Research On Empirical Wavelet Transform And Dispersion Entropy Based Fault Diagnosis Of Rolling Bearing

Rolling bearings are one of the most widely used parts of rotating machinery and their work conditions have direct impact on rotating machinery.Affected by the factors such as working environment and work intensity,the bearing life has a strong uncertainty and the bearing failure may cause economic property loss or even casualties.Therefore,it is important to perform state detection and fault diagnosis of rolling bearings.Due to the complicated working environment,the rolling bearing vibration signal often has non-stationary and nonlinear characteristics.It is difficult for conventional signal processing methods to obtain effective and stable vibration information from the bearing vibration signal.Empirical Wavelet Transform(EWT)is a recently proposed signal analysis method based on Fourier spectral segmentation and wavelet transform.It can effectively decompose a non-stationary and nonlinear signal into several momocomponent signals.At the same time,in order to extract the nonlinear fault characteristics of the rolling bearing,the recently proposed dynamic index for measuring irregularity of vibration signal named dispersion entropy(DE)is applied to the fault feature characterization of rolling bearing.Compared with sample entropy and permutation entropy,DE has certain advantages in computational efficiency and information representation.The main research contents and results of the thesis are as follows:1.The EWT theory and its performance are studied.By analyzing synthetic signals,EWT is compared with empirical mode decomposition and local characteristic-scale decomposition methods.The results show that EWT is more effective in suppressing mode mixing and physical meaning.Since it is difficult for the original Fourier spectrum segmentation method to deal with the spectrum of rolling bearing signals,the spectral segmentation method is improved and an improved EWT method is proposed.Simulation signal analysis results verify the effectiveness of the improved EWT.2.To accurately extract the fault characteristics of rolling bearings,DE and related complexity theories are introduced.Comparing DE with sample entropy and permutation entropy by analyzing simulation signals and the results show that DE is faster than sample entropy in computation and is more reasonable than permutation entropy.Meanwhile,since the feature information is limited and the fault diagnosis stability is poor in single scale,the multiscale dispersion entropy is proposed based on traditional multiscale method.For the single-channel signal contains less information,the multichannel signal analysis method called refined multivariate multiscale dispersion entropy is proposed and applied to the fault diagnosis of rolling bearings.The proposed method is compared with multivariate multiscale entropy and multivariate multiscale dispersion entropy and the results show much more stability and a higher fault recognition rate of the proposed method.3.For the original signal,which is not conducive to the feature extraction component,the improved EWT is used to reconstruct the original signal,and the average energy index is used to select the component containing the dominant vibration characteristic from the EWT decomposition component.The reconstructed signal is analyzed by using the refinedcomposite multivariate multi-scale dispersion entropy.Comparing the proposed method with the existing methods,the results show that the refined multivariate multi-scale dispersion entropy feature extraction effect and fault recognition rate after signal reconstruction are significantly improved.In this paper the theory of EWT and DE were studied.On the basis of the improved EWT and multiscale dispersion entropy,,several new fault diagnosis methods for rolling bearing were proposed.The research provides new ideas and technical means for fault diagnosis of rolling bearings.