基于Levenberg Marquardt算法的列车最优黏着控制研究  

Research on optimal adhesion control of trains based on Levenberg Marquardt algorithm

作　　者：倪章棚 吴兵 肖广文 沈铨 姚林泉[1,2] NI Zhangpeng;WU Bing;XIAO Guangwen;SHEN Quan;YAO Linquan(College of Rail Transportation,Soochow University,Suzhou 215131;College of Mathematical Sciences,Soochow University,Suzhou 215006)

机构地区：[1]苏州大学轨道交通学院,苏州215131 [2]苏州大学数学科学学院,苏州215006

出　　处：《高技术通讯》2023年第10期1090-1099,共10页Chinese High Technology Letters

基　　金：国家自然科学基金(51605318);江苏省高校自然科学基金(16KJB580008)资助项目。

摘　　要：电空制动是轨道车辆应用最广泛的黏着制动方式,其制动性能主要受制于轮轨间的黏着状态。在复杂低黏着条件下,传统制动控制系统面临的最大问题是无法使黏着时刻保持最优。因此,基于轮轨黏滑特性和车辆动力学理论,本文首先建立以黏着观测器为核心的蠕滑寻优模型;其次提出以Levenberg Marquardt(L-M)算法为核心的神经网络控制器,完成最优黏着控制系统;最后使用Matlab/Simulink平台分别对基于多交替轨面和实验低黏着轨面的列车黏着控制进行仿真模拟,并与传统比例积分微分控制器(PID)作用下的黏着情况做对比。结果表明,即使面对具有不同特性的低黏着轨面,轮轨黏着在控制系统作用下都能迅速维持在当前轨面下的最优值,有效缩短了制动距离和时间。相比传统PID控制,本文提出的控制系统在制动时间和制动距离上同比减小4.9%与4.1%,调控能力更强,适用于低黏着和大蠕滑下的列车制动工况。Electro-pneumatic braking is the most widely used adhesion braking method for rail vehicles,and its braking performance is mainly subject to the adhesion characteristics between wheels and rails.Under the complex and low adhesion conditions,the biggest problem is that it cannot keep the wheel/rail adhesion level be in the optimum adhesion utilization.Therefore,based on the wheel/rail stick-slip characteristics and vehicle dynamics theory,firstly,a creep optimization model with adhesion observer as the core is established;secondly,a neural network controller based on Levenberg Marquardt(L-M)algorithm is proposed to complete the control system;and finally,the Matlab/Simulink is used to simulate the train adhesion control based on multiple alternative rails and experimental low adhesion rails respectively,and it is also compared with the the results of the proportional-integral-differential(PID)controller.The results show that even in the face of low adhesion with different characteristics,the wheel/rail adhesion can quickly maintain the optimal value under the current rail with the actions of control system,reducing the braking distance and time effectively.Compared with PID method,the control system proposed in this paper reduces the braking time and braking distance by 4.9%and 4.1%respectively with stronger control ability.And it is applicable to the train braking conditions under low adhesion and large creep.

关 键 词：低黏着 高速列车 轮轨接触 神经网络控制器 Levenberg Marquardt(L-M) 

分 类 号：U284.48[交通运输工程—交通信息工程及控制]