齿轮箱故障诊断中的正交匹配追踪算法

Gearbox fault diagnosis based on orthogonal matching pursuit algorithm

为通过齿轮箱的振动信号进行故障诊断,应用正交匹配追踪算法对振动信号进行处理.齿轮箱的振动信号包含了齿轮箱运行状态特征,但同时也掺杂了大量噪声信号,总体呈现出非平稳性.齿轮箱故障诊断的关键是从齿轮箱的振动信号中剔除冗余信息,用少量特征信息准确的表达信号,完成对信号中故障特征的提取.传统的频域分析法,只能从频域图上定性的判断故障,无法做到定量判断.正交匹配追踪算法是一种定量提取特征的方法,在傅里叶正交基下对振动信号进行时域向频域的映射,在频域上定量的得到主要特征,再根据主成分分析思想,提取出3组主要特征点,将已知故障分类的信号特征与待检测信号的特征进行对比,通过频域的位置和幅值的两次比较,判断故障状态,实验证明该方法可以准确的判断出齿轮箱从正常状态到100%磨损的5个不同形态的特征,完成对齿轮箱的故障诊断和分类.

To implement fault diagnosis using the vibration signal of gearbox, the vibration signal is processed by the orthogonal matching pursuit （OMP） algorithm. The vibration signal of gearbox contains running characteristics and mixes a lot of noises, and presents non-stationary signal feature. It is important to remove redundant information, extract major information to express signal. Traditional frequency domain analysis method can only qualitatively judge fault from the frequency domain figure, cannot quantitatively compute the fault. Orthogonal matching pursuit is a quantitative method to extract characteristics. In this paper the OMP method is used to map time domain to frequency domain based on the Fourier orthogonal basis, and the major characteristics are obtained in the frequency domain. According to the principal component analysis （PCA）, it extracts three sets of major characteristics, then the characteristics of known fault classification signal are compared with the characteristics of the detected signal. By comparing the location and the fault of the vibration state can be judged. The experiment results show that the method can accurately judge normal state to the 100% wear, and complete the the characteristics of five different forms of the gearbox from the fault diagnosis and classification of the gearbox.