Clogging Progression Prediction of Permeable Pavement Laboratory Model Using Artificial Neural Networks

Abstract

As urbanization increases, impervious surfaces expand and this results in significant changes to urban hydrology. These impervious surfaces result in stormwater runoff that carry pollutants along its path to nearby waterways. Implementation of Low Impact Development (LID) techniques, and permeable pavement systems specifically, are commonly used to manage stormwater runoff and improve quality of water resources near urban areas. Hydrological performance of the permeable pavements, however, deteriorates over time mainly due to the sediment clogging on the surface. The effectiveness of permeable pavements and ultimately the captured runoff volume can be correlated to the extent of clogging on the surface. The clogging progression rates vary based on location, site characteristics, and rain events variables. Twenty one laboratory Permeable Interlocking Concrete Paver (PICP) models with different combinations of slope, gap size, and joint filling material were built and exposed to theoretical stormwater events such that correlations could be established between the physical system components and the progression of surface clogging. This study utilizes a neural network model to predict the clogging progression rates by different PICPs characteristics. The results indicate that the model is capable of predicting the extent of clogging along the length of permeable pavement with 98% accuracy. By predicting the precise cumulative rainfall depth based on the clogging length and the PICP specifications, the hydrologic operation for each configuration and at any rainfall depth is accessible. By better understanding the effects of pavement characteristics and choosing the most efficient pavement configuration, systems could be better designed to reduce clogging and more efficient maintenance schedules could be defined.