A parametric study of laser beam welding of dissimilar AISI 316L and AA6061 joints

This study evaluates the continuous wave laser beam welding of AISI 316L and AA6061 in an overlap joint arrangement. Experiments are conducted using a Central Composite Design under the response surface methodology approach to examine the impact of laser power, welding speed and laser spot diameter on tensile shear strength and microhardness. The maximum tensile shear strength of 49 MPa is achieved at 1500 W laser power, 350 mm/min welding speed and 1 mm laser spot diameter, whereas the highest microhardness of 275 HV is acquired at 1600 W, 400 mm/min and 1.5 mm spot diameter due to the formation of hard Fe-Al IMCs. Gaussian Process Regression models, optimised through Bayesian optimisation and validated using leave-one-out cross-validation (LOOCV), demonstrated excellent predictive performance, achieving cross-validation coefficients of determination (R2) of 0.9845 for tensile shear strength and 0.9825 for microhardness with low prediction errors. Model interpretability is enhanced using Explainable Artificial Intelligence, where SHAP analysis quantified parameter contributions and Partial Dependence Plots illustrated their marginal effects. XRD analysis confirmed the presence of brittle Fe–Al intermetallic phases. The proposed experimental machine learning framework provides a valuable insight for process optimisation and performance prediction in dissimilar laser-welded AISI 316L–AA6061 joints.