Vibration Fault Identification Based on Coupled EMD and Energy Distribution

The vibration fault identification is a significant process of the reliability analysis for large rotating machine, and the accuracy of identification results depends on the effectiveness of extracted fault symptoms directly. For the vibration symptoms extraction, the time-frequency domain analysis has been widely utilized. Empirical Mode Decomposition (EMD) is one of the most typical time-frequency method, but its performance is effected by the mode mixing problem. In this paper, a novel coupled EMD is proposed, and utilized to the vibration signal decomposition. In the CEMD, the differential vibration signal was obtained and decomposed by EMD, and then, the results were integrated and couple decomposed by EMD. After decomposition, the energy distribution vectors were extracted as the fault symptoms. At last, the obtained vibration symptoms were taken as input, and the classifier was utilized to identify the different kinds of faults for the large rotating machine. The proposed method was applied into the actual fault identification for the rolling element bearings, and the results demonstrated that, CEMD can improve the mode mixing problem significantly, while the energy distribution symptoms has the advantages of high universality and strong separability, and the vibration faults can be identified more accurately with higher efficiency.