Cross-scale prediction of macroscale viscoelasticity and fatigue of asphalt mixtures from meso-scale material compositions

Cross-scale prediction enhances the efficiency and reliability of performance prediction by linking macroscale performance to meso-scale material compositions. However, existing cross-scale models calibrated under specific meso-scale conditions can only predict macroscale viscoelastic responses at the same conditions, limiting extensibility across materials and load conditions. The cross-scale prediction of macroscale viscoelastic fatigue based on meso-scale coefficients remains insufficiently explored. This study aims to address this transferability issue within the investigated domain and predict asphalt mixtures’ viscoelastic and fatigue performance at the macroscale from meso-scale material compositions. First, uniaxial compressive dynamic modulus tests were conducted on asphalt mortar to extract meso-scale viscoelastic parameters. Discrete element models of asphalt mixtures incorporating these parameters and realistic aggregate distributions reconstructed via digital image processing were established for two gradations (AC-16 and AC-25). The models accurately predicted macroscale dynamic moduli within the tested temperature–frequency range, with differences within 8% compared with experiments. Then, a discrete element fatigue model (DEFM) was developed by implementing a J-integral based Paris’ law to model particle-to-particle crack growth at the meso-scale. Using a single set of Paris’ law coefficients (A and n) calibrated at a specific loading condition, the model can predict fatigue life across various stress levels and loading regimes within the investigated domain. Results demonstrate that, for the studied mixtures, the Paris’ law coefficients A and n at the meso-scale are independent of stress level and frequency but depend on temperature and gradation: higher temperatures increase A and decrease n, while coarser gradation lowers A and raises n.