A Supervised Data-Driven Approach for Wind Turbine Pinion Gear Defect Identification

Abstract

The health monitoring of the wind turbine pinion gear has been widely performed to ensure safe operation of the wind turbine. For years, vibration sensors have been routinely installed on wind turbines, claiming some success in preventing accidents. Gear defect recognition based on vibration, however, remains a challenge because the signature of the defect of a gear is usually masked by background noise pollution of the external variable transmission path of the vibration sensor installation. In this paper, the empirical mode decomposition (EMD) method is used to analyze the radial vibration measurement of 3-MW wind turbine gears in the case of a failure in a planetary gear. The vibration signal of EMD involves the subsequent envelope analysis of several intrinsic mode functions (IMFs). The appropriate components are used to characterize the fault feature selection of the gear. The analysis results show that the method has successfully demonstrated the gear failure characteristics and revealed the effectiveness of the EMD scheme.