RBF近似模型在汽车碰撞变复杂度建模中的应用  

Application on Variable Complexity Models for Vehicle Safety Based on RBF Meta-modeling Technique

作　　者：谢晖[1,2] 陈龙[1,2] 李凡[2] 

机构地区：[1]湖南大学汽车车身先进设计制造国家重点实验室,长沙410082 [2]湖南大学机械与运载工程学院,长沙410082

出　　处：《机械科学与技术》2016年第10期1624-1628,共5页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家科技支撑计划项目(2012BAF12B20);国家"863"项目(2013AA040605);2013长沙市科技计划重点项目(K1308116-11);2013湖南省科技支撑计划重点项目(2014GK4015)资助

摘　　要：针对汽车正面碰撞过程中考虑到车身后部部件结构几乎不发生明显变形的特点,采用最优拉丁超立方试验设计,利用较少的样本点数据在简化模型和高精度模型中建立一个差值补偿近似模型,在此模型和简化模型基础上构造新的样本点数据,利用RBF近似模型建立新的近似模型。利用该方法对以汽车前部7个主要吸能部件的板厚作为设计变量,以整车质量和整车碰撞加速度峰值aB作为优化目标函数的汽车100%正面碰撞的多目标优化问题进行仿真研究,结果表明该方法在保证模型精度的同时能够快速地得到优化值。Aiming at the complexity characteristics of automobile safety optimization and considering the less deformation of the part of the rear structure,a compensation response surface model was created by generating optimal latin square design through the small sample data spanning the simplified model and high accuracy model.And then,a new RBF（ radical basis function） response surface model was established on the basis of new test data which was received through the simplified model and compensation response surface model. The method is applied to vehicle 100% frontal impact multi-objective optimization design for the lightest quality and minimum body B column peak acceleration through selecting 7 parts which influence greater than the others of the front structure. The results show that the method can ensure the model precision,rapid convergence to the optimal solution.

关 键 词：汽车安全性 变复杂度 RBF神经网络 多目标优化 

分 类 号：U462.2[机械工程—车辆工程]