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 the characteristics of five different forms of the gearbox from the normal state to the 100% wear

and complete the fault diagnosis and classification of the gearbox.