基于RBF神经网络模型的车门多目标轻量化设计

Multi-objective Lightweight Design of Vehicle Door Based on RBF Neural Network Model

以某轻型客车车门为研究对象,在HyperMesh中建立有限元模型,对车门进行刚度和模态分析。以部分因子试验设计所得到的关键零件厚度为设计变量,采用哈默斯雷试验设计方法进行样本数据设计,使用RBF神经网络模型拟合车门质量、一阶模态频率、二阶模态频率、上扭转刚度、下扭转刚度、侧向弯曲刚度及下沉刚度响应的近似模型。在近似模型基础上,以车门质量最小和一阶模态频率最大为优化目标,以车门刚度和二阶模态频率为约束,应用多目标遗传算法进行优化设计,实现了车门轻量化最优目标。

Taking a light bus door as the research object,its finite element model was set up in HyperMesh software.The stiffness and modal analysis of the door were carried out.Taking the thicknesses of the key parts obtained from the partial factorial design of experiment as the design variables,the sample data was designed by the method of Hammersley test design.RBF neural network model was used to fit the approximate models of door mass,first order modal frequency,second order modal frequency,upper torsional stiffness,lower torsional stiffness*lateral bending stiffness and sinking stiffness response.Based on the approximate models and multi objective genetic algorithm,the optimization goal for a lightweight door was gained with minimizing door mass and maximizing first order modal frequency as optimization objectives,door stiffness and second order modal as constraints.