Experimental Performance and Machine Learning-Aided Design Tools for Modeling Unaged Warm Cup Lump Rubber-Modified Asphalt with Consideration of Global Warming Potential

When warm mix asphalt (WMA) technologies are used to generate asphalt mixes at lower mixing and compaction temperatures, moisture degradation becomes even more pressing. The impact of Evotherm on the moisture damage resistance of cup lump rubber-modified bitumen (CMB) was investigated using surface free energy (SFE), work of adhesion, work of cohesion, and a modified Lottman test. The laboratory findings were integrated with an optimizable decision-support system to promote sustainable consumption in line with the Sustainable Development Goals (SDGs). This study investigates the use of machine learning (ML) tools to predict the moisture damage performance of warm cup lump rubber-modified bitumen (WCMB) in environmentally friendly asphalt mixtures. Three Bayesian Optimization (BO)-optimized ML models were adopted to improve prediction accuracy. The models were evaluated based on their training time and prediction speed. Results showed that the BT-BO model achieved superior performance, with an RMSE of 10.856, an MAE of 5.4814, and an R² of 1 in the testing phase. Although its prediction speed (6,700 observations per second) was lower than that of GPR-BO and KSVM-BO (35,000 observations per second), it outperformed GPR-BO and KSVM-BO, with RMSEs of 273.04, MAEs of 147.7, and R² values of 0.91. Thus, establishing BT-BO as the most reliable model for predicting WCMB moisture damage performance. These findings reduce the need for trial-and-error laboratory testing by enabling accurate prediction of moisture damage resistance in WCMB mixtures using ML models. Thereby contributing to the reduction in energy consumption and greenhouse gas emissions, supporting the development of sustainable flexible pavements. Future work sbould include exploring ensemble learning and deep learning techniques to enhance model accuracy and extend the analysis to real-world conditions.