A soft computing technique for predicting flexural strength of concrete containing nano-silica and calcium carbide residue

The paper aims to explore the application of intelligent soft-computing techniques, including Gaussian Process Regression (GPR), Support Vector Regression (SVR), Ensemble Tree (LSBOOT, LES), and Multiple Linear Regression (MLR) for the prediction of flexural strength (σf) of PC containing calcium carbide residue (CCR) and nano-silica (NS). CCR was incorporated into the concrete mixtures to partially replace some portion of cement at 0%, 7.5%, 15%, 22.5%, and 30% by weight. At the same time, NS was added to the mixtures at 0–4% (interval of 1%) by weight of cement. The experimental tests evaluated the basic mechanical properties, modulus of elasticity (MOE), and water absorption. The proposed models were trained on the datasets obtained from the experimental result of concrete mixtures cured for 7 and 28 days. The results indicate that the soft computing techniques GPR, SVR, LSBOOST, (LES) and MLR predict the σf with high accuracy (NSE > 0.75) with GPR having the highest accuracy (NSE=0.9999, nRMSE= 0.0001, MAE=0.0001, MAPE= 0.0236, RRMSE=0.0258) in the validation phase. The Artificial Intelligence (AI) based models have improved the linear model (MLR) performance by 8.06%, 6.20%, and 4.5%, respectively, in the validation stage for GPR, SVR, and LES.