High Precision Thickness Control in a Cold Rolling Mill using a Non-Linear Roll Stand Deflection Model

Re-rolling is considered the last processing step in the cold rolling process, which has a substantial influence on the surface quality and the associated properties of the surface. In the case of steel materials, a height reduction of less than 5% is selected to avoid unnecessarily restricting the formability for subsequent forming operations. For this reason, low forces act in this stage of the process, which is non-linearly related to the deflection of the roll stand. In this paper, a nonlinear roll stand deflection model is identified by combining the Iteratively Reweighted Least Squares algorithm (IRLS) and the Gaussian Process Regression (GPR). This approach is designed to achieve a self-calibration of the roll gap, reduce the sensitivity to measurement errors and enable a highly accurate estimation of the stand deflection. This model is then used in a model-based control of the strip thickness that adjusts the roll gap and is evaluated in an experiment by providing a trajectory of the desired strip thickness. The experimental results show a high-precision tracking performance with a tolerance of 4 mu m.