Prediction of mechanical properties in graphene-reinforced aluminum nanocomposites using machine learning and molecular dynamics simulations

Abstract

Graphene-reinforced aluminum (Gr/Al) nanocomposites offer exceptional mechanical properties for aerospace, automotive, and electronics applications. Precise estimation of their characteristics, including ultimate tensile strength (UTS) and Young’s modulus (YM), remains challenging due to complex atomic interactions and computational limitations of traditional methods. This study proposes a novel machine learning framework combining Molecular Dynamics (MD) simulations, Adaptive Fast Desensitized Kalman Filter (AFDKF), Diffusion Variational Graph Neural Network (DV-GNN), and Arctic Tern Optimizer (ATO) for efficient and accurate mechanical property prediction. Important variables such as graphene alignment, volume fraction, chirality, and ambient temperature are captured by the method. DV-GNN achieves a prediction accuracy of 99.9%, significantly outperforming existing ML models. The framework also demonstrates low error rates, fast computation, and scalability, providing a robust computational tool for intelligent design of high-strength, lightweight Gr/Al nanocomposites.