A NEW FAST FUZZY ALGORITHM

Abstract

This pager deals with an implementational aspect of fuzzy control algorithms in a distributed control system (DCS) which is widely used in process industry. A new fast fuzzy algorithm is proposed in order to simplify the computer program of direct reasoning methods by treating controller inputs as fuzzy singletons and replacing both the consequence and the defuzzification by an equivalent crisp function. The structure of the fast fuzzy algorithm appears to be a hybrid reasoning method, but. the algorithm can be categorized into a direct one because the function used in the algorithm is determined by transforming the direct reasoning algorithm, not from designer's intuition. Three direct methods are compared with the fast fuzzy algorithm: Mamdani's, Larsen's and Mizumoto's methods. It turns out that all three direct methods can be expressed as the fast fuzzy algorithm without the loss of performance. The algorithm has been successfully implemented in an existing DCS without changing the existing hardware system for an industrial batch polymerization reactor.