Gaussian Process Regression for Constitutive Modeling of Austenitic Stainless Steel Under Various Strain Rates and Temperatures

This study presents an advanced framework for developing a mechanical behavior model of a structural material based on experimental data. In particular, mechanical tensile tests on cylindrical specimens of A2-70 austenitic stainless steel were conducted at different temperatures at quasi-static, intermediate, and high strain rates making use of an electromechanical testing machine, a servo-hydraulic machine, and a direct-tension split Hopkinson tension bar. The neuronal Gaussian process regression (GPR) algorithm based on the experimental strain, strain rate, and temperature of each test is proposed to predict the constitutive stress–strain behavior of austenitic stainless steel at hand, with the stress data used as target. The adopted GPR framework demonstrated considerable accuracy in reproducing the experimental results, proving to be a reliable tool for modeling the tensile mechanical behavior of structural materials.