Gear fault diagnosis based on resonance-based sparse signal decomposition of motor current

Abstract

Local gear faults will generate impact torques in the gear transmission system. Previous studies indicated that the fluctuating torque leads to a multicomponent modulation in stator current. So the motor current signature analysis can realize the non-invasive and remote diagnosis of gear faults. Instead of using the traditional methods which decompose a signal based on the frequency, this paper puts forward a gear fault diagnosis method using resonance-based sparse signal decomposition which distinguishes the signals by their oscillation property to extract fault features in motor current. Firstly, the stator current is amplitude demodulated using envelope analysis to eliminate the effect of current fundamental wave. Then, the demodulated signal is nonlinearly decomposed into a high and a low oscillatory component by the resonance-based sparse signal decomposition method. Finally, envelope spectrum analysis is carried out with the low-oscillatory component to extract fault features. The proposed method is verified on the gear fault diagnosis platform. And its effectiveness over the traditional discrete wavelet transform is also shown.