Multivariable Robust Model Predictive Control of a Laboratory Chemical Reactor

Optimal control of chemical reactors represents a challenging task, as their behaviour is non-linear, asymmetric, and highly affected by various uncertainties. The paper presents multivariable control of a laboratory chemical reactor, in which acetic acid reacts with sodium hydroxide and the neutralization process runs in the reaction mixture. The challenge is to optimize the set-point tracking of pH in the presence of parametric uncertainties. The control inputs are the volumetric flow rates of acid and base. The robust model predictive control was designed to fulfil the control task. To make the optimization problem more tractable, the LMI-based formulation of the convex optimization problem is implemented. The optimization problem has a special structure to ensure offset-free control performance and satisfaction of constraints on control inputs. Moreover, optimal control actions are calculated considering the multiple input system. As only the controlled output is measured, the state observer was designed and used. The main contribution of the paper is the experimental investigation of the LMI-based and state-observer-based robust model predictive control with integral action designed subject to input constraints. Control performance of the laboratory neutralization chemical reactor satisfies all control requirements.