A Mechanical Vibration Fault Diagnosis Method of Generator Circuit Breaker Based on Deep Residual Network

The frequent opening and closing characteristics of generator circuit breakers (GCB) in pumped storage power plants lead to their mechanical failure as the main cause of failure. The present paper proposes a deep residual network (DRN)-based mechanical vibration fault diagnosis method for GCB, and converts the vibration signal into a three-dimensional time-frequency image by using the so-called Hilbert-Huang transform (HHT). Then ELU activation function is used instead of ReLU to achieve the improvement of DRN. Finally, the adam optimization algorithm is introduced to better match the time-frequency image features, which in turn improves the fault diagnosis rate.