Simulative tests, such as asphalt pavement analyzer (APA), Hamberg rut tester etc. have been widely used to evaluate the performance of asphalt mixtures. However, simulative tests to evaluate the performance of the mixtures cannot give fundamental properties of Asphalt Concrete (AC) due to the complex stress and strain fields. On the other hand, due to the availability of high-performance computing systems and software, numerical techniques are gaining popularity. This dissertation presents a computational simulation method of the APA tests in order to evaluate the rutting potential of asphalt mixtures based on actual microstructure reconstructed from X-ray tomography images. In the study, the microstructure of AC is obtained through the analysis of X-ray images, which included the digital information of the microstructure for the scanned specimen. In the simulations the three phases, mastic (asphalt binder with mineral filler), aggregates, and voids are assigned with different material properties. Aggregates are modeled as an elastic material, and air voids are removed during the loading steps. The adopted two-layer model is only used to represent the rate and temperature dependent behavior of the mastics.

The parameters are obtained with inverse methods. Based on the sensitivity analysis of the parameters, an iterative procedure is performed to optimize the parameters using the experimental measurement and results of the model simulations. A parametric study is also conducted to study the effect of major parameters such as the stiffness ratio of the networks on the macro response of the model. The simulation results obtained shows a good agreement with the experimental results.

The dissertation also presents a method to measure micro strains in asphalt mixture. An automated procedure using tomography images to reconstruct three-dimensional particles is developed. The translations of the particles are obtained from the coordinate differences of particles' mass centers before and after the APA testing. The micro and macro strains in the mixture are calculated based on the particle translations. A good correlation is found between measured strains and experimental result.