Air void effect on an idealised asphalt mixture using two-dimensional and three-dimensional discrete element modelling approach

Abstract

In this study, an idealised asphalt mixture was modelled with the discrete element method for both two-dimensional (2D) and three-dimensional (3D) cases. The air voids were randomly generated and counted within the models to reach a specific air void level (e.g. 6%). The 2D models were used to compute the strain and stress responses when the specimens were subjected to a compressive load. Then, the moduli of the specimens were computed from the stress–strain curves. The 3D idealised model was generated using a number of layered 2D models. The air void distribution patterns were also studied with 2D and 3D randomly generated models at specific air void levels. The results showed that the modulus deviation increases when the air void level increases. In addition, the modulus deviations of the 3D models were found to be much lower than those of the 2D models. When comparing the modulus predictions from the 2D models with those from the 3D models, the research proved that the 3D models yielded higher moduli than the 2D models. The average of the predicted modulus difference between 2D and 3D models was 26% at 10% air voids, and 7% at 4% air voids. When the air void increased from 0 to 10%, the modulus decreased by 30% in the 3D models, when compared with 48% in the 2D models. The 2D and 3D models predicted the same modulus for 0% of air voids. However, the 2D models under-predicted the mixture modulus, especially when the air void level was higher. In the 2D modelling of the asphalt mixtures, a large number of models were needed to achieve a reasonable prediction due to larger deviation, even at lower air void levels. At higher air void levels, the 3D models yielded a much higher prediction than 2D simulations.

Keywords:

• micromechanical model
• asphalt mixture
• discrete element
• three-dimensional model
• two-dimensional model
• air void

Acknowledgements

This study was partly based on the work supported by the National Science Foundation under grant CMMI-0701264. Any opinions, findings and conclusions or recommendations expressed in this study are those of the authors' and do not necessarily reflect the views of the National Science Foundation. This study was also partially supported by the State of Michigan Research Excellence Fund.

Notes

^1. [email protected]

^2. [email protected]

Additional information

Notes on contributors

Sanjeev Adhikari

1. 1. [email protected]

Qingli Dai

2. 2. [email protected]