Harnessing Machine Learning for Rutting Resistance Prediction in Waste Glass-Incorporated Cold Bitumen Emulsion Mixtures

Waste glass (WG) use in pavement construction has gained attention for its potential to strengthen the rutting resistance of cold bitumen emulsion mixtures (CBEMs). This study evaluates the mechanical performance of CBEM modified with varying WG contents and employs machine learning models to predict rutting resistance, quantified by the flow number. Experimental findings indicate that replacing 100% of fine aggregate with WG increases the flow number by 32% compared to conventional CBEM, attributed to improved particle interlocking, enhanced stiffness from hydration reactions, and supplementary cementitious compounds in WG. Bulk density and air voids strongly influence the flow number, with correlations exceeding 0.90. Among machine learning models tested, the random forest (RF) model indicates the best predictive accuracy ( R 2 = 0.96 , RMSE = 81.03 , MAE = 64.01 ), surpassing gradient boosting machine (GBM), support vector machine (SVM), and k -nearest neighbors (KNN). These results highlight WG’s efficacy in improving CBEM rutting performance.