The gearbox is one of the critical components in a wind turbine, which is responsible for about 15~20% of its maintenance costs. It?ˉs very necessary to predict gear degradation development and judge the fault type which can help us to make a maintenance program previously. As a common spectrum analysis method, Fourier transform can extract the useful fault information of wind turbine gearbox. Wavelet Neural Network (WNN) has been proposed as a strategy for the non-stationary signal processing and represents powerful ability in fault prediction. In this paper, the relation matrix between critical gearbox fault and vibration frequency is summarized. Based on the relation matrix, three faults and normal condition of gearbox are simulated as the experimental sample. Then, amplitudes of main frequency are extracted by spectrum analysis. The amplitudefrequency data are not stationary and Dual Wavelet Neural Network (DWNN) is applied for prognosis based on these data. Finally, DWNN is compared with WNN and the prognosis result is presented.
展开▼
