this paper designs construction model of data base based on learning algorithm of C4.5 by combining with characteristics of railway signal system in China, meanwhile, it makes research and design on inference machine, explanation mechanism and human interface in expert system of fault diagnosis. It puts forward knowledge representation based on decision tree and application model of acquisition method, designs the complicated fault diagnosis algorithm based on decision tree. Through one fault case, this paper makes demonstration on effect of main modules in fault diagnosis for the designed expert system on fault diagnosis.