Simulation of an expert model-based adaptive controller

Abstract

Model-based adaptive controllers have been practiced with numerous successes. The controller is formed in a online discrete optimal controller and implemented in control computer. Because of the fast and accurate calculation capability of microcomputer, this type of controller has reached their limits. To explore the potentiality of model-based adaptive controller, we investigate the adaptive controller with an expert system for selection of identifiers. The model-based adaptive controller usually uses a recursive least squares identifier. This kind of identifier requires a lot of calculations. An alternative for adaptive function can be using an rule-based expert system to decide the need of updating process time series model. In addition, we can use a simplified recursive least squares identifier. This paper presents the formulation of this type of controller. Moreover, the simulations are carried out to test the practicality of such controller. The effect of such rule-based adaptation plus model-based optimization and controller formulation will be presented by accumulated loss versus sampling periods. The improvement of mean and standard deviation of controlled variable indicates the sophistication of combination of artificial intelligence and computation power of control computer.