Researchers Use AI to Enhance Asphalt Durability with Graphene Oxide

Researchers at the University of Transport Technology in Vietnam have developed a novel approach to predict the viscoelastic characteristics of asphalt modified with graphene oxide (GO). This advancement is crucial as traditional asphalt pavements struggle to meet the performance demands posed by increasing traffic loads and the adverse effects of climate change. The study, titled “Application of Extreme Gradient Boosting in Predicting the Viscoelastic Characteristics of Graphene Oxide Modified Asphalt at Medium and High Temperatures,” utilizes artificial intelligence to streamline the assessment of asphalt properties.

Conventional methods for determining key viscoelastic indicators—complex modulus (G*) and phase angle (φ)—in GO-modified asphalt are often complex, expensive, and time-consuming. To address these challenges, the research team constructed an extreme gradient boosting (XGB) model aimed at predicting G* and φ at medium and high temperatures. The model draws on two datasets from previously published experiments, comprising 357 samples for G* and 339 samples for φ, incorporating nine input parameters representing various asphalt binder components.

Model Performance and Comparative Analysis

The results indicate that the XGB model demonstrates exceptional predictive accuracy for both G* and φ. Evaluated using the coefficient of determination (R2) and the root mean square error (RMSE), the model achieved R2 values of 0.990 for G* and 0.9903 for φ, with corresponding RMSE values of 31.499 and 1.08. This level of performance significantly exceeds that of five other machine learning models, including artificial neural networks and decision trees, reaffirming the XGB model’s superior accuracy.

In addition to predictive performance, the research included an analysis using Shapley additive explanations (SHAP) values to evaluate the impact of input parameters on the viscoelastic properties of asphalt. This analysis identified the initial properties of the asphalt, GO content, and mixing temperature as the most critical factors influencing the material’s performance.

Implications for Asphalt Engineering

The implications of this research extend beyond theoretical advancements. By leveraging machine learning techniques, the study paves the way for more efficient and cost-effective methods of evaluating asphalt modified with graphene oxide, which is known for its potential to enhance mechanical properties. As the demand for resilient infrastructure grows, this innovative approach could play a vital role in developing pavements that withstand the challenges posed by increasing traffic and environmental factors.

The full study, authored by Huong-Giang Thi HOANG, Hai-Van Thi MAI, Hoang Long NGUYEN, and Hai-Bang LY, is available for further reading at: https://doi.org/10.1007/s11709-024-1025-y.

This research not only enhances the understanding of GO-modified asphalt but also highlights the potential of machine learning in materials science, promising a more resilient future for road infrastructure worldwide.