Virtual rolling quality system for cold rolling

A prototype of virtual rolling quality system for a 4 stands cold rolling process is created. In this system virtual sensors and models are used to calculate quality features based on process information and analyzed history data. The virtual-sensor technology is developed for estimating of hard-to-measure variables and further on to define methods to automatically detect and hierarchically classify deviations in rolling values and strip quality related to thickness, flatness, profile, etc. The analysis of the process data is done by means of classical rolling theories, statistical methods and databased methods such as artificial neural networks to identify and quantify the main effects on strip quality. The flatness and hot strip thickness profiles are decomposed into independent components by means of Chebyshev polynomia coefficients. The strip quality related measurements are also collected to the appropriate extent from up- and downstream processing lines.