Comparison of empirical and neural network hot-rolling process models

Steel manufacturers are under pressure to improve their productivity levels by optimizing their process parameters to create maximum efficiency and quality levels. One of the keys to achieve this goal is the automation of the steel making process. Automation using artificial intelligence techniques applied to the hot rolling process is a potentially important steel manufacturing technique. The mathematical modelling of the process has been recognized as a desirable approach for designing mill equipment and ensuring productivity and service quality. However, such an analysis is generally very complex and time-consuming. Thus, there is considerable interest in developing simple and effective techniques to obtain accurate rolling force, torque, and slab temperatures. In this paper, a neural network (NN) is used to model the roughing stage of a plate hot rolling process. The NN approach can predict steady-state values of the force, torque, and slab temperature. The proposed NN model is compared with the classical empirical models commonly used in industrial practice. Experimental data obtained from the Ereǧli Iron and Steel Factory in Turkey was used in developing the models.