Fault feature extraction of localised rub-impact fault based on vector-bispectrum and Teager energy operator

Bispectrum analysis is able to suppress the Gaussian noise and analyse quadratic phase coupling in the signals more accurately. However, traditional bispectrum has the problem of losing fault information at low frequency, and may not comprehensively reflect nonlinear information of the fault, which leads to weak diagnostic ability, especially for local rub-impact fault. For this reason, a new feature extraction method called vector-bispectrum energy operator is presented in this paper, which adopts vector-bispectrum based on Teager energy operator and is used to obtain the fault feature of local rubbing rotor. The results of the experiment show that the feature of the local rubbing fault extracted by Teager energy operator based vector-bispectrum is very noticeable, and the fault classification through SVM has a high recognition rate.