A dynamic fuzzy reasoning method for adaptive diagnostic systems design

Abstract

Tackles some critical issues in diagnostic problem solving by addressing the conceptual design of automatic diagnostic systems able to cope with complex dynamic processes. The dynamic switching fuzzy system model introduced by M.H. Smith (1994) is used as a generalized framework which dynamically changes the reasoning method by tuning the operators and/or defuzzification methods. This technique is applied to tune different fuzzy diagnostic models introduced by M. Ulieru (1995). The dynamic switching fuzzy diagnostic model (DFSDM) proposed in this paper can adapt online to drastic changes in the behavior of the diagnosed process. Tutorial investigations on an inverted pendulum clarify the ability of this method to cope with unexpected, unknown disturbances while simulation results on an electromechanical process prove the ability of this technique to manage critical faults in continuous dynamic systems.