The relationship between the multi-source information fusion and the fault diagnosis is briefly presented, and the general approach of multi-source information fusion based fault diagnosis is introduced. From the viewpoint of fusion structure and algorithms, the classified presentation is given about the multi-source information fusion based fault diagnosis method, and the diagnosis principle and the research status are also described, respectively. It is pointed out that the information fusion fault diagnosis method can be divided into hierarchy, multi-level and combination of neural network information fusion fault diagnosis from the aspect of fusion architecture. From the fusion algorithms, it can be divided into the Bayesian theory, DS evidence theory, fuzzy sets theory, rough sets theory and artificial neural network fault diagnosis methods of multi-source information fusion. Finally, some future development trends of information fusion based fault diagnosis methods are given.