Model predictive control design for multivariable processes in the presence of valve stiction

Abstract

This paper presents different formulations of Model Predictive Control (MPC) to handle static friction in control valves for industrial processes. A fully unaware formulation, a stiction embedding structure, and a stiction inversion controller are considered. These controllers are applied to multivariable systems, with linear and nonlinear process dynamics. A semiphysical model is used for valve stiction dynamics and the corresponding inverse model is derived and used within the stiction inversion controller. The two-move stiction compensation method is revised and used as warm-start to build a feasible trajectory for the MPC optimal control problem. Some appropriate choices of objective functions and constraints are used with the aim of improving performance in set-points tracking. The different MPC formulations are reviewed, compared, and tested on several simulation examples. Stiction embedding MPC proves to guarantee good performance in set-points tracking and also stiction compensation, at the expense of a lower robustness with respect to other two formulations.