Welding is a critical manufacturing process widely employed in industry for joining two or more materials through localized melting and subsequent solidification. Among the various welding techniques, Gas Metal Arc Welding (GMAW) is extensively used due to its high efficiency, versatility, and suitability for joining both ferrous and nonferrous materials. Optimizing GMAW process parameters is essential for improving weld quality, minimizing defects, and enhancing structural integrity in industrial applications. However, existing studies often rely on either statistical methods or machine learning approaches independently, with limited comparative analysis of their predictive capabilities, particularly within a simulation-based framework. This study aims to analyse and optimize key GMAW process parameters and to evaluate the predictive performance of Response Surface Methodology (RSM) and Artificial Neural Network (ANN) models. Finite element simulations are performed using Simufact Welding software to investigate the influence of welding current, arc voltage, and welding speed on output responses, including peak temperature, welding-induced deformation (distortion), and maximum residual stress. The simulated data are further analyzed using RSM to develop predictive mathematical models and examine interaction effects among input parameters, while an ANN model is implemented to enhance prediction and validation. The results indicate that both approaches are effective in modelling the process; however, RSM demonstrates superior predictive accuracy, as evidenced by a lower root mean square error (RMSE) compared to the ANN model. The key finding of this study highlights the effectiveness of RSM as a reliable and accurate tool for optimizing GMAW process parameters within a numerical simulation framework.
Ishak et al. (Sun,) studied this question.