Resumen
Permeable Interlocking Concrete Paver (PICP) systems provide onsite stormwater management by detaining runoff and removing contaminants. However, a major problem with PICPs is the significant maintenance cost associated with their clean-out to restore the original functionality, which discourages landowners and municipalities from adopting the systems. A combination of laboratory experiments and machine-learning techniques are applied to address this challenge. A total of 376 laboratory experiments were conducted to investigate four independent variables (cleaning equipment speed over the pavement, air speed in the cleaning jets, top opening width of the cupule, and filter media gradation) that affect the cleaning of PICPs. The Buckingham Pi-Theorem was used to express the four main input variables in three dimension-less parameters. This current investigation provides a novel understanding of variables affecting the sustainable and economically feasible maintenance of PICPs. A new model is derived to more accurately predict the percentage of mass removal from PICPs during clean-out using a machine-learning technique. The Group Method of Data Handling (GMDH) model exhibits high performance, with a correlation coefficient (R2) of 0.87 for both the training and testing stages. The established simple explicit equation can be applied to optimize the maintenance costs for industrial applications of Regenerative Air Street Sweepers for sustainable and cost-effective PICP maintenance. Pavements with larger surface areas are found to have lower maintenance costs ($/m2/year) compared to the ones with smaller surface areas. This study estimates $0.32/m2/year and $0.50/m2/year to maintain pavements with larger (5000 m2) and smaller (1000 m2) surface areas, respectively.