Fault Diagnosis for Conventional Circuit Breaker Based on One-Dimensional Convolution Neural Network

The vibration signal generated by the operating mechanism of conventional circuit breaker contains abundant mechanical state information. Aiming at traditional fault diagnosis methods that need to realize signal feature extraction based on feature selection, a fault diagnosis model based on one-dimensional convolutional neural network is proposed. In the diagnosis model, multiple convolutional neural networks are designed according to the type and degree of faults, and the network is set as a large convolutional kernel to enlarge the receptive field region; the raw vibration signal is used as the model input for training, and the corresponding fault type and degree are output after hierarchical diagnosis. The experimental results show that the model can automatically extract the fault signal features, effectively complete the fault diagnosis of the contact system for the conventional circuit breaker, and has good generalization ability. The model in this paper has a higher comprehensive diagnosis recognition rate compared with other methods, reaching 98.84%.