Nonlinear spectrum feature fusion diagnosis method for RV reducer of industrial robots

Abstract

RV reducers of industrial robots may be possibly subjected to wear and vibration impact and invalidated for keeping the repeatability due to long-time dynamic load in their life cycle. Any failure of RV reducers may decrease the reliability and lead to incredible loss due to the unexpectedly shutdown of the manufacturing system. Failures can cause nonlinear interference when occurring in RV reducers and pose a challenge to RV reducer fault diagnosis. Therefore, it is important to take proper measures to diagnose the failure in RV reducers and reduce the influence of nonlinear interference. Nonlinear spectrum features fusion method is proposed for diagnosing the faults of RV reducers in industrial robots. For the purpose of extracting fault features more effectively, nonlinear output frequency response function is employed to obtain nonlinear frequency spectrum for exactly describing the nonlinear mechanism of faulty phenomenon. Moreover, enhanced evidence theory with similarity measure is proposed to compute the basic probability assignment functions of evidences according to the similarity between different modes for improving the identification accuracy of different faults. RV reducer experiments with different faults are performed to simulate robot operating conditions for obtaining normal and fault data under a variable speed working condition. Among the 180 sets of test data including normal conditions, the diagnostic accuracy of the proposed nonlinear feature fusion method is 92.22% and greatly preferable to that of the ordinary frequency spectrum method. Experimental results demonstrated that the nonlinear spectrum method can effectively extract nonlinear features and the introduction of similarity can effectively improve the diagnostic accuracy in the proposed feature fusion method.