A learning rule-based control system

In this paper the concepts of knowledge acquisition and learning are reviewed and combined to build a rule-based estimation and control system. The methodology is demonstrated for controlling a chemical process, using actual laboratory results. The factors affecting learning are highlighted with emphasis on the learning environment. Both off-line and on-line learning from examples are discussed. Closed loop control using the rule-based system are demonstrated to yield acceptable setpoint tracking, both in steady state accuracy and speed.