It has been proposed to use the Multilevel Flow Modeling (MFM) by M. Lind as a framework for functional knowledge representation for qualitative reasoning in a complex process system such as nuclear power plant. To build a knowledge base with MFM framework makes it possible to represent functional characteristics in different levels of abstraction and aggregation.
A pilot inference system based on the qualitative reasoning with MFM has been developed to diagnose a cause of abnormal events in a typical PWR power plant. Some single failure events has been diagnosed with this system to verify the proposed method. In the verification study, some investigation has been also performed to clarify the effects of this knowledge representation in efficiency of reasoning and ambiguity of qualitative reasoning.