A Machine Learning Approach for Predicting and Optimizing the Dynamic Stability of Basalt Fiber Reinforced Asphalt Concrete
Dynamic Stability (DS) is an important indicator of the rutting resistance of asphalt concrete and commonly used to evaluate high-temperature performance. However, experimental DS testing for basalt fiber reinforced asphalt concrete (BFRAC) requires laboratory work and repeated mixture preparation. This study develops a BFRAC-specific data-driven framework for predicting and optimizing DS using four tree-based machine learning algorithms combined with the Barnacles Mating Optimizer. The models were trained, tested using a database of 143 experimental samples compiled from published studies, with nine mixture-related input parameters. Cross-validation was used during model development, and Monte Carlo simulations were performed to evaluate the effect of random data partitioning on model performance. Among the evaluated models, Light Gradient Boosting (LGB) produced the highest testing performance (R2 of 0.984). The selected LGB model was then coupled with BMO for constrained mixture-design optimization, providing candidate BFRAC designs within the material and proportioning ranges represented by the database
Authors
- May Huu Nguyen (ORCID: https://orcid.org/0000-0003-3577-6602)
- Hai-Bang Ly
- Thanh-Hai Le
- Huong-Giang Thi Hoang
Institutions
- Hiroshima University (JP)
- Higashihiroshima Medical Center (JP)
- University Of Transport Technology (VN)
Publication Details
- Journal
- Canadian Journal of Civil Engineering
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1139/cjce-2026-0169
- Primary Topic
- Asphalt Pavement Performance Evaluation
- Type
- article
- Field-Weighted Citation Impact
- 0.00