Numerical Modeling of Heat-Affected Zone in the GMAW Process

Low-carbon steel St37-2 is widely used in the bus-production industry. The gas-metal arc welding (GMAW) is strongly applied to join steel components due to its ease and low cost. Heat-affected zone (HAZ) is the area between the weld and the base metal, where the welded joint may have the lowest toughness and, hence, it has always been a matter of interest for numerous researchers. Our aim is to study the effects of GMAW parameters on the width of the HAZ. The Taguchi orthogonal array, signal-to-noise (S/N) ratios, and the analysis of variance (ANOVA) are used to investigate and analyze the effect of the welding parameters on the HAZ width.


Introduction
The process of gas-metal arc welding (GMAW) is an important component in numerous industrial operations [1]. This process has various advantages over the other melting welding methods. High welding speed, large metal deposition, and spatter free welding are some of its advantages. In addition, it is applicable to a wide variety of commercial metals and alloys, such as carbon steel, stainless steel, copper, and aluminum. Furthermore, it is a mechanized method and allows us to use robots [2][3][4]. These advantages have motivated many researchers to study the GMAW process in detail [5]. The heat-affected zone (HAZ) is a nonmelted area adjacent to the weld metal in fusion welding processes, which undergoes numerous microstructural changes as compared with the base metal. In several studies [6][7][8][9][10][11], it was indicated that the HAZ may have the lowest toughness in the welded joint and, hence, the importance of HAZ was emphasized. As the use of welded steel structures is intensified, it becomes clear that the HAZ is susceptible to various types of cracking and, especially, cold cracking, which is attributed to the formation of a very susceptible HAZ microstructure [12]. Thus, in analyzing all these problems, it makes sense to note that the minimization of HAZ width should be quite helpful. As far as we know, there is relatively little information about the minimization of HAZ width in low-carbon steels. Therefore, in the present work, we make an attempt to investigate and analyze the effect of GMAW parameters on the HAZ width in St37-2 steel.  applied to identify the most significant factor and predict the optimal setting of parameters [13]. In this work, the "smaller is better" criterion was selected for the HAZ width and the S/N ratios were calculated. This means that the optimal level for a factor is the level that results in the lowest value of the S/N ratio in the experimental region. As a verification of the S/N ratio results, the ANOVA was performed to determine the effect of different parameters on the output variable. Furthermore, the percentage contribution of each control factor was determined according to the results of ANOVA [14]. Finally, a mathematical model was developed and utilized to predict the HAZ width. For this purpose, a linear regression model using the Minitab 16 software was developed for finding the relationship between the predictor variables and the response variable.

Results and Discussion
The results obtained in the nine experimental tests are shown in Fig. 1. According to this figure, the average HAZ width changed from 1.23 to 3.04 mm for various welding conditions. The S/N ratio was calculated for each experiment and tabulated in Table 3. The average values of the S/N ratios for the four control factors on each level were calculated and presented in Table 4. The larger the difference between their maximum and minimum values, the stronger the relative effect of the parameter on the response [15].
It follows from the Delta values presented in Table 4 that the most significant parameter affecting the HAZ width is the welding speed and the next is the wire-feeding speed. The optimal condition for the minimal HAZ width (according to the average S/N values of the process parameters on three levels) is V-2 (20 V), WFS-1 (3 m/min), WS-3 (8 mm/sec) and Alpha-1 (25 deg). The main effect plots generated by the Minitab 16 software in Fig. 2 show the variations of the HAZ width with changes in the input factors. The plot of welding speed is more divergent than for the other factors and, hence, it can be assumed that the welding speed is the most significant factor that affects the HAZ width and the next factor is the wire-feeding speed, while the arc voltage and torch angle have less pronounced effects on the HAZ width. The ANOVA is a statistical tool used to analyze the S/N ratios [16]. It shows the percentage contributions of given input parameters to the measurable output parameter [17]. The ANOVA values for the S/N ratios and the regression model are calculated at a 95% confidence level by using the Minitab-16 software and tabulated in Table 5.
This analysis is carried out for a significance level (α = 0.05). It follows from the ANOVA that the welding speed is the most prominent factor that exerts the maximum influence on the HAZ width with a percentage contribution of 50%, while the next factor is the wire feeding speed with a percentage contribution of 31%. The percentage contribution indicates the relative power of a factor in reducing variation. For a factor with higher percentage contribution, small variations have a great influence on the performance [18].   The coefficient of determination (R-squared) in the ANOVA is a measure of strength of the linear relationship between the experimental and predicted values [19]. In the regression model, the R-squared is a statistical measure of how well the regression line approximates the actual data points [20]. In other words, it indicates the ability of a model to make predictions. The R-squared value is always between 0 and 100% and, in general, the higher the R-squared value, the better, which indicates a better fit to the regression model. The R-squared value equal to 91.59% in the present work is desirable and represents a good fit; moreover, the data closely follow a straight line. The normal probability plot of residual for the HAZ width is shown in Fig. 3. It is easy to see that the data closely follow the straight line. If the value of p in ANOVA is lower than the significance level (α = 0.05), then this means that the factor corresponding to this value of p is statistically significant for the regression model. Hence, the value of p lower than 0.05 for the welding speed and the wire feeding speed indicates that, unlike the arc voltage and torch angle, they have statistically significant effects on the response of HAZ width at the 95% confidence level. A low value of p (< 0.05) means that we can reject the null hypothesis. Conversely, a higher (insignificant) value p suggests that the changes in the predictor are not associated with changes in the response [21]. It follows from the ANOVA of linear regression for the HAZ width that the p -value in the regression equation indicates that the regression model is significant. The F-ratio (named after Fisher [22]) is used to determine the significant factors that affect the response. The higher the F-value for a factor, the greater the effect of variations of this factor on the response. The highest F-value (24.0029) verifies that the welding speed is the most effective factor as indicated in Table 4 and the next factor is the wire feeding speed with an F-value of 14.7565.

CONCLUSIONS
A Taguchi orthogonal array L9 was used to design experiments in the present work and the effect of the parameters of GMAW on the HAZ width in St37-2 steel was investigated and analyzed. According to the accumulated results, the most significant parameter that affects the HAZ width is the welding speed and the next factor is the wire feeding speed. At the same time, the arc voltage and torch angle exert weaker effects on the HAZ width. The ANOVA results show that the welding speed affects the HAZ width with a percentage contribution of 50% and the next factor is the wire feeding speed with a percentage contribution of 31%. In addition, the optimal condition predicted in the present work for the minimum HAZ width is as follows: V-2 (20 V), WFS-1 (3 m/min), WS-3 (8 mm/sec) and Alpha-1 (25 deg).