Proposal of benchmark problem based on real-world car structure design optimization

Abstract

A benchmark problem based on a real-world car structure design optimization1 is proposed. The benchmark problem is constructed by using a response surface method from the design optimization result of a car structure design optimization problem. Because this benchmark problem bases on actual car structure design optimization result conducted by Mazda, it contains features of real-world design optimization problems. Objectives are the minimization of the total weight of the different type of cars and maximization of the number of common gauge (standard plate thickness) parts among the different type of cars. This benchmark problem has 148 discrete design parameters and 36 design constraints for simultaneous design optimization of two types of cars and 222 discrete design parameters and 54 design constraints for simultaneous design optimization of three types of cars. Thus, it is a scalable and multiobjective design optimization problem with many and discrete design parameters and many and severe constraints. The benchmark problem is available on our website (http://ladse.eng.isas.jaxa.jp/benchmark/).