A fuzzy-logic-based approach to building efficient fuzzy rule-based expert systems

Abstract

Knowledge-based expert systems have been increasingly applied to diverse engineering problems. These systems are mainly based on the compiled knowledge represented as production rules pertaining to the problem domains. In reality, the available information on any problem is almost always imprecise, incomplete and ill-defined; the linguistic variables need be defined as fuzzy variables which are mapped into appropriate numerical domains. Consequently, a fuzzy rule-based expert system (FRBES) is becoming attractive in problem solving. In such an expert system, however, a number of redundant rules as well as illogical or unnecessary interconnections among them frequently exist, thereby rendering the system unduly cumbersome and ineffective. In this paper, a fuzzy-logic-based approach is proposed for evaluating and simplifying the rule base of an FRBES. Fuzzy networks are generated from the rules in the original rule base, and a systematic and in-depth analysis of the resultant networks is conducted by means of fuzzy logic. Such an analysis gives rise to a new rule base derived from the original one; the former is always logically correct and structurally simpler than the latter. The efficacy of the proposed approach is demonstrated by applying it to the design of two FRBESs: one for cyanide waste minimization in an electroplating plant and the other for fault detection in the operation of hazardous waste incineration facilities.