Abstract: A faulty diagnosis system based on multi-grade fuzzy neural network was presented which uses a multi-grade method to diagnose the fault of small satellite.In this multi-grade diagnosis a fuzzy clustering method was used in the first grade to estimate which module works abnormally.In the second grade fuzzy radial basis function neural network(RBFNN) was used to confirm which component or part was error.At last an experiment was done with this faulty diagnosis method according to a small satellite faulty simulation system .An experiment was done with this faulty diagnosis method according to a small satellite faulty simulation system at last.The result of the diagnosis shows no difference with the fault set beforehand in the simulation system.The result indicates that in a small satellite faulty diagnosis system,this multi-grade method can improve the integration and accuracy of the system,and greatly reduce the complexity of calculation in the process of the diagnosis.