IMPRECISE KNOWLEDGE IN EXPERT-SYSTEMS - A SIMPLE SHELL

Expert knowledge is often pervaded by imprecision and/or uncertainty, and this vagueness is considered more and more as a part of the expert ise. Methods such as probability theory, error calculation, Bayesian a pproach or certainty factor have been used for a long time in order to deal with uncertainty or imprecision. Among these methods, fuzzy logi c has the particularity To allow the management of uncertainty and imp recision in a unique framework. One of the interests of fuzzy logic in expert systems is that it handles in a single frame both imprecision and uncertainty, for instance through the generalized modus ponens app roach. Shells allowing to implement this approach are usually quite he avy to handle and a simple shell that implements these concepts has be en developed in order to provide students with a user-friendly and pow erful environment. The main originality of this shell, called FLESH (F uzzy Logic Expert Shell on Hypertext) is that it associates various se mantics with the implications involved in the production rules. Furthe rmore, the simplicity of its interfaces has enabled the use the system for two years for educational purposes. Two expert systems developed with this shell are shortly described: a decision support system for p roduction activity control and an expert system, for ball-bearing sele ction and assembly.