Parameter Estimation for Online Calculation Model of Hot-rolled Strip Crown

Abstract

Online crown calculations are essential for effective shape control of hot-rolled strip steel, the accuracy of which is often limited by modeling assumptions and the complexity of operating conditions. Model parameters significantly influence calculation accuracy, requiring process engineers to make regular adjustments based on their experience. As customization, small batch sizes, and multiple specifications become mainstream in production and supply modes, reliance on manual experience is increasingly insufficient. This paper aims to enhance the accuracy of online crown calculations for hot-rolled strip steel. Initially, based on the mechanism model incorporating gain coefficients and formula derivation, a relational expression between the gain coefficients for each stand in the tandem rolling mill group and the final crown is established, transforming the discrete issue of correcting influential factors into a unified problem of solving gain coefficients, thus improving correction efficiency. To ensure the reliability of the estimated results, three forms of constructing the constraint equation are compared: no constraint, limiting the proportional crown difference, and proportional exit crown. Utilizing rolling production data, two parameter estimation methods—normal linear regression and robust regression—are explored to determine the gain coefficients. Results indicate that combining the constraint based on proportional exit crown and robust regression will provide more reliable gain coefficients and significantly enhances the accuracy of online crown calculations, improving precision by over 50% compared with the model before correction.