A review of bearing fault diagnosis based on weakly supervised learning

In the era of big data, mechanical intelligent fault diagnosis is more intelligent due to the large-scale collection of data. However, fault data is often difficult to obtain, resulting in small or even no samples. This article reviews the research on weakly supervised learning in the field of fault diagnosis. For few shot learning methods, it is divided into data augmentation method and feature extraction and transfer method. Metric-based zero-shot fault diagnosis using pre-trained word vectors to calculate similarity. Finally, this paper summarizes and prospects the future research of weakly supervised learning.