Self-Tuning Control of a Difficult Process

Abstract There are two principal criteria for a ‘universal’ self-tuner: the scope and power of the ‘tuning-knobs’ associated with its performance, and its robustness with respect to the assumptions made in its derivation. A new method called ‘Generalized Predictive Control’ is assessed by a comparative simulation study to show its superiority over Generalized Minimum-Variance, LQG and Pole-Placement self-tuning algorithms. The simulations demonstrate that GPC is effective and robust for a range of typical plant models without the restrictions imposed by the other methods, and it is capable of good control of particularly difficult processes such as nonminimum-phase oscillators.