Digital reconstruction and geometrical model updating of large-scale bridge structures: a case study of Baijusi bridge

Maintaining and modernising bridge structures is crucial to ensuring their structural integrity and public safety. This paper presents an innovative, semi-automated approach to transforming 3D point cloud data of bridges into precise, editable computer-aided design (CAD) drawings. The proposed methodology achieves high precision in mapping essential bridge components, accurately recognising and delineating key structural features such as supports and deck structures. Testing shows that the automatically generated CAD models achieve an Intersection over Union (IoU) metric of up to 91% for critical structural outlines, with an F1-measure indicating dimensional inaccuracies of about 8% and an 83% success rate in identifying pivotal elements. In addition, the approach demonstrates exceptional efficiency, generating an editable CAD drawing in DXF format in under 1 min. The proposed framework is also used for updating finite element (FE) models for bridges, enabling more advanced structural analysis and proactive maintenance protocols. This combination of automated CAD modelling with FE model refinement provides a novel methodology for streamlining the evaluation and preservation of bridge infrastructure, making a significant contribution to the field of structural engineering.