Leveraging molecular dynamics and machine learning to predict impact performance in polycrystalline magnesium alloys

GY Chen and XY Liu and Y Zhang and D Lin and PL Mao, SOLID STATE COMMUNICATIONS, 403, 115961 (2025).

DOI: 10.1016/j.ssc.2025.115961

There is currently a lack of research on the effects of grain spatial orientation distribution and shape characteristics on the mechanical properties of nanocrystalline magnesium alloys. To provide a means of studying such problems, this paper employs molecular dynamics simulations to construct a dataset that incorporates the spatial distribution and orientation features of grains within the model. Using 12 different machine learning methods, In this study predict the material's high-velocity impact response and analyze the predictive performance of various machine learning algorithms on this dataset. Additionally, through feature selection and segmented training sets, In this study demonstrate the capability of machine learning methods to perceive grain characteristics such as spatial distribution in this dataset. This validates the feasibility and effectiveness of applying machine learning methods to study such data. Furthermore, In this study offer recommendations for employing machine learning techniques in conjunction with datasets that include grain spatial distribution and orientation characteristics. By analyzing molecular dynamics datasets, In this study also predict the high-velocity impact response of over a thousand magnesium alloy compositions, shedding light on the mechanical properties of magnesium alloys under high-velocity impact.

Return to Publications page