Fault Diagnosis of Roller Bearing based on Hybrid Feature Set and Weighted KNN

Aiming at the problem that the roller bearings early fault features are faint that difficult to be effectively identified,a fault diagnosis method of roller bearing based on hybrid feature set and weighted K- nearest- neighbor（ KNN） is proposed. Firstly,those early fault features of roller bearing are calculated based on the signal processing method in time domain,frequency domain and time- frequency domain to construct hybrid feature set. Then,those hybrid feature set are inputted into weighted K- nearest- neighbor for roller bearing early fault identification. The experimental results show that this proposed rolling bearing fault diagnosis method can effectively extract more sensitive early fault features,and the structure is stable,the diagnosis precision is high. It can be applied in the roller bearing real- time on- line monitoring.