Feedback Linearization + Contractive MPC: Stability Analysis / Application to a Polymerization Process

In this work a control technique for nonlinear systems which deals with process constraints in a computationally efficient way is proposed. The technique is a fusion of two control methods: input/output feedback linearization (FL) and contractive model predictive control (CNTMPC). In this configuration, CNTMPC guarantees closed-loop stability and robustness but it is a hard problem to solve on-line because it is usually subject to nonlinear and state dependent constraints. In order to deal with such constraints, different approximate methods are proposed. As a result of some of these approximations, the CNTMPC problem can be reduced to a quadratic programming problem (QP). Although approximate, these methods are implemented in a way through which satisfaction of the original constraints is guaranteed, while generating an optimization problem with the implementation and computational levels of difficulty nearly equivalent to those of the case where the plant and the constraints are linear. The controller is successfully applied to a batch solution polymerization reactor in both the nominal and robust cases.