1. Introduction
Welding processes are the most selected joining processes in the manufacturing and assembling of many components because of their good degree of reliability and high production velocity [
1]. Another advantage that motivates this preference is the economic feasibility of the various types of welding compared with other manufacturing processes [
2,
3]. Gas Metal Arc Welding (GMAW) is one of the most applied and preferred techniques in the industry [
4] because of its advantages, namely the capability of all-position welding and good quality of welds. This welding method is versatile, as it can be used in semi-automatic and fully automatic modes, in consonance with the requirements of each type of application [
5,
6].
Even though there are plenty of advantages in welding processes, one critical problem is that they can often produce high levels of defects, such as shrinkage and distortions [
7]. During the welding process, usually non-uniform expansions and contractions occur among the weld and the surrounding regions, and these effects cause distortions [
8]. One factor that contributes to the distortions’ appearance is the high temperature of the welded area and its thermal expansion, which is restricted by the surrounding areas where the metal is at a lower temperature, causing compressive stress [
9], while high values of tensile stresses are generated in the weld bead [
10].
A lot of research has been conducted to overcome these problems and find strategies to control the appearance of the welded components due to distortions. Distortions often increase production costs and time, especially when their values outpace the accepted limits [
11]. In order to achieve a minimal occurrence of the distortion effects and residual stresses, it is necessary to analyze the parameters used in the welding process and also the fabrication conditions [
12], particularly the material specifications, level of heat input, stiffener arrangements, joint shapes, welding type and continuity, initial distortions and welding sequences, and heat treatments before and after the welding [
13,
14].
The objective of this work was to investigate the influence of three different welding sequences of the GMAW process on angular distortion in butt-joint welds, performed on S235 steel plates by metrological assessment.
2. Influence of Welding Parameters on Distortions and Residual Stress Levels
According to D. Radaj [
15] and T. Schenk [
16], low residual stress levels occur when deformations are not restricted, and high residual stresses emerge when deformations are restricted (see
Figure 1). However, the level of residual stresses could be higher than yield stress and, at the limit, it can exceed the ultimate stress [
17]. So, in the last years, some researchers have been studying the optimal parameters to minimize residual stresses [
18]. These studies have been conducted using numerical [
19] and experimental approaches [
20] to achieve the appropriate balance between the level of residual stresses and the distortion value [
21].
The most pronounced type of distortion in butt-welded plates is angular distortion [
23]; it happens more frequently than other types of distortions, such as tailing and bending, and that is the reason why it is considered the most significant type of distortion [
24].
The angular distortion is a rotation of the structure around the welding line [
16]. When the transverse shrinkage is not uniform in the thickness direction, the angular distortion occurs in a butt joint [
23,
25], so the welded component is distorted in angular directions around the weld interface, as it is shown in
Figure 2 [
26].
The welding parameters must be selected considering the results to be achieved, and the choice of those parameters can also influence the distortions on the welded component. Many studies investigated the influences of the welding parameters in such distortions.
According to Vyas et al., if the voltage and the current are increased, bigger distortions are likely to occur, and the distortions decrease when the welding and feeding velocities increase [
27].
According to Narwadkar A. and Bhosle S, minor distortions are provided with higher gas flow rates, and they are increased with higher values of voltage and current [
28].
Ramani S. and Velmurugan V. concluded that the angular distortions can be directly proportional and influenced by the voltage and the torch travel angle. Contrarily, an opposite effect on distortions can be led with the increase in the length of the electrode and the feed rate [
29].
Deng et al., based on numerical simulations, verified that the heat input has a significant influence on the welding distortion. Large heat input is apt to resulting in buckling distortion in thin-plate panel structures. A reduction in heat input is an effective method of decreasing buckling propensity [
30].
The study conducted by Sakri A. et al. found, through simulations using FEA with experimental validation, that the angular distortions increased with the rise of the angle of V-preparation [
31].
Long e al. verified in their studies that the largest transverse shrinkage occurs at the middle section of the length of the plate and it is gradually reduced to the starting and ending edges of the welding line [
32].
A different investigation about the influence on joint gap distortions, number of passes, and time gap between passes was performed by Kumar P. [
33] and Kumar A. [
34]. With the experiments carried out, it was possible to conclude that an increase in joint gap and number of passes leads to an increase in distortions. On the other hand, the distortions decrease when the time gap between the passes increases.
The Taguchi method was used by Soni S. and Aggarwal N. in research to investigate angular distortions. This investigation concluded that the increase in the current, plate length, and electrode diameter brings an increase in distortions. Concerning the time gap between the passes, it affects the angular distortions oppositely [
11].
4. Results and Discussion
First, the results obtained for each of the experimental replicates were compared for SM, BM, and PM. The parametric
t-test method for comparison of means was applied to each pair of data for each experiment (
Table 6). The results show no statistically significant differences in any of the cases. The same conclusion is drawn when non-parametric tests are applied (Kruskal–Wallis and Mood’s median test) (
Table 7). These tests allow conclusions to be drawn even when the data do not follow a normal distribution and there is no homogeneity of variance.
From the parametric and non-parametric comparisons (
Table 6 and
Table 7), it is possible to conclude that there are no statistically significant differences between the three repetitions of each experiment (evaluated as a whole). A maximum range (difference for each cell between each pair of repeated experiments) was also calculated. The mean range for SM was 0.530 mm, for BM 0.259 mm, and for SP 0.343 mm. For this reason, the averages of vertical displacements can be used for each experiment (SM, BM, and PM).
The vertical displacement averages obtained from the three samples can be seen in
Figure 10, which represent the three analyzed sequences: (a) the symmetrical method, SM, (b) backward method, BM, and (c) single-pass method, SP.
Figure 11a,b indicates the sections with the most expressive distortions (sections A and H and Sections 4 and 5, respectively). Through the comparative graphs of SM × BM × SP, it was possible to analyze the deformation patterns and the amplitudes of the displacements led by welding by comparing the sections. The maximum mean vertical displacement was observed in Section 5 for all welding sequences studied in this work.
To verify whether the differences in displacements between the sequences are statistically significant, a one-way ANOVA was performed with the default significance level of 95%.
The sum of vertical displacements was performed and became a single variable for each sample (
Table 8). Prior to performing ANOVA, Levene’s test was applied to sample data to verify the homogeneity of variances. The test resulted in the null hypothesis being true in order to conclude that the variances are homogeneous and that it is possible to use the parametric ANOVA test for this data sampling. All the vertical displacements measured were summed and transformed into only one variable for each sample.
According to what is indicated in
Table 9, the
p-value obtained with the ANOVA was 0.011, showing the null hypothesis as a result (
p < 0.05). Therefore, it is concluded that there is a significant difference in the final distortion of the samples between the welding sequences.
The sequence averages were also calculated (
Figure 12), and it was verified that the symmetrical method (SM) sequence deformed around 9.8% less than the single-pass (SP) sequence and 12.0% less than the backward method (BM).
In order to make multiple comparisons and verify between which sequences would exist meaningful differences, the Tuckey test was used. According to
Table 10, it is possible to conclude that there are significant differences between the distortions that occurred when comparing the sequences SS × SR and the sequences SS × SC, obtaining a p-value of 0.012 and 0.033, respectively. On the other hand, for the interaction between BM × SP, the Tukey test showed that there are no significant differences between the averages of the displacements of the sequences, with a
p-value equal to 0.625 (
p > 0.05).
5. Conclusions
In order to not only prevent but also control and correct deformations that can occur in welding processes, it is necessary to study and comprehend which mechanisms and parameters can have any influence on distortions.
As a result of the experimental and theoretical considerations of this work, it was possible to conclude that through the statistical tools, namely ANOVA and Tuckey test, it was verified that there is a significant difference in the distortions between these welding sequences and that this difference is only for the symmetrical method sequence, SM, when compared with the backward method (BM) and single pass (SP).
In the welding snares on the 3 mm plates, there was a greater displacement in Sections 4 and 5 of all specimens, these being the regions closest to the shaft where the welds occurred. For the SS and SR sequences, A5 was the point of greatest displacement (8.01 mm and 9.22 mm, respectively). However, for the SC sequence, the point of the highest displacement peak was A5 (7.52 mm), with a value of 6.1%, 18.4% lower when compared with the points of SS and SR, in that order.
In conclusion, the symmetrical method (SM) was the sequence that least distorted and single pass (SP) was the sequence with the most symmetrical distortions. In average displacement values, the SM sequence deformed around 9.8% less than the single pass (SP) and 12% less than the backward method (BM).