Stacking Model-based Method for Traction Motor Fault Diagnosis

An ensemble learning method is proposed to improve low classification accuracy and weak anti-noise ability in traction motor fault diagnosis. This paper adopts stacking model to stack SVM, Xgboost and Lightgbm. Comparison between stacking and commonly used single model is further conducted in the cases of the noise and without noise. The results show the advantages of the stacking model-based method.