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

Blade icing caused by low-temperature environments results in the degradation of wind turbine power performance. As there is no obvious influence on the performance of wind turbines in the early stage of blade icing, it is difficult to detect the early icing state, so there will be inaccurate labels in the process of data collection. To address these challenges, this paper proposes a novel semi-supervised blade icing detection method based on a tri-training algorithm. In the proposed method, extreme gradient boosting tree (XGBoost) is used as the base classifier. A tri-training algorithm is used to integrate three base classifiers and the integrated model generates a pseudo-label for unlabeled data. In addition, we introduce Focal Loss as the loss of the base classifier in the proposed model, which solves the problem of class imbalance caused by the fact that the wind turbine is operating under normal conditions in most cases. In order to verify the effectiveness of the proposed blade icing detection method, experiments are implemented on the collected Supervisory Control and Data Acquisition (SCADA) data. The experimental results show that the proposed method effectively improves the ability to identify blade icing. Compared with other methods, it has better classification performance, robustness, and generalization.