• SDPs were used in signal feature processing to increase the identification precision of motor bearing diagnoses. • An ACGAN was used to address the problem of poor motor bearing diagnoses caused by data insufficiency. • Optimization algorithms were introduced to automatically optimize hyperparameters and improve the performance of the proposed diagnostic model. Bearing failures are the most common type of motor faults. They often occur during the early stages of motor faults. Randomly testing for bearing failures and analyzing the features of such failures and the internal components of bearings play a key role in preventing severe motor faults. This study focused on servomotor signals. Diagnostic analyses were conducted to identify motor faults under data insufficiency. To enable machine learning models to diagnose and identify failures, multiple time lag coefficients (τ) and signals sampled at different sampling rates were used to draw symmetrized dot patterns (SDPs). The distribution of these SDPs and the selection of the diagnostic model were evaluated so as to enable the diagnostic model to operate at its maximum efficiency. Subsequently, a generative adversarial network was used to perform data augmentation so as to enable the neural network model to be trained and provide precise diagnoses under data insufficiency. Finally, an optimization algorithm was used to optimize the diagnostic model and analyze its results and perform real-time model verification. The experimental results indicated that the data augmentation performed by the generative adversarial network effectively improved model prediction, and the optimization algorithm optimized the model with augmented data to further enhance model prediction. In summary, the performance of the proposed diagnostic model under real-time conditions was verified, indicating that the proposed diagnostic model can be used for different types of equipment.
Kuo et al. (Sun,) studied this question.