Sustainable process optimization and machine learning-driven evaluation of lap shear strength in dissimilar SS304–SS420 resistance spot-welded joints

The growing adoption of lightweight, corrosion-resistant structures has increased the use of dissimilar stainless-steel joints in transportation, manufacturing, and structural applications, necessitating reliable process optimization for sustainable resistance spot welding (RSW). This study investigates the welding behaviour of 1.2 mm thick SS304 and SS420 stainless-steel sheets using a mixed-level full-factorial design to evaluate the influence of welding current (7–10 kA), welding time (6–8 cycles), and electrode force (3–6 kN) on joint performance. Apparent lap-shear stress was determined from the measured bonded nugget area to provide a representative assessment of joint strength. Analysis of variance indicated that welding current was the most influential parameter, contributing 87.44% of the total variation, followed by welding time (8.84%) and electrode force (4.56%). The optimum welding condition of 10 kA, 8 cycles, and 5 kN achieved a maximum failure load of 8.58 kN with an apparent lap-shear stress of 437 MPa. Macrostructural observations revealed a transition from interfacial fracture to pull-out failure with indicating improved nugget development and joint integrity. Predictive models based on support vector regression, decision tree, gradient boosting, and random forest were validated using leave-one-out cross-validation and independent experiments, with random forest providing the highest predictive accuracy.