Analysis and Design of the Knowledge Representation for the Implementation of a Distributed Reasoning

Abstract

The representation of knowledge and the reasoning comprise fundamental parts of the design of an expert system. In the case under study the real world is represented by the different events that can arise in the operation of a nuclear plant. These events are characterized by a set of elements interrelated through expert reasoning. We have chosen fuzzy cognitive maps (FCM) for the representation of the expert reasoning front to a fault event, because it handles the uncertainty in the decision making, through the interrelation among its elements. In addition, the FCM emphasizes a representation that allows parallel distributed reasoning (PDR). In our work we detail the analysis and design that was developed to build the causality matrices that give origin to the distributed reasoning, and ultimately have a representation in the FCMs.

La representación del conocimiento y el razonamiento forma parte fundamental del diseño del sistema experto. En el caso de estudio, el mundo real está representado porlos distintos eventos que se pueden manifestar en la operación de la planta nuclear. Dichos eventos se caracterizan por un conjunto de elementos interrelacionados a través del razonamiento experto. Hemos elegido los mapas cognoscitivos difusos (MCD) para la representación del razonamiento experto frente a un evento de falla, debido a que éstos, logran manejar la incertidumbre en la toma de decisiones, a través de la interrelación de sus elementos, además de potenciar el razonamiento distribuido paralelo (RPD). En este trabajo se detalla el análisis y diseño de las matrices causales que dan origen al razonamiento distribuido, y que finalmente tienen una representación en los MCD.