IMPLEMENTATION OF INTELLIGENT SELF-ORGANIZING CONTROLLERS IN DSP CONTROLLED ELECTROMECHANICAL DRIVES

A detailed design procedure for fuzzy neural controllers in electromec hanical drives is presented. The proposed procedure has one or two sta ges. The first stage yields the number of fuzzy rules and the rules th emselves for the fuzzy neural controller. Initial estimates for centre s and widths for membership functions associated with the controller i nput and output variables are also obtained, From this information a f ive-layer fuzzy neural controller is configured and the resulting cont roller may be further tuned using a backpropagation-type of algorithm. Applications of fuzzy neural controllers to DC and AC drives are pres ented. Results are first presented for a separately excited DC drive i ncorporating a combined speed and armature current fuzzy neural contro ller. Results are also presented for the response of a vector controll ed induction motor drive with a fuzzy neural controller replacing the PI speed controller. Minimum numbers of fuzzy sets and rules have been established by the self-organising methodology. The drive responses o btained using fuzzy neural controllers are compared to the correspondi ng drive performances obtained using PI-type speed controllers. It is shown that a high dynamic performance can be obtained from drives inco rporating fuzzy neural controllers and that the ad hoc tuning procedur es associated with the implementation of PI and conventional fuzzy con trollers are eliminated.