Banca de DEFESA: Lara Monalisa Alves dos Santos

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : Lara Monalisa Alves dos Santos
DATE: 24/02/2025
TIME: 09:00
LOCAL: Híbrida - Auditório da FAU, Prof. Jayme Golubov, https://www.ppgfau.unb.br/component/agenda/agendas
TITLE:

Building inspection of flat roof systems: Use of digital image acquisition technologies and automated processes.

KEY WORDS:

Maintenance; Waterproofing system; Degradation; Photogrammetry; Drone; Computer Vision; Deep Learning.

PAGES: 274
BIG AREA: Ciências Sociais Aplicadas
AREA: Arquitetura e Urbanismo
SUMMARY:

Building inspection processes should be conducted periodically to assess the condition of structures and plan maintenance activities effectively. For roofing systems, building inspections can be associated with difficult access, safety risks due to working at height and the time it takes to carry out the inspections, as well as requiring technical knowledge based on sensory analyses. On flat roofs, waterproofing membranes are often used extensively to support installations and equipment, which increases the difficulty of on-site inspection and identification of pathological conditions. In this context, repairs and replacements to elements or components of the waterproofing slab, or equipment placed on it, present a complex situation to be investigated. A systematic literature review identified a gap in the integrated approach to the inspection of waterproofing slabs with different levels, taking into account the digital processing stages for the generation of orthoimages and the application of computer vision techniques for detecting equipment and automated stain segmentation. In this sense, this research questions how digital technologies and automated processes can contribute to maintenance management in flat roof systems with different types and levels. The main objective of this thesis is to promote the improvement of building inspections of flat roof systems by means of digital technologies for image acquisition and automated processes, with a focus on the development of scenarios applicable to monitoring and maintenance management. Based on a qualitative approach, the research combined different methodologies, integrating the bibliographical, exploratory and experimental stages. These approaches were used to characterize the systems in question, carry out field surveys, apply digital technologies and conduct experimental procedures on automated processes. The proposed methodology integrates the use of a drone for visual inspection using images, the construction of a database for the digital processing of these images and the application of computer vision techniques. The sample selected for the studies consists of the flat roofs of the Brazilian Army Headquarters (QGEx) and four other buildings of the University of Brasilia (UnB), located on the Darcy Ribeiro Campus: Central Library (BCE), Central Science Institute (ICC), Child and Adolescent Unit (UCA) and Rectory. The results obtained show satisfactory performance in the application of digital image processing and convolutional neural networks for the detection of equipment that interferes with maintenance operations and the segmentation of stains that affect the waterproofing of roofs. During the inspections, the drone's field of vision is different from the inspector's field of vision, showing that digital technologies can contribute to the visual inspection of stains, as these stains indicate degradation processes that accelerate water infiltration and compromise the performance and durability of the flat roof. With this work, we hope to contribute to an integrated methodological approach as a strategy for accelerating preventive and corrective maintenance actions in maintenance management and roof system monitoring.

COMMITTEE MEMBERS:
Externa à Instituição - HELIARA APARECIDA COSTA - UFT
Externa à Instituição - CLAUDIA TEREZINHA DE ANDRADE OLIVEIRA - USP
Interno - 404714 - JOSE MANOEL MORALES SANCHEZ
Externo ao Programa - 3212150 - LENILDO SANTOS DA SILVA - nullPresidente - 1851456 - VANDA ALICE GARCIA ZANONI