Multi-Objective Optimization of Automotive Stamping Process Based on Orthogonal Experimental Design and BPNN-NSGA-II-TOPSIS Framework

To address the multi-objective optimization of stamping process parameters for vehicle components, a hybrid framework coupling Dynaform simulations, Orthogonal Experimental Design, BPNN, NSGA-II, and TOPSIS was proposed. Taking car door accessories as a case study, OED was utilized to investigate the interacting effects of blank holder force, friction coefficient, stamping speed, and die clearance on the maximum thinning and thickening rates. The simulated data trained a BPNN to construct the highly non-linear mapping relationships. Subsequently, the NSGA-II algorithm generated the Pareto optimal frontier, and TOPSIS objectively selected the best compromise process parameters: a blank holder force of 43,113 N, a friction coefficient of 0.13, a stamping speed of 2496 mm/s, and a die clearance of 1.0 mm. Applying this combination effectively constrained the maximum thinning and thickening rates to 41.4% and 16.8% respectively. The BPNN relative prediction errors against numerical validation were merely 0.0691% and 6.9930%, fully verifying the high fidelity and effectiveness of the proposed optimization methodology.