Fault Diagnosis Method for EMU Traction Motor Based on Machine Learning

A fault diagnosis method based on machine learning was proposed in this paper. Based on the high-density on-board data of Electric Multiple Units (EMU), data cleaning and feature processing methods for various parameters such as temperature and current were studied. Additionally, according to the parameter characteristics of different parts, three fault diagnosis methods for the drive-end bearing, non-driven-end bearing and motor body were proposed. Furthermore, a fault diagnosis model for the drive-end bearing was built based on the classification algorithm; A fault diagnosis method for the non-driven-end bearing based on hypothesis testing was established by temperature characteristic statistics; Then, the isolation forest algorithm was used to diagnose the traction motor body fault. Finally, the test results showed that the accuracy and recall rate of the models were acceptable. As a result, the fault diagnosis method proposed in this paper can not only effectively diagnose all kinds of traction motor fault that had been tested, but also locate the fault.