Simultaneous structure design optimization of multiple car models using the K computer

Abstract

A design tool for simultaneous structure design optimization of multiple car models is developed and applied to simultaneous design optimization of MAZDA CX-5 (SUV car), MAZDA6 wagon (large car), and MAZDA3 hatchback (small car) to demonstrate feasibility of the developed design tool. This problem is a multiobjective design optimization problem where the objectives are minimization of total weight of three car models and maximization of number of common thickness parts. To find the nondominated designs efficiently, an evolutionary multiobjective optimization algorithm named Cheetah is adopted. For analysis of the obtained nondominated and dominated designs, an interactive data mining tool iSPM is used. To evaluate crash performances accurately, a finite-element software LS-DYNA is used for each crash mode. To overcome difficulty in computational time, the K computer is used, where the required computational time is more than 360 hours using 73,728 cores of the K computer. As a result, many designs that outperform the initial design developed by Mazda Motor Corporation are found. In addition, the obtained nondominated designs reveal the trade-off between the two design objectives of the simultaneous structure design optimization problem of multiple car models.