Nuclear power plant is a complex and huge system. There are interrelated, coupling and nonlinear relations between the systems. The rapid location, detection and maintenance of nuclear power plant failures is an important way to improve the safety and reliability of the system. Precise and fast fault detection device has become an important part of nuclear power plants. In this paper, a fault diagnosis strategy based on the fault propagation model based on the directed graph analysis for nuclear power plant is proposed. The fault diagnosis model is suitable for the single fault and multi fault situation of the system. According to the nuclear power system operation process, operation conditions, complex dynamic characteristics, and the qualitative logic relation between variables to maintain the basic characteristics of stable, the proposed diagnosis model for fault diagnosis, diagnostic reasoning method have incomparable advantage compared with other graph models. Besides, for the modules or units involved in the nuclear power system, a knowledge supplement based on probability and statistics is applied to achieve blind spot diagnosis and location.
