Comparison of condition monitoring techniques in assessing fault severity for a wind turbine gearbox under non-stationary loading

Abstract

Abstract In the present study, condition monitoring techniques such as vibration analysis, acoustic signal analysis and lubricating oil analysis are performed to early detect two failures of a wind turbine gear box such as tooth chip breakage and tooth root crack. A laboratory scale model of a three-stage spur gear system is tested under stationary and non-stationary loads and the response is captured through accelerometers and microphones. Wavelet analysis is performed to extract the features from the signals. Various statistical features are extracted and the dominant features are selected from it by using a decision tree algorithm. The selected features are then used for computing the classification accuracies through support vector machine. It has been observed that vibration signals at stationary loads and acoustic signals at non-stationary loads detect the early failure accurately.