Physically Based Strength Simulation and Multi-Objective Design of Permeable Concrete Pavement Composition for Urban Stormwater Management System

The strength of permeable concrete pavement (PCP) decreases as porosity increases; its water permeability depends on this porosity. Using the PRISMA 2020 guidelines, 28 open-access studies (337 mixtures) were selected that reported measured porosity, strength at 28 days, and filtration coefficient. Strength was predicted by a neural network, whose loss function incorporated the relationship between strength and porosity, its monotonically decreasing trend, and physical constraints. In the test studies, R2 was 0.74; when excluding one study at a time, it was 0.31 (the network without physical terms yielded 0.43 and 0.25, and linear regression yielded 0.49 and 0.45); the differences between the models are not statistically significant. The contribution of porosity to the SHAP prediction (4.9 MPa) is an order of magnitude higher than that of the water-to-cement ratio and aggregate parameters. The filtration coefficient cannot be extrapolated to new studies; it was estimated by calculating the Boltzmann lattice equations for 500 virtual samples with a correction factor of β = 0.020. Using the NSGA-II method, mixtures with strengths of 10.0–22.1 MPa, porosity of 15–26.5%, and a carbon footprint of 311–397 kg CO2-eq/m3 were obtained. In the EPA SWMM model for a 5-hectare watershed, a surface treatment using any of these mixtures reduces peak discharge by 96–100% during storm events with a return period of 2–50 years; permeability affects the result only when k < 0.01 cm/s. The composition should be selected based on strength and carbon footprint, while hydraulic efficiency should be ensured by the surface area of the pavement and the storage capacity of the subgrade. The results were obtained through calculation and require experimental verification.