基于多状态混合模型的城际列车走行部轴承智能故障诊断方法研究

As fundamental components of intercity trains, rolling element bearings always work under harsh conditions, such as high speed, heavy load. Due to the complex structure and various components of bogie, vibration-based condition monitoring is vulnerable in interfering signals, leading to a difficulty in feature extraction of bearings. Moreover, bearing fault diagnosis is challenging in engineering applications, because it involves different signal processing techniques, which requires the demand on the user’s expertise. To address the above problems, i.e. multi-source mixture and intelligence, this project focuses on the fault-induced mechanism and intelligent fault diagnosis method by means of the real signal measured from the intercity train bogie. First, inspired by the association analysis, a coupling effect among different failures would be constructed to reveal the fault-induced mechanism of multi-source mixture. Second, a multi-state mixture model controlled by the unknown hidden variables is proposed by modelling the vibration signal itself. In addition, the proposed method addresses the following objectives at once, in the same algorithm, such as fault detection, feature extraction and bearing signal reconstruction. Finally, to achieve the intelligent fault diagnosis, a data-driven self-running algorithm is provided which intends to simplify the parameter setting procedure. This proposal aims to improve the performance of incipient fault diagnosis in cases of multi-source mixture, while providing an intelligent method to reduce the demand on the user’s expertise. It is also a valuable reference for engineering applications in the field of automated bearing fault diagnosis of high-speed trains.

滚动轴承作为城际列车运行基础零部件，常常工作于高转速、重载荷等恶劣工作环境，由于列车走行部结构复杂、部件较多，振动监测易受干扰信号影响，导致轴承故障特征难以识别。此外，故障诊断方法涉及不同信号处理技术，对使用者的专业知识和工作经验要求较高，给工程转化带来极大困难。针对上述多源混叠和智能化等难点问题，本项目以全实物高铁轴承数据为基础，将从轴承故障演化机理和智能故障诊断两个层面展开研究：首先，研究故障动态演变规律，利用关联分析方法建立失效模式间的耦合关系。其次，依据真实数据自身的特性，建立由隐含变量控制的多状态混合模型，实现多源信号混叠情况下的故障报警、特征提取、信号重建等功能。最后，研究数据驱动的自启动算法，通过简化手动选参过程，实现智能化的故障诊断方法。本研究有望提高多源混叠工况下对轴承早期异常的诊断能力，并且减轻对基层技术人员的专业背景要求，为最终实现工程应用提供重要理论依据和有用参考。

轴承是城际列车的基础部件，同时也是易损部件，其工作状态对列车安全有着重大的影响。本项目针对复杂工况下城际列车走行部轴承健康监测、故障诊断的实际需求，提出了集早期故障报警、耦合故障诊断等功能的智能诊断框架，为优化我国城际列车运维策略提供理论依据。针对列车复杂工况的特点，搭建了基于列车轴承高转速、重载荷等特殊工况的全实物轴承实验平台，通过设置不同轴速、不同（方向和大小）载荷的工况，获取了包括内圈故障（严重故障、中度故障）、外圈故障（严重故障、中度故障）、滚子故障（严重故障、轻微早期故障）、保持架故障（严重故障）和正常轴承的振动信号，构成列车轴承故障数据集。针对多源混叠和智能化等难点问题，项目以全实物轴承数据为基础，研究了故障动态演变规律，依据真实数据自身的特性，建立了由隐含变量控制的时频域多状态混合模型，实现了多源信号混叠情况下的故障报警、特征提取、信号重建等功能。基于全实物轴承故障数据集，逐一验证以上研究内容的有效性、合理性，并优化了模型的超参数。最后，通过“城轨列车旋转机械部件故障诊断算法研制”课题，获得了地铁线路上的真实监测数据。结合工程现场需求，不断优化上述理论模型，开发了适用于现场振动监测信号的方法，初步完成了从理论创新到工程实践应用，满足了城轨列车在线监测和故障诊断告警的实际需求。