APPLICATION OF EEMD-SVD TECHNIQUE BSOA-SVM METHOD TO FAULT DIAGNOSIS FOR ROLLER BEARING

Abstract

This study proposes a new method to diagnose roller bearing failure, in which we use EEMD (Ensemble Empirical Mode Decomposition) method combined with the singular value decomposition method (SVD) and support vector machine. (SVM). In addition, the parameters of SVM are selected through Backtracking  Search Optimization Algorithm to make BSOA-SVM clasifiers. First, the accelerator vibartion signal is decomposed into IMF (Intrinsic Mode Function) by EEMD method. These IMFs are analyzed separately to create input vectors for SVM classifiers. Finally, BSOA-SVM classifiers are used to classify fault roller bearing patterns. Experimental results show that the proposed method can classify the conditions of roller bearings with higher accuracy and lower time when compared with other methods.