Uncertainty analysis in expert systems

Abstract

If knowledge-based systems are to make any significant impact in civil engineering practice, they must be able to deal with uncertainty in a satisfactory manner. The way this has been done in the past has lacked an adequate theoretical underpinning. It is the purpose of this paper to present an analysis which shows that the operations of fuzzy and probability logics are special cases on an interval of possibilities. A new method of support logic is briefly presented. The method follows the Horn clause representation of prolog. It does not rely on a closed world form of knowledge representation and can allow the modelling of total uncertainty. A computer program, FRISP (Fuzzy Relational Inference with SuPport logic), is briefly introduced.