Life cycle-oriented digital twin for civil infrastructure assets: A natural language processing model-assisted review

With the rapid advancement of information and intelligent sensing technologies, digital twins are emerging as a key technology for civil infrastructure management throughout the lifecycle. Existing review studies are mostly confined to specific phases or technical domains, lacking systematic lifecycle-oriented perspectives. This study presents a lifecycle-oriented review of digital twins for civil infrastructure assets using a data-driven framework. A dataset of 26,706 publications is constructed following a PRISMA-guided process, and an NLP-assisted classification method based on Sentence-BERT is developed to categorize studies across four lifecycle phases. The results reveal rapid growth since 2015, driven by advances in AI, IoT, and sensing technologies. Research is concentrated in the design phase, while operation and maintenance applications are expanding, and demolition-related studies remain limited. The analysis highlights key technologies, research evolution, and existing challenges, including data interoperability and cross-phase integration. Future directions are proposed toward intelligent, integrated digital twin systems for lifecycle management.