基于宽度学习的自适应相位分割转辙机故障诊断算法研究  

Research on fault diagnosis algorithm for adaptive phase segmentation switch machine based on broad learning

作　　者：朱可桢 胡金杰 刘亮 ZHU Kezhen;HU Jinjie;LIU Liang(Hunan CRRC Times Signal&Communication Co.,Ltd.,Changsha Hunan 410100,China;Jiangsu CRRC Mechanical&Electric Technology Co.,Ltd.,Nantong Jiangsu 226001,China)

机构地区：[1]湖南中车时代通信信号有限公司,湖南长沙410100 [2]江苏中车机电科技有限公司,江苏南通226001

出　　处：《现代城市轨道交通》2025年第4期78-84,共7页Modern Urban Transit

摘　　要：转辙机是轨道交通基础设施系统中的关键机械部件,在日常运行中会产生疲劳损伤、性能退化现象,进而影响列车的行车安全。考虑到转辙机部件耦合关联性强、各运行阶段机理迥异等特点以及故障样本数据缺失和不平衡等问题,文章通过分析故障与外部征兆间的内在关联关系,揭示转辙机各阶段的耦合关系,建立转辙机运行阶段自适应分段算法,提取转辙机不同阶段的多维无量纲特征,构建基于宽度学习算法的故障诊断模型,形成自适应相位分割的转辙机故障诊断算法。通过实验分析表明:该诊断方法的故障辨识准确率达到98%以上,可精确识别转辙机不同运行阶段的各种故障模式,并具备诊断单故障和不同阶段并发故障的能力。The switch machine is a critical mechanical component in rail transit infrastructure systems.It is prone to fatigue damage and performance degradation during daily operations,thereby impacts operation safety of trains.Considering the strong coupling correlations between switch machine components,the diverse mechanisms at different operational stages,as well as issues like missing and imbalanced fault sample data,this article analyzes the intrinsic relationships between faults and external symptoms,reveals the coupling relationships at various stages of the switch machine,establishes an adaptive segmentation algorithm for the operational stages of the switch machine,extracts multidimensional dimensionless features for different stages of the switch machine,establishes a fault diagnosis model on the basis of broad learning algorithm and formulating a fault diagnosis algorithm of the switch machine with adaptive phase segmentation.Experimental analysis shows that the fault identification accuracy of this diagnostic method reaches over 98%and it could accurately identify various fault modes across different operational stages of the switch machine,with capabilities for diagnosing single faults and concurrent faults at different stages.

关 键 词：轨道交通 转辙机 故障诊断 宽度学习 

分 类 号：U284.92[交通运输工程—交通信息工程及控制]