Assessing the compressive and splitting tensile strength of self-compacting recycled coarse aggregate concrete using machine learning and statistical techniques

Abstract

The construction industry is adopting high-performance materials due to technological and environmental advances. Researchers worldwide are studying the use of recycled coarse aggregates (RCA) as a partial alternative to natural coarse aggregates in concrete due to their sustainability and environmental benefits. This study compares the predictive abilities of three different machine learning techniques in evaluating the mechanical properties of 28-day-old self-compacting concrete (SCC) incorporating RCA to better understand how design parameters affect the mechanical properties of SCC containing RCA. The study used a range of statistical methodologies and machine learning algorithms, such as ANN, SVM, and M5P trees, to examine the relationship between design elements and accurately forecast the mechanical characteristics of concrete. The ANN model exhibited notable superiority in effectively forecasting compressive strength (CS) and splitting tensile strength (STS) compared to other models, with uncertainty bands of 15.038%-21.154% for CS and 15.701%-19.008% for STS. Moreover, all the uncertainties were under the threshold of 35%. Notably, the water-cement ratio emerged as the most crucial parameter in accurately predicting the mechanical properties of SCC. Finally, the parametric evaluation conducted revealed that CS and STS are inversely proportional to water-cement ratio and aggregate-cement ratio, whereas, are directly proportional to the water-binder ratio, and water-solids percentage.