AI-Assisted Reconstruction and Geometric Assessment of the Las Cucharas Railway Bridge: An Applied Case Study
Year: 2026
Type: article
Abstract: This study reconstructs the Las Cucharas railway bridge in Chile and examines how sampling, alignment, and model coverage affect geometric agreement. Codex assisted with interpreting the source material and preparing Python and Blender scripts. Conventional fitting, registration, and distance calculations produced the numerical results. The component assembly contains 1073 editable objects, including 204 main members with fitting records. Against 250,000 common cloud observations, the Revit reference and component model have median distances of 0.0692 and 0.1042 source units. Revit is closer for 58.59% of observations, although the distributions cross. For 2603 observations on visible steel supports, the respective medians are 0.0214 and 0.0545. Revit estimates of the nominal 75 m support spacing are 74.99 and 75.01 source units on the two sides. Two plausible orientations preserve the median ordering but change results at the masonry ends. In a separate surface reconstruction check, 100 of 200,000 queries account for 96.29% of the sum of squared distances. Stored source rows and nearest triangles allow these discrepancies to be examined individually. The Unity application supports annotation review in Revit through the BIM Collaboration Format (BCF) and displays sensor records with freshness indicators. A subset audit maps 14 native Revit elements to the assessed and Unity geometry. An instrumented extension transferred three test points through BCF to BIMcollab in Revit. All three retained their topic and component identities, and their point coordinates remained within a predefined 1 mm digital transfer tolerance. The results provide a documented component reconstruction and reproducible comparisons for railway documentation. They establish agreement within the available digital record. A quantitative assessment of AI efficiency, physical accuracy, and performance across further annotation cases remains for future work.
Source: Applied Sciences
DOI: 10.3390/app162211487
Institution: Pontificia Universidad Católica de Valparaíso
Language: English
Subfield: Geology
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,896
OpenAlex ID: W7236808886
