【作者】 吕鹏；

【导师】 周强；

【作者基本信息】 陕西科技大学 ， 模式识别与智能系统， 2012， 硕士

【摘要】 电动机是所有大型机械设备的动力源装置。为了保证工农业生产的正常进行，需要对电动机的健康状况进行密切监视。当电动机发生故障时，将会表现出各种异常现象，其中发出的噪声信号的差异是比较明显的特征之一。本文针对电动机在不同状态下发出的噪声的不同，以不同的噪声信号的特征值作为电动机故障诊断的依据。分析和对比了噪声的分类方式、性质以及各种特征提取方法，得到自然界中的任何噪声都是有色噪声以及小波变换，特别是小波包变换，是提取有色噪声特征的有效工具的结论；对伪白噪声和粉红噪声进行了深入的分析和研究，提出了伪白噪声的白化模型和粉红噪声的ARMA模型生成法，并且利用本文提到的有色噪声的特征提取方法对构建的两个模型进行仿真，验证了模型的效果；利用在有色噪声的研究中取得的理论成果进行电动机的故障诊断。本文设计了一个将信号的检测方法，信号的分析与特征提取方式以及神经网络故障诊断技术相结合的电动机故障诊断系统。其基本过程是：由信号采集模块对电动机产生的有色噪声信号进行检测；利用小波包分解技术对采集到的信号进行分析和特征提取；再利用RBF神经网络技术对提取的信号的特征进行识别和判断，对电动机的运行状态进行决策。以型号为Y90S—4的中型电动机为实际对象，利用该故障诊断系统对模拟设置的几种常见的电动机故障信号进行实时检测和诊断，并进行了相关的仿真实验。由Matlab仿真实验的结果可知，该系统在电动机故障诊断过程中取得极好的效果。同其他故障诊断系统相比，它具有许多优势，例如诊断速度快，精确度高，结构简单，成本低廉等。因此，它具有更好的发展前景和广阔的市场空间，可以被大量地应用于工农业生产实践中。更多还原

【Abstract】 The motor is the power source device of all large mechanical equipment. Inorder to ensure the normal production of industry and agriculture, the healthstatus of the motor need to be closely monitored to avoid the occurrence of thefault. When the motor is at fault, it will show all kinds of anomalies. Duringthese anomalies, noise signal is one of the more obvious characteristics.In this paper, in view of the difference of the noise emitted by the motor indifferent states, some characteristic values of different noise signal can be usedas the standard of the motor fault diagnosis. Analyzing and comparing noiseclassification, nature and a variety of feature extraction methods, the conclusioncan be drawn that any noises which exist in nature are colored noise and wavelettransform, especially wavelet packet transform, is an effective tool to extractcolored noise characteristics; doing some further analysis and study aboutpseudo white noise and pink noise, putting forward pseudo white noisewhitening model and pink noise ARMA model generation method, and using thecolored noise feature extraction method which is mentioned in this paper tosimulate these two models, so as to test their effect; according to these theoreticalresults obtained in the study process of colored noise, doing the research of themotor fault diagnosis.This paper designs a motor fault diagnosis system which combines signaldetection methods, signal analysis and feature extraction methods as well asneural network fault diagnosis technology. Its basic process is: using the signalacquisition module to detect the colored noise signal generated by the motor; byusing wavelet packet decomposition technology to analyze and extract acquiredsignals’ features; then making use of RBF neural network technology to identifyand judge the character of the extracted signal, and doing some decision-makingabout the motor running state.By using a medium-sized motor whose model is Y90S―4as the actual object, this fault diagnosis system is applied to detect and diagnosis severalcommon motor fault signals of the simulation setting in real time, and somerelated simulation experiments can be done. According to the simulation resultsof Matlab, it is shown that this system can achieve excellent effect. Comparingwith other fault diagnosis system, it has many advantages, such as fasterdiagnostic speed, higher accuracy, simpler structure, lower cost and so on.Therefore, it has better development prospect and broader market space and canbe greatly used into the practice of industrial and agricultural production.更多还原