【作者】 张芳芳；

【导师】 张世荣；

【作者基本信息】 武汉大学 ， 控制理论与控制工程， 2018， 硕士

【摘要】 机车变压器油泵是牵引变压器强迫油循环冷却系统的核心动力设备,变压器油泵故障会影响机车的稳定运行。对变压器油泵的运行状况进行实时监测和故障诊断,是保证机车牵引变压器正常运行的主要措施。因此本文将油泵远程故障监测作为研究课题。本文通过分析油泵的常见故障及故障机理,选择电机定转子位移作为故障监测的敏感信号。在国内外学者研究的基础上进行创新,设计出在电机定子两端安装感应线圈的位移监测方案。使用8个线圈分时复用形成6对差动输出分别监测油泵垂直径向、水平径向和轴向位移。结合油泵故障监测的实际需求,选择基于知识的故障诊断方法,以最小二乘支持向量机为故障分类算法。该算法需要样本少,诊断速度快,泛化能力强,且分类正确率高。在算法研究基础上,设计了变压器油泵远程故障监测系统。整个系统分为三个功能块:现场监测仪,云平台数据中心和客户端中心。现场监测仪获取线圈对输出,并设计了放大、数字滤波、均方根计算等信号处理环节;使用定位模块获取油泵位置信息。监测仪通过3G模块将油泵定转子的位移信息和位置信息发送到云平台数据中心。对云平台数据库库结构进行了设计,并采用模块化的思想设计了通信程序和基于最小二乘支持向量机的故障诊断程序。客户端包括APP客户端和WEB客户端,用于显示油泵的实时运行信息及位置信息,同时实现对油泵的管理操作。以实际油泵试验台为对象开展了现场试验,获取了油泵9种运行工况下故障和非故障状态数据。使用前7种工况下的数据进行最小二乘支持向量机分类模型的训练,使用全部9种工况的数据进行测试,测试结果证明最小二乘支持向量机能够准确辨识学习工况及非学习工况下的油泵故障状态,准确率可达100%。测试结果表明本文设计的故障监测系统能够完成油泵故障的远程监测,完成预定设计目标。更多还原

【Abstract】 Oil pump is the core power equipment of the forced oil circulating cooling system for the locomotive transformer.Transformer pump failures affect the stable operation of the locomotives.Real-time monitoring and fault diagnosis of the transformer oil pumps are the main methods to ensure the normal operation of the locomotive traction transformers.Therefore,the remote fault monitoring for the oil pumps is chosen as the research topic of this thesis.The common faults and fault mechanisms of oil pumps are analyzed,then the displacement between rotor and stator is chosen as the sensitive signal for fault monitoring.Based on the research results from the literatures,a displacement monitoring scheme is proposed,which employs induction coils installed on both ends of a motor stator as the sensors.In the proposed scheme,8 detection coils are used.Through time division multiplexing of the 8 coils,6 differential outputs are generated to detect the displacements in vertically radial direction,horizontally radial direction and axial direction,respectively.Considering the actual needs of oil pump fault monitoring,a knowledge-based fault diagnosis method is proposed,where least squares support vector machine is used as the fault classification algorithm due to its less requirement for samples,faster speed,excellent generation ability and high accuracy.Based on the algorithm research,a remote fault monitoring system for transformer oil pumps is designed.The fault monitoring system composes three parts:on-site monitoring device,cloud data center,client center implemented by APP and WEB.The outputs of the coil pairs are acquired by on-site monitoring device,then the signal processing,such as amplification,digital filtering and root mean square calculation,are conducted.Positioning module is used to get the location information of oil pumps.The displacement and location information of oil pumps are then sent to cloud data center through 3G module by on-site monitoring device.On cloud platform,the database structure is designed,and a communication program and a fault diagnosis program based on least squares support vector machine classification algorithm are designed with modularization idea.The client center includes an APP client and a WEB client.Client is designed to present the real-time operation information and position information of the oil pump,at the same time it realizes the management operation of the oil pumps.The field test was carried out with an actual oil pump.The operation data of the oil pump under normal and faulty conditions in 9 operating conditions are obtained first.Samples from the first 7 operating conditions are used to train the least east squares support vector machine classification model,and samples from all 9 operating conditions are used to test the classification model.The testing results prove that the least squares support vector machine can accurately identify faults of the oil pumps under both trained operating condition and non-trained operating condition with 100%accuracy rate.The test results show that the remote fault monitoring of oil pumps is completed by the fault monitoring system designed in this thesis.The scheduled goals are thus accomplished.更多还原