Rolling bearing faults diagnosis based on empirical mode decomposition: Optimized threshold de-noising method

The faults of rolling bearings frequently occur in rotary machinery, therefore the rolling bearings fault diagnosis is a very important research project. The vibration signal is usually noisy and the information about the fault in the early stage of its development can be lost. A threshold de-noising method based on Empirical Mode Decomposition (EMD) is presented in this paper. Firstly, the signal is decomposed into a number of IMFs using the EMD decomposition. Secondly the algorithm based on the energy to determine the trip point is designed for IMF selection, then, by comparing the energy of the selected IMFs with excluded IMFs, singular selected IMFs are treated with soft threshold function, and finally the de-noised signal is obtained by summing up the selected IMFs, it is proved that the best IMFs can be summed up and properly de-noised by the proposed method. The results show the effectiveness of the proposed technique in revealing the bearing fault impulses and its periodicity and amelioration the sensibility of scalar indicator for real rolling bearing vibration signals.

I. Introduction

Rotating machinery having components such as bearings provides a good example of condition monitoring. Specifically, misalignment, fatigue, looseness, and contamination, which can become major causes of bearing faults. Typically, fault-induced signals from rotating machinery involve periodical impulses that are masked by environmental noises. The spectral signatures of good and defective bearings have been ascertained, and widely explored in a variety of literatures. One of the most popular applications of such a method is noise elimination. Empirical mode decomposition (EMD) is a recently proposed method to analyze non-linear and nonstationary time series by decomposing them into intrinsic mode functions (IMFs) [8]. EMD based de-noising methods require a robust threshold to determine which IMFs are noise related components [6]. In this study, we propose optimized threshold de-noising method based on EMD. Firstly, the determination of trip point is designed for IMF selection, then, by comparing the energy of the selected IMFs with excluded IMFs. Secondly, the singular selected IMFs are treated by soft threshold function, and finally the de-noised signal is obtained by summing up the selected IMFs. Experimental results show that this method can extract fault characteristics of roller bearing effectively.