MACHINE-LEARNING-ASSISTED OPTIMIZATION OF DIS-SIMILAR STAINLESS STEEL RESISTANCE SPOT WELDING PARAMETERS FOR ENHANCED JOINT STRENGTH

ABSTRACT: Resistance spot welding (RSW) is extensively applied in high-volume manufacturing because of its short processing cycle and compatibility with automated production lines; however, achieving reliable joints in dissimilar stainless steels remains difficult due to variations in thermal conductivity and electrical resistance at the weld interface. This investigation aims to enhance joint strength in resistance spot welding of AISI 316L stainless steel and Duplex Stainless Steel 2205 using an integrated framework that combines structured experimentation with predictive statistical modeling. A full factorial design involving 27 welding trials was conducted to analyze the individual and interaction effects of welding current, weld time, and electrode tip diameter on tensile shear strength. Parameter ranking and dimensional reduction were performed through Taguchi analysis and Principal Component Analysis, while polynomial regression modeling was employed to forecast weld performance and determine optimal operating windows. The results identified electrode tip diameter as the most influential factor, followed by weld time and welding current. The developed regression model demonstrated strong predictive reliability with a coefficient of determination (R² > 0.98), indicating close agreement between experimental and predicted values. The optimal parameter combination of 9 kA welding current, 9 cycles weld time, and a 6 mm electrode tip diameter produced a maximum tensile shear load of 17.6 kN. The study confirms that integrating systematic experimental planning with predictive analytics significantly reduces trial-and-error effort, improves weld quality, and enables efficient parameter selection, thereby supporting data-driven and intelligent manufacturing practices for dissimilar material joining applications.