Research on planetary gearboxes feature selection and fault diagnosis based on EDT and FDA

This paper developed an automatical and accurate method in diagnosing the wear fault of planetary gearboxes. In this method, 19 features are extracted to characterize the gear health statuses. Then, select features which are sensitive to the changed health status and unrelated to the varied work condition. The process of feature selection is based on Euclidean distance technique (EDT). Finally, the Fisher discriminant analysis (FDA) is utilized to diagnose the wear fault. The effectiveness of this methodology is demonstrated using vibration data which obtained from a planetary gearbox test rig.