Neural Network Modeling of a Plate Hot-Rolling Process and Comparision with the Conventional Techniques

Abstract

The force, torque and slab temperature models of each pass in the plate hot-rolling process are established in this paper. In the first, two different experimental models of a plate hot-rolling are represented. The structures of these models are in the two different forms of the neural network and predict the steady-state values of force, torque and slab temperature. First of these algorithms is the classic back-propagation algorithm (CBA) and the second is the fast back-propagation algorithm (FBA). In the second part, the proposed neural networks models are compared with the classical empiric models commonly used in the rolling practice and each other. The experimental data obtained from Ercgli Iron and Steel Factory in Turkey was used for developing the models.