Cloud-Edge Collaborative Fault Diagnosis Method Based on Deep Q-Network for Rotating Machinery

In the actual, operating conditions of rotating machinery often change due to different tasks. Because of the delay in cloud model updates, the model cannot effectively match with the real-time data at the edge. To solve this problem, a new cloud-edge collaborative fault diagnosis method for rotating machinery is proposed in this paper based on deep <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{Q}$</tex> network (DQN). In this method, Transformer is introduced into the DQN to learn fault features under unknown operating conditions and assign pseudo-labels to real-time data. Meanwhile, A reward function based on cosine similarity is designed to evaluate the above process. The DQN is optimized through online learning, updating the cloud model, and then deployed to the edge through model slicing and ensemble to achieve real-time fault diagnosis under varying operating conditions. The rolling bearing fault diagnosis experiment is implemented to verify the effectiveness of the proposed method. The results show that our method can adaptively update under varying operating conditions, and the updated model achieves higher fault diagnosis accuracy compared to the non-updated model.