Modeling and Adaptive Feedforward Control in Continuous Casting
Sprache des Titels:
This thesis addresses the mold level control problem in continuous casting of steel, and contributes a novel control strategy to improve the quality and quantity of the final products.
In fact, in many continuous casting plants a disturbance effect called ‘dynamic bulging’ appears in particular for steel grades with medium carbon content (peritectic steel grades) and for higher casting velocities. Dynamic bulging causes large, almost periodic disturbances of the mold level, which significantly reduce the quality of the final products and can even cause instability that leads to an interruption of the continuous casting process. In some cases lowering the casting velocity may help, although it will decrease the productivity. However in general conventional approaches are not sufficient and tend to fail due to the many unknown parameters and the large time delays of the plant.
The aim of this thesis was to develop a new control method that allows a good suppression of mold level oscillations due to dynamic bulging without slowing down the process, thus achieving an increase in quality, casting speed and throughput. This was accomplished using two elements: a precise analysis of the disturbances and the use of a novel control structure.
Based on this analysis a model of the plant including the disturbance effects, in particular those due to dynamic bulging, has been developed. Besides the mold level, the model includes the roll motor currents of the strand as an additional important source of information on the dynamic bulging state. This additional information has proven to play an essential role in the estimation of the disturbances and has then been used in a new feedforward predictive control scheme added to the standard feedback controller.
For the prediction of the disturbances, a parameter varying internal model observer for frequency varying periodic oscillations has been developed, which partly relies on the casting speed, a known parameter.