Multivariate Wavelet Denoising Method Based on Synchrosqueezing for Rolling Element Bearing Fault Diagnosis

Abstract

The raw vibration signal of rolling element bearing carrying a great deal of information representing the mechanical equipment's health conditions, but the impulsive signal of interest (SOI) is usually hidden in heavy noise, and denoising technique is great significant to the fault diagnosis. High resolution time-frequency algorithms, such as the wavelet based synchrosqueezing transform, have a wide range of applications in removing noise, and multichannel sensor technology has highlighted the requirement for multivariate denoising. In this paper, a multivariate wavelet denoising method based on synchrosqueezing is proposed. The mutual modulated oscillations of multivariate data is identified by partitioning the time-frequency domain, and a modified universal threshold is employed to remove the noise components while to retain SOI. Numerical simulations and experimental investigations are included to illustrate the feasibility and performance of utilizing the novel method to process faulty signal of rolling element bearing.