主办：陕西省汽车工程学会
ISSN 1671-7988  CN 61-1394/TH
创刊：1976年
汽车实用技术

汽车实用技术 ›› 2025, Vol. 50 ›› Issue (21): 31-37.DOI: 10.16638/j.cnki.1671-7988.2025.021.006

• 设计研究 • 上一篇    

基于 RBF 神经网络的车门耐撞性优化设计

冷川 1，刘家员 2   

  1. 1.重庆人文科技学院 机电与信息工程学院； 2.澳汰尔工程软件（上海）有限公司
  • 发布日期:2025-11-04
  • 通讯作者: 冷川
  • 作者简介:冷川（1992－），男，硕士，讲师，研究方向为汽车被动安全及车身轻量化
  • 基金资助:
    重庆人文科技学院科学研究项目青年项目（CRKZK2021008）；重庆市教委科学技术研究计划项目青年项目 （KJQN202401803）

Optimization Design of Door Crashworthiness Based on RBF Neural Network

LENG Chuan1 , LIU Jiayuan2   

  1. 1.School of Mechanical and Information Engineering, Chongqing College of Humanities Science and Technology; 2.Altair Engineering Software (Shanghai) Company Limited
  • Published:2025-11-04
  • Contact: LENG Chuan

PDF

摘要： 文章应用径向基函数（RBF）神经网络与多目标遗传算法，对汽车车门进行高效的耐 撞性优化设计。在 HyperMesh 软件中建立车门侧面碰撞的有限元模型，对其进行耐撞性分析。 以车门关键部件厚度为设计变量，采用哈默斯雷试验设计方法生成分布合理的样本点进行计 算，对计算结果应用 RBF 神经网络方法构建了车门质量、吸能值、最大碰撞力和侵入量的近 似模型。以车门吸能值最大与质量最小为目标，以车门最大碰撞力与侵入量为约束，应用多 目标遗传算法对近似模型进行优化设计。优化后的车门在最大碰撞力与侵入量均减少的情况 下，吸能值提高了 29.33%，质量减少了 8.5%，耐撞性得到提高，且设计时间得到有效缩短。

关键词: 车门；耐撞性；优化设计；RBF 神经网络；多目标遗传算法

Abstract: In this paper, radial basis function (RBF) neural network and multi-objective genetic algorithm are applied to optimize the crashworthiness of automobile doors. The finite element model of the side impact of a car door is established in HyperMesh, and its crashworthiness is analyzed. Taking the thickness of the key components of the door as the design variables, reasonably distributed sample points are generated by Hammersley experiment of design method for calculation. The RBF neural network method is used to construct the approximate models of the door mass, energy absorption value, maximum impact force and intrusion. With the maximum energy absorption value and minimum mass of the door as objectives, and the maximum impact force and intrusion of the door as constraints, the multi-objective genetic algorithm is used to optimize the approximate models. Under the condition that the maximum impact force and the amount of intrusion are reduced, the energy absorption value is increased by 29.33%, the mass is reduced by 8.5%, the crashworthiness is improved, and the design time is effectively shortened.

Key words: car door; crashworthiness; optimization design; RBF neural network; multi-objective genetic algorithm