Wind turbine intelligent gear fault identification

Abstract

This paper aims to present the development of a frameworkfor monitoring of wind turbine gearboxes and prognosis ofgear fracture faults, using vibration data and machine learningtechniques. The proposed methodology analyses gear vibrationsignals in the order domain, using a shaft tachometerpulse. Indicators that represent the health state of the gearare algorithmically extracted. Those indicators are used asfeatures to train diagnostic models that predict the health statusof the gear. The efficacy of the proposed methodology isdemonstrated with a case study using real wind turbine vibrationdata. Data is collected for a wind turbine at various timesteps prior to failure and according to the maintenance reportsthere is enough data to form a healthy baseline. The data isclassified according to the time before failure that the signalwas collected.The learning algorithms used are discussed andtheir results are compared. The case study results indicatethat this data driven model can lay the groundwork for a robustframework for the early detection of emerging gear toothfracture faults. This can lead to minimisation of wind turbinedowntime and revenue increase.