Data-driven prediction of concrete strength by machine learning: hybrid-fiber-reinforced recycled aggregate concrete

Purpose The purpose of this study is to predict the compressive strength of hybrid-fiber-reinforced recycled aggregate concrete (HFRRAC) using machine learning (ML) models. The aim is to address the challenges of optimizing HFRRAC mix designs by modeling the complex and non-linear relationships between mixing ingredients and compressive strength. Design/methodology/approach A data set comprising 634 samples of HFRRAC mix proportions and corresponding compressive strength values was compiled from recent research studies. A total of 11 ML models were used for compressive strength prediction, and the models’ performance was evaluated by using metrics such as R², mean squared error(MSE), mean absolute error (MAE) and root MSE (RMSE). Feature importance analysis was also performed to identify key predictors. Findings Light gradient boosting machine demonstrated the highest accuracy with an R² of 0.969, an MSE of 14.04, an MAE of 2.76 and an RMSE of 3.75, making it the most efficient model for predicting HFRRAC compressive strength. CatBoost was identified as the second-best performer. Feature importance analysis highlighted water-to-binder ratio, coarse aggregate-to-binder ratio, fine aggregate-to-binder ratio, recycled aggregate replacement rate and curing age as critical factors influencing compressive strength. Originality/value This study demonstrates the application of ML techniques to predict HFRRAC compressive strength with high accuracy, offering a reliable approach to mix design optimization.