Abstract: In order to decrease the probability of malfunctions in nuclear power plant, a real-time expert system to be applied to malfunction diagnosis was designed. Based on the expert system theory the system converts the expert knowledge for diagnosing failures into the rules stored in database, and it can display real-time information of the abnormal symptoms, perform real-time diagnosis of malfunctions and suggest the operation actions related to malfunctions, etc. The results indicate that several typical malfunctions in nuclear power plant are diagnosed automatically and the corresponding operation schedules are given out by present expert system.