Online prediction of work roll thermal expansion in a hot rolling process by a neural network

Profile and shape control is utilized to maintain the dimensional quality of a rolled strip. Prediction of work roll thermal expansion is an important element in controlling the strip profile and flatness in a modern, high-speed rolling mill. In this study, a full three-dimensional analytical model based on the finite difference method under transient condition that already was developed is used to calculate the temperature and the thermal crown of a work roll. The work roll temperature field and thermal expansion were obtained at any instance during the process. The results were verified against the actual work roll temperature data measured in Mobarakeh Steel Co. as a real model of a hot rolling process. The computation time of this model using a quad-core 2.8-GHz computer was more than 15 s. Due to the long computation time of the accurate analytical model, the online application of this model was unfeasible. Hence, the results of the analytical model were used to train a neural network. The developed artificial neural network (ANN) model was used to predict the thermal crown expansion of the work roll. The developed model is realized to be accurate enough while its computation time for rolling of each slab is measured to be less than 0.1 s, making it possible for online application.