Optimization of the structural rib layout for the stiff tailgate based on multi-objective Bayesian optimization

Abstract

Structural ribs are widely used to strengthen structures. To lighten the product and increase rigidity, it is important to efficiently place ribs, but in the industry, it is often determined by the empirical judgment of engineers. In this study, we propose an Alternate Multi-Objective Bayesian Optimization framework for structural rib layout optimization. Multi-Objective Bayesian Optimization, an extended framework in Bayesian optimization, is not suitable for dealing with high-dimensional areas due to large computational costs and memory problems depending on the number of training data. The proposed Alternate Multi-Objective Bayesian Optimization framework divides the input parameters into several low-dimensional areas and then optimizes each low-dimensional area alternately. The framework was applied to strengthen the tailgate of the Sport Utility Vehicle (SUV) with structural ribs using simulation data obtained from the Abaqus FEM software. It was confirmed that the optimal structural rib layout parameters simultaneously minimize mass and displacements. It will be useful in the industry because it chooses the optimum design depending on the dataset without human intuition with only a small number of data.