Abstract: Bispectrum analysis method is wildly used in the rotor fault diagnosis because it can distill the nonlinear characters of the signals. But the general bispectrum analysis is used to take the single channel signals as the studied objects, it cannot embody the nonlinear properties of the rotor systems integrally and the problems of the information omission exist, and furthermore the analysis conclusions obtained may be different because of the difference of the information coming from the two-channel signals at the same cross section. In order to solve the problem, the new method named as the vector-bispectrum signal analysis is put forward based on the full vector-spectrum analysis method. Vector-bispectrum is the full vector-bispectrum analysis method that the amplitude information is fused coming from the two-channel signals at the same cross section and each phase information is retained, which can really reflect all kinds of information of the rotor running and satisfy the agreement demands of the analysis conclusions. The definition and algorithms of the vector-bispectrum are given. The simulation and the fault diagnosis experiment of the gearbox are studied. The results show that the nonlinear properties contained by the signals can be more completely reflected, the analysis conclusions are of agreement and reliability, and thus the accuracy of the intelligent fault diagnosis is also improved.

Key words: Bispectrum, Fault diagnosis, Full vector-spectrum, Information fusion, Vector-bispectrum