A fault diagnosis method for rolling element bearings based on ICEEMDAN and Bayesian network

As commonly used components in rotating machinery, rolling element bearings (REBs) can fail due to complex working conditions and high-speed rotation. The failure of bearings may cause great damage. It is necessary to identify the faults of bearings to prevent property losses and heavy casualties. This paper proposes a fault diagnosis approach based on improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and Bayesian network. The intrinsic mode functions (IMFs) extracted by ICEEMDAN algorithm are applied to construct feature vectors based on the energy entropy, and then the fault diagnosis model of the bearing is constructed by Bayesian network. The influence of load and sampling frequency on diagnostic accuracy of the bearing with different fault types is studied in this paper. And the research results show that the ICEEMDAN-BN method can improve the uncertainty reasoning ability and accuracy of the developed fault diagnosis model.