Materials Today Communications: Sciencedirect

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Materials Today Communications 22 (2020) 100836

Contents lists available at ScienceDirect

Materials Today Communications


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

Increasing the efficiency of computational welding mechanics by combining T


solid and shell elements
D.G. Karalis
Hellenic Navy, Hellenic Naval Academy, Mechanics & Materials Division, Marine Materials Laboratory, Hazjikyriakou Avenue, Piraeus 185 39, Greece

A R T I C LE I N FO A B S T R A C T

Keywords: This paper emphasizes the benefits encountered when solid and shell elements are combined in a single 3-D
Thermal analysis finite element analysis (FEA) thermomechanical model - simulating a typical problem of computational welding
Thermal stress analysis mechanics. In order to clearly obtain the aforementioned benefits, FEA models containing only solid elements
Welding are compared with FEA models containing a combination of solids and shells. More specifically, in the current
Residual stress
paper the post-weld heat treatment of a typical marine structure is simulated by using four different 3-D finite
Residual distortion
Solids
element analysis models. In the first model (model I), thermoplastic solid elements were used, whereas in the
Shells second model (model II), a mix of thermoplastic and thermoelastic solids was employed. In order to investigate
whether the combination of solids with shells can substantially increase the computational efficiency, and
therefore reduce the duration of the analyses and the model sizes, two more models were set up by combining
thermoplastic and thermoelastic solid elements with thermoelastic shells: constant (model III) and varying as-
pect ratio (model IV). As expected, these four models presented equivalent results as far as the calculation of the
thermally induced temperatures, displacements and stresses was concerned however, severe differences in
computational efficiency were observed. The aim of this paper is to quantitatively present the benefits en-
countered from the combination of solids with shells in thermal and thermal stress analyses and to discuss the
computational efficiency increase that was observed as a result of this element scheme. Important numerical
issues like position of the solid to shell transition area, coupling of solids and shells, and aspect ratio of the
elements, which affected the overall efficiency, are also discussed. As a result, increase in computational effi-
ciency was observed when solid elements were combined with shells of varying aspect ratio, the latter modeling
the far field area of the weld. This combination proved to be very beneficial in computational welding mechanics
and is expected to facilitate the simulation of thermal and thermal stress analyses of large scale welded structures
using the finite element method.

1. Introduction for the design of the weld, the selection of the welding sequence in
multi-pass welding and the quality of the welded structure.
Simulation of welding or post-weld heat treatment of large struc- The necessity of detailed results, especially at the vicinity of the
tures, such as ship sections, currently is a very difficult task. The fusion zone, where large gradients through the thickness exists as far as
strongly nonlinear simulation which is normally carried out using a the thermal, mechanical and microstructural phenomena are con-
staggered approach, involves the solution of the thermal and mechan- cerned, makes the use of 3-D solid elements inevitable. As a con-
ical problem, thus, transient temperature analyses and transient sequence, the number of degrees of freedom (DOFs) that are required to
thermal stress analyses have to be carried out [1]. Most often, these represent the modeled structure quickly reaches the computational
analyses are carried out using the finite element method [2]. The limits, making modeling of welding of large-scale structures (e.g.
thermal analysis, besides the temperature distribution, provides the marine structures) an unachievable task. Typical examples from the
extension of the heat-affected zone, heating and cooling rates, and the published work of the last decade, where 3-D solid elements have been
final microstructure and hardness of the material, which are very im- used in welding simulations are listed in the reference section [3–16]. It
portant parameters for the design of the weld or the post-weld heat is therefore obvious, that in order to increase the computational effi-
treatment. On the other hand, the mechanical analysis provides the ciency, which means to speed-up the analysis and minimize the com-
residual stresses and distortions which are very important parameters puter resources, the simulated problem size has to be reduced. Notice

E-mail address: KaralisDimitris@TEEmail.gr.

https://doi.org/10.1016/j.mtcomm.2019.100836
Received 18 July 2019; Received in revised form 29 November 2019; Accepted 6 December 2019
Available online 09 December 2019
2352-4928/ © 2019 Elsevier Ltd. All rights reserved.
D.G. Karalis Materials Today Communications 22 (2020) 100836

that reduction of the problem size does not necessarily imply modeling
of a physically smaller structure, but modeling of the structure with less
computational cost (less DOFs).
One way to decrease the DOFs in computational welding mechanics
is to combine solid elements with shell elements in a single finite ele-
ment analysis model formulated in 3-D space.
Encouraging examples of this technique in welding simulations are
presented in [17–19]. In [17], the authors utilized fine brick elements
for meshing the main computational zone of a 3-D FEA model, whereas
the rest of the model was meshed by coarse shell elements. In [18,19],
the authors developed 3-D FEA models in order to predict welding
distortions and residual stresses but besides the use of solid-only ele-
ment models they also combined solids and shells in order to increase
the computational efficiency. They concluded that this combination of
solids with shells resulted in reduction of the total computational time Fig. 2. The modeled structure containing a plate and a welded stiffener.
by 42% and 40%, respectively.
Note that as there are no severe restraints on the aspect ratio of the
yet which quantitatively presents the benefits and drawbacks (with
shell elements; fewer shell elements can be used to model large areas of
respect to the computational efficiency), as well as a methodology of
the structure where large through-thickness temperature gradients do
combining the use of solid elements and shell elements compared to
not exist. Thus, a dense mesh of thermal solid elements and thermo-
models using only solid elements in Computational Welding Mechanics.
plastic solid elements can be used to represent the area close to the
The aim of this paper is to quantitatively present the benefits en-
fusion zone, whereas a sparse mesh of thermal shells and thermoelastic
countered from the combination of solids with shells in thermal and
shells can be used in order to model the far field region.
thermal stress analyses and to discuss the computational efficiency in-
A problem that arises from the combination of solids with shell
crease. Important numerical issues like position of the solid to shell
elements is the compatibility across the coupling region. In thermal
transition area, coupling of solids and shells, and aspect ratio of the
analysis, both elements have similar DOFs, the nodal temperatures. In
elements, which affected the overall efficiency, are also discussed.
mechanical analysis, solid elements have three translational DOFs per
node, whereas the shells also have additional rotational DOFs. As a
result, the coupling compatibility has to be solved prior to setting up a
2. Preprocessing of the models
model which combines both types of the elements. One way to solve the
coupling problem of shells and solids is to use two additional beam
Four different finite element analysis models were set up simulating
elements [20] that are connected perpendicularly to the node of the
the post-weld heat treatment of the welded structure depicted in Fig. 2.
shell (see Fig. 1a). A second way is to use the penetration method [21],
The specific structure, which is typical in marine applications, refers to
where the shell elements penetrate into the solids at the area of cou-
a plate and a welded stiffener.
pling (see Fig. 1b). Alternative ways entail using fictitious shell planes
For every model, the thermal analysis and the post mechanical
perpendicular to the original shell plane or via multipoint constraint
analysis (thermal stress analysis) were carried out in a staggered ap-
(MPC) equations [22] (see Fig. 1c).
proach. Thus, the calculated temperatures of every time step of the
As far as the author is aware, there has not been a paper published
thermal analysis were imported into the mechanical model, in order to

Fig. 1. Coupling of shells and solids: (a) using additional beams, (b) using penetrating shells, (c) using perpendicular shells.

2
D.G. Karalis Materials Today Communications 22 (2020) 100836

Fig. 3. Model I containing only solid elements. Yellow, green and cyan colored Fig. 5. Model III containing both solid and shell elements. Yellow, green, and
elements refer to thermal solids in thermal analysis and thermoplastic solid cyan colored elements refer to thermoplastic solid elements, whereas the brown
elements in the mechanical analysis. (For interpretation of the references to colored elements refer to thermoelastic solid elements in the mechanical ana-
colour in this figure legend, the reader is referred to the web version of this lysis. Grey colored elements refer to thermoelastic shell elements in the me-
article.) chanical analysis. All elements refer to thermal elements in thermal analysis.
(For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
calculate the thermally induced displacements, strains and stresses. The
four different models that were set up are described below.
Model I. The specific model was set up using thermal solids for the model. Model III is depicted in Fig. 5. Similarly to Models I and II, the
thermal analysis and thermoplastic solid elements for the mechanical dimensions of all elements (except those describing the area very close
analysis. The dimensions of all elements (except those modeling the to the fusion zone) were equal to 5 mm x 5 mm x 5 mm (aspect ratio
area very close to the fusion zone) were equal to 5 mm x 5 mm x 5 mm 1:1:1) for the solids and 5 mm x 5 mm for the shells (aspect ratio 1:1).
(aspect ratio 1:1:1). The latter model is considered as the “reference” The thickness of shells was set equal to 10 mm. At the transition areas,
model. Model I is depicted in Fig. 3. the penetration depth of shells into the solids was set equal to 2 ele-
Model II. Model II is similar to Model I. The only difference is that ments (penetration depth will be discussed later). Model III aims at
in the mechanical analysis, a very big part of the structure was modeled quantitatively assessing the benefits obtained by the combination of
using thermoelastic solid elements instead of thermoplastic solid ele- solid elements and shells in 3-D computational welding mechanics.
ments. More specifically, in Model II thermoelastic solid elements were Notice that in order to make a fair comparison, solid and shell elements
used to model the material that did not experience heating higher than of the same dimensions (5 mm x 5 mm x 5 mm and 5 mm x 5 mm re-
ΔT = +18 K (see transition areas in Fig. 4, ΔT level will be discussed spectively) and aspect ratios (1:1:1 and 1:1, respectively) were used in
later). The aim of Model II was to confirm that the solid-to-shell tran- Models I, II and III.
sition areas were properly positioned by comparing the results of Model IV. The specific model is similar to Model III. The only dif-
Models I and II and to investigate any benefit obtained as a result of the ference is that shell elements of geometrically varying aspect ratios
combination of thermoelastic and thermoplastic solid elements. (from 1:1 up to 1:7) were used. Thus, compared to Model III the number
Model III. The specific model is based on Model II and combines of DOFs was severely reduced. The aim of Model IV was to confirm that
shells with solid elements. More specifically, in the mechanical analysis, the use of shell elements of varying aspect ratio, modeling the far field
the thermoelastic solid elements of Model II were replaced by ther- region of the weld, in combination with solid elements modeling the
moelastic shell elements modeling the mid-plane of the plate. In the area at the vicinity of the fusion zone, is beneficial in computational
thermal analysis, thermal shells replaced the thermal solids, accord- welding mechanics. Model IV is depicted in Fig. 6.
ingly. In order to provide exactly the same nodes for the application of
the boundary conditions with respect to the models I and II, a narrow
band of solid elements were also coupled with shells at the edges of the 3. Models summary

The number and the element types that were used in the four dif-
ferent models are summarized in Table 1.

Fig. 4. Model II is similar to Model I and contains only solid elements. Yellow,
green and cyan colored elements refer to thermoplastic solid elements, whereas
the brown colored elements refer to thermoelastic solid elements in the me-
chanical analysis. All elements refer to thermal solids in thermal analysis. (For
interpretation of the references to colour in this figure legend, the reader is Fig. 6. Model IV is similar to Model III and contains solids (aspect ratio 1:1:1)
referred to the web version of this article.) and shell elements of varying aspect ratio (up to 1:7).

3
D.G. Karalis Materials Today Communications 22 (2020) 100836

Table 1
Number of elements, element types, dimensions and aspect ratios used for every model.
Thermal analysis Mechanical analysis Aspect ratios and dimensions

Model I 31400 solids 31400 thermoplastic solids Solids 1:1:1


(reference) (31400 elements in total) (31400 elements in total) 5 × 5 × 5 mm3
Model II 31400 solids 9800 thermoplastic solids Solids 1:1:1
(31400 elements in total) 21600 thermoelastic solids 5 × 5 × 5 mm3
(31400 elements in total)
Model III 11800 solids 9800 thermoplastic solids Solids 1:1:1
10800 shells 2000 thermoelastic solids 5 × 5 × 5 mm3
(22600 elements in total) 10800 thermoelastic shells Shells 1:1
(22600 elements in total) 5 × 5 mm2
Model IV 11800 solids 9800 thermoplastic solids Solids 1:1:1
3900 shells 2000 thermoelastic solids 5 × 5 × 5 mm3
(15700 elements in total) 3900 thermoelastic shells Shells up to 1:7
(15700 elements in total)

4. Simulation parameters depicted in Fig. 8.


In Fig. 9, transient Von Mises stress (t = 50 s) and transient dis-
The entire duration of the post-weld heat treatment simulation, placement magnitude (t = 50 s) of the four models are depicted.
which was assumed to be carried out by a moving arc, was equal to The residual stresses and residual displacement magnitudes
500 s with the first 100 s simulating the moving arc and the remaining (t = 500 s) of the four models are depicted in Fig. 10.
400 s simulating the cooling phase. The heat input was applied on the In Fig. 11, the histories of the temperatures and the Von Mises
basis of the double ellipsoid distribution of the arc [23,24] of which stresses of the fusion zone surface which is located at mid-length of the
dimensions were obtained from the size of the modeled fusion zone. treated weld are depicted for all four models.
The arc velocity was set equal to 5 mm/s and the arc efficiency was The transient displacement magnitudes of the most distorted nodes
properly selected in order to heat the fusion zone up to 707 K (434 °C). (see red colored areas of the displaced models in Fig. 10), are depicted
As far as the boundary conditions are concerned, for the thermal part, in Fig. 12.
both convection and radiation were applied to all free surfaces. More
specifically, the convection coefficient was set equal to h = 10 W/m2K 6. Discussion
for the treated surface of the fusion zone and h = 5 W/m2K for the rest
surfaces. The emissivity was set equal to ε = 0.9 for the treated surface From the results presented in Figs. 8–12, it is concluded that the
of the fusion zone and ε = 0.8 for the rest of the surfaces. The ambient four models present very similar results in engineering terms. More
temperature was set equal to Tamb =298 K for both convection and specifically:
radiation. The boundary conditions for the mechanical analysis were
applied to the nodes of the solids in all models as depicted in Fig. 7. 1 The four thermal models provided almost identical results as far as
Temperature dependent thermal and mechanical properties were the transient and the final temperatures are concerned (see Figs. 8
assigned for all elements simulating the steel structure. The yield stress and 11). Regarding transient stress and transient displacement
of the steel was set equal to 250 MPa at room temperature and isotropic magnitude, the three models provided very similar results, pre-
hardening was applied [20]. Eight-node 3-D solids and four-node 3-D senting negligible deviations (see Figs. 9 and 12).
shells were used in all simulations [21]. Material and geometry non- 2 The histories of temperatures and Von Mises stresses of the fusion
linearities were taken into account. All analyses were executed using zone surface presented no significant differences among the four
the same desktop computer (Intel Pentium 4 CPU 3.0 GHz, Intel models (see Fig. 11). The same was also observed for the histories of
82945 G Express Chipset family, single HDD). the displacement magnitudes of the most distorted nodes (see
Fig. 12). A small difference (+3.2%) of the displacement magni-
5. Results tudes among models II, IV and I, III was observed.
3 Almost identical results were obtained as far as the residual stresses
Transient and residual results of the four models are depicted in and residual displacements are concerned (see Figs. 10, 11 and 12).
Fig. 8–10. More specifically, the temperature distributions during the More specifically, the four models present almost the same long-
treatment (t = 50 s) and after cooling (t = 500 s) of the structure are itudinal bending and angular distortion after the end of the treat-
ment.

Thus, one can conclude that the four models are equivalent as far as
the calculation of the thermally induced temperatures, displacements
and stresses are concerned. The question that arises is whether this
combination of shells and solids in a single thermomechanical 3-D
model (Models III and IV) provides a more efficient method for the
analysis and design of the welded structure by means of the finite
element method. In order to answer this question, the benefits as well as
a few drawbacks that were encountered during the numerical simula-
tion of the four different models are discussed below.

6.1. Analysis duration and model size

Fig. 7. The boundary conditions used for the mechanical analysis in all models; Two of the biggest concerns in computational welding mechanics
the “T” symbol refers to translational restriction along the specified direction. are the duration of the simulation and the final model size.

4
D.G. Karalis Materials Today Communications 22 (2020) 100836

Fig. 8. Temperature distributions during the treatment (t = 50 s) and after cooling (t = 500 s) of the structure.

In the case of Model I (reference model), the thermal analysis lasted the final model size was reduced to 6.84 GB.
4 h and 11 min while the final model size was equal to 1.54 GB. The Model IV, which is the model of the greatest interest in the current
duration of the thermal analysis of Model II was equal to 4 h and 22 min work, was the fastest among the other three models, requiring 1 h and
while its size was the same as Model I, equal to 1.54 GB. The duration of 49 min for the thermal analysis and 23 h and 23 min for the mechanical
the thermal analysis of Model III was equal to 2 h and 2 min, thus sig- analysis. The size of Model IV was also strongly reduced, as the thermal
nificant reduction compared to the reference model was observed. model size was equal to 0.71 GB while the mechanical model size was
Severe reduction was also observed as far as the final model size is equal to 4.71 GB. The results are summarized in Table 2 and Figs. 13
concerned, which in Model III was equal to 0.98 GB. The results, and 14.
however, were controversial for the case of the mechanical analysis. From the results presented in Table 2 and Figs. 13 and 14, the fol-
Model I took 30 h and 20 min to complete and required 9.56 GB of disc lowing is concluded:
space. The mechanical analysis of Model II lasted 39 h and 25 min,
whereas the final model size was equal to 9.54 GB. Unfortunately, the 1 In the thermal analyses, there was a significant benefit regarding the
duration of the analysis of Model III was equal to 54 h and 45 min but analysis duration and the final model size when thermal shell

5
D.G. Karalis Materials Today Communications 22 (2020) 100836

Fig. 9. Transient Von Mises stress and transient displacement magnitude of the structure (t = 50 s).

elements were combined with thermal solid elements (compare analyses durations and model sizes (compare Models I and IV). Here
Models I and III where elements of the same dimensions and aspect it has to be emphasized, that this substantial storage benefit (up to
ratios have been used). On the other hand, in the mechanical ana- 54.2%) is very important in case of convergence study analyses
lyses the same combination resulted in severe increase of the [25], where multiple executions have to be performed for both the
durations of the analyses, but strong reduction of the final model thermal and the mechanical analyses. As a result, Model IV proved
sizes (again, compare Models I and III). Thus, computational effi- to be the most efficient model among the four developed in the
ciency increase was not accomplished for the mechanical analyses current work.
between Models I and III.
2 Model II did not provide any benefit regarding the computational The author would like to mention that the evolution of computer
efficiency with respect to Model I. performance over the last four decades increasingly allowed researchers
3 The combination of solids with shell elements of varying aspect ratio to simulate large 3-D structures using solid elements. Typical examples
(from 1:1 up to 1.7) provided indisputable benefits regarding of large scale welding simulations using only solid elements are listed in

6
D.G. Karalis Materials Today Communications 22 (2020) 100836

Fig. 10. Residual Von Mises stress and residual displacement magnitude (t = 500 s, displaced scale factor = 10).

[26–32]. As the processing speed of the computers, as well as the their difference up to ΔT = +18 K (see Fig. 4). Sensitivity analyses which
storage media capacity are expected to increase in the forthcoming were carried out (and are not presented herein) showed that moving
decades, more complicated FEA models are expected to be developed, this transition area towards the fusion zone, in order to further increase
aiming to solve increasingly more complicated systems of equations. the area that is modeled with thermoelastic shell elements and thus
Thus, any attempt to adopt new flexible modeling techniques, like the increase the computational benefit, resulted in different transient and
one proposed in the current paper, in order to reduce the problem size, residual response between Models I and II. This is attributed to the
seems to be in the right direction. higher temperatures (higher ΔT) that are realized closer to the fusion
zone and the fact that the material exceeds thermoelasticity at these
regions. As a result, prior to developing Model IV, the analyst should at
6.2. Solid to shell transition area least set up Model II in order to locate the boundary between the
thermoplastic and thermoelastic behavior of the material.
In the presented models, the transition area between the solids and
the shells was positioned at the material that experienced temperature

7
D.G. Karalis Materials Today Communications 22 (2020) 100836

Fig. 11. History of temperatures and Von Mises stresses of the fusion zone at mid-length for the four different models.

Fig. 12. Transient displacement magnitudes of the most distorted nodes (see also Fig.10).

Table 2
The duration of the analyses (CPU-time), the final model sizes of the four
models and the percentage of difference with respect to the reference model.
Thermal analysis Mechanical analysis

Model I 4 h 11 min, 30 h 20 min,


(reference) 1.54 GB 9.56 GB
Model II 4 h 22 min (+4.4%) 39 h 25 min (+29.9%)
1.54 GB (0%) 9.54 GB (−0.3%)
Model III 2 h 2 min (−50.6%) 54 h 45 min (+80.4%)
0.98 GB (−36.2%) 6.84 GB (−28.4%)
Model IV 1 h 49 min (−56.5%) 23 h 23 min (−22.9%)
0.71 GB (−54.2%) 4.71 GB (-50.7%)

Fig. 13. Analysis duration versus the number of elements used.

8
D.G. Karalis Materials Today Communications 22 (2020) 100836

Data availability

The raw/processed data required to reproduce these findings cannot


be shared at this time due to technical or time limitations.

CRediT authorship contribution statement

D.G. Karalis: Conceptualization, Methodology, Software,


Validation, Investigation, Resources, Writing - original draft, Writing -
review & editing.

Declaration of Competing Interest

There are no interests to declare.

Fig. 14. Model size versus the number of elements used. References

6.3. Shell penetration depth [1] D. Radaj, Heat Effects of Welding, Springer-Verlag, Berlin, Heidelberg, 1992.
[2] L.-E. Lindgren, Finite element modeling and simulation of welding part 1: increased
complexity, J. Therm. Stress. 24 (2001) 141–192.
In the current simulation the penetration depth of the shell elements [3] F. Kong, R. Kovacevic, 3D finite element modeling of the thermally induced residual
was set equal to a two-element depth (see Figs. 1 and 6). One element stress in the hybrid laser/arc welding of lap joint, J. Mater. Process. Technol. 210
(2010) 941–950.
penetration depth analyses (not presented herein) resulted in different
[4] L. Gannon, Y. Liu, N. Pegg, M. Smith, Effect of welding sequence on residual stress
results between Model I and Models III and IV. Thus, several sensitivity and distortion in flat-bar stiffened plates, Mar. Struct. 23 (2010) 385–404.
analyses have to be carried out prior to finalizing Model IV, in order to [5] C.-H. Lee, K.-H. Chang, Prediction of residual stresses in high strength carbon steel
optimize the depth penetration of the shells. pipe weld considering solid-state phase transformation effects, Comput. Struct. 89
(2011) 256–265.
[6] C. Heinze, C. Schwenk, M. Rethmeier, The effect of tack welding on numerically
calculated welding-induced distortion, J. Mater. Process. Technol. 212 (2012)
6.4. Aspect ratio of the shells 308–314.
[7] W. Jiang, Y. Zhang, W. Woo, Using heat sink technology to decrease residual stress
in 316L stainless steel welding joint: finite element simulation, Int. J. Press. Vessel.
Notice that as mentioned in section §1, one of the biggest benefits Pip. 92 (2012) 56–62.
when using shells is that there are no severe restraints on their aspect [8] D. Camilleri, N. McPherson, T.G.F. Gray, The applicability of using low transfor-
mation temperature welding wire to minimize unwanted residual stresses and
ratio. Thus, a sparse mesh of shells (Model IV, see Fig. 6) allows the
distortions, Int. J. Press. Vessel. Pip. 110 (2013) 2–8.
severe reduction of DOFs and provides a very efficient numerical si- [9] D. Deng, H. Murakawa, Influence of transformation induced plasticity on simulated
mulation. The latter was clearly confirmed from the results presented in results of welding residual stress in low temperature transformation steel, Comput.
§5 and §6. Here it has to be noted that this sparse mesh of quadrilateral Mater. Sci. 78 (2013) 55–62.
[10] M. Islam, A. Buijk, M. Rais-Rohani, K. Motoyama, Simulation-based numerical
shells does not provide a convenient future basis for the application of optimization of arc welding process for reduced distortion in welded structures,
the external operating loads [25] (e.g. distributed nodal hydrostatic Finite Elem. Anal. Des. 84 (2014) 54–64.
pressures or thermal loads etc.). As a result, Model IV is very efficient, [11] N. Yadaiah, S. Bag, Development of egg-configuration heat source model in nu-
merical simulation of autogenous fusion welding process, Int. J. Therm. Sci. 86
at least for the calculation of the thermally induced temperatures, (2014) 125–138.
displacements and stresses. [12] A.A. Bhatti, Z. Barsoum, H. Murakawa, I. Barsoum, Influence of thermo-mechanical
material properties of different steel grades on welding residual stresses and angular
distortion, Mater. Des. 65 (2015) 878–889.
[13] S. Shadkam, E. Ranjbarnodeh, M. Iranmanesh, Effect of sequence and stiffener
7. Conclusion shape on welding distortion of stiffened panel, J. Constr. Steel Res. 149 (2018)
41–52.
[14] E.D. Derakhshan, N. Yazdian, B. Craft, S. Smith, R. Kovacevic, Numerical simulation
On the basis of the results presented in the current paper, the fol-
and experimental validation of residual stress and welding distortion induced by
lowing is concluded: laser-based welding processes of thin structural steel plates in butt joint config-
uration, Opt. Laser Technol. 104 (2018) 170–182.

• Increase in computational efficiency without compromising the ac- [15] C.-l. LI, D. Fan, X.-q. Yu, J.-k. Huang, Residual stress and welding distortion of Al/
steel butt joint by arc-assisted laser welding-brazing, Trans. Nonferrous Met. Soc.
curacy of the results was obtained when shell elements of varying China 29 (2019) 692–700.
aspect ratio were used to model the far field region of the weld, in [16] B. Kumar, S. Bag, Phase transformation effect in distortion and residual stress of
combination with solid elements modeling the area at the vicinity of thin-sheet laser welded Ti-alloy, Opt. Lasers Eng. 122 (2019) 209–224.
[17] M.R. Forouzan, S.M.M. Nasiri, A. Mokhtari, A. Heidari, S.J. Golestaneh, Residual
the fusion zone. The benefit in computational efficiency pertains to stress prediction in submerged arc welded spiral pipes, Mater. Des. 33 (2012)
reduction of the analysis duration and model size. The reduction of 384–394.
the analysis duration in the simulated case of the current paper was [18] M. Perić, Z. Tonković, A. Rodić, M. Surjak, I. Garašić, I. Boras, S. Švaić, Numerical
analysis and experimental investigation of welding residual stresses and distortions
equal to 56.5% for the thermal analysis and 22.9% for the me- in a T-joint fillet weld, Mater. Des. 53 (2014) 1052–1063.
chanical analysis (average 39.7%), whereas the model size reduc- [19] J. Shen, Z. Chen, Welding simulation of fillet-welded joint using shell elements with
tion was equal to 54.2% and 50.7% (average 52.4%), respectively. section integration, J. Mater. Process. Technol. 214 (2014) 2529–2536.

• The combination of solids with shells in computational welding [20] L. Andersen, Residual Stresses and Deformations in Steel Structures, PhD Thesis
Department of Naval Architecture and Offshore Engineering, Technical University
mechanics requires preliminary investigation and optimization re- of Denmark, 2000.
garding the position of the solid to shell transition area and the [21] ALGOR® Documentation, Algor Inc., 2007.
[22] N. Osawa, K. Hashimoto, J. Sawamura, T. Nakai, S. Suzuki, Study on shell–solid
coupling of the shells with solids (penetration depth in the current
coupling FE analysis for fatigue assessment of ship structure, Mar. Struct. 20 (2007)
case). 143–163.
[23] J. Goldak, A. Chakravarti, M. Bibby, A new finite element model for welding heat
This combination of solids with shells lightens the computational sources, Metall. Trans. 15B (1984) 299–305.
[24] J. Goldak, M. Bibby, J. Moore, R. House, B. Patel, Computer modeling of heat flows
cost and facilitates the simulation of thermal and thermal stress ana- in welds, Metall. Trans. 17B (1986) 587–600.
lyses of large scale welded structures using the finite element method.

9
D.G. Karalis Materials Today Communications 22 (2020) 100836

[25] D.G. Karalis, N.G. Tsouvalis, V.J. Papazoglou, D.I. Pantelis, A practical flow dia- (2012) 1705–1715.
gram for the solution of complex non-linear thermo-mechanical numerical models, [29] G.S. Brar, C.S. Singh, FEA of residual stress in cruciform welded joint of hollow
Comput. Mater. Sci. 95 (2014) 288–301. sectional tubes, J. Constr. Steel Res. 102 (2014) 44–58.
[26] J. Guirao, E. Rodriguez, A. Bayon, F. Bouyer, J. Pistono, L. Jones, Determination [30] D.-f Fu, C.-q. Zhou, C. Li, G. Wang, L.-x. Li, Effect of welding sequence on residual
through the distortions analysis of the best welding sequence in longitudinal welds stress in thin-walled octagonal pipe-plate structure, Trans. Nonferrous Met. Soc.
VATS electron beam welding FE simulation, Fusion Eng. Des. 85 (2010) 766–779. China 24 (2014) 657–664.
[27] J. Guirao, E. Rodriguez, A. Bayon, J. Pistono, L. Jones, FEM simulation of a small EB [31] Y. Li, K. Wang, Y. Jin, M. Xu, H. Lu, Prediction of welding deformation in stiffened
welded mock-up and new sequence proposed to improve the final distortions, structure by introducing thermo-mechanical interface element, J. Mater. Process.
Fusion Eng. Des. 85 (2010) 181–189. Technol. 216 (2015) 440–446.
[28] J. Wang, M. Shibahara, X. Zhang, H. Murakawa, Investigation on twisting distortion [32] Y. Zhang, Y. Wang, The influence of welding mechanical boundary condition on the
of thin plate stiffened structure under welding, J. Mater. Process. Technol. 212 residual stress and distortion of a stiffened-panel, Mar. Struct. 65 (2019) 259–270.

10

You might also like