Study And Application On The Second Generation Wavelet Transform In Fault Diagnsis Of Rotating Machine

Wavelet analysis all has fine localization characteristic property in the time domain and frequency region, therefore it is better than Fourier analysis.At present, wavelet transform is applied widely in rotating machine fault diagnosis. The 2nd generation wavelet transform inherits the advantages of tradition wavelet transform, and its computation is carried out on the time domain, the window functions are not generated by the dilation or compression of a mother wavelet no longer. In addition, the 2nd generation wavelet transform has these merits such as simple algorithm structure, less computation, less memory. In this paper, the 2nd generation wavelet is used to carry out the fault diagnosis of the rotor and ballbearing.(1) Denoising the rotor fault signals based on the 2nd generation wavelet transform has been studied. When the traditional wavelet denoising methods are used to denosise and process the rotor signals, the rotating rate and the signal sampling frequency have very great effect on the choosing of the number of the wavelet decomposition layers.So, it is difficult to achieve the process of denoising automatically. In this paper, aiming at the above problem, the 2nd generation wavelet are full used, and a wavelet automatical denoising method based on scale transform is advanced, and the original signals are sampling repeatedly according to the certain method. Meanwhile, the wavelet soft threshold denosising method are used together to achieve the denoising. This method can eliminate the negative effects of the rotating speed and the sampling frequency, and the number of the wavelet decomposition layers can be automatically choosed. In the whole process of denoising, all steps can be accomplished automatically. Finally, the fault datas gathered from the rotor test-bed and the aero-engine rotor experimental rig on imbalance, misalignment, oil-film whorl and rubbing conditions are used to do the denoising analysis, and the approving results are obtained.(2) The feature extraction of rotor fault signals based on the 2nd generation wavelet transform is studied. Aiming at the problem that the wavelet frequency band energy characteristics are affected by the changes of the rotating speed and the sampling frequency, in this paper, according to the scale transform theories, the original signals are sampling repeatedly, and the 2nd generation wavelet transform is used for the signals sampled. Afterward, the obtained signals are decomposed to fixed layer so as to obtain the frequency band characteristics of the original signals. This method can eliminate the negative effects of the rotating speed and the sampling frequency, and the frequency band energy characteristics extracted have the unitive physical significance. Finally, the the rotor test-bed and the aero-engine rotor experimental rig are used to obtain 4 kinds fault data samples on imbalance, misalignment, oil-film whorl and rubbing conditions of rotor system, and the feature extraction has been done. Afterward, the structure self-adaptation neural network is constructed for the diagnosis and analysis of the fault data samples, and the very high distinguishing rate is obtained.(3) the ball bearing fault signal analysis based on the 2nd generation wavelet transform is studied. Firstly, the ball bearing fault characteristics are analyzed, and the 2nd generation wavelet transform is used to decompose the vibration acceleration signals of ball bearing fualts to different scales, and the resonance frequency band is extracted. Afterward, the Hirbert transform is used to demodulate the signals, and the frequency analysis of the signals demodulated has been done to obtain the wavelet spectra from which the fault characteristic informations of ball bearings are obtained. Finally, the analysis and validation have been done by using the ball bearing fault data which have been gathered form the ball bearing fault test-bed belong to the electric engineering laboratory of Case Western Reserve University of America. The results show that the method is very effective.