Fault feature extraction method for marine shafting based on empirical wavelet transform-spectral kurtosis

Abstract

Objectives In order to solve the problems of redundant information and low fault feature extraction rate in the vibration signal of marine shafting, a fault feature extraction method based on empirical wavelet transform-spectral kurtosis (EWT-SK) is proposed. Methods First, the original signal is processed by empirical wavelet transform (EWT) to eliminate the excessive vibration components. This method solves the inherent defects of empirical mode decomposition (EMD) such as the endpoint effect and mode aliasing. Second, the modal function is reconstructed based on kurtosis and correlation coefficient, highlighting fault information and improving the signal-to-noise ratio. Finally, the optimal bandpass filter parameters are obtained by spectral kurtosis and used to design filters, then the filtered signal envelope is demodulated to realize fault diagnosis. Results According to the results of case analysis and verification, in the aspect of signal decomposition, EWT has higher stability and efficiency in feature extraction, enabling it to ensure the integrity of shafting fault information. In the aspect of the denoising effect, after using the EWT-SK method, the kurtosis value of the fault signal is 4.761 6, the number of the correlation coefficient is 0.708 8 and the signal-to-noise ratio is 3.762 4, which is better than EMD and variational mode decomposition (VMD). Conclusions The EWT-SK method has good feature extraction ability and noise suppression ability, making it suitable for the fault diagnosis of marine shafting.