Rolling element bearings are essential components in rotating machinery, it is of great importance to detect the bearing fault as earlier as possible during the operation of machinery. In recent years, variational mode decomposition (VMD) has been widely used in signal decomposition and feature extraction from non-stationary signals. However, the determination of the decomposition layers and quadratic penalty parameter of VMD is still puzzling. In this paper, an improved VMD and optimized frequency band entropy (OFBE) method is proposed. Firstly, the energy entropy maximum principle is utilized to select the decomposition layers and quadratic penalty parameter. After that, a detection method of OFBE based on the principle of maximum kurtosis is presented to determine the bandwidth parameters. Finally, envelope analysis is employed for the detection of fault-related frequency components based on the obtained sub-band signal. The performance of the proposed approach is validated by faults in different parts of the bearing. Results show that the new methodology yields a good accuracy in bearing fault diagnosis. Keywords: Fault diagnosis; bearing; VMD; entropy; envelope analysis
