Combined first-principles calculation and machine learning for the strength of segregated grain boundaries in aluminum

Grain boundary (GB) segregation plays a critical role in determining the structural stability and mechanical properties of nanocrystalline aluminum alloys. In this study, first-principles calculations combined with interpretable machine learning were carried out to investigate the segregation behavior of solute elements and their effect on GB strength. A dataset containing five GBs and 52 solute elements was constructed for the prediction of GB strength, incorporating descriptors related to GB structure, solute–matrix interaction, and the intrinsic solute properties. Machine learning with leave-one-GB-out validation achieves robust strengthening/weakening classification and quantitative regression using a low-cost feature group dominated by tabulated solute properties.