Research shows that multi-fractal can not only exhibit the singular probability distribution form of the fractal
signal completely, but also increase the fine level of signal geometrical characteristics and local scaling behavior. Based
on multi fractal dimension calculation of time-frequency matrix of vibration signal of rolling bearing in this paper, energy
distribution characteristics of time-frequency domain of vibration signal could be extracted, then adaptive fuzzy
neural network (ANFIS) was used in signal classification. Experiments showed that this method can realize fault classify
of rolling bearing effectively, it is feasible in engineering application.