Research On Gearbox Fault Diagnosis Based On Adaptive Processing Of Vibration Signal

As an important component of modern machinery and equipment, a research of fault diagnosis on gearbox has important practical significance. At present, there are many gearbox fault diagnosis methods, but due to the bad work environment and the non-stationary characteristics of the gear vibration signal, the single method is sometimes difficult to extract effective gear fault characteristic information. Therefore, this paper combines various methods to recognize the fault of gearboxes. The main research contents are as in the following:1. Research on gearbox fault diagnosis based on local mean decompositionThe performance of local mean decomposition(LMD) and empirical mode decomposition(EMD) algorithm in time-frequency analysis is studied. A new time-frequency analysis method, called self-adaptive wigner-ville distribution based on local mean decomposition(LMD-AWVD), is proposed to effectively analyze non-stationary vibration signal. The vibration signal was firstly decomposed into several product functions(PFs) by LMD, then calculating the correlation coefficients of PFs and according to this choose the principal PF components which contain most of fault information. Finally, WVD of the principal PF components are calculated and the complete time-frequency distribution can be obtained by adding them together. The analysis results of experiment signals indicate that the LMD-AWVD method can be used in the gear fault diagnosis effectively.2. Research on gearbox fault diagnosis based on intrinsic time-scale decompositionImproved algorithm of intrinsic time-scale decomposition(ITD) is studied. In view of some defects of ITD method, an ensemble intrinsic time-scale decomposition(EITD) method is proposed. The de-noise performance of mathematical morphology filtering for the vibration signal is studied. The good qualities of singular value decomposition(SVD) in feature extraction are analyzed. Combining EITD method and mathematical morphology and SVD, a morphological singular value entropy is put forward. Aiming at fault feature extraction and classification problem of the non-stationary gear vibration signal, a fault diagnosis method based on morphological singular value entropy and support vector machine is put forward.3. Research on wind turbine gearbox fault diagnosis based on intrinsic time-scale decompositionAs it is difficult to extract effective fault feature of wind turbine gearbox, a new fault diagnosis method of wind turbine gearbox is proposed. This method firstly uses the cubic spline interpolation to fitting baseline control points and decomposes the vibration signal into several proper rotation components(PRCs), then wavelet packet transform is applied to decompose PR component of the biggest correlation coefficient and the energy distribution of wavelet packet coefficients are computed. Wavelet packet coefficients of relatively bigger energy are selected to reconstruct PR component. Finally, the correlation dimensions of the reconstructed PR component are calculated and the fault diagnosis results of vibration signals are obtained. Practical examples show that the diagnosis method which is put forward in this paper can extract gear fault features and identify gear fault patterns effectively.