M Yuvaperiyasamy, Dilli Ganesh V., M Kalaimani, Sabari K., Arunkumar K.
Friction stir welding (FSW) is a solid-state welding process commonly used to join high-strength aluminium alloys due to its ability to produce high-quality welds with few defects and high mechanical performance. This research examined the effects of the main FSW process variables on the mechanical properties of AA2050 aluminium alloy joints through an integrated statistical and Machine learning framework. The key process variables were rotational speed, traverse speed, and axial load, which were optimized with the help of Response Surface Methodology based on the BBD consisting of fifteen experimental runs. Tensile tests and Vickers microhardness tests were conducted to evaluate the mechanical performance of the welded joints. ANOVA was used to assess the statistical significance of the derived regression equations, and the developed models exhibited coefficients of determination (R2) of 0.8567 for tensile strength and 0.8267 for microhardness, indicating moderate-to-good agreement between experimental and predicted values. Multi-response optimization using the desirability function predicted an optimal parameter combination of 989.53 rpm rotational speed, 48.42 mm/min transverse speed, and 7.85 kN axial load, resulting in predicted responses of 298.67 MPa tensile strength and 150.25 HV micro hardness, with an overall desirability of 0.968. Furthermore, machine learning models were employed to validate and assess the predictive performance of the developed RSM models. The microstructural examination revealed significant grain refinement in the stir zone due to dynamic recrystallisation, resulting in a more homogeneous microstructure and improved mechanical performance. The results demonstrate that integrating RSM-based optimization with machine learning provides a reliable framework for optimizing FSW parameters and enhancing the performance of high-strength aluminium alloy welded joints.