Physics-Informed Reality Modeling: Behavior-Aware Reality Capture Representation of Civil Infrastructure

Digital Twins (DTs) of civil infrastructure are intended to represent not only 3D geometry but also dynamic structural behavior. However, current practice still lacks a practical means of embedding such behavior into reality-captured 3D mesh models, leaving even highly detailed representations behaviorally static. To address this need, this study proposes Physics-Informed Reality Modeling (PIRM) for transforming static 3D reality-captured structural meshes into behavior-aware digital representations. PIRM uses a skeleton-to-mesh framework consisting of structural skeleton creation and placement, mesh binding and weighting, and skinning-based deformation transfer to convey modal motion to the 3D mesh. Within this framework, two complementary motion-definition branches are established: Continuous Analytical Articulation (CAA), using idealized mode-shape functions, and Discrete Nodal Articulation (DNA), using experimentally identified modal amplitudes and phase relationships. The methodology was demonstrated on two pedestrian bridges using photogrammetry-based 3D meshes and measured vibration data processed through output-only modal identification. The evaluation qualitatively examined mesh deformation continuity and geometric fidelity, while computational efficiency was assessed using processing time across multiple skeleton node-density and mesh-resolution configurations. Results showed that PIRM can translate structural dynamic information into qualitatively continuous, interpretable, and mesh-resolved structural motion, while DNA better preserved phase-varying behavior in modes with appreciable complex modal content. Overall, the study advances DTs by transforming reality-captured models into behavior-aware digital representations of civil infrastructure.