This paper describes the design and simulation with real time data of a Model Predictive Controller (MPC) with a dynamic model for a reheating furnace in a Hot Strip Mill. In the new design, the temperature at the end of a zone is defined as target. A predictive model predicts the expected temperature for all slabs. The differences between the predictions and targets are the inputs for the zone temperature controller. The predictor is chosen to be a so-called ARX model, which has been identified by data from a test slab, which went through the furnace with thermocouples inside. Simulations were executed with actual data from a full day of operation, using a Furnace Response Model, which has been identified by the same data. Finally the improvement is discussed using performance criteria defined by the mill operational requirements.
展开▼