Research On Fault Diagnosis Method For Rotating Machine Based On LMD And HSMM

With the continuous improvement of science and technology, the work intensity of the rotating machine is increasing. It has a very important and realistic significance for rotating machine to do condition monitoring and fault diagnosis. In this way it can not only guarantee the equipment system to operate safely, reliably and efficiently, but also can avoid the huge economic loss and the occurrence of major accidents.For traditional time-frequency analysis methods (such as short-time Fourier transform, wavelet and wavelet packet transform) have the problems of time resolution and frequency resolution are mutual restraint, and signal decomposition is lack of adaptability, this paper introduced the latest time-frequency analysis method of Local Mean Decomposition (LMD). By the same way, due to the traditional pattern recognition methods (such as artificial neural network) limited to static pattern recognition problems, this paper introduced the rapid developing dynamic pattern recognition method-Hidden Semi-Markov Model (HSMM).By using LMD method, a complex multi-component signal can be self-adaptively decomposed into a series of Product Functions (PF), each of which has a clear physical meaning. Each PF component is obtained by multiplying a pure frequency modulated signal and an envelope signal. By combining the instantaneous frequency and instantaneous amplitude of all PF components, we can get the complete time-frequency distribution of original signal. In addition, HSMM can make the model and achieve classification for the dynamic information in a time span statistically. It is particularly suitable for the signal with large amount of information, non-stationary, poor repeated characteristics. What’s more, with few training samples and high training speed, the diagnostic accuracy of HSMM is high, which means that HSMM has a good ability of pattern classification. Therefore, this paper used the method of combining LMD and HSMM to research the condition monitoring and fault diagnosis of rotating machine.First of all, this paper discussed the general situation of the development of fault diagnosis technologies for rotating machine, and introduced the related concepts, basic theory and algorithm of LMD method. Compared with the EMD method, the LMD method was proved to have the advantages of dealing with non-stationary signals through theoretical comparison and analysis of simulation signal. On this basis, this paper put forward a kind of feature extraction method based on the combination of wavelet packet denoising and LMD decomposition. The process can be divided into four steps:first, used wavelet packet to reduce the effects of noise, followed by LMD decomposition, and then analyzed the correlation of the decomposed PF components and the original signal, at last, selected the effective PF components to extract feature parameters in time domain and frequency domain. The effectiveness of the proposed method is verified by the simulation analysis and the actual signal data processing.Then, the condition monitoring and fault diagnosis method of mechanical equipment based on HSMM was studied in this paper, and some problems in the basic algorithm were improved. A fault diagnosis method of rotating machine based on LMD and HSMM was proposed, and it was applied to the fault diagnosis of rolling bearing. The experimental results showed that the training speed of HSMM was fast, the recognition accuracy was high, and using the method of combination of LMD and HSMM could effectively identify the running state of rolling bearings and could ensure the timeliness and accuracy of fault diagnosis.Finally, this article further used the method of combination of LMD and HSMM to identify the condition of mechanical seal end face film thickness, and achieved a more ideal recognition result, which could verify the method was effective and applicable to be used in the condition monitoring and fault diagnosis of rotating machine. In order to verify the HSMM had the advantages to do the condition monitoring and fault diagnosis of rotating machine, BP neural network was used in the condition recognition of mechanical seal end face film thickness with the same signal characteristics. By comparing the recognition results of HSMM and BP neural network, it was concluded that, the training speed of HSMM was faster than BP neural network, and the recognition accuracy was higher, HSMM is more applicable to be used in the condition monitoring and fault diagnosis of rotating machine.