Prediction and optimization of chloride ion penetration resistance in sustainable concrete incorporating supplementary materials using ANN and RSM

Abstract The penetration of chloride ions into recycled aggregate concrete (RAC) causes degradation of durability by corroding internal reinforcement. The incorporation of supplementary materials represents a promising strategy for improving resistance to chloride ion penetration; however, experimental evaluation of their effects on chloride transport properties is often time-consuming and costly. To address these challenges, this study develops a data-driven framework using reliable predictive models of artificial neural networks (ANN) and response surface methodology (RSM), to predict and optimize chloride ion penetration resistance of RAC concrete incorporating supplementary materials in terms of chloride ions diffusion coefficient (D Cl − ) considering a sustainable approach. To this end, a dataset comprising 729 RAC mixtures, including ground granulated blast furnace slag (GGBFS), waste crumb rubber (WCR), and recycled coarse aggregate (RCA), with varying water-to-binder ratio (W/B), binder content (BC), and curing age (CA), was analyzed. The results reveal that the ANN model exhibits superior predictive accuracy compared to RSM, as evaluated using different statistical metrics. However, RSM demonstrates acceptable proximity to superior model of ANN, confirming RSM reliability. The RSM also provides a mathematical formula that demonstrates the nonlinear relationship between input and output variables. Furthermore, the optimization results of RSM indicate that using GGBFS = 40% (wt% of BC), WCR = 2.7% (wt% of BC), and RCA = 3.5% (wt% of coarse aggregate), with the lowest W/B ratio and maximum values of BC and CA, achieve the lowest D Cl − while maximizing the use of waste supplementary materials during the production of RAC mix design.