Fault Diagnosis Based on MI-DANN for Traction Motor Under Multi-working Conditions

A domain adaption neurual network with mutual information(MI-DANN)-based fault diagnosis method is proposed for traction motor under multi-working conditions. First, the domain shift from a working condition data (source) to others working conditon data (target) is conducted by DANN. Then, the MI is introduced into the DANN to select the most discriminative fault features from source and target domain. The experimental results show that the proposed method achieves higher diagnostic accuracy compared to the traditional DANN-based fault diagnosis method.