| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 25.79 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A inspeção e monitorização de estruturas de betão armado ainda implica atualmente tempos de execução alargados e um número considerável de meios humanos e mecânicos, os quais acarretam custos consideráveis para as entidades que têm a seu cargo estas estruturas. Com o intuito de automatizar este processo e torná-lo mais célere e económico, pretende-se aperfeiçoar ferramentas baseadas em redes neuronais convolucionais associadas ao processamento de imagens captadas por intermédio de veículos aéreos não tripulados, permitindo que este método seja uma mais-valia e uma nova forma de pensar e executar a inspeção e monitorização de estruturas em betão amado. Esta dissertação tem como objetivo a identificação e caracterização de uma das anomalias de estruturas de betão armado, as fissuras. No decorrer deste estudo foram efetuados diversos testes de validação da metodologia aplicada e os resultados obtidos permitem concluir que as ferramentas desenvolvidas garantem uma abordagem robusta e resultados com elevado grau de fiabilidade. No futuro pretender-se-á alargar e adaptar a metodologia apresentada a outras anomalias e patologias comuns nas estruturas de betão armado.
The inspection and monitoring of reinforced concrete structures currently still imply extended execution times and considerable human and mechanical resources, which translate into considerable costs for the entities in charge of these structures. In order to automate these processes and make them faster and more economical, a processing tool for drone captured images based on convolutional neural networks is presented, allowing this method to be an asset and a new way of envisioning and carrying out inspection and monitoring of concrete structures. The development of this dissertation pretends to give a contribution on the identification and characterization of one of such anomalies present on reinforced concrete structures, namely the cracking. During this study, many tests were conducted, hinting on the robustness and reliability of the results. So future improvement of these tools will include extending this method to other anomalies which can be included in this method
The inspection and monitoring of reinforced concrete structures currently still imply extended execution times and considerable human and mechanical resources, which translate into considerable costs for the entities in charge of these structures. In order to automate these processes and make them faster and more economical, a processing tool for drone captured images based on convolutional neural networks is presented, allowing this method to be an asset and a new way of envisioning and carrying out inspection and monitoring of concrete structures. The development of this dissertation pretends to give a contribution on the identification and characterization of one of such anomalies present on reinforced concrete structures, namely the cracking. During this study, many tests were conducted, hinting on the robustness and reliability of the results. So future improvement of these tools will include extending this method to other anomalies which can be included in this method
Description
Keywords
Inspeção Monotorização Betão armado Estruturas Redes Neuronais Convolucionais (RNC) Unmanned aerial vehicles (UAV) Anomalias Patologias Fissuras Inspection Monitoring Reinforced concrete Structures Convolutional Neural Networks (CNN) Anomalies Pathologies Cracks