Research on rolling bearing fault detection in the initial stage of degradation based upon optimized NLM and TKEO

Abstract

During the operation of rolling bearings, vibration signals contain abundant state information, which exhibits strong nonstationarity and nonlinearity. It is always arduous to detect the initial damage point during the lifetime. Non-local means (NLM) algorithm can suppress noise and highlight the components of the fault impact, but the problem lies in the determination of parameters which directly affect the result. In this paper, we proposed a signal processing method combined NLM optimized by Fruit fly Optimization Algorithm (FOA) and Teager Kaiser energy operator (TKEO) to detect the initial stage degradation of bearings. First of all, the proposed optimal NLM algorithm is used to denoise the bearing vibration signals which are gathered in the initial stage of bearing degradation. Then, the TKEO algorithm is applied to suppress the non-impulsive components and the periodic impulsive characteristics of the denoised signals are enhanced simultaneously. Furthermore, the analysis of the frequency components in the Teager energy spectrum is conducted to detect whether the bearings are abnormal or not. Experimental and comparative analyses are presented to validate the proposed method in the end.