Multi-scale algorithm for controllable virtual aggregate generation in mesoscale modeling
Dong Feng
Abstract
The morphological features of aggregates fundamentally influence the physical and mechanical behavior of asphalt concrete (AC) mixtures. However, effectively capturing and controlling these features—particularly across coarse, medium, and fine morphological scales—remains an unresolved scientific challenge. This limitation has significantly hindered the advancement of high-fidelity mesoscale modeling and the development of predictive tools for AC mixture design, structural assessment, and long-term performance analysis. To bridge this gap, the key novelty of this study is the development of a scientifically rigorous and computationally efficient multi-scale algorithm capable of generating virtual aggregates with precisely controllable morphology. Coarse scale geometry is defined using ellipsoidal primitives perturbed by random noise to mimic natural variability; angularity is constructed at the medium scale through α-shape and local enhancement; and at the fine scale, surface textures are embedded using Perlin or Simplex noise. The algorithm allows independent adjustment of key morphological indices—flatness, elongation, true sphericity, angularity, and surface roughness—enabling comprehensive and scale-specific morphology control. Quantitative validation against a 3D-scanned real aggregate database confirms strong agreement, with Bhattacharyya coefficients of 0.9710, 0.9432, and 0.9499 for critical indices, underscoring the method’s high morphological fidelity. As a demonstration of practical relevance, the algorithm is applied to construct AC mixture models, and subsequent dynamic simulations verify the mechanical reliability of the generated structures. Collectively, these findings highlight the algorithm’s scientific significance as an innovative and scalable solution for virtual aggregate generation, laying a robust foundation for accurate mesoscale simulation and performance-driven design of AC materials.