Prediction Of Strength Properties Of Ternary Blended Concrete By Using Artificial Intelligence And Machine Learning Techniques

Abstract

Concrete is the most frequently used material in development due to its high pliancy, economy, security, and outstanding durability. It must be sufficiently strong to endure various loads, with compressive strength being its most vital mechanical attribute. The current study investigates binary and ternary mixed concrete blends with silica fume, ceramic powder, bagasse ash, and alccofine to determine compressive and flexural strength. Results from compressive strength tests indicate that mixes containing superfine alccofine exhibit higher strength. Additionally, the impact of supplementary cementitious materials on surface morphology was examined using scanning electron microscopy. This study employs linear regression, K-Nearest Neighbors (KNN), and Bayesian-optimized extreme gradient boosting (BO-XGBoost) to estimate the compressive strength of ternary mixed concrete. The predictive models were validated using the coefficient of determination (R²), mean absolute error (MAE), and mean square error (MSE). Linear regression and BO-XGBoost models demonstrated high accuracy in predicting outcomes, with R² values of 0.883 and 0.880, respectively, compared to 0.736 for the KNN model. Furthermore, normalized feature importance analysis identified the input variables that significantly affect compressive strength, highlighting the importance of CaO and SiO₂ in predicting the compressive strength of ternary mixed concrete.