A NEW ROLLER BEARING FAULT DIAGNOSIS METHOD BASED ON MVMD-RMS AND DE-LSSVM

Abstract

This research presents a new roller bearing fault diagnosis method based on least square support vector machine (LSSVM) with parameters optimized by the Differential Evolution (DE) algorithm, namely DE-LSSVM.  First, the Multivariate Variational Mode Decomposition (MVMD) method decomposed the roller bearing acceleration vibration signals into component functions. Second, those functions extracted initial feature matrices by the root mean square method. Third, these values serve as input vectors for the DE-LSSVM classifier. Experimental results showed that the proposed method had a lower test error and a less computational time than other methods using the same collected data.