Multi-Response Optimization of AHSS Spot-Welding Parameters: Desirability, Trade-offs and Sensitivity in a Synthetic Case Study

Abstract

An optimum welding schedule is conditional on the response model, the admissible search region and the decision preferences used to compare candidates. This paper examines those dependencies using the explicitly synthetic DP780-labelled and DP980-labelled response surfaces contained in an AHSS spot-welding thesis. Tensile-shear and cross-tension loads are represented by fitted quadratic models in welding current, weld time and electrode force. A discrete search evaluates 28,701 points inside the convex hull of a three-factor Box–Behnken design before applying assumed width and indentation constraints. Equal-weight geometric desirability selects 9.20 kA, 310 ms and 4.00 kN for the DP780-labelled scenario, with fitted loads of 8.866 and 4.106 kN. For the DP980-labelled scenario, it selects 9.10 kA, 295 ms and 4.00 kN, with fitted loads of 9.611 and 4.245 kN. These are conditional numerical selections, not qualified process settings. Preference-weight changes alter the selected combinations, especially force in the first scenario. Baseline comparisons, interior alternatives, local perturbations and an additional worst-neighbourhood calculation expose differences between nominal performance and decision stability. The analysis retains the source model’s adverse cross-tension lack-of-fit result and does not treat additional random draws as physical confirmation. Its contribution is a reproducible decision audit that makes preferences and feasibility assumptions visible. Experimental qualification remains necessary before any schedule can be recommended for actual AHSS joining.