|Ph.D Student||Zeibak-Shini Reem|
|Subject||Comparative Geometric Analysis of a Building Information|
Model(BIM) using Point Cloud Data of Damaged
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Rafael Sacks|
|Professor Sagi Filin|
|Full Thesis text|
Earthquakes strike with no warning and often cause massive damage to urban facilities including buildings and infrastructure. Following such events, the need arises to rapidly and safely gather information to support search and rescue operations. Terrestrial laser scanning (TLS) can accurately depict such information and provide detailed description of buildings’ facades. In addition to the post-event mapping, information defining the pre-event state of the building is required to enable damage detection and evaluation. As semantically rich damage representation is needed to support recovery activities, building information technology (BIM) is used to comprehensively compile the physical and functional characteristics of buildings and their components.
The broad concept underlying this research is to leverage the synergy of TLS and BIM to develop a semi-automated model that facilitates the damage inspection process by generating semantically rich models of damaged buildings. As building components might change their location, orientation and shape dramatically following a catastrophe, the main challenge is to identify and match corresponding, yet different, elements. The goal of this research is to reconstruct an ’as-damaged’ BIM of the building's facades using its ’as-built’ BIM and the point cloud data (PCD) acquired from laser scanning of the damaged facade. Given that this is the first known attempt to achieve automated compilation of ’as-damaged’ BIM models, the scope is restricted to regular buildings with reinforced concrete elements. Moderate-to-heavy damage is studied; consideration of missing parts and separation of object parts where they entirely disconnect from one another is excluded.
An innovative concept of fusing a BIM-to-Scan process to reconstruct topology and a Scan-to-BIM process to reconstruct geometry was proposed. The first step of the proposed framework is a top-down approach for reconstructing the topology between structural frame members and locating them in the scan. This is achieved by scaling the ‘as-built’ structural grid into the scan extent, associating between voids and structural grid cells and void-driven displacement of elements in their original dimensions. The second step is a bottom-up approach for refining geometry and modeling local damages to individual components. An iterative process is performed in which planar segments are extracted from the ’as-damaged’ scan, comparable faces from the deformed BIM (the result of the first step) are prepared, best-match pairs of BIM faces and scan segments are identified using a tailored utility function, and damaged building components are reconstructed by decomposing objects into solid shells using a heuristic algorithm.
The results proved feasibility and efficiency of the proposed approach. The first step was applied on two synthetic full-scale datasets of damaged RC frame buildings and showed its ability to overcome large displacements and successively place components in their new location and orientation. The second step was performed on individual damaged specimens that were scanned in the field and illustrated the ability to model real damage patterns on RC components using real PCD. More cases of damaged buildings will need to be tested in the future to seek evidence for robustness of both steps of the proposed framework as a whole.