2019 Prognostics and System Health Management Conference (PHM-Qingdao)

1 / 2019-04-29 17:30:27

Fault Diagnosis of Wind Turbine Drive Train using Time-Frequency Estimation and CNN

drive train of wind turbine,Wigner Ville Distribution,convolutional neural networks

Based on the rapid development of wind power, drive train monitoring system is designed to determine the health of wind turbine and to reduce the operating costs by allowing a predictive maintenance strategy. This paper presents a novel method of fault detection using vibration signal. The method is composed of two steps. First, the Wigner Ville Distribution with window function is used to extract time frequency feature from every acquired signal. Then, the convolutional neural networks served as tracking detection model is applied on all available time frequency feature. The spectral peaks and the spectral structures are constantly learned, which are utilized to generate the diagnosis system of drive train. The proposed method is tested on signals from a drive train of wind turbine test rig. Experiments indicate that the proposed recognition algorithm is an effective way to diagnosis the faults between gearbox and rolling bearing.