Machine learning applied to predict the flow curve of steel alloys

This study aims to employ machine learning, specifically artificial neural networks (ANNs), to predict the flow curve of hot-deformed steel alloys. The method involved creating a dense ANN with two hidden layers, trained with data from 70 steel classes, including information on chemical composition, temperature, and strain rate. The results indicate robustness and good generalization capability, with a mean absolute error of 11.4 MPa and a mean squared error of 10.3 MPa. The model demonstrates an R² value of 0.98, highlighting its effectiveness in explaining variability in the data. The conclusions underscore the feasibility of ANNs in describing the mechanical behavior of steel alloys, providing an efficient and rapid tool for metal forming projects, with potential for future research and innovations in this field.