Image-Based Finite Element Specimen Replication for Quantifying Microstructure-Driven Variability in Asphalt Concrete

Abstract This study introduces an image-based finite element specimen replication framework to quantify the impact of microstructural arrangement on the variability of asphalt concrete response. The method integrates vectorised, image-derived geometry reconstruction, two-phase finite element modelling, and controlled numerical redistribution of aggregate particles. High-resolution scanned specimen images were processed to extract aggregate boundaries and generate explicit aggregate-matrix geometries. Coarse aggregates were modelled as linear elastic inclusions, while the surrounding binder-mastic matrix was represented using linear viscoelastic Prony series parameters obtained from dynamic shear rheometer measurements. Synthetic specimens were generated by randomly redistributing aggregate particles while preserving their original shapes and overall gradation, allowing the effect of microstructural arrangement to be isolated under controlled numerical conditions. The framework was evaluated using resilient modulus and repeated creep tests for two asphalt mixtures with different aggregate and binder systems. The simulations closely reproduced the average resilient modulus response and captured the general trend of repeated creep deformation, particularly during the early and intermediate loading stages. The coefficient-of-variation-based assessment showed that the numerical specimen replication framework accounted for approximately 78-79% of the laboratory variability in resilient modulus, but only about 13-18% of the variability in the repeated creep slope parameter. These results indicate that stiffness variability is strongly influenced by aggregate arrangement and internal mixture structure, whereas repeated creep variability is governed mainly by testingrelated factors and unmodelled damage or viscoplastic mechanisms. The proposed framework provides a computational procedure to distinguish microstructure-driven variability from external laboratory scatter in heterogeneous asphalt concrete.