Key points are not available for this paper at this time.
Mesostructure-based simulation technology offers an effective approach to modeling the mesomechanical responses of asphalt concrete and studying its mesomechanisms. However, current exploration depth and utilization level of the simulation data are both very limited. The intricacies of the vast mesomechanical response data necessitate the development and adoption of advanced data-processing methods that can gain valuable insights. To this end, this study proposed a data-mining framework that integrates a three-dimensional (3D) random aggregate method and the finite-element (FE) method for mesoscopic simulation of asphalt concrete. This framework mainly consists of two stages. In the first stage, a random aggregate method is used to establish 3D mesostructures of asphalt concrete, and FE simulations are performed on these mesostructures to model their mesomechanical responses. Based on the simulated responses, several procedures for data export, processing, and mining are developed and sequentially performed in the second stage. The data-export procedure extracts the mesomechanical responses from the FE result files and stores them into a database for efficient management. The required response data are accessed via SQL and then processed preliminarily using statistical and computational methods. Machine-learning methods and variable importance analysis techniques are used to conduct in-depth mining analyses on the processed data. The efficacy of this framework was demonstrated by applying it to investigating the effects of the volume fractions of coarse aggregates (larger than 2.36 mm) of different sizes on 129 simulated mesostructures of asphalt concrete with a nominal maximum aggregate size of 13.2 mm. The results indicated that the coarse aggregates with the particle sizes between 4.75 and 9.5 mm may make the highest contribution to the overall dynamic modulus of asphalt concrete. The coarse aggregates with larger size differences could cause more significant stress concentrations in the surrounding asphalt mortar.
Wei et al. (2024) studied this question.