Two-Stage Multi-Scale Fault Diagnosis Method for Rolling Bearings with Imbalanced Data

Intelligent bearing fault diagnosis is a necessary approach to ensure the stable operation of rotating machinery. However, it usually difficult collect data under actual working conditions, leading serious imbalance in training datasets, thus reducing effectiveness data-driven diagnostic methods. During stage augmentation, multi-scale progressive generative adversarial network (MS-PGAN) used learn distribution mapping relationship from normal samples with transfer learning, which stably generates at different scales for dataset augmentation through training. diagnosis, MACNN-BiLSTM method proposed, based on attention fusion mechanism that can adaptively fuse local frequency features and global timing extracted input signals multiple achieve diagnosis. Using UConn CWRU proposed achieves higher accuracy than achieved by several comparative methods Experimental results demonstrate generate high-quality spectrum extract features, better classification accuracy, robustness, generalization.