AI-Driven Multi-Objective Optimization Of Low-Carbon Pervious Concrete For Stormwater Quality, Infiltration And Groundwater Recharge

Abstract

The rapid urbanization has resulted in a rise of impervious surfaces and has increased stormwater runoff, urban flooding and erosion of natural groundwater recharge. Pervious concrete is a promising nature-based engineering solution that allows rain to pour through an interlocking pore network while providing a load-bearing pavement. But permeability in general reduces mechanical strength and the amount of cementitious material is increased along with permeability but not carbon content. Here we propose an integrated experimental and explainable artificial intelligence approach to the development and optimization of low-carbon pervious concrete for stormwater management. In this study, pervious concrete is studied with recycled concrete aggregate (RCA) and other cementitious materials including fly ash and ground granulated blast-furnace slag (GGBS). The mixtures are evaluated in terms of density, porosity, compressive strength, flexural strength, water absorption rate, hydraulic conductivity, clogging resistance and stormwater pollutant removal. The experimental database is used to develop machine learning models such as Random Forest, XGBoost, LightGBM and artificial neural networks. Model performance is evaluated using coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The best model is selected and the analysis of the model by SHAP to identify which mixture and hydraulic parameters have an impact on the model performance and direction. A multi-objective optimization framework based on NSGA-II is implemented to maximize the mechanical performance, infiltration and pollutant removal efficiency with minimal embodied carbon. The resulting Pareto-optimal solutions provide a variety of sustainable mix designs rather than a single deterministic mixture. A set of optimum mixtures is tested in order to confirm the reliability of the proposed data-driven framework. With this study, we provide a link between sustainable concrete, stormwater management, groundwater recharge, and explainable artificial intelligence which will help move the city towards climate-resilient urban infrastructure.