An algorithm is proposed to monitor the large-area deep wear faults of the yaw bearing raceway of wind turbine.The accelerometer (vibration sensor) is selected to collect the bearing state data,and the missing value interpolation is implemented to complete the bearing state data;then the data collected is denoised by the improved wavelet modal maxima denoising algorithm;Based on this, the joint sub-domain adaptive adversarial migration network establishes a classification model of the bearing fault feature extraction,extracts the wear and tear fault features of the bearing,and constructs a diagnostic model to classify and recognize the extracted features;At the same time,according to the degree of wear of the bearing to set the wear severity level,establish early warning mechanism,and after the classification of the failure to carry out the wear level scoring.The experimental results show that when the algorithm is utilized to complete the bearing wear monitoring,the denoising effect of the bearing state data is obvious, and the monitoring accuracy is very high.