基于微分进化算法的列车吸能方管结构耐撞性多目标优化设计  

Multi-objective optimization design of train crashworthy tube and analysis of energy absorption based on differential evolution algorithms

作　　者：舒冬[1] 姚松[1] 

机构地区：[1]中南大学轨道交通安全教育部重点实验室,湖南长沙410075

出　　处：《铁道科学与工程学报》2011年第2期92-96,共5页Journal of Railway Science and Engineering

基　　金：国家科技支撑计划项目(11504-XF-034)

摘　　要：采用正交试验设计方法及碰撞计算程序,得出了均匀分散并整齐可比的试验数据点。提出了利用无约束微分进化算法来构建不同的回归模型,并结合方差分析技术选出最优模型的方法,解决了传统单目标优化设计只能对其中一个目标进行优化和传统响应面法利用最小二乘得出回归模型过程中因矩阵奇异而导致拟合精度达不到要求的问题。采用动态调整控制参数策略对标准微分进化算法进行了改进,提高了该算法求解多目标问题的效率和寻优能力。利用约束条件下的改进微分进化算法快速有效的求出最优解,最终优化解的各个目标函数值均得到改善。数值算例表明:该方法具有较高的精度和工程应用价值。Using the orthogonal design method and LS - DYNA which is a kind of collision calculation software to get the dispersed and comparable experiment data points, the problem of traditional single - objective optimization design was solved. As it optimizes one objective and the defect of low accuracy caused by matrix singular while calculating the regression model using the least squares in the traditional response surface method, the method of differential evolution without constraints was quoted for the construction of the different regression models to choose the best model by the method of analysis of variance. Moreover, the parameter of dynamic adjustment control algorithm was used to overcome the defects of the standard differential evolution algorithm which is not suitable for solving multi - objective optimization problem, individual evaluation and convergence problem. The efficiency of solving multi - objective problem and searching capability was enhanced too. Thus, the optimal solution was calculated quickly and efficiently with the constrained differential evolution algorithms. All of the target values were improved. Numerical example indicates that this method has high precision and engineering practicability.

关 键 词：碰撞模型 优化设计 多目标 微分进化 

分 类 号：U260.2[机械工程—车辆工程]