Artificial Intelligence Framework for Concrete Compressive Strength Prediction

Abstract

This study develops and validates a robust Artificial Intelligence (AI) framework for predicting concrete compressive strength. A hybrid dataset of 1,274 samples was established by combining 244 locally tested specimens with 1,030 data points from a previously published source. Eight input parameters, including cement, blast furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate, and curing age, were used. Python libraries were employed with two key preprocessing steps: a logarithmic transformation of concrete age to address the non-linear strength gain behavior and MinMax scaling to normalize input variables. The performance of the ANN was compared with Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models. All machine learning models demonstrated strong predictive capability, with the XGBoost achieving the best performance, yielding a Mean Absolute Error (MAE) of 0.106 MPa and a Coefficient of Determination (R²) of 0.999 on the independent test dataset. The proposed framework offers a highly accurate and interpretable tool for practical applications in quality assurance, concrete mix optimization, and data-driven decision-making within the construction industry.