Abstract: In order to lower production costs and create lightweight components, manufacturers are increasingly focusing on the combining of different materials. Welding with tungsten inert gas (TIG) is the chosen method for achieving a defect-free, reliable joint with a notably pleasing appearance. However, it's important to note that when dealing with materials that possess distinct chemical, physical, and thermal properties, special care is required to ensure the resulting joint is robust and rigid. This study delves into an investigation and modelling of TIG welding parameters concerning the size of pores and its impact on weld quality. The Response Surface Methodology (RSM) model has produced a numerical optimal solution consisting of a current of 200.72A, voltage of 20V, wire diameter of 2.40mm, and wire feed speed of 20m/s, which will yield a weld pore size of 0.195185. This solution was deemed optimal, boasting a desirability value of 93.9%, based on the design expert's evaluation. In parallel, an Artificial Neural Network (ANN) was employed in this study. For training purposes, 70% of the data was utilized, with 15% allocated for validation and the remaining 15% for the actual testing. The results have culminated in the creation of a regression plot, demonstrating the correlation between the input variables and the target variable, which yielded R2 values of 0.82928. Upon a comprehensive evaluation of the results, the Artificial Neural Network emerged as the superior predictive model compared to the Response Surface Methodology because the ANN output fits closer to the experimental than that of RSM. Thus, the approaches effectively optimized and predicted the weld pore size.
Keywords: TIG welding, weld Pore size, RSM, ANN