Prediction of high strength ternary blended concrete containing different silica proportions using machine learning approaches

Abstract

The most often utilized material in construction is concrete. High plasticity, good economy, safety, and exceptional durability are a few of its characteristics. Concrete is a type of structural material that needs to be strong enough to withstand different loads. The compressive strength of the concrete members is the most crucial mechanical characteristic of concrete because of its brittleness. Furthermore, concrete with ternary blended cementitious materials is a sophisticated composite material. The present study explores the binary and ternary blended concrete mixes with silica fume, ceramic powder, bagasse ash, and alccofine, to determine the compressive and flexural strength. Results of compression strength tests show that mixes, including ultra-fine alccofine, have higher strength. Furthermore, the impact of additional cementitious materials on the surface morphology was examined using scanning electron microscopy on the various mixes. This study investigates using linear regression, KNearest Neighbors (KNN), and Bayesian-optimized extreme gradient boosting to estimate the compressive strength of ternary blended concrete (BO-XGBoost). Using the coefficient of determination (R2), mean absolute error (MAE), and mean square error (MSE), the prediction results are further validated. In comparison, the linear regression and BO-XGBoost models show high accuracy towards the prediction of the outcome as indicated by its high R2 value equal to 0.883 and 0.880, respectively, while the R2 value for KNN is 0.736. Additionally, normalized feature importance was included in determining the input variables that significantly influenced compressive strength. The sensitivity of the model indicates CaO and SiO2 shows significant importance to predict ternary blended concrete compressive strength.

Keywords

• Alccofine
• Sustainable material
• KNN
• Machine learning