Optimizing compressive strength of hybrid fiber-reinforced recycled aggregate concrete: Experimental investigation and ensemble machine learning approaches

Evaluating the compressive strength (CS) of fiber-reinforced recycled aggregate concrete (FRRAC) with varying recycled aggregate (RA) content is crucial for sustainable construction applications. However, accurately predicting the CS of FRRAC compositions with different fiber lengths and fiber ratios remains challenging due to the variability in RA properties. In this study, experimental tests and machine learning (ML) models were employed to assess and predict the CS of hybrid FRRAC. First, an experimental program investigated the effect of RA replacement (50 % and 100 %) and hybrid fiber reinforcement using glass fiber (GF) and polyvinyl alcohol fiber (PVAF) in different ratios (0.05 %, 0.1 %, and 0.15 %) after 28 days of curing. The results indicated that hybrid fibers (GF + PVAF) at 0.1 % content significantly improved CS, achieving 67.8 MPa for 50 % RA and 66.4 MPa for 100 % RA. Second, ML models, including support vector regressor (SVR), Random Forest (RF), Gradient Boosting Regression (GBR), and Extreme Gradient Boosting (XGBoost), were developed to predict hybrid FRRAC CS using a dataset of 283 entries, including data from the literature. XGBoost achieved the highest prediction accuracy, with R2 values of 0.987 on training and 0.956 on testing. Feature importance analysis using SHapley Additive exPlanations (SHAP) identified water content, ordinary Portland cement (OPC), and fly ash (FA) as key factors influencing CS. Overall, the study demonstrates that combining experimental and ML approaches can effectively optimize hybrid FRRAC compositions for engineering applications, supporting the development of sustainable and high-performance concrete.