Surrogate Model and Multi-objective Evolutionary Algorithm Applied to Automotive Stamping

Abstract

Automotive parts stamping is a process of forming metallic parts that are used in the manufacture of automobiles, such as the Internal Tailgate. The problem is characterized as a multi-objective optimization problem, as it involves the optimization of multiple antagonistic objectives simultaneously. In this study, we used the Extra Trees multivariate regression algorithm to characterize the problem and use it as a surrogate model for the Non-dominated Sorting Genetic Algorithm II. Our goal is to explore the possibilities of simultaneously minimizing Fracture, Insufficient Elongation, and Wrinkling. In addition, we apply data correlation analysis and machine learning model sensitivity analysis tools to assess the quality of the surrogate model and allow preliminary decision-making. The tools of data correlation analysis and sensitivity analysis of the Machine Learning model are applied to assess the quality of the surrogate model and allow prior decision-making. The results achieved when using the proposed approach with these auxiliary tools help to establish an efficient model, identify suitable stamping parameters, and reduce the costs associated with conducting empirical research conducted by experts in the field.