Physics driven incremental meta-learning rolling bearing fault diagnosis under few-shot emerging conditions
Traditional deep learning fault diagnosis methods rely on sufficiently labeled training data, which limits their ability to adapt to the continual emergence of new fault data. Although continual learning methods can alleviate catastrophic forgetting to some extent, the model’s diagnostic performance is still significan...