No abstract

Currently, intelligent fault diagnostics of rotating machinery have significantly contributed. However, real-world labeled data obtained from high-value equipment such as gas turbine units, pumps, and other rotating components are occasionally insufficient. This article proposes an unsupervised deep transfer learning model that can directly extract features from the data itself, hence reducing the number of samples required. The well-designed neural network with domain-specific antagonism mechanism(NND-SAM) would align features between the source and target domains and then make data-driven decisions more efficiently. The parameter-free gradient reversal layer (GRL) is used as an optimizer, therefore considerably reducing the cross-domain discrepancy and accelerating convergence. The average multi-classification accuracy under transferable conditions reaches 97 % and 91% after being conducted on two cases of fault diagnosis. Moreover, the F-score of the system improves by more than 4 % while simultaneously reducing the amount of time required in comparison to the existing models The results reveal that the suggested strategy is suitable for the challenging unlabeled dataset and represents a significant improvement over existing unsupervised learning techniques.