A Data-Driven Approach to Detect Mechanical Faults in Wind Turbine Gearbox

Online condition monitoring systems play an important role in preventing catastrophic failure, reducing maintenance costs, and improving the system reliability. In this paper, wind turbine gearbox mechanical fault detection system is developed. An adaptive filtering technique is applied to separate the impulsive components from the periodic components of the vibration signals. Then different features of the periodic components and impulsive components are extracted. An extreme learning machine based classifier is designed and trained by using the features extracted from simulated vibration data of wind turbine gearbox. Simulated vibration signals of wind turbines gearbox are used to demonstrate the effectiveness of the presented methodology.