Nonlinear model based control with application to polymerization reactors.

Abstract

The main focus of this thesis is the development of continuous and discrete-time nonlinear model based controllers that effectively and efficiently address two major limitations in the area of nonlinear process control. The first limitation lies in the construction of stable nonlinear inverse-based or trajectory-following control algorithms for nonminimum-phase systems. To this end, both continuous and discrete-time nonlinear controller designs are presented which utilize synthetic outputs that are statically equivalent to the process outputs and make the system minimum-phase. A systematic procedure is outlined for the construction of a particular class of statically equivalent outputs that lead to prescribed transmission zeros. The selected outputs are then used to synthesize a model-state feedback controller for continuous-time and a model-algorithmic controller for discrete-time, both of which guarantee zero steady-state error between the original process outputs and their set points. The proposed concepts are illustrated with a nonminimum-phase chemical reactor where a series/parallel reaction is taking place. It is shown through simulation results that the proposed controllers lead to excellent output set point tracking and regulatory behavior in closed-loop. The second significant limitation presented in this thesis is the construction of nonlinear controllers for multirate systems. This situation is resolved through the synthesis of a nonlinear multirate model-algorithmic controller that incorporates all available data and has specific closed-loop asymptotic stability criteria. In a simulation case study, the multirate controller is applied to a continuous-stirred tank polymerization reactor to regulate the reactor temperature and number averaged molecular weight of the polymer. The performance of the controller is tested for several sets of controller parameters, where it is shown that the controller effectively tracks step changes in the output set points under active input constraints. The nonlinear multirate model-algorithmic controller is also experimentally implemented on a lab-scale polymerization reactor. The controller performs well in the presence of modeling errors for both steady-state operation and tracking of set point changes. Lastly, to improve the robustness of the algorithm, a feedforward/feedback version of the multirate model-algorithmic controller is derived to efficiently reject the effect of disturbances on output measurements.