Modeling Time-Dependent Behavior of Pavement Drainage Using Linear System Identification and Neural Network Techniques

Abstract

Inadequate drainage continues to be a major cause of problems associated with long-term structural integrity and performance of roadway pavements. To reduce the impact of these drainage-related problems, it is customary to provide measures to prevent water from entering the pavement system and to enhance the drainage capability of the pavement base to rapidly move the water that inevitably finds its way into the pavement system. The performance of different drainage systems in the field is not clearly known. Oklahoma Department of Transportation (ODOT) has been collecting rainfall and outflow information at sites with edge drains since 1992. The sites have different types of surfaces, base courses, and edge drains. The data collected by ODOT at these sites were used to develop two types of numerical models to predict the outflow–time history using rainfall–time history as the input. One model is based on linear system identification theory, and the other model is based on an artificial neural network. The development of these models is presented, and the model predictions are compared with the measured field data. The efficiency of the drainage systems, including the AASHTO criteria for the drainage time, at these sites is compared by using the numerical models and synthetic, but the same, rainfall events.