Predicting Dynamic Modulus of Asphalt Mixtures Using Explainable Artificial Intelligence Techniques: A Design Optimization Application
Abstract
Dynamic Modulus (DM) is a critical characteristic for the design of Hot Mix Asphalt (HMA), which is currently determined through complex empirical models. This study applies advanced explainable Artificial Intelligence (AI) techniques to predict DM, enhance the understanding of the factors affecting it, and support design optimization applications. An extensive experimental dataset was used that included a variety of aggregate gradations, different types of bituminous binders, volumetric characteristics, and test conditions in terms of test temperature and loading rate. Eight different predictive models were developed, ranging from simple linear regression to advanced machine learning models, including Support Vector Machines (SVM) and ensemble techniques, using a random selection scheme for cross-validation. Comparison of model performance indicated that the best performing model was the Cubic SVM, with an R2 = 0.96 and RMSE = 2.22 × 105 psi, representing an approximate 39% reduction in error compared to the baseline linear regression. Explainable AI analysis revealed that phase angle was the most significant parameter in predicting DM, followed by aggregate gradation and binder characteristics. Overall, the findings show that the DM of HMA is influenced by complex nonlinear interactions among multiple factors. The AI-based models provided higher predictive accuracy and insights into interactions among material properties, supporting the development of optimized asphalt mix design guidelines for durable pavements.
© 2026 Muhammad Ajmal, Uneb Gazder, Muhammad Junaid, Rashed Ismaeel, published by University of Žilina
This work is licensed under the Creative Commons Attribution 4.0 License.
