• A machine learning framework was developed to predict CDI and TDI using asphalt mix design parameters. • The MLP model showed the highest accuracy for CDI prediction with an R² value of 0.92. • The Decision Tree model performed best for TDI prediction with the lowest RMSE and MAE. • SHAP analysis identified aggregate gradation as the most influential factor in densification behavior. Workability and long-term performance of asphalt mixtures are vital for durable pavements. Conventional evaluation of Compaction Densification Index (CDI) and Traffic Densification Index (TDI) using Superpave Gyratory Compactor (SGC) data is labour-intensive and fails to capture nonlinear relationships among mix variables. Although machine learning (ML) has been increasingly applied in pavement engineering, most studies address single indices and lack interpretability. This study bridges that gap by developing an explainable ML framework to simultaneously predict CDI and TDI from key mix parameters aggregate gradation, nominal maximum aggregate size (NMAS), and binder type. A dataset of 151 samples, compiled from published studies and laboratory databases, was divided into 70% training and 30% testing subsets. Three ML models Decision Tree, Random Forest, and Multi-Layer Perceptron (MLP) were developed and compared. The MLP achieved the highest predictive accuracy for CDI (R² = 0.92), while the Decision Tree slightly outperformed others for TDI (R² = 0.50). The study further aims to interpret model predictions using SHAP (SHapley Additive exPlanations) to identify the influence of individual input variables such as gradation and NMAS on compaction behaviour. The proposed interpretable framework enables rapid and reliable estimation of densification indices, offering a practical tool for optimizing asphalt mix designs, minimizing laboratory effort, and enhancing pavement quality and service life, thereby supporting Sustainable Development Goal 9 (Industry, Innovation and Infrastructure), SDG 11 (Sustainable Cities and Communities), and SDG 12 (Responsible Consumption and Production).
Kumar et al. (Wed,) studied this question.