Enhancing pavement management systems through a data-driven digital twin framework

Digital Twin (DT) technology combines virtual representations of physical infrastructure with real-world data to support continuous condition assessment and proactive decision-making. This study develops a data-driven DT framework to improve pavement performance prediction for network-level Pavement Management Systems (PMS), using biennial condition data from 2009 to 2023 for 57 Louisiana interstate control sections (about 936.6 centerline miles). The Pavement Condition Index (PCI) and International Roughness Index (IRI) were modelled using Extreme Gradient Boosting (XGBoost), with pavement age, structural thickness, traffic and climate as predictors. Models were trained on 2009–2021 data and then retrained with 2023 data to evaluate their adaptability. The updated model increased the coefficient of determination (R2) by 13.8%. An interactive, web-based tool allows users to explore historical conditions, forecast performance and assess maintenance strategies. An Equivalent Uniform Annual Cost analysis showed that DT-enabled decisions can reduce costs by up to 48.1% by avoiding delayed rehabilitation.