Multiscale approach for modeling hot mix asphalt

Hot mix asphalt (HMA) is a granular composite material stabilized by the presence of asphalt binder. The behavior of HMA is highly influenced by the microstructure distribution in terms of the different particle sizes present in the mix, the directional distribution of particles, the distribution of voids, and the nucleation and propagation of cracks. Conventional continuum modeling of HMA lacks the ability to explicitly account for the effect of microstructure distribution features. This study presents the development of elastic and visco-plastic models that account for important aspects of the microstructure distribution in modeling the macroscopic behavior of HMA. In the first part of this study, an approach is developed to introduce a length scale to the elasticity constitutive relationship in order to capture the influence of particle sizes on HMA response. The model is implemented in finite element (FE) analysis and used to analyze the microstructure response and predict the macroscopic properties of HMA. Each point in the microstructure is assigned effective local properties which are calculated using an analytical micromechanical model that captures the influence of percent of particles on the microscopic response of HMA. The moving window technique and autocorrelation function are used to determine the microstructure characteristic length scales that are usedin strain gradient elasticity. A number of asphalt mixes with different aggregate types and size distributions are analyzed in this paper. In the second part of this study, an elasto-visco-plastic continuum model is developed to predict HMA response and performance. The model incorporates a Drucker- Prager yield surface that is modified to capture the influence of stress path direction on the material response. Parameters that reflect the directional distribution of aggregates and damage density in the microstructure are included in the model. The elasto-visco-plastic model is converted into a numerical formulation and is implemented in FE analysis using a user-defined material subroutine (UMAT). A fully implicit algorithm in time-step control is used to enhance the efficiency of the FE analysis. The FE model used in this study simulates experimental data and pavement section