An adaptive neural network model for predicting the post roughing mill temperature of steel slabs in the reheating furnace

The walking beam furnace and roughing mill of a hot strip mill were studied. A novel control method using measurement data gathered from the production line is proposed. The model uses adaptive neural networks to predict the post roughing mill temperature of steel slabs while the slabs are still in the reheating furnace. It is possible to use this prediction as a feedback value to adjust the furnace parameters for heating the steel slabs more accurately to their pre-set temperatures. More accurate heating enables savings in the heating costs and better treatments at rolling mills. The mean error of the model was 5.6 °C, which is good enough for a tentative production line implementation. For 5% of the observations the prediction error was large (>15 °C), and these errors are likely to be due to the cooling of the transfer bar following unexpected delay in entry into the roughing mill.