Robust adaptive control of the mold level in the continuous casting process using multiple models

In the continuous casting of steel, mold level control is fundamental for obtaining high productivity and high quality. Using conventional methods, it is difficult to achieve both stability and performance robustness because of different classes of disturbances and parameters uncertainties in the process. This paper presents a multi-model adaptive control architecture based on the so-called RMMAC methodology. With the help of precise definition of robust performance requirements, the number of models, estimators and controllers are merely derived. More importantly, the combination of robust non-adaptive mixed-μ synthesis and stochastic hypothesis testing concepts enables controller performances prediction as well as online monitoring process parameters which could be used by operators to take corrective actions. The generated signals are likewise useful for understanding the physical phenomena in the process.