基于多级特征提取的高速列车走行部故障诊断与状态预测

Fault diagnosis and state prediction of running gear system is an effective way to ensure safe, reliable and efficient operation of high-speed trains. Based on the background of high-speed train running gear system, this project takes into account the problem of the hidden behavior, state retention, uncertainty disturbances and real-time data. The theory, algorithm and application of feature extraction, fault diagnosis and state prediction of running gear under actual working conditions are studied. Firstly, multi-level feature extraction is carried out for multi-source operation data to analyze hidden fault features. Secondly, the uncertainty disturbance and state retention are studied. The parameter estimation model of disturbance factor is proposed and the adaptive effect of feedback is analyzed to realize the fault diagnosis of the running gear system. Then, considering the rapid change of real-time data, a multi-layer intelligent structure based on priority scheduling strategy is proposed to enhance the real-time state prediction ability of the running gear system. Finally, the main theoretical results are divided into two stages of validation. In the first stage, a simulation is carried out on the parameters test platform of the running gear of CRRC Changchun railway vehicles. In the second stage, engineering application analysis is conducted on the travelling part system of the operating train under actual working conditions, so as to achieve the double verification of theory and engineering. The research results not only have important theoretical significance, but also can provide necessary technical support for safe and efficient operation of high-speed train running gear system, which has potential engineering application value.

对走行部系统进行故障诊断与状态预测是保障高速列车安全、可靠、高效运行的有效手段之一。本项目以高速列车走行部系统为应用背景，考虑其隐藏行为、状态滞留、不确定性扰动和数据实时性等问题，开展实际工况下走行部特征提取、故障诊断与状态预测的理论、算法和应用研究。首先，对多源运营数据进行多级特征提取，分析隐藏故障特征；其次，研究不确定性扰动与状态滞留问题，提出扰动因子参数状态估计模型并分析自适应反馈效果，实现走行部系统的故障诊断；然后，考虑实时数据的快速变化，提出基于优先级调度策略的多层智能结构，增强走行部系统实时状态预测能力；最后，将主要理论成果分为两阶段验证，先利用中车长客走行部参数试验台进行模拟仿真，后通过实际工况下运营列车的走行部系统进行工程应用分析，达到理论与工程的双验证。研究成果不仅具有重要的理论意义，还可以为运营列车走行部系统安全、可靠、高效运行提供技术支撑，具有重大的工程应用价值。

项目以实际高速列车走行部系统为研究对象，对系统特征提取、故障诊断与状态预测问题进行了理论与应用研究。通过将深度学习、宽度学习、控制理论等方法的融合与改进，解决了实际系统中故障隐藏行为、状态滞留、不确定性扰等问题对特征提取、故障诊断、状态预测的实时计算带来的影响。具体研究内容分为三部分，（1）基于运营数据的特征提取方法：引入了多元统计、流形学习等技术，改进了特征提取的算法结构，提高了实际数据降维与特征分析的效率和精度。（2）面向故障分析的故障诊断方法：分析了实际故障的传递特性、围绕噪声、干扰等影响因素设计了有效的故障诊断框架，降低了传统算法的漏报率和误报率。（3）考虑实时性能的状态预测方法：结合了定量信息与定性知识，提出了不同模式的实时状态预测框架，增强了预测方法对实际系统的适用性。（4）高速列车走行部系统的实际应用研究：项目从高速列车走行部系统的实际问题出发，设计的理论方法均已得到了列车系统的有效性验证与实用性验证。整体实现了从数据采集、数据处理、故障诊断、状态预测的全过程管理，课题的研究对于保障高速列车安全稳定运行具有重要的意义。根据课题的总体研究计划、研究目标与任务需求，该项目已完成了预期设置的研究内容并取得了一定的成果包括：负责人以第一作者身份发表SCI论文18篇，会议论文3篇，授权专利1项。在理论与实际成果的支撑下，负责人获得吉林省拔尖创新人才（三层次）、2021年吉林省科技进步二等奖。所获得的理论与应用成果成功应用于多家列车相关企业获得应用证明5份。未来将继续持续稳步的推进相关工作。

内容获取失败，请点击重试