Abstract: In order to achieve gearbox fault online detection, the method of gearbox vibration signal online identification based on recursive AR model parameters identification is presented. The different condition vibration signals of the laboratory gearbox are detected. Then the order of auto-regression model and the initial values of model parameters are determined by the optimal instrumental variable approach for these detect signals. On these foundations, recursive parameter online identification based on Kalman filter is implemented taking coefficients of the auto-regression model as the states variables. Meanwhile, gearbox fault online detection is realized depends on the 2-norm of model parameters change. And two kinds of faults that peeling off the bearing outer ring and gear wear are analyzed. Results show that their 2-norm of model parameters change for two failures happened mutation.

Key words: recursive parameters identification, gearbox, vibration signal, Kalman filter, fault online detection

摘要： 为实现齿轮箱故障的在线检测，提出基于递推AR模型参数辨识的齿轮箱振动信号在线辨识方法。对实验室的齿轮箱进行不同工况下振动信号的检测，利用最优辅助变量法确定其自回归模型的阶次和模型参数的初值，以自回归模型系数作为状态变量，采用Kalman滤波器技术进行在线递推参数辨识。实验结果表明，该方法中参数变化量的2-范数会发生突变，能检测出齿轮磨损和轴承外圈剥落的故障。

关键词: 递推参数辨识, 齿轮箱, 振动信号, Kalman滤波器, 故障在线检测

CLC Number: 

• TP301.6