Online mechanical property prediction and automatic product release for hot rolled steel strip products using artificial neural network models.

This article focuses on the development, integration, and industrial deployment of artificial neural network (ANN) models to predict mechanical properties—yield strength (Rp0.2), ultimate tensile strength (UTS), and elongation (A80)—of hot rolled strip steel products manufactured by Tata Steel UK. Two steel grades were studied: DD11, a low carbon steel for cold forming, and S355MC, a high strength low alloy (HSLA) steel. The ANN models use 17 input variables including chemical composition and hot rolling parameters, and after six months of implementation at the Port Talbot hot mill, predicted mechanical properties closely matched physical tensile test results within product specifications. The study also proposes product release criteria based on input data distribution, temperature uniformity, and prediction confidence to ensure reliable use of ANN predictions in production, highlighting that prediction accuracy decreases for inputs outside the training data range and emphasizing the need to improve model extrapolation capabilities for broader industrial application.