An improved EMD method for fault diagnosis of rolling bearing

Empirical mode decomposition (EMD) is powerful in analyzing the vibration signals of bearings. However, the accuracy of its decomposition result is heavily dependent on the choice of envelope interpolation algorithms. In this paper, we propose a bandwidth method to select the envelope in EMD called bandwidth based EMD (BEMD). Since there is an optimization process in envelope selection, the scale mixing problem caused by envelope interpolation can be significantly weakened. The proposed BEMD method is demonstrated to be effective in rolling bearing fault diagnosis by comparing with existing improved EMD.