Abstract: In order to resolve the problems of acquiring knowledge and reasoning in traditional expert system,a new diagnosis system based on the combination of the neural network and expert system was proposed.The principal component analysis method was adopted to simplify the training pattern of neural network,and optimize the structure of the network.The technique of neural network ensemble was used to resolve the difficulties of choosing the number of network middle layer and the threshold of judgement.The reasoning process and the interpreting methods for diagnosis results were given.The technology was applied to fault diagnosis of a satellite position control system,and the result proved that the diagnosis efficiency and accuracy was increased.

Key words: Neural network ensemble, Fault diagnosis expert system, Principal component analysis, Attitude control, Satellite