The thicknesses of vehicle front parts were defined as variables, and 100 points of vehicle front crash finite element models were sampled by the Latin hypercube design of experimental method. The Kriging model, LSR model and RBF model were then applied to construct the approximate models of the weight of front structure, the maximal acceleration of B pillar and the maximal absorbed energy of front structure according to the FEM simulation results. In the optimization of maximal acceleration of B pillar, the weight of front structure, the maximal absorbed energy of front structure and the thickness of vehicle front parts were constrained. Then, an ASA algorithm and searching the best result in a large number of samples in three spaces were adopted to minimum the maximal acceleration of B pillar and find an optimal thickness of the front parts. The study shows that the precision and efficiency of this method is good and can meet the crashworthiness engineering requirements.

crashworthiness, Kriging model, least squares regression(LSR) model, radial basis function(RBF) model, Latin hypercube, adaptive simulated annealing(ASA)

以汽车前部结构主要零件的厚度为变量，采用拉丁超立方试验设计方法生成100个汽车正面碰撞有限元仿真模型样本数据并进行计算，对计算结果应用Kriging模型、最小二乘响应面模型、径向基函数模型构建前部结构质量、B柱加速度最大值和前部结构最大吸能相对于各部件厚度的三种近似模型。以B柱加速度最大值为目标，约束前部结构最大吸能、前部结构零件质量及各零件厚度，利用模拟退火算法和三种空间密集撒点优化搜索方法，最终得到一组最优的前部零件厚度组合，使得B柱加速度最大值最小。研究表明，该方法计算精度和效率较好地满足了耐撞性工程设计的需求。

耐撞性, Kriging模型, 最小二乘响应面模型, 径向基函数模型, 拉丁超立方, 自适应模拟退火算法