Gear Fault Diagnosis Based on Dual Parameter Optimized Resonance-Based Sparse Signal Decomposition of Motor Current

Abstract

Motor current signature analysis (MCSA) provides a nondestructive and remote approach for a gear fault diagnosis. However, in addition to the fault-related components, motor current in the faulty gear system also contains the eccentricity-related components and gear meshing-related components, which contaminate the fault features and increase the difficulty of fault diagnosis. To extract fault features from these interferences, this paper proposes the dual parameters optimized resonance-based sparse signal decomposition (RSSD) method, which can decompose a complex signal into a high- and low-resonance component with two sets of overcomplete wavelet bases. After the decomposition, the fault-related components, which have short duration, will exist in low-resonance component. The novelty is that the wavelet bases related parameters, Q-factors, and decomposition levels are chosen automatically based on artificial bee colony algorithm to obtain the optimal decomposition results instead of chosen subjectively. Kurtosis of the low-resonance component is employed as optimization index. The proposed method is then verified on the gear fault-diagnosis platform, which consists of two permanent magnet synchronous motors and a pair of gears with transmission ratio of 3:2, and its effectiveness over some existing methods under different operating conditions is also validated.