Computational Multi-objective Optimization of Friction Stir Welding Parameters for AZ31B Mg/6061-T6 Al Dissimilar Joints: A Meta-modeling Approach Based on Literature Data

This study proposes a computational framework for multi-objective optimization of friction stir welding (FSW) parameters in dissimilar AZ31B magnesium and 6061-T6 aluminum joints. Given the experimental complexity and cost associated with parameter optimization for dissimilar material systems, we developed predictive meta-models for peak temperature, residual stress, and distortion using response surface methodology calibrated against published experimental data. The temperature model demonstrated strong predictive capability (R² = 0.834, MAPE = 1.88%) when validated against twelve independent configurations from literature. A multi-objective genetic algorithm was subsequently employed to identify Pareto-optimal welding parameters that minimize thermal exposure, residual stresses, and geometric distortion. The analysis yielded 25 non-dominated solutions, with the compromise solution achieving 25.7% reduction in predicted residual stresses compared to baseline parameters while maintaining thermal stability. Optimal parameters were identified as: rotational speed 700 rpm, welding speed 61 mm/min, shoulder-to-pin ratio 2.51, and plunge force 6728 N. The results indicate that lower heat input conditions favor stress minimization in Mg-Al dissimilar joints, consistent with established thermo-mechanical principles. While experimental validation of the stress and distortion predictions remains necessary for absolute quantification, the methodology provides a cost-effective computational tool for preliminary process design and parameter screening, potentially reducing the experimental burden in industrial FSW development.