A fault diagnosis method based on multi-agent reinforcement learning for small sample data

The fault diagnosis method of rotating machinery has always been a research hotspot, and the fault diagnosis method based on deep learning has a great dependence on the sample size, which cannot maintain its advantages in the fault diagnosis of small data industry. Aiming at this problem, this paper applies the method of multi-agent reinforcement learning in the field of image classification to the field of fault diagnosis. Aiming at the disadvantage that the reward function in this method cannot provide timely feedback information for classification, a new reinforce reward is introduced. Under its guidance, the agent can adjust the strategy faster, which significantly improves the learning efficiency and exploration efficiency of the agent. The accuracy of the method on small sample data is improved by more than 6 %, which provides a more practical solution for the field of industrial fault diagnosis.