基于神经网络的变截面梁抗撞性分析及优化设计

Crashworthiness optimization of a thin-walled rail with variable section based on artificial neural network

通过比较不同截面形状薄壁梁碰撞吸能特性曲线,提出了一种有效的变截面薄壁梁结构,并将该结构应用于某车架前纵梁的碰撞模拟中,以提高其动态吸能能力。将人工神经网络引入抗撞性优化设计中,选取变截面梁的主要设计参数作为研究对象,将有限元分析与试验设计、神经网络和遗传算法等结合起来,对变截面梁各结构尺寸进行了抗撞性优化设计。建立了变截面梁结构的总吸能神经网络预测模型,并采用遗传算法进行了优化求解。最后将优化好的变截面薄壁梁结构应用于整车40%偏置碰模拟中,结果表明A柱的加速度峰值显著降低,整车的被动安全性得到提高。

Here-in,an effective structure of a thin-walled rail with variable section was proposed by comparing a variety of thin-walled rails with different section features.Then,the thin-walled rail was used in the front of a vehicle's frame in order to improve its dynamic crashworthiness characteristics of energy absorption.Several design parameters of the structure were used to optimize the rail's structure.Optimal design method was presented and utilized to obtain the optimal crashworthiness design of the thin-walled rail with variable section.The methodology adopted in this research made use of design of experiments(DOE),finite element analysis(FEA),artificial neural network(ANN)and genetic algorithms(GA).The forecasting model for energy absorption was created using ANN method.The ANN model was interfaced with GA method to find the optimal parameter values and the optimal energy absorption.Then,the optimal results were verified through the finite element analysis of the thin-walled rail.In the end,the optimal thin-walled rail with variable section was used in some vehicle's 40% offset-barrier impact model and the passive safety of the vehicle was improved a lot.