Nonlinear modeling and computational homogenization of asphalt concrete on the basis of XRCT scans

• We study effective mechanical properties of asphalt concrete.
• We propose a novel algorithm for asphalt concrete microstructure generation.
• We analyze X-ray Computed Tomography (XRCT) data.
• We discuss morphological and material nonlinearities.

This paper provides a methodological framework to investigate the effective mechanical properties of asphalt concrete. We, therefore, use numerical tools based on morphological X-ray Computed Tomography (XRCT) data from asphalt concrete specimens. Asphalt concrete is a multi-component material with spatially varying constituents, but in contrast to many other microstructures used in materials science, the partial microscopic material bulk properties of the constituents of asphalt concrete are accessible by physical testing and, therefore, can be considered as well investigated and known. The information gained by the XRCT is used to create artificial Statistical Volume Elements (SVEs) for our numerical investigations. We apply a discrete particle simulation to generate a densely packed sphere model with a pre-defined particle size distribution (PSD) as a first representation of the mineral filler particles. This model serves as the starting point for a weighted Voronoi diagram. Finally, the volume fractions are adjusted by a stochastic shrinkage process of the Voronoi cells. The artificial microstructures are, a priori, generated in a periodic manner and, therefore, possible boundary layer effects during computational homogenization are minimized. The SVEs are considered to be statistically similar to the real structure and serve as its best possible representation. Besides the SVE generation, this paper focuses on the constitutive description of the bituminous binding agent, which we interpret as a viscoelastic fluid. In our analysis of the results we concentrate on the upscaling properties of morphological and material nonlinearities.