Estimating local pavement performance and remaining service interval using neural networks-based models and automation tool

Abstract

This study introduces an integrated approach to enhance county pavement management, emphasising operational efficiency in determining the Remaining Service Interval (RSI) for rigid and flexible pavements. It establishes a robust methodology for systematically processing raw county road data through dynamic segmentation and summarisation to create a structured pavement database. It also incorporates innovative approaches and input configurations in employing Artificial Neural Networks (ANNs) to predict current and future county pavement performance indicators, including International Roughness Index (IRI), rutting, transverse, and longitudinal cracks, even with limited data. Evaluation of the ANN models on independent county road databases exhibited high prediction accuracies (0.86 < R 2 < 0.99), varying with specific performance indicators. The study results in an automation tool for expediting road performance estimation over multiple years. This tool seamlessly integrates the ANN models, empowering county engineers to make data-driven decisions and optimise resource allocation for effective pavement management, achieving significant cost savings.