Intelligent geometric deformation analysis method of railway bridges driven by digital twin

Abstract Effective geometric deformation analysis of bridge structures plays a crucial role in enhancing the resilience and safety of bridge infrastructure. Digital twin technology serves as the core link between physical structures and virtual models, providing unified support for information organization, state awareness, and intelligent decision-making throughout the entire lifecycle. This paper systematically reviews the evolution of three core technologies in the intelligent analysis of geometric deformation in railway bridges. First, in information management, the research has evolved from data collection and static recording to semantically connected and inferable knowledge graphs, marking a paradigm shift from ‘data storage’ to ‘knowledge-driven’ analysis. Second, in diagnosis and prediction, the focus has shifted from empirical and statistical models to integrated approaches that combine data-driven and physics-based models, with an emphasis on fine-grained deformation diagnosis and trend prediction. Third, in visualization, the development has moved from two-dimensional engineering drawings and static displays to integrated 3D visualization and enhanced interaction based on building information modeling and geographic information system, expanding functionality from ‘information display’ to ‘decision support.’ Furthermore, by aligning and integrating these three domains, this study establishes a digital twin-driven intelligent analysis framework for railway bridge geometric deformation, built on information organization, centered on diagnostic and predictive modeling, and expressed through interactive visualization. The framework is applied and validated through typical case studies. The study aims to provide a systematic reference for cross-regional and cross-scenario monitoring and intelligent management of bridge geometric deformation, promoting an intelligent transformation of the field from ‘experience-driven’ to ‘data–model collaborative-driven’ development.