Prediction of Mechanical Properties of Low Carbon Steel in Hot Rolling Process Using Artificial Neural Network Model

Abstract

This work deals with the prediction of mechanical properties of hot rolled steel slab in the hot rolling mill to avoid the manual working of preparing tension test samples in the mechanical testing lab. The time consumption for testing is avoided and the cost of product is decreased.A model for predicting mechanical properties of low carbon steel has been developed and Feed Forward Back Propagation (FFBP) as one type of algorithm of the Artificial Neural Network has been applied to the prediction system. Yield strength (YS), Ultimate tensile strength (UTS) and Elongation(EL) are the basic mechanical properties of low carbon steel are predicted as a function of thermo-mechanical process parameters. These properties mainly depend on the input parameters such as Dispatch Temperature (DISTEMP), Transfer-Bar/Rolling Temperature (TBART), Finishing Temperature (FINT), Coiling Temperature (COILT) and Carbon Equivalent (CEQ). The FFBP is a supervised system that requires a lot of input and output data pairs for training process. The data are acquired from Indian Public Sector Steel Company and preprocessed before training.Performance of the model is evaluated by the Normalized Root Mean Square Error (NRMSE) and the Coefficient of Correlation (R). The NRMSE and the R values of both training and validation parts show excellent values. Therefore, the model using the FFBP algorithm is appropriate to predict the mechanical properties of the hot rolled low carbon steel.