In the research of fault diagnosis methods for electric machinery equipment, Deep learning-based diagnostic techniques can greatly reduce the interference of human experience, and thus have been widely applied. However, these methods assume that data have the same distribution, while in practice, rotating machinery often operates under different conditions, which makes this assumption difficult to meet and reduces the generalization ability of the model. In order to identify defects in electrical industrial equipment, this article proposes a method for exploiting acoustic signals. By analyzing the acoustic signals during operation, fault diagnosis of electric machinery equipment is carried out. The domain mixup method is used to increase the model's generalizability by reducing the distributional distance between data from various domains. Frequency analysis, time-frequency analysis, and feature fusion methods are used to detect faults in fan, bearing, and gearbox in electric machinery equipment. The results of the trial demonstrate that the model is highly accurate and available for diagnostics.