Fault damage degrees diagnosis for rolling bearing based on Teager energy operator and deep belief network

Considering that the traditional classifiers' generalization ability is not strong in the early fault diagnosis of rolling bearings, the fault diagnosis method based on Teager energy operator (TEO) and deep belief network (DBN) were put forward. Firstly, the instantaneous amplitudes of the vibration signal were calculated by TEO, and the instantaneous energies of the signal were extracted. Then the characteristic vectors were constituted with the instantaneous energies. DBN classifiers were used to identify the faults of rolling bearing. For different types of fault diagnosis, DBN structure parameters were adjusted according to the classification error rate of training sets. Using the bearing fault experiments' data of American West Storage University, the classification accuracy of SVM and KNN was compared. The results show that the suggested methods are more effective and stable for the identification of rolling bearing fault diagnosis in various situations. © 2017, Central South University Press. All right reserved.