Traditional Pavement Management Systems (PMS) are increasingly hindered by aging infrastructure and reliance on reactive, manual inspection-based maintenance strategies. While Digital Twin (DT) technology offers a transformative pathway toward predictive management, existing implementations remain fragmented, often lacking holistic integration between Building Information Modeling (BIM), heterogeneous data sources, and advanced Artificial Intelligence (AI) analytics. To bridge this gap, this study develops and validates a comprehensive, web-based Intelligent Decision Support System (IDSS) founded on a novel four-layer Digital Twin architecture (Physical, Communication, Model, and Service layers). The methodology orchestrates a robust automated pipeline that integrates a sophisticated sequential AI processing chain—comprising generative data augmentation (Diffusion Models), real-time distress detection (YOLOv12) and semantic segmentation (DeepLabV3+), long-term performance prediction (LSTM), and meta-heuristic maintenance optimization (Grey Wolf Optimizer)—within a high-fidelity 3D BIM environment powered by Autodesk Platform Services. Results from a pilot implementation on the KT54 motorway in Finland demonstrate the system’s capability to automate objective pavement condition assessment with high accuracy, substantiated by rigorous cross-dataset validation against independent field data and public datasets to ensure robust generalizability beyond the specific case study. The finalized interactive platform provides road authorities with unprecedented situational awareness, enabling real-time 3D monitoring, visualization of future deterioration trends, and the generation of cost-effective, proactive maintenance schedules under budget constraints. This research contributes a validated, scalable framework for transitioning infrastructure management from reactive approaches to a data-driven, predictive digital twin paradigm.
Talaghat et al. (Sat,) studied this question.