Fault Diagnosis of Bearing Based on Empirical Mode Decomposition and Decision Directed Acyclic Graph Support Vector Machine

When faults of bearing happen, vibration signal of rotation machine always behave in complex form of modulation. The EMD can adaptively decompose signal according to the physical meaning of signal. The SVM has been used in many fields including fault diagnosis because of its excellent learning performance and favorable generalization capability. In this paper, energy eigenvector of frequency band is got through EMD. Fault diagnosis of bearings is realized by DDAGSVM. The most excellent model parameters are selected based on LOO. The final results indicate that the method based on EMD and DDAGSVM can effectively discriminate different faulty states of bearings.