高压异步电机转子故障智能诊断方法研究

发布时间：2018-03-15 20:34

本文选题：高压异步电机　切入点：电机转子故障　出处：《长沙理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着现代科学技术的进步和电力系统规模日趋庞大,电机在现代工业生产中充当着越来越重要的角色。电机故障不仅会损坏电机本身,影响整个系统的正常运转,严重的情况下甚至危及人身安全,造成巨大的经济损失和重大的社会影响,因此对电机故障的诊断具有重要意义和工程实用价值。研究感应电机的监测和故障诊断技术,对预防感应电机故障的发生,及时发现并消除故障,保证感应电机可靠运行,提高生产效率都具有十分重要的意义。论文在阐叙国内外电机故障诊断研究现状的基础上,对高压异步电机转子的基本结构和转子故障进行了简单的分析。在此基础上提出了两种基于不同故障诊断机理的电机转子故障诊断方法：“质朴型-贝叶斯网络拓扑模型”和“小波神经网络的故障诊断网络模型”。由于电机系统故障信息中存在许多不确定性,电机转子出现故障时,在传统方法的基础上,通过综合样本信息和先验信息,建立基于转子系统故障类型和对应的故障征兆的贝叶斯网络模型。贝叶斯网络作为目前推理领域和不确定性知识表达的的有效模型之一,能高效的推理和表达不确定性知识和概率推理。同时,电机的振动故障信往往包含大量的短时突发、时变的成分,而傅里叶变换采用的方法是将信号从频域和时域整体角度出发,缺乏时频局部性,不能准确的对这些非平稳随机信号分析,故达不到故障信号特征提取的要求。小波神经网络就是在研究电机振动信号的基础上建立的故障诊断模型。采用小波时频分析技术对电机故障信号消噪滤波并提取故障特征,然后用BP神经网络进行故障识别,最终达到故障诊断的目的。最后对异步电机典型故障转子不对中、转子质量不平衡和轴承摩擦等故障用上面提到的两种不同的诊断方法建立模型进行故障诊断。结果证明了论文所设计的两种方法能够对高压异步电机的故障有效地进行诊断,提高了电机故障诊断的准确性。
[Abstract]:With the progress of modern science and technology and the increasingly large scale of power system, motor plays an increasingly important role in modern industrial production. Motor failure will not only damage the motor itself, but also affect the normal operation of the whole system. Under serious circumstances, even endangering personal safety, causing huge economic loss and great social impact, it is of great significance and practical engineering value to diagnose motor fault. The monitoring and fault diagnosis technology of induction motor is studied. It is of great significance to prevent the occurrence of the fault of induction motor, to find and eliminate the fault in time, to ensure the reliable operation of the induction motor and to improve the production efficiency. The basic structure of rotor and rotor fault of HV asynchronous motor are simply analyzed. On the basis of this, two methods of rotor fault diagnosis based on different fault diagnosis mechanisms are proposed: "simple Bayesian network" The fault diagnosis network model of wavelet neural network. Because there are many uncertainties in the fault information of motor system, When the motor rotor is in trouble, on the basis of the traditional method, by synthesizing the sample information and the prior information, A Bayesian network model based on rotor system fault type and corresponding fault symptoms is established. Bayesian network is one of the effective models of reasoning field and uncertain knowledge representation. It can efficiently infer and express uncertain knowledge and probabilistic reasoning. At the same time, the vibration fault letter of motor often contains a large number of short-time burst and time-varying components. The method of Fourier transform is to analyze these non-stationary random signals from the whole angle of frequency domain and time domain without time-frequency localization. Wavelet neural network is a fault diagnosis model established on the basis of studying motor vibration signal. Wavelet time-frequency analysis technology is used to filter and extract fault feature of motor fault signal. Then BP neural network is used to identify the fault, and finally the purpose of fault diagnosis is achieved. Finally, the typical fault rotor of asynchronous motor is misaligned. The fault diagnosis of rotor mass imbalance and bearing friction is based on two different diagnosis methods mentioned above. The results show that the two methods designed in this paper can effectively diagnose the faults of high voltage asynchronous motor. The accuracy of motor fault diagnosis is improved.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM343

【参考文献】