Development of hybrid ANN-GA and RSM models for optimizing mechanical properties in SS400-SUS304 welds

Dissimilar metal welding, such as joining SS400 and SUS304, presents significant challenges due to differences in metallurgical and thermal properties, often leading to internal stresses, cracking, and reduced joint durability. These disparities compromise mechanical performance, necessitating advanced optimization techniques. This study applies a hybrid optimization approach integrating Artificial Neural Networks (ANN), Genetic Algorithms (GA), and Response Surface Methodology (RSM) to optimize Gas Metal Arc Welding (GMAW) parameters for enhanced tensile strength and hardness. Experimental conditions involved welding currents of 80–120 A, speeds of 200–400 mm/min, and a shielding gas composition of 95 % argon and 5 % oxygen. The ANN-GA model demonstrated superior predictive accuracy, with RMSE values of 236.43 N for tensile strength and 2.27 HV for hardness. Maximum tensile strength and hardness of 16,302 N and 192.08 HV, respectively, were achieved at specific parameter combinations. The ANN-GA-optimized conditions of 107.10 A and 336.38 mm/min provided a balance between strength and hardness, significantly improving weld quality. RSM revealed interactions among welding parameters but exhibited lower predictive accuracy than ANN-GA. This study highlights the effectiveness of ANN-GA in capturing nonlinear welding relationships, offering a robust framework for optimizing welding processes. The optimized parameters enhance weld quality, reduce costs, and improve operational efficiency in industrial welding applications