【作者】 余博；

【导师】 贾利民；

【作者基本信息】 北京交通大学 ， 交通安全工程， 2014， 博士

【摘要】 我国正处在大规模城轨线路建设和开通的高峰期,近年来城轨列车晚点乃至事故的教训说明轨道交通安全保障技术需要由被动安全向主动安全转化。基于实时状态监测的运维保障是轨道交通主动安全保障的基础,其技术的开拓创新是提高轨道交通系统的安全运营能力水平和加强故障处置能力的发展方向。本文研究的面向运维的城轨列车在途监测体系结构中几项关键技术,涵盖了状态监测网络构建,基于列车信号和传感器数据的列车运维关键部件故障诊断方法。通过实际应用,实现了列车运营管理的网络化,同时将维保业务的工作模式从基于计划与事故的维修,转变为基于车辆状态的故障早期即开展的维护工作。本文旨在为我国城轨列车安全工程理论及建设提供理论依据,主要研究工作如下：(1)系统地研究了面向运维的城轨列车在途监测系统体系结构。分析了面向运维的城轨列车在途监测系统的构建目标、需求分析和功能结构。提出并使用当量故障数累计频率分析法辨识了广州地铁二、八线列车运维关键系统,研究确定了列车在途监测关键对象。通过建立适用体系结构,对本文研究的关键技术进行了说明。(2)优化了城轨列车状态监测网络传输。建立了树状结构的列车状态监测网,提出了列车视频传输优化方法,为列车在途故障诊断提供了安全有序的数据基础。通过建立网络容量与延时模型,对基本拓扑进行的比较显示了树状网络的优越性。进而,为满足城轨列车在途监测环境下多传感器大数据的安全传输,设计并实现了树状结构的城轨列车状态监测网物理结构、路由和传输方法。在基于以太网的树状网络建模与仿真表明,在使用网络通讯协议(TCP/IP)协议的树状结构下,相对于指数分布(EXP)流量模型和使用传输流控制协议(TCP),使用恒定流码流量(CBR)模型与用户数据报协议(UDP)能够获得更好的性能。(3)针对面向列车安全运维的三大关键系统：制动、客室车门与二系悬挂进行了故障诊断方法研究。首先分别对三个系统建立了工作模型,通过分析,我们选用不同的方法进行故障诊断。对于制动系统,使用基于模型设计的残差发生器和估计器的方法进行故障诊断具有良好的准确性以及防误报的能力。对于二系悬挂,我们选用数据驱动的方法进行故障诊断,包括使用小波包能量矩作为故障特征,使用最小二乘支持向量机(LSSVM)作为故障分类器,使用改进的粒子群算法进行联合特征选择。不同列车运行速度下的仿真结果表明,空气弹簧故障的识别率最高,横向阻尼器次之。对于客室车门系统,先使用基于主元分析(PCA)的参数估计方法进行故障检测,再使用粗糙集故障诊断方法进行故障分离,在长时开关门模式的实验中获得了较高的诊断准确率。(4)以广州地铁2、8号线的实际应用为背景,建立了面向运维的城轨列车在途监测系统示范工程。通过对应用系统进行充分的调研,研究了故障等级划分的方法,完成了应用系统结构和功能设计。在J2EE平台的实现,验证了本文研究成果的实用性。更多还原

【Abstract】 ABSTRACT:Our country is in peak period of large-scale urban rail line construction, over the years, urban rail train delays and accidents show that rail transit safety technology needs to be transformed from passive to active. Operation and maintenance based on realtime conditions are the basis for rail transit active safety, whose pioneering and innovative is the develop direction of urban rail safety engineering.The key technices showing in this paper cover realtime monitoring network construction and fault diagnosis of urban rail train key systems. Through practical application, we realize networked train operation and maintenance system to transform the maintenance business from planed and accidents based model to conditions based pre-maintenance model.The main purpose of this paper is to provide theoretical basis for safety engineering in Chinese urban rail train, details of the main work are as follows:Firstly, system structure of operation and maintenance oriented urban rail trains realtime monitoring has been constructed. We analyse the construction goals, demands and functions for system construction. By using the equivalent failure number cumulative frequency analysis with data of Guangzhou metro lines, we identify the key monitoring objects for urban rail train operation and maintenance. Through the establishment of a suitable system structure, key technologies are describedSecondly, studies have been made to set up the tree structure of urban rail train conditions monitoring. We put forward an optimization method of train video transmission. In order to meet the demand of large data transmission of city rail train with multi-sensors, urban rail train conditions monitoring network of tree structure has been designed and implemented by physical structure and data transmission methods design. Through the establishment of the network capacity and the delay model, the basic topological comparison shows the superiority of the tree network. The modeling and simulation of tree network based on Ethernet shows, compared to the EXP flow model with TCP, the transmission performance of the CBR flow model with UDP is better.Thirdly, for the three key systems we have identified in chapter2, train brake, passenger doors and secondary suspension, we studied fault diagnose methods. At the beginning, we establish mathematic models for the three systems, through analysis we choose different ways for fault diagnosis. For braking system, the use of residual generator and residual estimator has a feature of accurate rate and anti false positives. For secondary suspension system, we use the data driven fault diagnosis method, including the use of wavelet packet energy moment as the fault feature, the least square support vector machine as fault classifier, improved particle swarm algorithm for combined feature selection. The simulation under various train speeds results show that, air spring fault recognition rate is higher than transverse damper. For passenger door system, we use the parameter estimation method with PCA for fault detection, and use rough set for fault isolation. Door model experiment has achieved a high accuracy rate of diagnosis in long opening and closing profile.By taking the line2and line8of Guangzhou Metro as the application background, we build physical structure of this system, establish the system of operational oriented urban rail train realtime conditions monitoring system. After doing research with investigation, methods are found for fault classification and structure and function design of application system. The implementation of application system on J2EE platform shows the practicability.更多还原