Artificial Intelligence-Based Prediction of Permeable Pavement Surface Infiltration Rates

Abstract

Permeable pavements are a type of Low Impact Development technology that aim to reduce stormwater runoff at the source through infiltration and storage. Overtime, sediments carried by stormwater runoff degrade the performance of these pavements and can eventually completely diminish the infiltration capacity of the pavement system. Maintenance procedures have been developed for permeable pavements and it is becoming increasingly understood that these procedures are needed for sufficient long-term hydraulic performance. However, these procedures are expensive and are thus performed infrequently or not performed at all, leading to many permeable pavement systems that no longer perform at their designed infiltration capacity. The objective of this research is to develop data-driven models for modelling infiltration capacity and predicting clogging. Three types of data-driven numerical models, namely a linear regression, an artificial neural network, and a convolutional neural network were created to investigate the relationship between the images of the pavement surface and its associated surface infiltration rate. Preliminary results indicate that some image properties serve as good predictors of clogging, and thus, surface infiltration rates. Further testing and calibration of the models are required to validate the numerical models. These models can be used to estimate the surface infiltration rates of permeable concrete pavements, leading to more widespread maintenance and thus, ensure the designed performance level is maintained.