Aggregate occupies at least three-quarters of the volume of asphalt mixture and can significantly affect the performance of pavement. The geometrical morphology influences the slippage and interlock among aggregates for resisting and distributing applied loads. In recent years, the discrete-element method (DEM) has been employed for simulation of asphalt mixture structure. This paper introduces an approach for simulation of aggregate and asphalt mixtures using parameterized shape and size gradation. Both the plane geometry factor (PGF) and the section aspect ratio (SAR) were employed to describe the three-dimensional (3D) geometric characteristics of aggregates. A numerical technique of aggregate models was implemented with probabilistic parameters depending on statistical results of PGFs and SARs. The 3D numerical model of asphalt mixtures was assembled with three different components, and was validated by uniaxial compression tests via comparison with the laboratory result. It was found that the PGF and SAR are appropriate to describe the three-dimensional features of aggregate shapes, because a simplified space object can be described by a two-dimensional (2D) graphical projection and a vector scalar corresponding to the space vector. Probability distribution curves of PGFs and SARs between coarse aggregates were in concordance with the Gauss-type function, because their correlation coefficients were all greater than 95%. It was verified that the developed clumping algorithm of aggregates was reasonable in terms of the shape and size gradation. Based on the parallel-bond model and Burger’s model, the results of virtual tests were in good agreement with those of laboratory uniaxial tests. The angularity (PGF) of aggregates has a beneficial effect on the strength and stability of asphalt mixtures, whereas the flat-elongated feature (SAR) has a negative effect on the strength and stability of asphalt mixtures.

The cumulative permanent strain of subgrade soils plays an important role in the pavement structure design and long-term pavement performance maintenance. In Southern China, the hot and humid climate increases the risk of becoming a saturated condition for subgrade soils. The saturated subgrade soils will generate significant cumulative permanent strain under long-term cyclic loading compared with unsaturated soils. This study investigated the impact of the moisture content, number of cyclic loading, deviatoric stress and confining pressure on the evolution of the cumulative permanent strain for subgrade soils by using long-term cyclic (10,000 cycles) triaxial tests. The test results indicated that the cumulative axial permanent strain increase with the increasing moisture content, as well as the increasing number of cyclic loading and deviatoric stress but decreases with the increasing confining pressure. In order to predict the permanent deformation of subgrade soils in Southern China, a new model for predicting the permanent deformation of subgrade soil, which contains number of cyclic loading, initial bulk stress, octahedral shear stress and saturation, was proposed and validated.

This study aims at developing an efficient and accurate methodology to estimate the resilient modulus of subgrade soils. First, a new resilient modulus model incorporating stress dependence and moisture dependence was proposed. Second, prediction models were developed to conveniently and accurately determine model parameters of SWCC and resilient modulus model. In order to characterise the moisture dependence of subgrade soils, the matric suction was added into the proposed model. The matric suction was measured by the pressure plate test and the soil-water characteristic curve (SWCC) was used to determine the matric suction value at any given moisture contents. In order to develop prediction models for model parameters of SWCC and resilient modulus model, the laboratory experiments and multiple regression analysis were conducted on 22 soil samples. A series of performance-related soil properties were measured and used to develop the coefficients prediction models. The developed coefficients prediction models using the performance-related soil properties have high R-squared values and were validated by comparing the measured and predicted values of resilient modulus. Therefore, when the basic physical properties of soils were obtained, the resilient modulus can be predicted for the subgrade soils at any given matric suctions and stress states.