Fault diagnose for rolling bearing based on higher order spectrum

Rolling bearing vibration signals are born with an apparent non-Guassianity, higher order spectrum is an effective way to analyze this kind of signals. The analyzation methods of bispectrum and bicoherence will be introduced into fault diagnose of rolling bearing in this article, these can be shown by giving the graphs of the bispectrum and bicoherence under three different working conditions which were normal, outer race local fault and inner race local fault have been extracted, it is found that there exists a mapping relation between the bispectrum distribution area and fault model. As a result, bispectrum distribution can be used as classification feature. It is difficult to get an accuracy fault classification by directly extracting the amplitude of bispectrum. The usage of binary bispectrum as eigenvectors contributes to constructing nearest template classifier, which can successfully achieve bearing pattern recognition and fault diagnose. The entire calculation process is simple, efficient and practical.