Deep Transfer Learning for Bearing Fault Diagnosis: A Systematic Review Since 2016

The traditional deep learning-based bearing fault diagnosis approaches assume that the training and test data follow same distribution. This assumption, however, is not always true for collected in practical scenarios, leading to significant decline performance. In order satisfy this transfer learning concept introduced by transferring knowledge learned from other or models. Due excellent capability of feature domain transfer, methods have gained widespread attention recent years. article presents a comprehensive review development since 2016. review, novel taxonomy proposed perspective target properties divided labels, machines, faults. By covering whole life cycle discussing research challenges opportunities, provides systematic guideline researchers practitioners efficiently identify suitable models based on actual problems encountered diagnosis.