A knowledge representation model for the nuclear power generation domain

Abstract

A knowledge representation model for the nuclear power field is proposed. The model is a generalized production rule function inspired by a neural network approach that enables the representation of physical systems of nuclear power plants. The article discusses some techniques currently used for knowledge representation of physical systems and argues that the proposed approach overcomes some aspects of the deficiencies that exist in these other techniques. An application to the Containment Spray System of the Comanche Peak Steam Electric Station is provided. The algorithm developed worked accurately, managed to handle multiple fault scenarios and gave correct and useful information. Also, none of the other knowledge representation techniques discussed seems to be capable of providing the level of detail and the advantages of the method proposed.