Machine learning for sustainable concrete incorporating marble waste and recycled aggregate

Purpose This study aims to evaluate the mechanical and durability performance of sustainable concrete incorporating Marble Fine Powder (MFP), Marble Coarse Powder (MCP) and Recycled Concrete Aggregate (RCA), and to develop machine learning (ML) models for accurate prediction of concrete properties. Design/methodology/approach A total of 250 experimental data points were generated from 50 concrete mix designs of M25 and M50 grades tested at five curing ages. Six ML models, namely Random Forest (RF), Support Vector Regression (SVR), CatBoost, Extra Trees, XGBoost and Gradient Boosting Machine (GBM), were developed to predict compressive strength (CS), split tensile strength (STS), flexural strength (FS), Rapid Chloride Permeability Test (RCPT) and sorptivity. Model performance was evaluated using R2, RMSE, MAE and MAPE. SHAP analysis and permutation importance were used for feature interpretation. Findings XGBoost demonstrated the best overall performance with a test R2 of 0.9978 for compressive strength and nearly perfect prediction accuracy for RCPT (R2 = 1.0000). The model produced a low RMSE of 0.598 MPa for CS prediction. SHAP analysis identified cement content and curing age as the most influential parameters affecting concrete performance. Among replacement materials, MFP showed the highest impact on strength and durability properties. The study also found that M25 concrete can safely incorporate up to 10% MFP, 20% MCP and 40% RCA, while M50 concrete can tolerate up to 60% RCA without significant loss in performance. Social implications Waste material utilization and cost-saving materials. Originality/value This study presents one of the first integrated ML frameworks for sustainable concrete containing both marble waste and recycled concrete aggregate. The research combines experimental investigation with explainable artificial intelligence techniques to provide reliable prediction tools for sustainable concrete mix design and optimization.