Aiming at the aero-engine fault, an abrupt fault diagnosis method based on affinity propagation clustering was proposed. Firstly abrupt fault historical monitoring data are used to establish faults database, then through affinity propagation clustering to find all exemplars of abrupt faults in the database, when need to diagnose new test data, through affinity propagation clustering to find the exemplar of the new collected data, and we can obtain the fault type by center matching with the centers obtained from the faults database. After applied this fault diagnosis method on the aero-engine abrupt fault diagnosis, the simulation and experiment results show that this method can diagnose abrupt fault, and compared with other methods, it took shorter time and produced lower error.