Multi-objective optimization of fiber laser welding parameters for 316L stainless steel

This study proposes a multi-objective optimization framework for fiber laser welding of 316L stainless steel, combining a stacked regression model with multi-objective particle swarm optimization. Key parameters - laser power, welding speed, and defocusing distance - were optimized to minimize carbon emissions while improving tensile strength and weld morphology. The stacked model, integrating random forest, extreme gradient boosting, and support vector regression with a Kriging meta-model, achieved high prediction accuracy, while local interpretable model-agnostic explanations and partial dependence plot were used for interpretability. Validation experiments confirmed that the optimized parameters reduced carbon emissions by 28.98%, increased tensile strength by 20.36%, and improved the depth-to-width ratio by 13.08%. The proposed method provides a concise and effective pathway toward low-carbon, high-quality laser welding, with clear potential for sustainable manufacturing applications.