Investigative studies for a knowledge-based intelligent adaptive generalized predictive controller

Abstract

This paper presents preliminary results on tuning the Generalized Predictive Controller (GPC) using an expert system technique. The closed-loop poles that result from applying the GPC to a first order plant are investigated. The plants are classified according to the pole-zero locations and the closed-loop behavior is studied in detail when the tuning parameters are varied. The tuning rules can be formulated from this knowledge and are implemented in the expert system to tune the GPC to satisfy the user specifications. Implementation using an expert system shell. NEXPERT, shows that this knowledge-based tuning procedure gives consistently good results.