When is constrained control necessary for large scale processes?

Abstract

It is of significant practical interest to identify which processes will benefit from the use of constrained consuggested that explicit constraint-handling may not be necessary for many large scale sheet and film processes provided that the controller is designed to be robust to mode1 trol algorithms such as model predictive control, and which will not. Explicit conditions are derived that identify whether constraint-handling is needed for a particular process. The conditions can be computed directly from a transfer function model, a simulation model (ordinary/partial differential equations), or experimental input-output data. The formulation considers the effects of measurement noise, process disturbances, model uncertainties, plant directionality, and the quantity of experimental data.It is of significant practical interest to identify which processes will benefit from the use of constrained control algorithms such as model predictive control, and which will not. This is especially important in the control of large scale systems, where the computations associated with the implementation of constrained control techniques such as model predictive control may not be feasible without an expensive upgrade of the existing control hardware. This is also of interest for application to high speed control systems such as controlling the idle speed in automobiles, where there is sigmficant pressure to minimize control hardware costs. In such applications, model predictive control should only be used if significant performance improvements can be achieved.