AR模型和分形几何在设备状态监测中的应用研究

USING AR MODEL AND FRACTAL GEOMETRY FOR CONDITION MONITORING OF WORKING MACHINERY

状态特征指标对机械的状态监测和故障诊断具有重要意义。本文提出应用机械设备工作状态下噪声信号自回归模型的关联维数来描述设备在不同工作状态下的特征 ,进而实现对状态的监测、识别和分类。文中通过实验证明 ,设备在相同工作状态下 ,噪声信号的AR模型参数具有相近的关联维数 ,在不相同状态下则有明显不同的关联维数。因此关联维数不仅可以作为状态监测与识别和分类的重要依据 ,而且可以作为其他特征提取方法的补充。

There are a wide variety of condition monitoring techniques currently used for the recognition and diagnosis of machinery faults. However, little research has been carried out about the occurrence and detection of chaotic behavior in time series noise signals. In this paper the AR model based on the noise signal of working machinery is established for the analysis of the correlation dimension of the typical noise signals. Correlation dimension of the AR model parameter is calculated to recognize the working condition and detect the faults. Finally, some experimental results from the air pumps show that there are distinct differences in the correlation dimension obtained for normal air pumps and those with low pressure, low vacuum and unstable faults in the same air pumps. The experimental results confirmed the feasibility and validity of the method proposed in this paper.