摘要：AbstractAlthough great improvements have been made in cold rolling over recent years, roughness and related properties such as the strip’s tribology and paintability have not been mastered yet. In order to obtain a predefined surface quality, a layered control loop is implemented in context of a cold rolling mill to track a given roughness reference. Feedback is provided by a non-contact roughness sensor, which is used for online estimation of the mill’s imprint model. In this context, a grid sort algorithm and Gaussian process regression are combined to estimate the nonlinear relationship between the measured roughness and the specific rolling force. The derived model is then used in conjunction with a nonlinear stand characteristic and cold rolling model to control the surface roughness. Here, a model-based controller is used which allows to track a desired roughness reference. An experimental test series is presented suggesting high tracking performance of the roughness controller. However, the test data further indicates that the strip thickness should be considered as an additional control variable. Subsequently, the potential of using the strip tension as a complementary actuator to the actuated roll gap is investigated, examining the feasibility of a combined strip thickness and roughness control.
关键词：KeywordsControloptimizationProcess modelingCold rollingRoughness controlNonlinear system identificationReal-time control
