Research And Application Of Fault Feature Extraction For Rotating Machinery

With the rapid development of modern industrial and scientific technology, rotating machinery in the industrial fields is also showing great changes. Especially the main and auxiliary equipments in the electric power plant is developing toward large-scale, automation, high efficiency, electromechanical integration direction, safety factors is increasing. Therefore, ensuring these rotating machinery safety and economic operation is to become the focus in feature extraction technology. Autoregression(AR) Model, Wavelet analysis, Short-Time Fourier Transform(STFT), Wigner-Ville Distribution (WVD) and Hilbert-Huang Transform(HHT) are studied in the paper.In AR Model, determining the best order number is the key to obtain power spectrum, so the order rule is mainly discussed. The resolution of autocorrelation, Burg and improved covariance method are compared when sampling points are different. Fault signals are used to estimate power spectrum to validate AR model application in feature extraction. The results show that AR Model can get better resolution and variance performance and smooth spectral lines, which can effectively extract fault feature.For wavelet analysis, the main type of wavelet function and wavelet transform are introduced. The main research for wavelet analysis are singularity signal in rotating machinery, a variety of mixed signal and signal with noise application. Fault signals are proposed by wavelet analysis, which is explained in feature extraction application. Finally, the results show wavelet analysis can be very good application in fault signal feature extraction of rotating machine.For time-frequency analysis technology, STFT, WVD and HHT are mainly studied. WVD and STFT are compared with Fourier transform. The characteristics of WVD and STFT are studied by examples. For HHT, the endpoint effect is suppressed by using cycle extension and symmetry extension for two kinds of method. In this paper, three kinds of time-frequency analysis methods are compared and applied in rotating machinery vibration signal of feature extraction.The results verifies that the time-frequency analysis technology can get the signal frequency information changing with time.Finally, rotating machine vibration signal analysis system is developed by using C++Builder and Matlab combination of methods. The system can estimate signal power spectrum using autocorrelation, Burg and improved covariance methods, and obtain signal STFT, WVD and HHT. STFT, WVD and HHT are to be realized through C++Builder mobilizing Matlab engine repository.