Predicting the Mechanical Properties of RCA-Based Concrete Using Supervised Machine Learning Algorithms.

Area of Science:

• Materials Science
• Civil Engineering
• Environmental Science

Background:

• Growing global construction demands deplete natural resources and increase waste.
• Recycled coarse aggregate (RCA) offers a sustainable solution to mitigate environmental impact in concrete production.
• Utilizing RCA in concrete is crucial for reducing construction waste and promoting eco-friendly practices.

Purpose of the Study:

• To predict the compressive strength (CS) and splitting tensile strength (STS) of concrete incorporating RCA.
• To evaluate the effectiveness of Decision Tree (DT) and AdaBoost machine learning (ML) techniques for predicting concrete mechanical properties.
• To analyze the influence of various input parameters on the predictive accuracy of the ML models.

Main Methods:

• Development and application of Decision Tree (DT) and AdaBoost machine learning (ML) models.
• Utilizing a dataset of 344 concrete mix designs with nine input variables, including RCA properties.
• Validation of models using k-fold cross-validation and statistical metrics (R², MSE, MAE, RMSE).

Main Results:

• Both Decision Tree and AdaBoost models demonstrated reliable performance in predicting concrete's CS and STS.
• Statistical validation confirmed the accuracy and robustness of the developed machine learning models.
• Sensitivity analysis identified key variables influencing the mechanical properties of RCA concrete.

Conclusions:

• Machine learning techniques, specifically DT and AdaBoost, are effective tools for predicting the mechanical properties of concrete containing RCA.
• The findings support the increased use of RCA in concrete as a viable and sustainable construction material.
• Accurate prediction of mechanical properties facilitates the wider adoption of eco-friendly concrete, contributing to resource conservation and waste reduction.