ARTIFICIAL NEURAL NETWORK MODEL FOR PREDICTION OF MECHANICAL PROPERTIES OF STEEL

Abstract

The prediction of mechanical properties is very important task in the quality control of steel, which is one of the world's most important engineering and construction materials. To the main factors affecting the mechanical properties of steel belongs thermomechanical treatment and the chemical composition. The chemical composition may vary for a special high-strength low-alloy steel to meet certain mechanical property requirements. This paper presents a machine learning approach for prediction of tensile strength of structural steels. A single hidden layer Artificial Neural Network (ANN) model was developed and validated based on industrial data set. The ANN model inputs were parameters of interatomic interaction, characterizing the chemical and structural state of the melt. The modeling results showed that the ANN model might be an attractive and useful tool for forecasting the mechanical properties of steel. It offers also a potential for the further derivative free optimization of chemical composition for special quality criteria.