Enhancing concrete strength for sustainability using a machine learning approach to improve mechanical performance

Area of Science:

• Civil Engineering
• Materials Science
• Data Science

Background:

• Growing environmental concerns necessitate sustainable construction materials.
• Recycled concrete aggregate (RCA) offers a viable alternative to natural aggregates, reducing waste and resource depletion.
• Inconsistent performance of RCA concrete hinders its widespread adoption.

Purpose of the Study:

• To develop machine learning (ML) models for predicting the mechanical performance of RCA concrete.
• To identify critical parameters influencing the compressive strength (Fc) and split tensile strength (STS) of RCA concrete.
• To support the integration of RCA in construction through reliable performance prediction.

Main Methods:

• A comprehensive dataset of 583 RCA concrete samples was compiled from existing literature.
• Three ML models—Extreme Gradient Boosting (XGBoost), Decision Tree, and K-Nearest Neighbors (KNN)—were trained and evaluated.
• SHAP analysis was employed to determine feature importance for strength prediction.

Main Results:

• XGBoost exhibited superior performance, achieving R² values of 0.86 for Fc and 0.88 for STS.
• The water-to-cement ratio and cement content were identified as the most significant factors affecting RCA concrete strength.
• Decision Tree showed moderate accuracy, while KNN had limited predictive capabilities.

Conclusions:

• ML models, particularly XGBoost, provide a reliable method for predicting RCA concrete performance.
• Understanding key influencing factors enables optimization of RCA concrete mix designs for enhanced sustainability.
• This approach facilitates the adoption of greener construction practices by providing engineers with predictive tools.