| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 12.48 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A norma NP EN 1090-2:2020 estabelece requisitos para a execução de estruturas metálicas, ao
oferecer diretrizes detalhadas para garantir uma execução correta, salvaguardando a
integridade estrutural e a segurança da construção final. Com principal foco nas tolerâncias de
montagem, a presente dissertação tem como principal objetivo apresentar uma metodologia
automatizada para verificação de desvios geométricos em edifícios industriais metálicos
utilizando tecnologias digitais e técnicas de inteligência artificial. As regras normativas são
avaliadas usando dados de nuvens de pontos derivados de modelos digitais 3D de uma
estrutura, como o modelo As-Is e o modelo As-Design. O modelo As-Is é obtido através da
integração de fotogrametria e laser scanner terrestre, enquanto o modelo As-Design é
adquirido a partir do modelo estrutural do projeto. O aspeto inovador desta metodologia
assenta na aplicação do algoritmo YOLOv8 para identificar automaticamente a secção
transversal de elementos estruturais utilizando técnicas de segmentação por instâncias e
Oriented Bounding Box (OBB). Os desvios geométricos entre os dois modelos são calculados e
analisados para garantir o cumprimento das regras normativas estabelecidas pela NP EN 1090-
2:2020. Os resultados deste estudo indicaram que a metodologia proposta é robusta o
suficiente para ser aplicada em qualquer tipo de estrutura de aço, independentemente da sua
complexidade.
The quality requirements of metallic structures are established by EN 1090-2:2020, which provides detailed guidelines to ensure proper execution, thereby safeguarding the structural integrity and safety of the final construction. With the focus on assembly rules, the main objective of this dissertation is to present an automated methodology for verifying geometric deviations in metallic industrial buildings using digital technologies and artificial intelligence techniques. The normative rules are evaluated using point cloud data derived from 3D digital models of an asset, the “As-Is” and “As-Design” models. The “As-Is” model is obtained through the integration of photogrammetry and Terrestrial Laser Scanning, while the “As-Design” model is acquired from the project’s structural model. The innovative aspect of this methodology is based on the application of the YOLOv8 algorithm to automatically identify the cross-section of structural elements using instance segmentation and “Oriented Bounding Box” (OBB) techniques. Geometric deviations between the two models are computed and analysed to ensure that normative rules established by EN 1090-2:2020 are accomplished. The results of this study indicated that the current methodology is robust enough to be applied on any type of steel structure, even the most complex ones.
The quality requirements of metallic structures are established by EN 1090-2:2020, which provides detailed guidelines to ensure proper execution, thereby safeguarding the structural integrity and safety of the final construction. With the focus on assembly rules, the main objective of this dissertation is to present an automated methodology for verifying geometric deviations in metallic industrial buildings using digital technologies and artificial intelligence techniques. The normative rules are evaluated using point cloud data derived from 3D digital models of an asset, the “As-Is” and “As-Design” models. The “As-Is” model is obtained through the integration of photogrammetry and Terrestrial Laser Scanning, while the “As-Design” model is acquired from the project’s structural model. The innovative aspect of this methodology is based on the application of the YOLOv8 algorithm to automatically identify the cross-section of structural elements using instance segmentation and “Oriented Bounding Box” (OBB) techniques. Geometric deviations between the two models are computed and analysed to ensure that normative rules established by EN 1090-2:2020 are accomplished. The results of this study indicated that the current methodology is robust enough to be applied on any type of steel structure, even the most complex ones.
Description
Financiado
Keywords
EN 1090-2 OBB Fotogrametria LiDAR TLS YOLOv8 Photogrammetry