The Modelling of Lateral Spread in Hot Strip Finishing Mill

This paper concerns with modelling of the lateral spread in the hot strip finishing mill. Correlation analysis is perfomed to obtain knowledge about the most significant variables affecting the lateral spread in the finishing mill. The results from correlation analysis are used to select proper input variables for modelling. Also, expert knowledge of Ruukki personnel is used throughout the study. Neural networks and linear regression models are built. In this study, some new variables have been introduced into the model and the results have improved significantly. The estimation error depends slightly on time and a simple adaptation scheme is incorporated into models to compensate for this time-dependency. The results with adapted models show about 20% better accuracy than the original results.