As there is some probability to get a wrong result with the vibration signal at the special time, a novel fault diagnosis method based on time informa tion fusion is proposed. The specific algorithm has been implemented and provid ed in this paper by using the wavelet transform and probabilistic model. A quant itative discriminant fault indicator, Fault Comparability Degree, is also propo sed. Various failures on a rotor have been carried out. The experiment results s h ow that the Failure Comparability Degree is able to distinguish different fau lt categories, and thus improved diagnostic accuracy can be achieved.