机车前端薄壁吸能管仿真模型模糊参数的支持向量回归反求

Support vector regression based method for the inverse problem of fuzzy parameters in the simulation model of thin-walled energy-absorbing tubes installed in the locomotive front end

为了获得影响耐撞性结构有限元计算精度的准确模型参数,提高冲击仿真的准确性,提出一种基于支持向量回归(support vector regression,SVR)模型进行参数优化反求的方法。以一种机车前端防爬结构中的预压薄壁吸能圆管为研究对象建立有限元模型,进行台车冲击试验验证仿真模型准确性。通过拉丁超立方试验设计驱动有限元模型进行少量计算获得数据集,有限元模型中的模糊参数为输入变量,计算与试验载荷的差异为目标响应,通过SVR方法构建映射关系,并采用增强精英保留遗传算法(strengthen elitist genetic algorithm,SEGA)对超参数进行优化,确定SVR模型最佳配置;通过该最优SVR模型再次使用SEGA优化反求,获得最佳模糊参数组合。使用这组参数组合设置有限元模型,其仿真结果相较初始计算耐撞性指标和载荷曲线匹配程度都得到了提高。研究结果为有限元模型中模糊参数的准确设定、碰撞仿真的精度提升提供了一种新的思路。

In order to obtain accurate model parameters influencing the finite element calculation accuracy of crashworthiness structures and improve the precision of impact simulation,a method based on support vector regression(SVR)for solving parameter inverse problem was proposed.A finite element model was established for the preloaded thin-walled energy-absorbing circular tube installad in the locomotive front anticlimbing structure.The accuracy of the model was validated through tests.A Latin hypercube experimental design was employed to drive the finite element model with a small amount of calculation to obtain a dataset.The fuzzy parameters in the finite element model were used as input variables,the difference between the calculated and experimental loads was taken as the target response.The SVR method was applied to construct a mapping relationship,and the strengthen elitist genetic algorithm(SEGA)algorithm was utilized to optimize the hyperparameters.The optimal SVR model was then used again with SEGA optimization for solving inverse problem,and obtaining the best combination of fuzzy parameters and then using them to configure the finite element model.The simulation results show an improvement in the matching degree of crashworthiness indicators and load curves compared to the initial calculations results.The study provides a new approach for accurately setting fuzzy control parameters in finite element models and improving the accuracy of collision simulations.