Approach to fault feature extractions of rolling bearing via EEMD and full-vector envelope spectrum

Misjudgments and missed judgment widely occur during the fault detections of rolling bearing due to the fact that single-channel vibration signal information is often collected incomprehensively. In order to recognize bearing faults as possible, a method is proposed that features the combination of Ensemble Empirical Mode Decomposition (EEMD) and full-vector envelope spectrum through the following steps. Firstly, the two homologous double-channel fault signals of bearings undergo EEMD decomposition individually. Then intrinsic mode functions (IMF) with the maximum and secondary kurtosis values at all directions are selected as the reconstructed signals. Finally the reconstructed signals are subjected to full-vector envelope fusion by the use of full-vector envelope spectrum so that the fault feature frequency of bearings can be extracted. By the use of the present method, the real vibration state of rolling bearing were reflected objectively, and the fault feature frequencies of rolling bearing were extracted effectively for the purpose of recognizing fault types, as the experiment results showed.