In this paper, the probabilistic neural network (PNN) is applied to accomplish the task of fault characteristic classification in test and fault diagnosis for missile system. In this way, artificial intelligence (AI) is introduced into the application research of electronic system test and diagnosis. Further more, a novel reduction method based on the stretchable constraint clustering algorithm is adopted in the process of training pattern reduction in the PNN. The proper and effective reduction can make the diagnostic PNN more time efficient and easier to understand.