Rolling bearing fault diagnosis using improved LCD-TEO and softmax classifier

A novel rolling bearing fault diagnosis method based on improved local characteristic-scale decomposition (LCD), Teager energy operator (TEO) and softmax classifier is proposed in this paper. First, vibration signals are decomposed into several intrinsic scale components (ISCs) by using improved LCD; second, TEO and fast Fourier transform (FFT) are respectively used to extract instantaneous amplitude (IA) and frequency spectra of ISC1s, and then FFT is again employed to obtain spectra of IA; third, energy ratio of the resonant frequency band against the total, frequency entropy (FE) in the spectra of ISC1s and several amplitude ratios in the frequency spectra of demodulated ISC1s are extracted as fault feature vectors, and principal components analysis (PCA) is applied for dimensionality reduction; finally, these feature vectors are taken as inputs to train and test softmax classifier. As a new non-stationary signal analysis tool, LCD can decompose adaptively a signal into series of ISCs in different scales and give good results in situations where other methods failed. However, there are two main issues in this method, end effect and mode mixing, possibly leading to unexpected results. In this paper, a slope-based method and noise assisted analysis are applied to restrain the problems respectively. Experimental results show the proposed method performs effectively for bearing fault diagnosis.