Improved Ensemble Empirical Mode Decomposition for Rolling Bearing Fault Diagnosis

Rolling bearing is an important part in mechanical system and faults occur frequently with vibration noise. Empirical mode decomposition (EMD) is a tool for nonlinear and non-stationary signals analysis. However, the major drawbacks of EMD are mode mixing problem, ensemble empirical mode decomposition (EEMD) provides a new tool for signal analysis, and it is an improved technique of EMD. In order to alleviate the mode mixing problem and choose useful IMFs, a method called EEMD and distributing fitting testing is proposed in this paper, and it is used in rolling bearing fault diagnosis. Firstly, using it for rolling bearing fault diagnosis, the fault signal is decomposed by EEMD. Then applying distributing fitting testing to choose components with truly physical meaning and the de-noised signal can be obtained. Finally, utilizing envelope spectrum to distinguish different faults. The results demonstrate the proposed method can sift useful IMFs and diagnose faults effectively, such as inner race fault, outer race fault. The advantage of the proposed method is suitable for rolling bearing diagnosis.