Machine learning techniques for estimating high–temperature mechanical behavior of high strength steels

Machine learning (ML) has emerged as a powerful tool for predicting the mechanical behavior of materials by analyzing stress-strain data from tensile tests. In this study, we present a novel ML-based framework for accurately predicting the high-temperature mechanical properties of high-strength steels (HSS) using reduction factor datasets. A key contribution of this work is the development of an extensive experimental dataset through tensile testing of HSS specimens with varying thicknesses and temperatures, further expanded with over 450 additional data points covering a diverse range of material conditions. Various ML models—including Lasso Regression, Gradient Boosting, Random Forest, Extreme Gradient Boosting, Support Vector Regression, and Adaptive Boosting were rigorously evaluated to determine their predictive capabilities. The GB model achieved the highest accuracy, with an adjusted coefficient of determination (R 2 ) exceeding 0.98, outperforming conventional regression approaches. The proposed framework was validated against experimental data, demonstrating strong agreement and confirming its effectiveness in capturing complex nonlinear material behavior. This study provides a significant advancement over traditional empirical methods by offering a data-driven approach for predicting HSS performance under elevated temperatures, facilitating the design of safer and more efficient structural components in high-temperature applications. • ML predicts steel properties at high temperatures, surpassing traditional methods by capturing non-linearity. • Tensile test data and 450 reduction factor curves create a diverse dataset, enhancing ML model accuracy for steel analysis. • Gradient Boosting achieved R 2 > 0.98, outperforming other ML models in capturing non-linear trends in steel properties. • ML predictions improve reliability and safety in structural designs exposed to high temperatures, overcoming method limits. • This study shows ML's potential to refine high-temp material predictions, driving innovation in materials science & engineering.