Multidisciplinary Lightweight Design of Automotive Crashworthy Structure Based on Improved Gaussian Process Model

Autobody lightweight design, on the premise of ensuring vehicle safety and comprehensive performance, is a high-dimensional and highly nonlinear Multidisciplinary Design Optimization (MDO) problem. However, there are problems such as low accuracy of the approximation model and poor convergence accuracy in engineering research. To solve the above problems, this article proposes an MDO design method combining an improved Gaussian Process Regression (GPR) model with a dynamic relaxation factor in Analytical Target Cascading method (DFATC). To improve the accuracy of the GPR approximation model, the combined kernel function is used instead of a single kernel function to have the advantages of multiple single kernel functions, and the particle swarm algorithm is used to adaptively optimize hyperparameters. To achieve high convergence accuracy, the dynamic relaxation factor is introduced to adaptively improve response deviation constraint in ATC to meet the convergence requirements. Finally, the crashworthy structure of the front body is designed using this novel MDO method, and the results show that the proposed method can solve complex engineering problems with high efficiency and accuracy.