AN EXTENSION OF NEWTON-TYPE ALGORITHMS FOR NONLINEAR PROCESS-CONTROL

This study extends the multistep, Newton-type formulation [Li and Biegler (1989) Chem. Eng. Res. Des., 67, 562-577] for nonlinear constrained process control. The computation of the control law is based on an augmented performance index, which improves the performance of the method over the original formulation. A number of modifications are also introduced in the computational algorithm to extend the range of problems, eliminate steady-state offsets, and extend the output predictive horizon to infinity. We show that the global stability properties of the original Newton controller are preserved by the present modifications. Moreover in the absence of active constraints, the method behaves as an extension of the linear quadratic regulator for, exhibiting therefore excellent stability properties for a large range of tuning parameters. The capabilities of the method are illustrated by application to various process examples.