Wind Turbine Bearing Fault Diagnosis Method Based on Multi-domain Feature Extraction

Abstract

Aiming at extracting fault features of wind turbine bearing vibration signals, a bearing fault diagnosis method is proposed that based on multi-domain features extraction and deep modeling. Firstly, the multi-domain features of the wind turbine bearing vibration signal are extracted, including time-domain features, frequency-domain features, and time-frequency domain features after Ensemble Empirical Mode Decomposition (EEMD) decomposition. Secondly, the Random Forest (RF) is used to reduce the data dimension of the multi-domain feature set, delete the features irrelevant to the classification, and improve the accuracy of fault diagnosis. Finally, a wind turbine fault diagnosis model is established based on DBN, and the model is verified with the CWRU data set and the actual operation data of Jiangxi wind farm. The experimental results show that the fault classification accuracy of the proposed model is above 94.4%, which is higher than the comparison method. The superiority of this method in the extraction of wind turbine bearing fault information is confirmed.