A diagnosis framework based on domain adaptation for bearing fault diagnosis across diverse domains

In the current research, the diagnosis process of fault diagnosis models is based on an assumption that the same feature distribution exists between training data and testing data. Regrettably, in real applications, datasets are often from diverse domains; in this case, the traditional diagnostic models lack adaptability. To address this issue, this work proposed a transfer diagnosis framework based on domain adaptation, in that the model trained by the labeled data can be applied to diagnose a new but similar target data. An improved domain adaptation algorithm-weighted transfer component analysis (WTCA) is embedded into this framework. Five fault datasets of bearing are used to demonstrate the applicability and practicability of the proposed diagnosis framework. The results show that the proposed diagnosis framework achieves high accuracy in the transfer tasks of bearing fault diagnosis across diverse domains.