基于案例域的列车关键设备服役状态辨识与预测方法研究

发布时间：2018-08-13 18:17

【摘要】：安全是轨道交通永恒的主题,尤其是在轨道交通事业集中建设和高速发展的全盛时期,安全问题更是全社会关注的焦点。轨道交通列车的正常服役是保障轨道交通系统安全高效运营的必要条件,而轨道交通列车能否正常运行直接取决于其中的运行安全关键设备的服役状态。但是,我国现有的轨道交通列车及其运行安全关键装备在线服役状态的辨识、预测、诊断和控制技术和手段远远不能满足轨道交通系统动态化系统化的主动安全保障需求。针对这一重大问题,急需研究并提出系统化的列车运行安全关键设备服役状态辨识和预测的方法。鉴于此,本文在进行了如下研究工作： 1.在了解分析列车关键设备状态监测、安全域估计理论和方法等相关的国内外研究现状的基础上,参考借鉴相关领域的已有成果,系统地提出了基于安全域理论的列车运行安全关键设备服役状态辨识方法。阐述了安全域的基本概念和内涵,说明了基于安全域估计进行状态辨识的基本原理；提出了基于安全域估计理论的状态辨识方法框架,给出了方法实施的通用步骤,并针对方法实现中的边界估计这一关键技术问题,根据研究对象是否便于建立数据模型提出了两条并行的技术路线：基于模型的边界估计技术和数据驱动的边界估计技术；对于本文使用的数据驱动的边界估计技术实现,提出了采用支持向量机的方法,并根据状态辨识需求给出了二分类的和多分类的支持向量机算法。 2.在基于安全域估计理论的状态辨识方法框架的基础上,提出了一种面向实时状态特征的安全域状态辨识方法。该方法在状态特征提取方面,首先采用了较新颖的局部均值分解方法将数字信号分解为多个分量,然后计算了信号分量的直接时域特征以及基于能量和熵的两类实时状态特征指标；以列车滚动轴承作为实例,分别利用不同工况环境下的数据对算法的辨识精度、鲁棒性和实时性进行了全面测试,实验结果表明基于实时状态特征的状态辨识方法的计算效率很高,而辨识精度和抗干扰性能方面表现一般。 3.从基于数据的统计分布特性提取状态特征方面考虑,提出了面向统计状态特征的安全域状态辨识方法。首先清晰地给出了基于统计状态特征提取和支持向量机的状态辨识方法,详细叙述了其实现步骤；然后,针对方法步骤中的统计状态特征提取问题,细致地阐述了基于主成分分析的统计状态特征提取方法；仍以列车滚动轴承作为实例,通过不同工况环境下的多组实验仿真,测试了方法的辨识精度、鲁棒性和实时性,实验结果表明基于统计状态特征的状态辨识方法具有极高的辨识精度和优越的鲁棒性,但该方法计算负担大执行效率低,实时性方面表现一般；最后简要介绍了作者参与的国家863重点项目中关于安全域状态辨识方法的现场应用系统的设计工作。 4.基于列车运行安全关键设备服役状态辨识方法的研究结果,进一步对基于状态监测的剩余寿命预测问题进行了探讨。叙述了几类常用的基于状态监测的剩余寿命预测方法,给出了各类方法的基本原理、优缺点和适应环境；详细讨论了能够同时融合可靠性信息和状态监测信息的比例风险模型,对基于比例风险模型的剩余寿命预测方法进行了详细阐述；基于滚动轴承的全寿命振动数据,进行了试验仿真,结果表明,相对于仅依靠状态监测信息的剩余寿命预测方法来说,本文提出的基于统计状态特征的比例风险模型能够精确地预测设备的剩余寿命。
[Abstract]:Safety is the eternal theme of rail transit, especially in the heyday of centralized construction and rapid development of rail transit. Safety is the focus of attention of the whole society. The normal service of rail transit trains is the necessary condition to ensure the safe and efficient operation of rail transit system, and the normal operation of rail transit trains is directly decided by the normal operation of rail transit trains. However, the existing on-line identification, prediction, diagnosis and control technologies and means of rail transit trains and their operational safety critical equipment in China are far from meeting the requirements of active safety assurance for the dynamic and systematic rail transit system. It is necessary to study and put forward a systematic method for identifying and predicting the service state of key equipment for train operation safety.
1. On the basis of understanding and analyzing the status monitoring of train key equipments, the theory and method of safety domain estimation, and referring to the achievements in related fields, a method of identifying the service status of train key equipments based on safety domain theory is proposed systematically. The basic principle of state identification based on security domain estimation is explained, the framework of state identification method based on security domain estimation theory is proposed, and the general steps of implementation of the method are given. There are two parallel technical routes: model-based boundary estimation and data-driven boundary estimation. For the implementation of data-driven boundary estimation, a support vector machine (SVM) method is proposed, and two-class and multi-class SVM algorithms are given according to the requirements of state identification.
2. Based on the framework of the state identification method based on the theory of security domain estimation, a real-time state feature-oriented security domain state identification method is proposed. Direct time domain feature and two kinds of real-time state feature indexes based on energy and entropy are used to test the identification accuracy, robustness and real-time performance of the algorithm using the data of different working conditions respectively. The experimental results show that the state identification method based on real-time state feature is effective. The rate is very high, but the identification accuracy and anti-jamming performance are general.
3. Considering the extraction of state features based on statistical distribution characteristics of data, a new method of state identification in security region for statistical state features is proposed. The problem of state feature extraction is discussed in detail, and the statistical state feature extraction method based on principal component analysis (PCA) is elaborated. Taking train rolling bearing as an example, the identification accuracy, robustness and real-time performance of the method are tested through a series of experimental simulations under different working conditions. The experimental results show that the state identification method based on statistical state feature is effective. The method has high identification accuracy and superior robustness, but its computational burden is heavy and its execution efficiency is low, and its real-time performance is general.
4. Based on the research results of the identification method of the service state of the key equipment for train operation safety, the problem of residual life prediction based on condition monitoring is further discussed. Proportional risk model which can fuse reliability information and condition monitoring information at the same time is proposed, and residual life prediction method based on proportional risk model is expounded in detail; full-life vibration data of rolling bearing is tested and simulated, and the results show that the residual life prediction method based on condition monitoring information is better than that based on condition monitoring information only. For example, the proposed proportional hazard model based on statistical state features can accurately predict the residual life of equipment.
【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：U298.1

【参考文献】