关键词: 类质心优化, 轴承故障, 恒定转矩, 特征频率变化, 轴承调制分量, 聚类指示矩阵

Abstract: When the modulation capture of the output signal of the wind turbine pitch bearing is incomplete, the probability of correct diagnosis is low when using the selected class centroid to determine the bearing fault. In order to improve diagnostic accuracy,research has been conducted on fault diagnosis technology for large wind turbine variable pitch bearings based on class centroid optimization. The total load torque of the bearing is characterized by the additional components of constant torque and characteristic frequency variation. By calculating the position of the mechanical angle of the bearing rotor under torque balance, the modulation components of the bearing at different positions are obtained, achieving complete capture of signal modulation. The Laplacian matrix is used to modulate and decompose the bearing signal into several base partitions,which are coordinated with the bearing fault clustering indicator matrix to select the optimized centroid. Under the elimination effect of the auxiliary domain, match the fault class centroid corresponding to the bearing output signal in the target domain to complete the fault diagnosis of large wind turbine pitch bearing. In the test results,the design technology achieved a correct diagnostic level of over 91. 67% for both the inner and outer ring fault samples of the bearing,improving diagnostic accuracy.

Key words: class centroid optimization,bearing malfunction,constant torque,characteristic frequency variation,bearing modulation component,fault clustering indicator matrix