Multi-scale performance prediction and interpretability study of asphalt mixtures based on molecular dynamics and XGBoost-Attention-MLP fusion model

2026-07-09

Zihang Xu *

Xue Han

Tao Xu *

DOI：10.1016/j.dibe.2026.100987

Original

Original request for help

Save

Abstract

En 中文

• Evolutionary characteristics of key microscopic parameters of asphalt mixture are clarified. • Synergistic effect of ternary interface notably improves the bonding energy and mechanics. • π-π stacking breaks polar aggregation of asphalt molecules and changes its diffusion mode. • Core indicators of asphalt mixture are predicted by fusion model with relative errors <2.0%. • Structure-function relationship between micro-mechanism and macro-property is quantified.

Keywords:

Asphalt mixture design

Multi-scale modeling

Graphene-modified asphalt

Machine learning fusion model

Molecular dynamics simulation

Explainable artificial intelligence

Multi-scale performance prediction and interpretability study of asphalt mixtures based on molecular dynamics and XGBoost-Attention-MLP fusion model

Abstract

Journal

Organization

Cited Papers

Citing Papers