Water Transmissible Pavement: A physics of Granular Sub-base Permeability through Road Dust Analysis using Machine Learning

Amidst the rapid urbanization and heightened infrastructure demands, contemporary cities are capitalizing on every available space, converting previously permeable land into impermeable surfaces. This transition obstructs the absorption of storm water, leading to intensified runoff. To counteract this challenge and address infrastructure requirements, Low Impact Development (LID) techniques have emerged, among which permeable pavement stands out as a widely adopted solution. Serving as a transient storage facility, permeable pavements store storm water within their Granular Sub-Base (GSB) or reservoir layer, thereby diminishing the size of storm drains and contributing to the implementation of Sustainable Urban Drainage Systems (SUDS). Nevertheless, the utilization of permeable pavements is commonly recommended for walkways, parking lots, or low-volume roads due to their susceptibility to clogging. This study delves into the potential for clogging in the reservoir layer, employing Machine Learning models such as Random Forest (RF), Gradient Boosting Regressor (GBR), Light Gradient Boosting Machine (LGBM), and Extra Trees (ET). The investigation incorporates data from 200 instances with varying GSB layers, thicknesses, and combinations of road dust particle sizes. The results reveal a robust correlation (R2 > 0.97) with experimental data, indicating that GSB-III demonstrates optimal clog resistance under high dust loads. The findings suggest that GSB-V and GSB-VI may be suitable for areas with dust loads below 200 gm/month. This research provides valuable insights for the development of clog-resistant permeable pavements tailored to moderate to high-volume roads.