Modelling compressive strength of sustainable concrete containing flyash, granulated ground blast slag, metakaolin, and recycled aggregate using genetic algorithm and support vector regression technology

This study proposes a genetic algorithm-support vector regression (GA-SVR) method to predict the compressive strength (CS) of recycled aggregate concrete (RAC), where GA optimises SVR hyperparameters. Using recycled aggregate and cement replacements like flyash, GGBS, and metakaolin promotes sustainable concrete development. Accurately predicting its compressive strength (CS) is urgently needed for engineering applications. Predicted results show GA-SVR outperforms KNN, LR, DT, ANN, SVR, XGBoost, and RF, achieving R2 of 0.951, MAE of 2.817 MPa, and RMSE of 3.848 MPa on the test set. Sensitivity analysis reveals age has the strongest positive impact on CS, followed by total cementitious material, cement, and GGBS. Recycled coarse aggregate size variation significantly affects CS. Specifically, a reasonable fine aggregate infiltration ratio (reducing RCA 10 down) and coarse aggregate adjustment (increasing RCA 20 down) improve strength prediction. The proposed model facilitates optimal design of RAC with supplementary cementitious materials.