Development of an Image-Based Multi-Scale Finite Element Approach to Predict Mechanical Response of Asphalt Mixtures

摘要

Image-based simulation of complex materials is a very important tool for understanding their mechanical behavior and an effective tool for successful design of composite materials. Asphalt concrete, as one of these multi-phase complex materials, is a composite of asphalt binder, air voids, and mineral aggregate particles. Simulation of asphalt concrete with numerical methods is not a new topic but is faced with many challenges. In addition to requiring tremendous computational cost, it is not clear yet how to effectively model the aggregate-to-aggregate contact behavior during loading and deformation. In this paper an image-based, multi-scale modeling is introduced to reduce the computational cost significantly by reducing the amount of elements in the numerical model. In this approach the "up-scaling" and homogenization of each scale to the next is critically designed to improve accuracy. In addition to this multi-scale efficiency, this study introduces an approach for consideration of particle contacts at each of the scales in which mineral particles exist. The FE analysis is performed in the study at four scales, asphalt binder, mastic, mortar, and asphalt mixture scale. The well-known finite element (FE) software ABAQUS is used to conduct FE simulations at all scales. The inputs are based on the experimentally derived measurements for the binder viscoelastic properties from which the binder constitutive model is implemented into the software via the user material subroutine (UMAT). For the scales of mastic and mortar, the artificially 2-dimensional (2D) images of mastic and mortar scales were generated and used to characterize the properties of those scales. Finally, the 2D scanned images of asphalt mixtures after elimination of fine aggregate particles is used to study the asphalt mixture behavior under uniaxial creep and recovery loading. Comparison between experimental results and the results from the model shows that the model developed in this study is capable of predicting the effect of asphalt binder properties and aggregate micro-structure on mechanical behavior of asphalt concrete under repeated creep loading.