结合长短期混合特征的均衡风缸控制系统健康状态评估方法  

Health Status Assessment Method Combining Short and Long Term Hybrid Features for Equalizing Reservoir Control System

作　　者：高殿柱[1,3] 彭军 王胜男[2,3] 张晓勇 杨迎泽 黄志武[1,3] GAO Dianzhu;PENG Jun;WANG Shengnan;ZHANG Xiaoyong;YANG Yingze;HUANG Zhiwu(School of Automation,Central South University,Changsha 410083,China;School of Computer Science and Engineering,Central South University,Changsha 410083,China;Hunan Engineering Laboratory of Railway Vehicle Braking Technology,Changsha 410075,China)

机构地区：[1]中南大学自动化学院,湖南长沙410083 [2]中南大学计算机学院,湖南长沙410083 [3]轨道车辆制动技术湖南省工程实验室,湖南长沙410075

出　　处：《铁道学报》2023年第8期57-67,共11页Journal of the China Railway Society

基　　金：湖南省自然科学基金(2021JJ30876)。

摘　　要：制动机均衡风缸控制系统具有部件种类多、退化过程复杂、工况多样等特点,给系统的健康状态评估带来挑战。提出一种结合长期和短期混合特征的健康状态评估方法,通过对均衡风缸系统关键部件进行寿命预测进行健康状态评估。结合系统工作模式切换对部件退化过程的影响提取累积物理动态指标,使用循环神经网络实现部件的长期累积寿命预测;在时间滑窗内提取局部时域和频域特征,表征系统短期波动对老化的影响,利用改进的轻量梯度提升机进行短期寿命预测;通过分段模型平均法进行长期和短期寿命预测结果的融合,得到健康状态评估结果。通过搭建制动机均衡风缸控制系统实验平台,进行加速老化实验,获取全寿命周期老化数据,对提出的混合健康状态评估方法进行仿真,验证所提方法的有效性。The characteristics of multiple components,complex degradation process and diverse operating conditions of equalizing reservoir control system(ERCS)of brake system pose challenges to the health state assessment of the system.Therefore,this paper proposed a health status assessment method combining the short-and long-term hybrid features,which evaluates the health status by predicting the lifetime of the key components in the system.The cumulative physical dynamic indicators were extracted by combining the effects of system mode switching on the component degradation process,and long-term cumulative life prediction of components was achieved using recurrent neural networks.Local time-and frequency-domain features were extracted within the time sliding window to characterize the effects of short-term fluctuations of the system on aging.An improved lightGBM was used for short-term life prediction.The short-and long-term lifetime prediction results were fused by piecewise averaged model to obtain the health status assessment results.The establishment of an experimental platform for the ERCS,implementation of accelerated aging tests,acquirement of the whole life cycle aging data and simulation of the proposed hybrid health status assessment method verified the effectiveness and accuracy of the proposed methods.

关 键 词：制动机 健康状态 循环神经网络 轻量梯度提升机 时间滑窗 

分 类 号：U264.911[机械工程—车辆工程]