The model of circuit breaker (CB) state identification cannot explain the correlation between fault characteristics and diagnosis results in electrical professional thinking. In this paper, a state assessment method of CB based on the vibration characteristic domain is proposed to realize an analysis of the fault characteristics of CB. The continuous wavelet transform (CWT) is used to describe the time-frequency graph of the vibration signal. The time-frequency characteristic set of the closing process is constructed. Then, after comparing the performance of different deep convolutional neural networks (DCNN) on the CWT time-frequency graphs, the state identification model of CB is proposed. By using the explanation technology——SHapley Additive exPlanations (SHAP), the fault characteristicdomains of the CB are constructed. Then the time-frequency scale representation of the fault characteristicdomain under different action sequences of the operating mechanism is described in detail. The experimental results show that the method can capture the fault characteristics domain on the time-frequency scale. Also, the model can still identify its fault characteristics for random fault states.