Towards greener roads: Evaluating the performance of sustainable grouts with reduced cement using hybrid machine learning models

Semi-flexible pavements are in the spotlight due to their enhanced durability and strength compared to conventional pavements. These pavements are made with a skeleton of open-graded asphalt mix filled with grouts recognized for their durability. The grouts consume a substantial amount of cement, which raises environmental issues. Consequently, this study evaluated the effect of replacing cement with additives such as supplementary cementitious materials (SCMs) and waste polyethylene terephthalate (PET) on 28-day compressive strength (CS_28d) of cementitious grouts in the presence of superplasticizer (SP). Using 231 experimental data points, hybrid machine learning models were developed to analyze seven inputs: water-cement ratio (W/C), flow, SP%, PET%, SCM%, 1-day (CS_1d), and 7-day compressive strength (CS_7d). The WolfNet model, an artificial neural network optimized with the gray wolf optimizer, outperformed others. Shapley additive explanations (SHAP) and partial dependence plots (PDPs) identified CS_7d and CS_1d as the most influential factors, with flow values between 10 and 18 seconds also significantly impacting CS_28d. A user-friendly graphical user interface (GUI) was developed for practical applications. Predictions using the GUI showed an increase in CS_28d as a result of SCM addition and a reduction in CS_28d after adding 10% PET alongside 1% SP. These findings were validated with a set of experimental tests and were found to be only 2.59% deviated from the predicted target. Overall, this study provides key insights into optimizing grout mixtures for semi-flexible pavement applications.