Extracting cyclo-stationarity of repetitive transients from envelope spectrum based on prior-unknown blind deconvolution technique

Abstract

Repetitive transients are typical symptoms of rotating faults and their extraction can be used to indicate early rotating faults. The main issue is to extract the cyclo-stationarity of repetitive transients which is insensitive to impulsive noises. The objective of traditional blind deconvolution methods is to recover impulsive signals from time domain. Considering the cyclo-stationarity of repetitive transients, a novel blind deconvolution (BD) technique, coined as prior-unknown blind deconvolution technique, is proposed to extract the cyclo-stationarity of repetitive transients. The proposed technique finds an inverse filter by minimizing an optimal criterion that is defined based on envelope spectrum (ES). Additionally, the proposed method does not require the prior knowledge of initialization for BD. Based on the proposed method and the generalized Lp/Lq norm, a demonstration about how to design a BD algorithm for extracting the cyclo-stationarity of repetitive transients is introduced. Results showed that the proposed method is superior to traditional time-domain BD algorithms in detection performance and its robustness.