中南大学学报（自然科学版）2024，Vol.55Issue(3)：1218-1230,13.DOI:10.11817/j.issn.1672-7207.2024.03.032
基于机器学习的长编重联动车组碰撞能量管理方案优化
Optimization of collision energy management for long series EMU based on machine learning
摘要
Abstract
In order to effectively relieve the pressure of railway transportation,the passenger capacity can be increased doubly by adopting the reconnection operation of EMUs.However,once a collision accident occurs,the huge collision energy will cause serious occupant injury and property loss.Therefore,the research on collision energy management of long series EMU has become a focus object.In this paper,two modes of collision energy dissipation,namely concentrated dissipation and uniform dissipation,were proposed.With the platform force and compression stroke of the energy absorption device at the head and middle car ends as design parameters,the optimal design of collision energy management of long series EMU was carried out based on KNN,MLS,RBF and RF machine learning algorithms.The results show that MLS and RBF are the best machine learning models for predicting the energy absorption of the head car and the variance of energy absorption of the middle car,respectively,with relative errors within 4%.The platform force of the energy absorption element of the head car and the middle car is the main parameter affecting the energy absorption of the head car,and the parameters of the energy absorption device of the middle car are the main parameters that affect whether the energy distribution of middle vehicle is uniform.In the concentrated dissipation mode,48.24%of the collision energy is absorbed by the front car and the reconnection interfaces,and 51.76%of the collision energy is absorbed by the middle car,and this energy distribution mode requires higher energy absorption at the front end of the front car.In the uniform dissipation mode,only 22.75%of the collision energy is absorbed by the head and reconnection interfaces,while 77.25%is absorbed by the middle interfaces.This energy distribution mode will increase the distance between cars and lead to the increase of train length.This two optimized collision energy management modes can ensure the integrity of the car body structure of long series EMU under the collision condition of 36 km/h,and the maximum of 120 ms average acceleration of the car body is 2.64g and 2.36g respectively.关键词
长编重联动车组/碰撞能量管理/多目标优化/NSGA-Ⅱ/机器学习Key words
long series EMU/crash energy management/multi-objective optimization/NSGA-Ⅱ/machine learning分类
交通工程引用本文复制引用
姚曙光,谢旻翰,李治祥,张鹏,董云辉..基于机器学习的长编重联动车组碰撞能量管理方案优化[J].中南大学学报（自然科学版）,2024,55(3):1218-1230,13.基金项目
湖南省自然科学基金资助项目(2021JJ30853) （2021JJ30853）
国家重点研发计划项目(2021YFB3703801-04)(Proiect(2021JJ30853)supported by the Natural Science Foundation of Hunan Province （2021YFB3703801-04）
Project(2021YFB3703801-04)supported by the National Key Research and Development Program of China) （2021YFB3703801-04）
