A fault diagnosis method for ship shaft system based on empirical wavelet transform and particle filtering

Abstract

In order to solve the problems of excessive redundant information and low fault feature extraction rate of ship shafting vibration signal, a fault feature extraction method based on empirical wavelet transform and particle filter is proposed. Firstly, the original signal is processed by empirical wavelet transform to separate the redundant vibration components, and solve the defects such as end effect and modal aliasing inherent in empirical mode decomposition. Secondly, intrinsic mode functions are filtered and reconstructed based on kurtosis, correlation coefficient and energy ratio to highlight fault information and improve signal-to-noise ratio. Finally, particle filter is applied to the reconstructed fault feature components to eliminate the residual non-linear and non-Gaussian noise in EWT, and the fault type is analyzed by the envelope spectrum. Through the analysis of the experimental bearing signal, the feature extraction ability and noise suppression ability of the method are verified, which can be effectively applied to the fault diagnosis of the ship shaft system.