439 / 2022-09-14 13:37:55

Partial Transfer Fault Diagnosis of Rotating Machines Based on Subclass Alignment Network

Fault Diagnosis, Transfer learning, Artificial Intelligent， rotating machine

Machine condition monitoring and fault diagnosis

With the rapid development of intelligent manufacturing and industrial big data, deep transfer learning has been widely applied in the field of intelligent diagnosis of rotating equipment. It is usually assumed that the source domain and the target domain data have the same label space in such methods. However, there are a large number of scenarios in the industrial field where the label space of the target domain is a subset of the label space of the source domain. The existing deep transfer learning is unavailable to eliminate the effect of negative transfer of the outer categories of the source domain on the classification performance of the target domain. To address this issue, a partial transfer fault diagnosis method based on the subclass alignment network is proposed. In the proposed method, Vision transformer is adopted as the basic network to extract the global feature information from both source domain and target domain data. Meanwhile, a weight balance mechanism is constructed to promote the marginal distribution alignment between the shared categories of the source domain and target domain data, and a metric learning approach is used to realize the conditional distribution alignment between the shared categories of the source domain and target domain data. The application on the bearing and gear faulty data verifies the superior performance of the proposed method.