Machine Learning Validated Mix Design for RCA–Graphene Oxide Modified M40 Concrete for Rigid Pavement Applications

This study investigates the mechanical and durability performance of M40-grade concrete incorporating 20% recycled coarse aggregate (RCA) and graphene oxide (GO) as a nano-modifier, supported by predictive machine learning (ML) modelling. Although RCA contributes to sustainability, its porous and weak interfacial microstructure typically reduces concrete performance. To overcome these drawbacks, GO was introduced at dosages ranging from 0–0.10%. Experimental testing revealed that GO significantly improved compressive, tensile, flexural strength, and durability characteristics up to an optimal dosage of 0.04–0.06%, where enhanced hydration, pore refinement and improved interfacial bonding were most evident. Beyond this range, strength and durability decreased due to GO agglomeration.Advanced ML algorithms—XGBoost, LightGBM, and CatBoost—were developed using the experimental dataset to validate trends and predict performance parameters. CatBoost achieved the highest accuracy across all mechanical and durability indicators with R² values between 0.955 and 0.979, demonstrating strong capability to model complex nonlinear interactions. The strong agreement between experimental and predicted results confirms the reliability of the dataset and highlights the potential of ML-assisted mix design for sustainable concrete. The findings demonstrate that controlled GO addition can effectively upgrade RCA concrete for structural and pavement applications while reducing dependency on natural aggregates.