Two-stage surrogate-based optimization of a vehicle composite subframe with modified fast flexible space-filling designs

In this article, an optimization procedure of composite structures is proposed and applied to a vehicle subframe redesign. The process is divided into two stages and surrogate models are built by Gaussian Process Regression. The modified Fast Flexible Filling designs is used to deal with the non-rectangular design region problem. And combined with the adaptive sampling method, uniform training and test points which increase the accuracy of surrogate models are obtained. The two established surrogate models are solved by non-dominated sorting genetic algorithm-II and particle swarm optimization algorithm, respectively. The redesigned subframe exhibits significantly weight reduction and performance improvement.