Fault diagnosis of bearing based on LMD and MSE

Abstract

This study proposes the method which is used for roller bearing feature extraction of local mean decomposition (LMD) and multi-scale entropy (MSE). LMD is firstly applied to decompose the measured vibration signal into a series of product functions (PFs), and then MSE is used to extract the feature vectors from the selected PF component. Lastly, the obtained feature vectors of roller bearing with different scale factors are the inputs of support vector machine (SVM) to fulfill the fault patterns identification. The results of fault diagnosis are demonstrated effective by the proposed method used in rolling bearing. The experimental results in rolling bearing fault diagnosis.