Jianhua WANG. Research on Four-Quadrant Input Over-current Fault Diagnosis of Electric Locomotive Based on Recursive 2-Classification Method[J]. Electric Drive for Locomotives, 2019,(2):74-79,138.
Jianhua WANG. Research on Four-Quadrant Input Over-current Fault Diagnosis of Electric Locomotive Based on Recursive 2-Classification Method[J]. Electric Drive for Locomotives, 2019,(2):74-79,138. DOI: 10.13890/j.issn.1000-128x.2019.02.016.
Four-quadrant over-current fault is one of the main faults of the HX,D,1 locomotive traction converter. Due to the class distribution of fault cause is unbalanced and there are a lot of over-current fault causes, four-quadromt over-current fault is difficult to distinguish and diagnose. A fault diagnosis method based on recursive 2-Classification was proposed. Based on K-Means and wavelet packet decomposition, feature extraction of data-related key fields was carried out. Reliable classification model was constructed by gradient boosting decision tree (GBDT) based on ensemble learning, and multi-label data was classified and diagnosed recursively by recursive 2-classification. Experiments on four-quadrant input over-current fault data files showed that this method could improve the classification and diagnosis of fault samples.