基于小波变换和神经网络的旋转机械故障诊断研究

发布时间：2018-08-31 13:47

【摘要】：旋转机械是航空、化工、电力等领域的关键设备，因此对其进行故障诊断研究具有重大的现实意义。随着振动检测和信号处理等相关技术的不断发展，以振动信号检测、处理与分析为基础的故障诊断技术已成为故障诊断领域一个重要的研究方向。同时基于神经网络的智能故障识别与诊断技术的研究，也为故障诊断技术的研究和应用开辟了一条崭新的途径。本文详细介绍了小波变换以及BP神经网络等相关内容。一方面，介绍了连续小波变换、离散小波变换、正交小波包变换等内容，并且分析了边沿效应的问题；另一方面，介绍了BP神经网络的基本原理，分析了标准BP神经网络的学习算法和存在的问题，研究了一种基于BP算法的小波神经网络，并通过仿真实例对BP神经网络与基于BP算法的小波神经网络性能进行了对比。同时，为了提取旋转机械故障特征，研究了基于连续小波变换极大模值法和基于最优正交小波包法提取故障特征值的理论及具体实现过程。最后，，通过多功能转子试验台模拟旋转机械的常见故障，运用连续小波变换的模极大值法及正交小波包变换提取转子系统常见故障的特征量，再将该特征量输入到BP神经网络中进行故障诊断，结果表明上述方法应用在转子系统故障诊断中能够取得较好的效果。
[Abstract]:Rotating machinery is the key equipment in the fields of aviation, chemical industry and electric power, so it is of great practical significance to study the fault diagnosis of rotating machinery. With the development of vibration detection and signal processing, fault diagnosis based on vibration signal detection, processing and analysis has become an important research direction in the field of fault diagnosis. At the same time, the research of intelligent fault identification and diagnosis technology based on neural network also opens a new way for the research and application of fault diagnosis technology. In this paper, wavelet transform and BP neural network are introduced in detail. On the one hand, the continuous wavelet transform, discrete wavelet transform and orthogonal wavelet packet transform are introduced, and the problem of edge effect is analyzed. On the other hand, the basic principle of BP neural network is introduced. This paper analyzes the learning algorithm and existing problems of standard BP neural network, studies a wavelet neural network based on BP algorithm, and compares the performance of BP neural network with wavelet neural network based on BP algorithm through simulation examples. At the same time, in order to extract the fault features of rotating machinery, the theory and implementation process of extracting fault eigenvalues based on the maximum modulus method of continuous wavelet transform and the optimal orthogonal wavelet packet method are studied. Finally, the common faults of rotating machinery are simulated on a multifunctional rotor test-bed, and the characteristic quantities of common faults of rotor system are extracted by modulus maximum method of continuous wavelet transform and orthogonal wavelet packet transform. Then the eigenvalue is input into the BP neural network for fault diagnosis. The results show that the above method can achieve good results in rotor system fault diagnosis.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】