.
The price of such a complex solution is the extended duration of calculations. This is the main reason
why the calculation of large welded structures with many welds uses other calculation techniques.
However, such a large amount of data, including distribution of temperature fields, hardness,
metallurgical phases, stresses and deformations makes it an ideal method by which the already
mentioned local effects of the welding process can be determined. Based on the results of these
analyses, it is also possible to conduct further simulations regarding the impact of external loads, i.e.,
forces, moments and pressures, as well as e.g., assessment of fatigue life of a structure based on the
Dang Van criterion [1–4,16,18].

2.2. Macro‐Bead Technique: Imposed Thermal Cycle

As it was already mentioned, calculations using the “transient” technique are practically
impossible to use, from the point of view of the amount of data generated during them, in the case of
numerical analyses of large or complex welded structures. In such cases, the “macro‐bead” (MBD)
technique is used [1,3,4,23,24], which is in a way an extension of the “transient” technique and
involves the use of a properly prepared thermal cycle immediately on one or several areas (elements)
of the model simultaneously (Figure 3).

Figure 3. Imposed thermal cycle set on selected nodes of the model in calculations using the macro‐
bead (MBD) technique.

In this technique, the actual welding trajectory is divided into sub‐areas so that the welding
order and directions are maintained. The number of these sub‐areas and the time step are defined on
Materials 2020, 13, 608 6 of 18

the basis of technological parameters of the welding process as well as the experience of the person
working with this method [1–4]. The standard procedure is to prepare the thermal cycle based on a
simple model “transient” analysis result (often using also 2D cross‐section models) or as a result of
thermocouple measurements. This cycle is then applied successively to the mentioned sub‐areas of
welds.
The difference is that in the case of the “transient” technique with a moving heat source, it is
needed as many time steps (calculation cards) as there are elements in the grid on the welding
trajectory. When applying the imposed thermal cycle, required are as many steps as the points
describing the given thermal cycle in each of the sub‐areas (Figure 4). In addition, the mesh can be
modified and divided into a significantly smaller number of finite elements because there are no
temperature gradients in the middle of the weld section at the time of the applied thermal cycle. They
only occur at the beginning and end of the bead (sub‐section). Therefore, in the MBD technique, the
mesh is only refined at the edges of the section of elements on which the thermal cycle is applied. In
their center, due to the lack of thermal gradient, the mesh is much thinner. Thus, for example, 100‐
time steps in the calculation with the “transient” technique can be replaced with a maximum of 20–
40 time steps in the MBD technique, which significantly reduces the calculation time (Figure 4).

Figure 4. Differences between the number of simulation time steps for the “transient” and MBD
analysis.

Based on the MBD technique, it is also possible to perform calculations using the “local–global”
method [1,3,4]. Its main idea is the assumption that the welding process leads to local changes in
stresses distribution and plastic deformations, while the effect of this on a global scale is a specific
state of deformations. The local effects of the welding process are determined in this case by means
of precise calculation models of single welded joints using the “transient” or MBD techniques. When
the joints are repetitive (welded with the same parameters), the calculations are significantly
accelerated. The results of local analyses are then transferred to the global model to determine the
total deformations of the structure. The limitation of this method, however, is that the results of the
analyses carried out on the global model are only structural deformations as well as internal forces
and moments under specific mounting conditions. The stresses level and distribution of individual
metallurgical phases are determined by local models obtained as a result of analyses carried out using
the “transient” or MBD technique [1,3,4,23,24].

2.3. Shrinkage Technique: Distortion Engineering

The application of the “local–global” method in the case of very large and complicated welded
constructions may have been problematic even despite the acceleration of calculations, which is
provided by the use of the MBD technique [1,3,4]. In this case, it is also possible to use the “shrinkage”
technique, which, despite many simplifications, allows to determine the deformation of the structure
in situations where the standard techniques and even MBD is not possible to carry out due to too
long calculations time. In accordance with the assumptions of the “local–global” method, in the case
of calculations carried out using the “shrinkage” technique, calculations of thermal phenomena and
related metallurgical changes are not carried out. Only the results of the analysis of mechanical
Materials 2020, 13, 608 7 of 18

phenomena occurring during the welding process are used. In a simplified way, it can be described
that the thermal cycle first causes an expansion of a certain area of the material as a result of its heating
and then shrinkage, which is caused by specific cooling conditions and heat dissipation from this
zone. If during the cooling process some areas of the joint are subject to shrinkage, it is the user’s task
to find a representative value of the force that causes this shrinkage. In order to find this
representative force causing the contraction, it is necessary to carry out the calibration process. For
this reason, a special local model is created (for individual, selected joints from the whole structure),
on which a classic “transient” analysis is performed and deformations are determined in specific
places of the model. Then, the same model is calculated using the “shrinkage” technique. The
deformations obtained as a result of the analysis are compared to the model from the classical analysis
and the input data are calibrated to obtain similar values. The parameter that changes the value of
the representative force for this calculation method is the radius of the pipe/sphere encompassing the
model grid nodes at a certain distance from the assumed trajectory of the weld (weld axis) (Figure 5).
All mesh nodes inside the area enclosed by its radius will be subjected to contraction. By increasing
the value of this radius, it is possible to set the conditions of the shrinkage impact in such a way that
the results of the local model of this technique come close to the results obtained in the local analysis
of the “transient” type. After the calibration, it can be transferred to the model of the entire structure.
In the case where the structure has a large number of the same joints (joint type and lateral dimensions
of the weld), due to its repeatability, the calculation process is significantly accelerated in time as it
was the case with the “local–global” method using MBD techniques [1–4,23,24].

(a) (b)
Figure 5. View of a fragment of the T‐joint model with (a) a representative shrinkage force and (b)
marked trajectory (weld axis) with shrinkage area (tube) [4].

Analyses carried out using the “shrinkage” technique allows for quick analysis of structure
deformations with a very large number of welded and welded joints, e.g., with a laser beam, as well
as optimization of the order of their execution in terms of minimizing deformations. The method
allows for an efficient search for a compromise between the clamping conditions (the degree of
stiffening of the structure) and the level of stresses and deformations as well as determining the
optimal welding sequence, which can additionally be a complicated, complex sequence of variable
fastening of elements and increasing rigidity of the structure as the assembly progresses [1,3,4,23,24].

3. Calculation Examples
To show the differences between the results and the duration of calculations obtained using
individual techniques, as well as to introduce the described issues, several examples of analyses were
carried out using the techniques described above. All calculations were made on a computer with an
Intel i5‐8750H processor and 16 GB of operational memory in VisualWeld 14.5 (SYSWELD 2019.0
solver produced by ESI Group, Paris, France).
Materials 2020, 13, 608 8 of 18

3.1. Numerical Analysis of a T‐Joint with one Fillet Weld Arc Welding Process
In order to compare the results obtained with the described techniques, a calculation model of
the T‐joint of sheets 4.0 mm thick and 200 mm long made of S355 steel GMAW welded with one fillet
weld was prepared. As a material database was used standard S355 database included in the
SYSWELD material database. The finite element (FE) model consisted of 26,172 elements and 20,923
nodes in the case of calculations using the “transient” method, and the heat source model used was
a double ellipsoid model. Heat source parameters were set as follows: energy per unit length, 380
J/mm; welding speed, 5.0 mm/s; arc heat efficiency, 0.7; length/width/depth of Goldak’s model,
8.0/4.0/2.0 mm. In the case of calculations with the imposed thermal cycle, the FE model consisted of
13,982 elements and 11,193 nodes, and in the case of calculations using the “shrinkage” method,
respectively, 4550 2D elements and 4464 nodes (Figure 6a–c). The thermal cycle used in the
calculation of the MBD technique and the calibration of the shrinkage method model was carried out
for auxiliary analysis using the 2D cross‐section model (Figure 7). Mesh was fined in the weld area
to increase calculation accuracy. Clamping conditions during welding were set to simulate welding
without any additional mounting. It means that were used three nodes in which the possibilities of
displacement in three, two and one of the directions were blocked (Figure 6d). This type of boundary
condition simulates a situation where the elements lie freely on the welding table without additional
clamping during the welding and cooling process. Boundary conditions correspond to heat
dissipation to the environment on each external surface and the heat transfer coefficient correspond
to the cooling conditions in the free air with an ambient temperature of 20 °C.

(a) (b)

(c) (d)
Figure 6. Finite element (FE) models prepared for the comparison of computational techniques for (a)
“transient”, (b) MBD, (c) “shrinkage” techniques and (d) clamping conditions of welded elements.

(a) (b) (c)

Figure 7. Temperature fields distributions comparison on the cross‐sections of arc welded T‐joint: (a)
calculated using the “transient”, (b) MBD technique and (c) an example of imposed thermal cycle
used in the MBD technique calculations.
Materials 2020, 13, 608 9 of 18

The calculated distortion distributions in all three techniques differed to some extent in the size
distribution, but the maximum values and the deformation tendency of the joint were similar (Figures
8–10). As was observed, the stresses values in the case of the “transient” technique were higher;
however, the maximum value is usually influenced in this case by the beginning and end of the weld,
while in the remaining areas they are much lower (Figure 8). It also results in a sense from the
definition of the start and end of the heat source model movement (certain inaccuracies related to the
definition of the amount of heat in these areas can be improved by using a power ramp). This
difference cannot be seen in the other two other techniques. However, when the “macro‐bead”
technique is used, there are characteristic peaks of the stresses values at the points where the
subsequent sections are joined at which the heat cycle is given (Figures 9 and 10). The precision of
stresses state calculations for these last two methods is much lower due to the simplifications used
and should rather be treated as approximate values.

(a) max. 0.78 mm (b) max 601 MPA

Figure 8. Distributions of (a) normal displacements and (b) von Mises stresses in arc welded T‐joint
calculated using the “transient” technique.

Figure 9. Distributions of (a) normal displacements and (b) von Mises stresses in arc welded T‐joint
calculated using the Macro‐Bead Distortion Engineering (MBD DE) technique.

Figure 10. Distributions of (a) normal displacements and (b) von Mises stresses in arc welded T‐joint
calculated using the “shrinkage” technique.
Materials 2020, 13, 608 10 of 18

Depending on the calculation method used, obtaining the result requires a certain time. A
comparison of calculation times is presented for all variants in Table 1. If the MBD technique is used,
it is also possible to carry out the “local–global” method (without the calculation of the temperature
and metallurgical phases distributions), therefore the time for this case is also given as Macro‐Bead
Distortion Engineering (MBD DE). It should also be added that the times needed to calibrate the
computational models in MBD and the “shrinkage” technique were not included. However, these
operations can be carried out on simplified models, e.g., 2D cross‐section (same as for the “transient”
heat source model calibration), which significantly reduces the computation time compared to the
variant using the classical “transient” technique. In the case for the complex structure wherein there
are many of the same joints/welds, a significant acceleration of calculations is visible because the
calibration on a separate model is made only for one, especially when the changes do not apply to
technological parameters (welding current, welding speed, etc.).

Table 1. Summary of the duration times of numerical analyses for tested calculation techniques of arc
welding of a T‐joint model with one fillet weld.

Transient MBD (MBD DE)1 Shrinkage

2h 44 min 2 s 12 min 10 s (119 s) 55 s
1 MBD DE—Macro‐Bead Distortion Engineering for the “local–global” approach using the MBD
calculation technique.

3.2. Numerical Analysis of the T‐Joint Multipass Arc Welding Process with Three Fillet Welds
In the case of multipass welded joints, it is not possible to use the models made in the shell
technique (made of 2D elements). However, it is still possible to use the MBD technique as a tool that
significantly reduces computation time allowing the analyses of more complex structures. Therefore,
a T‐joint model was prepared for S355 steel sheets 10 mm thick and 200 mm long, GMAW welded on
one side, with three fillet welds. In both computational techniques, the same model (without
changing the density of the mesh inside the sections, characteristic of the MBD technique) was used
as a computational model consisting of 49,780 elements and 41,814 nodes. For the material database,
the standard S355 database included in the SYSWELD material database was used. Boundary
conditions replies to clamping conditions during welding were set to simulate welding without any
additional mounting as it was described in the previous example and shown in Figure 11. Boundary
conditions corresponding to heat dissipation to the environment also were on each external surface
as free air cooling and the model was cooled to ambient temperature at a level of 20 °C. The calibration
of the heat source model was carried out on the 2D cross‐section model and on this basis, the heat
cycle for MBD calculation was also prepared (Figures 11 and 12). As the heat source model in the
“transient” technique was used Goldak’s heat source model. Heat source parameters were set as
(similar for each bead): energy per unit length, 450/350/390 J/mm; welding speed, 8.0 mm/s; arc heat
efficiency, 0.7; length/width/depth of Goldak’s model, 12.0/8.0/4.0 mm. The area of elements affected
by the heat source model (in the SYSWELD nomenclature called LOAD) was additionally determined
separately to get the right expected shape of the molten area of each bead.

Figure 11. A view of (a) the 2D cross‐section of the FE model, (b) imposed thermal cycle prepared for
the calculation by the MBD technique and (c) clamping conditions of welded elements.
Materials 2020, 13, 608 11 of 18

Figure 12. Comparison of temperature field distributions on the cross‐sections of the multipass T‐joint
calculated using the (a) “transient” and (b) MBD techniques.

The calculated displacements, similarly to the previous case, slightly differed in the size and
value distribution; however, they are similar to the place and direction of occurrence (Figures 13 and
14).

Figure 13. Distributions of (a) normal displacements and (b) von Mises stresses in arc welded
multipass T‐joint calculated using the “transient” technique.

Figure 14. Distributions of (a) normal displacements and (b) von Mises stresses in arc welded
multipass T‐joint calculated using the MBD technique.
Materials 2020, 13, 608 12 of 18

The calculated stresses values were also similar. As for the distribution of these stresses, also in
this example, the “transient” technique offers more precise results, while in the case of the “macro‐
bead” technique, the user also receives more general information about the order of height and
location of these stresses in the joint (Figures 13 and 14). In addition, in the case shown, no value
peaks are visible along the length of the joint because the heat cycle has been applied along the entire
length of each simulated beads (Figure 14).
The summary of calculation times by the compared techniques/methods are presented in Table
2. A comparison of thermal cycles of the welding process calculated by the “transient” and MBA
techniques showed considerable similarities in their course as well as values (Figure 15).

Table 2. Summary of the duration times of numerical analyses for tested calculation techniques of T‐
joint with three arc welded beads.

Transient MBD MBD DE1

1h 8 min 2 s 20 min 26 s 11 min 28 s
1 MBD DE—Macro‐Bead Distortion Engineering for the “local–global” approach using the MBD

calculation technique.

Figure 15. Thermal cycles in selected nodes of the model, calculated using the “transient” (moving
heat source) and MBD (imposed thermal cycle (ITC)) techniques.

3.3. Numerical Analysis of the Laser Welding Process of Pipes Butt Joints
When using laser techniques, it is also possible to use the described computational techniques.
Especially in this case, when the number of welds made using a laser beam is large and they are often
repetitive welds, optimization of the order of the execution of individual beads is an ideal task to
perform using the technique of “shrinkage” in the “local–global” approach. The use of the “transient”
technique due to the need to use a very dense mesh (due to the size of the laser beam focus point)
causes a drastic increase in the calculation time due to a large number of analysis time steps.
However, typical issues can also be solved using the MBD technique, which efficiently reduces this
time.
For numerical simulations of 70 × 5.0 mm pipes laser butt welding, a calculation model
consisting of 16,576 elements and 14,232 nodes was prepared. Simulations were done for two
different materials: AISI304 stainless steel and Inconel 625 nickel superalloy, which are also
implemented in the standard SYSWELD material database. The calculation model was made as a
symmetrical model (only half of the joint separated by a plane of symmetry was calculated to reduce
the calculation time). Calibration of the heat source model was carried out on the 2D cross‐section
model, and on this basis, the imposed heat cycle used further in the MBD calculation was also
determined (Figure 16a–c). Boundary conditions replies to clamping conditions were set as a
complete stiffening of the two nodes in every four places (quarters) of the tube as shown in Figure
16d. After cooling the element in the final phase of calculations, the element was released from its
Materials 2020, 13, 608 13 of 18

mounting. For the clamping boundary conditions, in this case, a symmetry condition was also
defined at the contact surface of the axisymmetric model. Heat dissipation to the environment was
set on all external surfaces of the model as the heat transfer coefficient corresponding to the cooling
conditions in the free air with an ambient temperature of 20 °C. In this case, described earlier, the
conical source with a Gaussian distribution was used as a moving heat source model in the
“transient” calculation technique. Heat source parameters were set as follows: laser beam power –3.0
kW, welding speed –39 mm/s, laser beam heat efficiency –1.0, ze –5.0 mm, zi –0 mm, re –1.0 mm, ri –
0.4 mm. For MBD calculation was set the thermal cycle obtained as a result of heat source calibration
on a simple 2D model.

Figure 16. Temperature fields distributions during the laser welding process of pipes butt joints: (a)
calculated with “transient”, (b) MBD technique, (c) imposed thermal cycle used in MBD technique
calculations and (d) clamping conditions during welding.

Comparison of the results calculated with both techniques showed similarities in the values and
the distributions of displacements and stresses in the analyzed welded joints (Figures 17–20). The
thermal cycles of the welding process calculated by the “transient” and MBD techniques showed a
high convergence of their course as well as values (Figure 21). The summary of calculation times
achieved by different calculation techniques/methods are presented in Table 3.
Materials 2020, 13, 608 14 of 18

Figure 17. Distributions of (a) normal displacements and (b) von Mises stresses in stainless steel laser‐
welded pipes butt joints calculated using the “transient” technique.

Figure 18. Distributions of (a) normal displacements and (b) von Mises stresses in stainless steel laser‐
welded pipes butt joints calculated using the MBD technique.

Figure 19. Distributions of (a) normal displacements and (b) von Mises stresses in Inconel 625 nickel
superalloy laser‐welded pipes butt joints calculated using the “transient” technique.
Materials 2020, 13, 608 15 of 18

Figure 20. Distributions of (a) normal displacements and (b) von Mises stresses in Inconel 625 nickel
superalloy laser‐welded pipes butt joints calculated using the MBD technique.

Table 3. Summary of the duration times of numerical analyses for tested calculation techniques of
laser‐welded stainless steel and Inconel 625 nickel superalloy pipes.

Transient MBD (MBD DE)1

AISI 304 45 min 8 s 5 min 1s 50 s
Inconel 625 33 min 12 s 3 min 48 s 55 s
1 MBD DE—Macro‐Bead Distortion Engineering for the “local–global” approach using the MBD
calculation

Figure 21. Thermal cycles in selected nodes of the model, calculated using the “transient” (moving
heat source) and MBD (imposed thermal cycle (ITC)) techniques for AISI 304 pipes laser welding
simulation example.

4. Conclusions
This article describes three calculation techniques for welding processes numerical simulations
using the finite element method. The described methods are both the “transient” method that gives
the largest amount of precise information and details. Due to their specificity, they require conducting
step‐by‐step calculations, which increases the calculation time, as well as the MDB or “shrinkage”
technique, which in turn allow for significant reducing the calculation time and allows conducting
analyses of complex welded constructions at the expense of quality and precision of obtained results.
Significantly longer calculation times in the “transient” technique result from the aforementioned
calculation method, where the moving heat source travels over the grid elements generating
Materials 2020, 13, 608 16 of 18

subsequent time moments (calculation cards) as well as the fact that all quantities describing the
welding process are calculated in these steps (temperature fields, phase distributions, heat fluxes,
etc.). In the case of complicated structures with a large number of welds/joints, this makes it
impossible to carry out analyses due to the real‐time needed for their implementation and the amount
of data generated in the resulting output file. However, analyses of this type result in a number of
important, precise data on the so‐called local welding process effects for the selected welded joint.
They also support the creation of simple (and therefore requiring shorter calculation times) models
for calibrating the input data used in the MBD or “shrinkage” technique. They can be both 2D cross‐
section models as well as fully three‐dimensional models, however, with less complexity or smaller
size (primarily length but sometimes also transverse dimensions).
As the presented examples have shown, the use of the MBD technique significantly reduces the
time of calculations due to the fact that in this technique the nodes of the mesh are affected by a given
thermal cycle consisting of several to several points (steps) instead of the amount resulting from the
density of the mesh and the size of the liquid metal pool (and hence also in some sense the dimensions
of the heat source model). The cost incurred by the user in the event of its application results in the
lower precision of the results obtained, caused by the specifics of thermal load (e.g., increases in
stresses values at the end of the section into which the weld bead is divided (Figure 9b). Some
differences also occur as a distribution of calculated quantities, although the maximum and minimum
values and their distribution are usually similar to “transient” calculations results (Figures 8–10, 13
and 14).
In addition, in the case of resignation from calculations of the temperature fields and
metallurgical phase distributions and the imposed heat cycle that was previously calibrated using a
simple model calculated using the “transient” technique, it is observed a significant reduction of the
calculation times. For the presented simulation examples, the calculation time reduction (without
taking into account the calibration stage of the heat source model, which must also be carried out for
the “transient” method at heat source calibration stage) is from a few to several times (Tables 1–3). In
the case of very complex structures, where the use of the “transient” technique is impossible due to
the aforementioned long calculation time, the use of the MBD DE or “shrinkage” technique supports
the very fast prediction of the deformations of the weld structure. In this case, the abandonment of
calculations of temperature field and metallurgical phase distribution and the use of shell elements
instead of 3D solid elements, on the one hand, limits the result to the approximate deformation
distribution, but on the other hand, it allows to quickly check different clamping variants and the
order and direction welding individual beads. It can, therefore, be said that in a situation when, after
proper calibration, the displacements distributions are similar, and the analysis mainly concerns
structure deformations, the use of this computational technique brings measurable time benefits. This
will be particularly evident when a large element or entire structure containing a large number of
welds/beads is being analyzed. The mentioned repeatability of the welding parameters, allows for
calibration on one model using the “transient” technique, and then transferring them to the structure
and finally calculating the total deformation using the MBD or “shrinkage” technique. In such cases,
the results of the analyses are significantly accelerated and this is often the only possible solution in
terms of hardware and time limitations.
However, not only the mesh density or the complexity of the structure affects the calculation
time. The type of material of the welded parts also affects the calculation time. It is because the whole
simulation process is divided into time cards. Their amount depends on the density of the mesh, i.e.,
the number of steps that the moving heat source and cards related to the next cooling of the welded
element to the set temperature must follow. Individual iterations in the calculation process of the
selected time card must meet the assumed convergence criterion. Only then does the solver move to
the next card. The number of iterations strictly depends on the non‐linearity of the calculations. In
turn, the non‐linearity of the calculations is greater when the plastic deformations in the material are
higher. Thus, the lower the yield strength of a material, the faster it gets into plastic deformations but
also convergence problems. This is very well visible on the presented example of with laser welding
of AISI 304 steel and Inconel 625 nickel superalloy pipes. The almost twice lower yield strength for
Materials 2020, 13, 608 17 of 18

AISI 304 steel makes it so that the calculation time is longer compared to the calculation time for
Inconel 625 nickel superalloy. The differences in the calculation time for both materials in the case of
the “transient” method are approximately 26% and 24% for the MBD technique (Table 3). The slight
difference in the calculation times when using the MBD DE technique is confirmed by the fact that,
in this case, they are not taken take into account with the distribution of metallurgical phases and
changes in material properties, and hence stresses and strains values.
In summary, in a situation where currently the main challenge for the industry is to reduce the
time needed for design work and research and development activities, the described modern
calculations methods create possibilities to increase efficiency and reduce the waiting time for the
final product. In addition, it is possible to detect positioning problems and incorrect design,
placement or use of fasteners. If we add the possibility of detecting the risk of defects during
assembly, there is a real advantage of using numerical analyses in the form of improving production
efficiency as well as significant economic benefits.

Funding: This publication was co‐funded by the statutory subsidy of the Faculty of Mechanical Engineering at
the Silesian University of Technology in 2019.

Conflicts of Interest: The author declares no conflicts of interest.

References
1. Welding simulation user guide, Sysweld manual ESI Group, Available online:
https://www.esigmbh.de/downloads/ESI/Dokumente/Welding/The_Welding_Simulation_Solution_21040
8 (access on February 2016).
2. Slováček, M.; Kik, T.; Use of Welding Process Numerical Analyses as Technical Support in Industry. Part
1: Introduction to Welding Process Numerical Simulations. Biuletyn Instytutu Spawalnictwa 2015, 4, 25–31.
3. Slováček, M.; Vaněk, M.; Kik, T.; Use of Welding Process Numerical Analyses as Technical Support in
Industry. Part 2: Methodology and Validation. Biuletyn Instytutu Spawalnictwa, 2015, 59, 38‐43.
4. Kik T. Metody analiz numerycznych procesów spawania złożonych konstrukcji. Technol. Rev. 2017, 89, 41–
45.
5. Kik, T.; Górka, J. Numerical simulations of laser and hybrid S700MC T‐joint welding. Materials 2019, 12,
516.
6. Janicki, D. Fiber laser welding of nickel based superalloy Inconel 625. Proceedings of SPIE, Vol. 8703, Laser
Technology 2012: Applications of Lasers, 2013, 87030R 10.1117/12.2013430.
7. Lisiecki, A. Titanium Matrix Composite Ti/TiN Produced by Diode Laser Gas Nitriding. Metals 2015, 5, 54–
69.
8. Stadnicki, J.; Wróbel, I. Numerical effectiveness of the simulation of an automotive body part stamping.
Adv. Mech. Eng. 2015, 7, 708434.
9. Chmielewski, T.; Golański, D. The role of welding in the remanufacturing process. Weld. Int. 2015, 29, 861–
864.
10. Tomkow, J.; Fydrych, D.; Rogalski, G. Role of bead sequence in underwater welding. Materials 2019, 12,
3372.
11. Ferro, P.; Porzner, H.; Tiziani, A.; Bonollo, F. The influence of phase transformations on residual stresses
induced by the welding process ‐ 3D and 2D, numerical models, Model. Simul. Mater. Sci. Eng. 2006 14,
117–136.
12. Wu, W.; Liang, N.; Gan, C.; Yu, G. Numerical investigation on laser transformation hardening with
different temporal pulse shapes. Surf. Coat. Technol. 2006, 200, 2686–2694.
13. Shanmugam, N.S.; Buvanashekaran, G., Sankaranarayanasamy, K.; Ramesh Kumar, C. A Transient Finite
Element Simulation of the Temperature and Bead Profiles of T‐joint Laser Welds. Mater. Des. 2010, 31, 4528–
4542.
14. Winczek, J.; Skrzypczak, T. Thermomechanical states in arc weld surfaced steel elements with heat of
melted electrode material taken into account. Arch. Metall. Mater. 2016, 61, 1277–1288.
15. Chiocca, A.; Frendo, F.; Bertini, L. Evaluation of Heat Sources for the Simulation of the Temperature
Distribution in Gas Metal Arc Welded Joints. Metals 2019, 9, 1142.
Materials 2020, 13, 608 18 of 18

16. Moravec, J.; Bradác, J.; Neumann, H.;Nováková, I., Grain size prediction of steel P92 by help of numerical
simulations. In the Proceedings of METAL 2013 22nd International Conference on Metallurgy and
Materials, Brno, Czech Republic, 15–17 May 2013; pp. 508–513.
17. Kong, F.; Ma, J.; Kovacevic, R. Numerical and experimental study of thermally induced residual stress in
the hybrid laser‐GMA welding process. J. Mater. Process. Technol. 2011, 211, 1102–1111.
18. Devaux, J.; Leblond, J.B.; Bergheau, J.M. Numerical study of the plastic behaviour of a low alloy steel during
phase transformation. In the Proceeding of 1st International Conference on Thermal Process Modelling and
Computer Simulation, Shanghai, China, 28–30 March 2000; pp. 206–212.
19. Goldak, J.; Chakravarti, A.; Bibby, A. A New Finite Element Model for Welding Heat Sources. Metall. Trans.
1984, 15, 299–305.
20. Dupont, J.N.; Marder, A.J. Thermal Efficiency of Arc Welding Processes. Welding Journal, The American
Welding Society: Miami, FL, USA, 1995; ISSN 0043‐2296.
21. Bradáč, J. Calibration of Heat Source Model in Numerical Simulations of Fusion Welding, Machines,
Technologies. Materials 2013, 11, 9–12.
22. Wu, C.S., Wang, H.G., Zhang, Y.M. A new heat source model for keyhole plasma arc welding in FEM
analysis of the temperature profile. Weld. J. 2006, 85, 284–291.
23. Slováček, M. Numerical simulation of welding, calculation and evaluation of distortions and residual
stresses. Ph.D. thesis, Brno Military Academy, Brno, Czech Republic, September 2005.
24. Slováček, M.; Diviš, V.; Ochodek, V. Numerical simulation of the welding process, distortion and residual
stresses prediction, heat source model determination. 58th Annual Assembly Meeting, Prague, Czech
Republic, 10−15 July 2005; pp. 1586–2005

© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).