Prediction of Structural Health of Pavement Preservation Treated Pavements Using Neural Networks

Abstract

Structural testing of pavements is required to ensure their strength, durability, and longer service life. Destructive testing methods are less costly but more time-consuming and pose a safety risk for road users and crew members. On the other hand, nondestructive testing is faster and less intrusive, but the initial cost is very high for procuring and purchasing equipment. In network-level management, testing every road segment within a network is not feasible when the testing process is time-consuming and resource-demanding. Therefore, a tool or methodology to predict the structural condition of the pavements from the visible surface distresses is much needed. The present study used a neural network-based model for this purpose. Also, the existing benchmarking for the structural condition index was not found sensitive enough to capture the changes in structural condition due to the application of pavement preservation treatments. Therefore, a modified benchmarking was proposed in this study. Finally, a decision tree was developed for application in a Pavement Management System (PMS) that incorporates the structural health condition of the pavement while assigning a particular maintenance or rehabilitation activity that can directly benefit the agency providing recommended rehabilitation activities based on the overall pavement health.