A systematic review of digital twins for civil infrastructure inspection

Abstract

Civil infrastructure worldwide faces growing pressure from aging assets, constrained inspection resources, and increasing exposure to extreme events. Conventional inspection remains largely manual and periodic, which limits the consistency, timeliness, and decision value of condition evidence. Digital twins (DTs) are increasingly positioned as a practical framework for transforming inspection by linking sensing, geometric reconstruction, semantic asset information, and computational models to support monitoring, diagnosis, and decision support. However, existing reviews are often domain-specific or technology-focused and rarely provide an inspection-centred synthesis of how DTs enable damage-scenario generation and anomaly detection. This systematic review synthesises 141 peer-reviewed studies published between 2019 and 2025 to characterise DT applications for civil infrastructure inspection, the enabling data and platform foundations, and the integration of DTs with anomaly detection and damage scenario generation. The literature has grown rapidly since 2021, with strong emphasis on bridges and buildings and expanding applications in offshore structures, pavements, pipelines, water networks, and tunnels. Approximately two-thirds of the reviewed studies integrate DTs with anomaly detection, most commonly through deep learning for vision-based defects and physics-based modeling for response-driven degradation, with increasing interest in hybrid physics-informed approaches. Reported benefits include improved diagnostic capability, earlier localization of abnormal trends, and more structured maintenance planning, while persistent barriers include uneven data quality, limited historical benchmarks, computational constraints, and interoperability challenges across building information modeling, geographic information systems, and finite element modeling workflows. This review identifies research priorities to advance predictive and interoperable digital twins, scalable real-time implementation, and explainable analytics that enable dependable field deployment.