A physics-informed neural network for predicting fatigue life of asphalt mixtures using the superpave indirect tensile test
Durability is crucial for asphalt pavement design, which necessitates an accurate prediction of the fatigue life of asphalt mixtures under complex environmental and loading conditions. A prevailing fatigue life prediction method, developed based on Hot Mix Asphalt fracture mechanics (HMA-FM), relies on a healing curve to characterise fatigue-healing properties that are highly sensitive to environmental factors. This study extends the HMA-FM framework by integrating a physics-informed neural network (PINN) to model the nonlinear dependencies of Indirect Tensile (IDT) test parameters on environmental factors. A novel PINN fatigue-healing (PINN-FH) model is proposed to predict the fatigue life of IDT samples. The key dataset for training and validating the PINN-FH model was quantified via Superpave IDT tests that control dynamic water exposure, aging duration, and temperature. Results demonstrate that the PINN-FH model accurately captures environmental effects on IDT fatigue life. Cross-dataset validation confirms the PINN-FH model’s generalisability.
Authors
- Zhiqiang Shu (ORCID: https://orcid.org/0009-0008-5759-4782)
- Lijun Sun (ORCID: https://orcid.org/0009-0008-6195-8622)
- Haoran Li (ORCID: https://orcid.org/0000-0003-4679-1322)
- Yu Yan
Institutions
- Tongji University (CN)
- Massachusetts Institute of Technology (US)
Publication Details
- Journal
- Road Materials and Pavement Design
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1080/14680629.2026.2761039
- Primary Topic
- Asphalt Pavement Performance Evaluation
- Type
- article
- Field-Weighted Citation Impact
- 0.00