A system optimisation design approach to vehicle structure under frontal impact based on SVR of optimised hybrid kernel function

Abstract

Frequent occurrences of road traffic accidents put forward more and more stringent requirements on the vehicle safety performance. To optimise the vehicle structure crashworthiness, the different optimisation design methods, including deterministic, reliability-based and robust optimisation, are performed simultaneously in this study. The support vector regression (SVR) model is employed to approximate responses between design variables and objectives, and the hybrid kernel function (HKF) is introduced to overcome the drawback of a single kernel function of SVR. Meanwhile, the particle swarm optimisation (PSO) algorithm is adopted to optimise HKF–SVR model parameters and improve the accuracy of the model. By combining the nondominated Sorting Genetic Algorithm II (NSGA-II) and the Monte Carlo Simulation (MCS), the proposed optimisation design approach is proven to be an efficient and systematic tool to guarantee the reliability and robustness of the vehicle structure safety design. These different optimisation design results are discussed and contrasted with initial design. The results show that the proposed approach not only improves the crashworthiness and lightweight of vehicle, but also increases the reliability and robustness of design parameters. Through reliable and robust optimisation, more conservative solutions can be generated.