The paper describes a fault diagnosis system for Nuclear Power Plant with a new human-interface system design method by combining the conception of Multilevel Flow Model and Ecological Interface Design to achieve the goals of quick diagnosis and easy understanding of the diagnosed result. With the modern man-made systems become more and more complex, the task to supervise, diagnose and control the system process also becomes more difficult and goes beyond operators¡Ç ability. Thus, a support system should be provided to operators by advanced information technology to cope with them. It is welcome that the operators need not do anything by themselves since they can get the diagnosis result from the computer. However, because of the serious consequence by the operators¡Ç mistrust to automatic support system, it should be also equipped with an important function that the operators can confirm to make sure that the given diagnosis result is true. For this purpose, the operating support system should not only ¡Ègive right diagnosis timely¡É but also ¡Èmake explanation for easy understanding¡É for the operators to confirm the fault cause. In order to achieve these goals, MFM model is conducted in the process diagnosis by intelligent computation and EID is used for operators to confirm the diagnosed result easily and quickly. A working flow of the diagnosis and confirmation is developed and a prototype system is developed based on a PWR simulator.