Predicting corrosion of recycled aggregate concrete under sulfuric acid rain using machine learning and uncertainty analysis

The growing use of recycled concrete aggregate (RA) in concrete has raised concerns about their corrosion, which affect their properties, particularly in harsh environments. This research utilizes a set of experiments to evaluate the properties of recycled aggregate concretes (RACs) subjected to acid rain using magnetic water (MW) and nano silica (NS). Furthermore, deep learning (DL) and support vector machines (SVM) are integrated with the experiments to predict the response of RACs. The results confirm that MW enhances the properties, particularly compressive strength (CS) and sorptivity coefficient (SC). However, the MW is less effective than NS. In contrast, RA replacement decreases the resistance to acid rain, as evidenced by reduction in CS. NS replacement also leads to the enhanced electrical resistivity higher than MW. The prediction results using DL and SVM further facilitate quantifying the level of importance of treatment measures, particularly over longer exposure periods, where DL markedly outperforms SVM. Noticeably, RA is the second major property after pH, controlling the response of RACs. Despite the positive effects of MW, its utilization under acid rain can only surpass NS replacement mostly enhancing SC. Nevertheless, pH, RA, and NS affect the acid rain resistance of RACs significantly more than MW.