Rolling Bearing Fault Diagnosis Based on Meta-Learning with Few-Shot Samples

As an essential component of mechanical equipment, the state of the rolling bearing has a substantial impact on the operation of the entire automatic system. The fault diagnostic technology based on deep learning surpasses the traditional fault diagnosis technology in many aspects and dramatically improves the accuracy of fault diagnosis but requires a massive amount of labeled data for training. Generally, it takes a lot of effort to obtain tagged data in a natural industrial environment. Therefore, this paper proposes a rolling bearing fault diagnosis method based on meta-learning, which applies the experience learned in the past to new tasks to use few-shot labeled rolling bearing fault samples for training to obtain reliable diagnosis accuracy. The results show that the proposed method can significantly improve few-shot rolling bearing fault samples' accuracy than other traditional methods.