Physics-Informed Machine Learning for Predicting Buckling Performance and Damage Evolution in Graphene Reinforced Basalt/Epoxy Composites

Basalt fiber-reinforced polymer composites are attractive for lightweight structural applications due to their high specific strength, thermal stability, and environmental sustainability. Their performance is limited by weak fiber–matrix interfaces, premature damage initiation, and insufficient compressive stability. This study presents an integrated experimental–computational framework to improve the structural performance of basalt/epoxy composites through dual-phase reduced graphene oxide (rGO) functionalization. Basalt fibers were coated with rGO using electrophoretic deposition, while rGO was simultaneously incorporated into the epoxy matrix through a hybrid ultrasonication–high-shear dispersion process. Mechanical characterization demonstrated that dual-phase functionalization significantly enhanced tensile, compressive, interlaminar shear, and fracture properties by improving interfacial bonding and matrix stiffness. The experimentally optimized composite was incorporated into a multiscale modeling framework in which representative volume element (RVE)-based homogenization was used to determine effective orthotropic properties for nonlinear finite element analysis. Progressive damage behavior under compressive loading was evaluated using Hashin failure criteria coupled with continuum damage mechanics. A comprehensive dataset comprising 12,852 laminate configurations was generated by varying geometric parameters and stacking sequences and was subsequently used to develop machine-learning surrogate models for prediction of critical buckling loads. An ensemble model combining linear regression and support vector regression achieved a coefficient of determination of 0.973. SHAP-based interpretation identified laminate length and width as the dominant parameters governing buckling resistance. Progressive damage analysis further revealed that dual-phase rGO functionalization delays damage initiation and promotes stable load redistribution. The proposed framework provides an efficient pathway for the design and optimization of lightweight, damage-tolerant composite structures.