欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

JASE ›› 2017, Vol. 08 ›› Issue (03): 246-251.DOI: 10.3969/j.issn.1674-8484.2017.03.004

• 汽车安全 • 上一篇    下一篇

基于改进EGO 算法的汽车40% 偏置碰撞优化设计

宋正超，章斯亮   

  1. 泛亚汽车技术中心有限公司，中国上海，201201
  • 收稿日期:2017-02-16 出版日期:2017-09-28 发布日期:2017-10-03
  • 作者简介:第一作者 / First author ： 宋正超(1979—)，男( 汉)，湖北，工程师。E-mail: yorksongster@163.com。

Optimizational design for 40% offset vehicle frontal crash based on a modified EGO algorithm

SONG Zhengchao, ZHANG Siliang   

  1. Pan Asia Technical Automotive Center Co., Ltd., Shanghai 201201, China
  • Received:2017-02-16 Online:2017-09-28 Published:2017-10-03

PDF

433

摘要：

为提升优化的精度和效率，对某多用途车(MPV) 车型进行整车正面偏置碰撞结构优化设计。以整车碰撞后侵量和变形量等为约束条件，考虑了序列样本对目标响应和约束响应的改进效果，建立了基于Kriging 模型的改进的高效全局优化(EGO) 算法和相应的序列采样优化流程。结果表明: 与不考虑序列采样的传统优化方法、Jones 经典EGO 序列采样算法和Schonlau 约束EGO 序列采样算法进行对比，该算法可以在最小的112 个样本规模下，得到误差小于8.42%的优化解，碰撞案例在减质量2.89 kg，且所有碰撞约束性能均满足要求的情况下，目标碰撞有效加速度从28.48 g 下降为26.77 g。从而，验证了该方法的准确性和效率。

关键词: 整车开发, 汽车碰撞, 近似模型误差, Kriging 模型, 高效全局优化(EGO) 算法, 改进的EGO 算法

Abstract:

An optimizational design was investigated at vehicle frontal crash cases for a multi-purpose vehicle (MPV) at 40% offset frontal crash to improve the optimization accuracy and efficiency. A modified Efficient Global Optimization (EGO) algorithm was built based on Kriging model considering the improved effect of
sequence samples of target response and constraints with a new established sequential sampling process at constraint conditions of intrusion and deformation in vehicle crashes. The results show that using the proposed modified EGO algorithm has a mininal sample number of 112 and an error of less than 8.42% with the target acceleration reducing from 28.48 g to 26.77 g in the case of crash, compared with the algorithm without considering sequential sampling, Jones classical EGO sequential sampling algorithm and Schonlau constraint EGO sequential sampling algorithm; while all the other constrained crash performances meet the requirements with reduced mass of 2.89 kg. Therefore, the accuracy and efficiency of the method are verified.

Key words: vehicle development, vehicle crash, metamodel prediction error, Kriging model, efficient global optimization (EGO) algorithm, modified EGO algorithm