The KNN and SVM-based 2-level comprehensive health indicators diagnosis method for detecting the failure of belt conveyor′s idlers

Belt conveyors are important transport equipment in coal industry. Yet, the methods currently available for diagnosis of the fault of such equipment are still technologically immature. For promoting the application of intelligent fault diagnosis method, the intelligent method targeting at the fault diagnosis of idlers, moving parts of a belt conveyor, is taken as the object of study, and a diagnosis method based on two-level comprehensive health indicator is proposed. The method extracts the Mel frequency ceptrum coefficient (MFCC) of the qudio sequence as a feature and calculates the primary-level health indicators based on the K-nearest Neighbor (KNN) classification algorithm to determine whether any fault occurs. The result of Support Victor Machine (SVM) classification algorithm is used to calculate the secondary-level health indicators, so as to realize identification of degree of failure of any idler and make final intelligent evolution of the failure level. Verification of the method is made using the sound data of idlers collected at Tunlan Coal Preparation Plant of Shanxi Coal & Electricity (Group) Co. Ltd., and through test with Matlab platform. As evidenced by verification result, the method can identify more effectively the occurrence of fault of idlers as well as the level of the failure, and is much higher in identification accuracy as compared to other methods.