基于对比学习的电液转辙机故障智能诊断方法  

Intelligent Fault Diagnosis Method for Electro-hydraulic Switch Machines Based on Contrastive Learning

作　　者：温伟刚[1] 刘洋[1] WEN Weigang;LIU Yang(School of Mechanical,Electronic and Control Engineering,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学机械与电子控制工程学院,北京100044

出　　处：《铁道学报》2024年第5期92-99,共8页Journal of the China Railway Society

基　　金：北京交通大学基本科研业务费(KMJBZY23003536)。

摘　　要：道岔转辙机是铁路运行调度的关键设备,由于其外部使用环境恶劣、内部设备结构复杂,使得转辙机在工作过程中容易产生不同的运行故障。针对广泛应用于高速铁路的电液转辙机,提出基于对比学习的电液转辙机故障智能诊断方法:对电液转辙机左右液压缸油压监测信号使用对比学习的思想正则化特征空间;使用实例级权重策略增强模型泛化能力;使用多种数据增强方法提高模型鲁棒性。最后通过电液转辙机的运行故障实验验证本方法的有效性与优越性。Switch machine,as the key equipment for railway operation scheduling,is prone to malfunction in the working process because of its harsh external environment and complex internal equipment structure.Aiming at the electro-hydraulic switch machines widely used in high-speed railways,this paper proposed a fault intelligent diagnosis method.The idea of contrastive learning was used to regularize the feature space based on the oil pressure detection signals of the left and right hydraulic cylinders of the switch machine.An instance-level weighting strategy was used to enhance model generalization.A variety of data enhancement methods were used to improve model robustness.Finally,through the operation fault experiment of the electro-hydraulic switch machine,the effectiveness and superiority of the electro-hydraulic switch machine fault intelligent diagnosis based on contrastive learning were verified.

关 键 词：电液转辙机 对比学习 故障诊断 智能诊断 神经网络 

分 类 号：TH17[机械工程—机械制造及自动化] U284.721[交通运输工程—交通信息工程及控制]