Diagnostic knowledge generation of nuclear power plants using knowledge compilers

This paper discusses a method to generate diagnostic knowledge of nuclear power plants, from commonly accepted physical knowledge and design information about plant configuration. This method is based on qualitative reasoning, which is advantageous to numerical information processing in the sense that system can explain why and how directly applicable knowledge is correctly generated, and that knowledge base is highly reusable and expandable because it is independent on detailed numerical design specifications. However, reasoning ambiguity has been found as the largest problem in applying the technique to nuclear power plants. The proposed approach mainly consists of a knowledge representation scheme, reasoning algorithm, and qualitative model construction method. (author). 4 refs, 8 figs, 1 tab