Constrained Model Predictive Control of a Continuous Annealing Furnace

Abstract

For the design of constrained model predictive control of a continuous annealing furnace both the modeling of an industrial furnace and the structure of an appropriate constrained predictive control scheme are presented. The model of the furnace comprises a partitioning of the furnace into suitable heating zones, a description of the spatially discretized moving steel strip, and a dynamic formulation of the heaters. The final process model is of linear state space form and utilizes the energy flow to the heaters as inputs, the ambient temperatures as disturbances, and the temperatures of the strip elements and the heaters as states. The model can be easily adapted to varying operation conditions since all important design parameters of the furnace and the strip can be modified on-line. Utilizing measurement data from an industrial furnace for the model validation shows excellent agreement. The constrained predictive control scheme explicitly considers the disturbance, allows for the inclusion of reference values for the inputs, and simultaneously guarantees constraints on the maximum temperature of the heaters. This structure is shown in simulations to reliably obey multiple and contradictory technological restrictions of the process without the need for lengthy parameter optimizations.