A Multi-Input Multi-Output Neural Network Model To Characterize Mechanical Properties Of Strip Rolled High Strength Low Alloy (HSLA) Steel

Abstract

High-strength low-alloy (HSLA) steels provide increased strength-to-weight ratios over conventional low-carbon steels. Because HSLA alloys are stronger, they can be used in thinner sections, making them particularly attractive for transportation-equipment components where weight reduction is important. Application of artificial neural networks (ANN) for predicting key mechanical properties of hot rolled steels has been increasingly used for process automation. A multi-input-multi-output (MIMO) neural networks model has been developed to predict the mechanical properties, namely, yield strength, ultimate tensile strength and percentage of elongation as a function of thermo-mechanical control processing parameters for strip rolling of HSLA steels. The input parameters in the ANN model are steel chemistry, ferrite grain size, cooling rate, rolling temperature, finish roll temperature and coiling temperature. Broyden-Fletcher-Goldfarb-Shanno (BFGS) network optimization algorithms have been incorporated to minimize the error for improved predictive capability. The model predictions of mechanical properties are found to be in excellent agreement with the measured values.