基于图滤波理论的齿轮箱故障特征提取应用研究

Application Research of Gearbox Fault Feature Extraction Based on Graph Filtering Theory

摘要: 风力发电机、汽车、轮船等机械设备都离不开齿轮箱。传统的故障特征提取方法有时难以正确检测故障。本文采用图信号处理方法，提出了基于图滤波理论的齿轮箱故障特征提取应用研究。首先，将齿轮箱的振动信号转化为路图，建立路图拉普拉斯矩阵(Laplacian Matrix)，得到图信号的特征值和特征向量；其次，进行图傅里叶变换(Graph Fourier Transform, GFT)，选择图常数滤波器和图理想带通滤波器对图傅里叶变换后的信号进行重构；最后，将重构信号再次进行图傅里叶变换，从而提取齿轮箱故障特征。研究结果表明：经过图信号滤波的信号相比没有经过图信号滤波的信号故障特征提取更明显，具有一定的故障诊断效果。

Abstract: Wind turbines, cars, ships and other mechanical equipment are inseparable from the gear box. Traditional fault feature extraction methods are difficult to detect faults correctly sometimes. Using graph signal processing method, this paper puts forward the application research of gearbox fault feature extraction based on graph filtering theory. Firstly, the vibration signal of gearbox is trans-formed into road diagram signal, where its Laplacian Matrix is established, then, extracting its ei-genvalues and eigenvectors; Secondly, the Graph Fourier transform (GFT) is established, the graph constant filter and the graph ideal band-pass filter are selected to filter the graph signal after Fou-rier transform, and then, the graph signal on convolution is reconstructed; Finally, the reconstruct-ed signal is transformed into graph Fourier transform, which extracts the fault features of gearbox from different angles. The results show that the signal filtered by graph signal is better than that without graph signal filtering, and the fault feature extraction is more obvious.