Abstract: In terms of mechanical fault diagnosis, factor hidden Markov model is the common tool for recognition at present. Infinite factor hidden markov model (IFHMM), which is a generalization of factor hidden markov model (FHMM), overcomes the deficiency of FHMM, i.e. chain number is often assumed. Here, IFHMM is introduced to fault diagnosis of rotating machinery, a fault diagnosis method based on IFHMM is propose, and compared with the FHMM recognition method. Finally, the proposed method had been successfully applied in the fault recognition of rotating machinery. The experiment results showed that the proposed method has obvious superior to the fault recognition method based on FHMM.