Predicting the Compressive Strength of Concrete Utilizing Machine Learning Techniques and Conventional Techniques

Abstract

In civil engineering, accurately determining the compressive strength of concrete is a crucial aspect of designing buildings. Precisely predicting this strength can lead to significant time and cost savings by quickly generating essential design data and reducing the need for trial mixes, thus minimizing material waste. This research employed two different types of soft computing approaches, specifically artificial neural network (ANN) and Random Forest (RF), to efficiently project the compressive strength (CS) of concrete to forecast the compressive strength of concrete reliably. The variables considered include age, cement content, fly ash, Blast Furnace Slag, water content, superplasticizer content, coarse aggregate, and fine aggregate. This study highlights the vast potential of cutting-edge machine learning models as a superior option for precisely predicting the compressive strength of concrete based on the concrete's components.  The statistical analysis results show that all of the machine learning models displayed outstanding predictive abilities, as demonstrated by their high coefficient of determination () values of 99.5% and 95.3%, along with low Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Error values of 1.177, 3.069, 0.387, 2.657, Additionally, the compelling findings suggest that the proposed models based on the RF and ANN techniques significantly outperformed those proposed using conventional approaches.