Roll contribution estimation from thickness variation

In the hot rolling process, defects in roll geometry often lead to periodic thickness variations in high-temperature steel, impacting downstream process control and final product quality. This paper introduces a novel signal processing approach to identify defective rolls using thickness variation measurement after the finishing mill. Proposed method isolates imprints from all rolls and ranks their individual contributions to final quality. The method is validated through simulations and real-world data from industrial strip mill. Sensitivity of the method is validated in different operating conditions with Monte Carlo simulations. The method is shown to exhibit robustness in high signal to noise ratio conditions and with signals that comprise of multiple disturbance sources simultaneously. An analysis of two real cases with steel strips measured on an industrial strip mill are presented. First case presents detection of change in wear trend when backup roll change is performed. In the second case a work roll malfunction at a roll stand is identified using the proposed method.