Self-Tuning Adaptive Multiple Model Predictive Control with application to pH Control process

In this paper, an automatic learning and self-tuning multi-variable adaptive Multiple Model Predictive Controller (MMPC) is implemented on a laboratory scale pH pilot plant. The variables of the process is regenerated with an optimized real time soft analyzer, and a self-organizing neural network is trained to model the process. Generalized Predictive Control based on Independent Model (GPCI) and Dynamic Matrix Control (DMC) are used as base controllers. In the presented theme the nonlinear modeling of the plant, model bank generation, model identification for each region, and controller selection/design are done automatically. A new disturbance rejection supervisor is used to improve the performance of the automatic controller in the presence of disturbances. Exhaustive implementation analyses are provided to demonstrate the abilities of the presented control scheme.