Influence of carbon content on the hydrogen resistance of steel based on first-principles and machine learning
The hydrogen embrittlement issue in ferrite steels has always been a hot topic of concern for researchers,particularly the unclear influence of different carbon contents on the hydrogen resistance of iron and steel materials.A high-precision machine learning force field(MLFF)for the iron-carbon-hydrogen system was constructed by com-bining first-principles calculations with machine learning algorithms.Molecular dynamics simulations were per-formed to investigate the diffusion behavior of hydrogen atoms in steels with different carbon contents.The high-precision MLFF was trained using a neural network(NN)model based on first-principles molecular dynamics(AIMD)results of multiple configurations.Various tests were conducted to ensure that the machine learning force field could accurately describe the statistical and dynamic properties of the iron-carbon-hydrogen system.Using this MLFF,molecular dynamics simulations were performed on ferrite steels with different carbon contents,and the hy-drogen diffusion coefficients were calculated.It was found that the hydrogen diffusion coefficient generally decreased with increasing carbon content,in good agreement with experimental results.The algorithm model established in this study can analyze the influence of carbon content on the hydrogen resistance of iron and steel materials,which is of significant importance for studying hydrogen-induced damage in steel materials and composition design.
first-principlesmachine learningmolecular dynamics force fieldhydrogen diffusionhydrogen-induced damage
NETL
NSTL
万方数据
