Multi Objective Optimization Of Vehicle Crashworthiness Based On Surrogate Model

Demand for vehicle safety is getting higher and higher with the increasing of car ownership. Vehicle safety is divided into active safety and passive safety. Vehicle can be actively avoid before the accident with active safety. But the vehicle should have a better passive safety performance that can ensure the safety of the passenger when the accident happens. So the optimization of vehicle crashworthiness is very important as the last line of defense for vehicle safety. In order to reduce the cost, the finite element method is used in the vehicle crash. But the simulation process is a nonlinear dynamic process and time consuming and cannot meet the requirement of iterative. So the finite element model should be replaced by surrogate model with high accuracy. It can reduce computational cost and improve computational efficiency. The main content of this paper is multi objective optimization of vehicle crashworthiness by surrogate model.The thickness of the absorbing energy panel is considered as design variable. The minimum of the acceleration of B-pillar and the footwell intrusion and the entire vehicle mass is considered as optimization objective. Based on the fundamental theory of surrogate model, the sample points are selected by using Latin hypercube design. The Gauss radial basis function and Multiquadric radial basis function and Kriging model are built with the sample points. The error analysis is used to the surrogate models and the model with the highest accuracy is selected. The Kriging model is used to fit the model of B-pillar acceleration and footwell intrusion. The Multiquadric radial basis function with c=4.5 is used to fit the model of the entire vehicle mass. These three models are used to instead of finite element model and the accuracy of them is in accordance with the engineering requirements. The multi objective optimization is then conducted using genetic algorithm and particle swarm. The Pareto optimal solution set is worked out and chosen by designer. Finally two sets of optimum combination are obtained by two kinds of optimization algorithm. The B-pillar acceleration is reduced 10.07% and the footwell intrusion is reduced 19.32% and the entire vehicle mass is reduced 0.9kg with multi objective genetic algorithm. The B-pillar acceleration is reduced 4.9% and the footwell intrusion is reduced 6.31% and the entire vehicle mass is reduced 4.9kg with multi objective particle swarm. The optimization effect is obvious. It has theoretical and engineering guiding significance to the optimization design of vehicle crashworthiness.