부산대학교 도서관

As an important part of rotating machinery, the failure of bearings will cause serious vibration and noise of mechanical equipment, which will affect the normal operation of the equipment and even lead to economic losses and casualties. To accurately and efficiently diagnose the working state and fault category of bearings, a new fault diagnosis method for rolling bearings based on the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), weighted permutation entropy (WPE) and adaptive chaotic particle swarm optimization back propagation (BP) neural network (ACPSO-BP) was proposed. CEEMDAN and WPE were used to extract fault features and optimize the feature vector by mean domain specification principles. ACPSO optimizes the convergence speed and recognition accuracy of the BP neural network by introducing an adaptive tent mapping interval. The experimental results on bearing data from Western Reserve University and actual wind turbine data show that the proposed diagnosis method can achieve high fault recognition accuracy with a small number of training samples.