Deep auto-encoder network for mechanical fault diagnosis of high-voltage circuit breaker operating mechanism

Abstract

Abstract To improve the accuracy of the mechanical fault diagnosis of the operating mechanism and fully exploit the characteristic information in the vibration signal of the high-voltage circuit breaker, a mechanical fault diagnosis method of the operating mechanism of the high-voltage circuit breaker based on the deep self-encoding network is proposed. First, the vibration signal of the switch operating mechanism is extracted, the wavelet packet conversion is performed, and the vibration signal of each frequency band is divided into equal times. The energy of the time–frequency subplane of the vibration signal is then calculated, and the time–frequency energy distribution is used as a switch. Finally, a breaker failure diagnostic model based on the deep self-coding network is established. Pretraining and tuning and a 126 kV high-voltage switch are used to simulate different types of faults and validate the method. Experimental results show that this method can acquire sample failure data and perform failure diagnosis, and the diagnosis accuracy rate reaches 97.5%. The deep self-coding network can fully pierce deep information on the switch vibration signal.