.png){kind=icon}
Return
A data-driven study on asphalt pavement roughness evolution using full-scale test track data
.png){kind=icon}
OA
AI
DOI：10.1016/j.ijtst.2026.03.001.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The roughness of asphalt pavements directly affects ride comfort and structural performance, making its long-term prediction essential for pavement management. This study investigates the evolution of international roughness index (IRI) across seven typical asphalt pavement structures using full-scale circular track data from RIOHTRACK. Statistical analysis reveals three characteristic deterioration patterns: stable (semi-rigid and rigid with thin pavements), cyclical fluctuation (thick and full-depth asphalt pavements), and nonlinear growth (semi-rigid base with 4 cm, 6 cm, 8 cm surfaces and inverted-base structures). Pavements with a strong-base–thin-asphalt-surface or medium-thickness configuration exhibit the most stable roughness development, whereas nonlinear growth pavements show accelerated deterioration beyond 40–50 million axle loads. Correlation analysis identifies cumulative axle load as the dominant factor in IRI progression, with temperature exerting limited or structure-specific effects. To capture these nonlinear and coupled influences, a hybrid BO-CNN-LSTM model was developed, integrating Bayesian optimization for hyperparameter tuning. The model achieved superior accuracy (root mean squared error: 0.011–0.029, mean absolute percentage error: 0.6%–1.8%) compared with traditional regression approaches, particularly for semi-rigid and thick asphalt concrete pavements. These findings provide both mechanistic insights and a data-driven tool to support proactive pavement maintenance and long-term performance optimization.
Keywords:
Asphalt pavement
Roughness evolution
Roughness prediction
Data-driven modeling
Journal
IF:
4.8
Papers:
1.5K
Citations:
1.5K
Organization
Cited Papers
No cited papers available