Aiming at abrupt failure of aero-engine, an aero-engine abrupt failure prognosis method is proposed by combining of multi-states MOG-HMM and Viterbi algorithm. First of all, according to the historical monitoring data of aero-engine, a multi-states MOG-HMM model is built, that it can determine the number of states, the state transfer matrix, the observation probability distribution and the ultimate abrupt failure state. Then new observation data are collected, through Viterbi algorithm to decode the data to obtain its current state. Finally, the abrupt failure can be predicted by by calculating the time from current state to the final abrupt failure state. Simulation and experimental results show that the method can realize the prognosis of abrupt failure, which also meet the requirements of standard predictor.