NSFC General Project

Title: A High-precision Digital Twin Modeling Method for Construction Scenes of Complex Underground Stations Guided by Knowledge

Sponsor: National Natural Science Foundation of China

Duration: 2026.01 – 2029.12

Key words: Virtual Geographic Environments; Underground Station Construction Scenes; High-precision Twin Modelling; Knowledge Graphs; Large Language Models

Abstract: High-precision digital twin scenes are the critical foundation for achieving the intelligent construction of complex underground stations. The construction of underground stations involves complex procedures, numerous influencing factors, and dynamic environmental conditions, leading to difficulties in characterizing scene object relationships, seamless integration modeling, and low update efficiency. This project aims to investigate the dynamic association representation patterns of all elements in construction scenes of complex underground stations, reveal the coupling mechanisms between multi-source data and scene objects of underground station construction, and innovate a high-precision digital twin modeling method for construction scenes of complex underground stations guided by knowledge. Primarily, the project mainly includes the following parts: (1) automatic construction of knowledge graphs for underground station construction scenes by integrating large language models, which accurately portray all elements and their spatiotemporal and semantic relationships involved in construction scenes of underground stations; (2) high-precision fusion modeling for construction scenes of underground stations guided by structured knowledge, which generates a high-precision digital twin model of construction scenes at the component level; (3) dynamic updates of underground station construction scenes by synergy of sensing Data and simulation models, which realizes virtual-real symbiosis between twin scenes and construction environments; (4) implementation of prototype system and experiment analysis. The expected results can advance theories and methods for geospatial digital twin modeling, provide key technical support for intelligent underground station construction, and demonstrate significant potential for broader applications in complex underground infrastructure construction, such as the Yaxia project.