Adaptive approximation-based multi-objective hybrid optimization method for dual-gradient top-hat structures

Abstract

In this article, an adaptive approximation-based multi-objective hybrid optimization method, integrating a multi-objective decision-making method, adaptive approximation method and hybrid optimization method, is developed. In particular, a multi-objective decision-making method developed by combining a weight strategy inspired by grey relational analysis and entropy analysis is applied to normalize multiple objectives; the trust region is introduced into the support vector regression model to solve the problem of the sequential sampling for the approximation model; a hybrid optimization algorithm combining the genetic algorithm pattern search algorithm is developed to further assist the surrogate model-based optimization. Finally, a novel dual-gradient structure with graded thickness and graded material property along the axial direction is applied to the crashworthiness optimization design. The results demonstrate that the developed hybrid method has faster optimization efficiency than the Pointer algorithm built into Isight commercial software, and the final optimized structure is superior to the original one.