Developing Individual-Rule Based Fuzzy InferencernMethods and Investigating the Equivalence Conditions

摘要

This paper provides an alternate approach to construct individual-rulernbased multiple-input single-output (MISO) inference methods for fuzzyrninference systems based on the techniques of aggregation of weightedrninformation instead of the usual approximate reasoning techniques. ThernIF-THEN rules of a fuzzy system are viewed as a class of weightedrninformation. The weighted t-norm and t-conorm operations are used tornaggregate the weighted information. Four classes of individual-rule basedrnMISO inference methods are obtained. Under certain conditions, thernproposed inference methods are equivalent to the inference methodrnproposed by Mamdani, Larsen and Zadeh. Meanwhile, new types ofrninference methods are obtained. Moreover, it is justified that the proposedrninference methods satisfy some basic conditions. Next, this paper reviewsrntwo functional-type fuzzy inference methods, including Takagi-Sugeno (T-S)rnfuzzy inference method, and functional-type simplified fuzzy inferencernmethod with weight (functional-type SFIW method). Then, this paperrninvestigates the equivalence conditions of the above functional-type fuzzyrninference methods with the proposed individual-rule based MISO inferencernmethods. This paper shows that under certain conditions, twornfunctional-type fuzzy inference methods are equivalent to the proposedrnindividual-rule based MISO inference methods.