Desenvolvimento de software para a estimativa da profundidade de carbonatação, vida útil e captura de CO2 de estruturas de concreto empregando RNA’s
Carregando...
Data
2015-12-07
Autores
Felix, Emerson Felipe
Título da Revista
ISSN da Revista
Título de Volume
Editor
Resumo
É crescente a busca por modelos matemáticos capazes de estimar e prever a frente de carbonatação do concreto, uma vez que a profundidade carbonatada está diretamente relacionada à vida útil das estruturas. Porém, devido a complexidade, ao número de variáveis envolvidas e a não-linearidade do problema, a determinação destes modelos torna-se uma tarefa árdua. Para contornar estas dificuldades, utilizam-se atualmente ferramentas computacionais que possibilitam a modelagem de problemas de forma otimizada, como é o caso das Redes Neurais Artificiais (RNA’s). Estas redes apresentam características semelhantes às biológicas, possibilitando o reconhecimento de padrões e comportamentos de forma eficiente, impulsionando sua utilização para modelar a profundidade carbonatada. Neste trabalho, visando criar um software para estudos de durabilidade e vida útil das estruturas de concreto sujeitas à carbonatação, modelou-se a degradação do concreto empregando RNA’s do tipo MultiLayer Perceptron implementadas em linguagem orientada à objetos C++. Ao todo foram implementadas 1200 redes com características distintas, das quais selecionou-se a que apresentava o melhor comportamento, obtendo correlação de 0,98, erro máximo de 3,8 mm e erro médio abaixo de 1,0 mm. Por meio de comparações com modelos da literatura, verificou-se a potencialidade da rede criada, possibilitando utilizá-la na geração de modelos secundários, para estimativa da vida útil de projeto, cobrimento mínimo da armadura e captura de CO2 devido à carbonatação do concreto, acoplados em um software denominado CARBEM. Concebido em linguagem JAVA, o software possuí interface gráfica agradável facilitando a interação máquina-usuário. Os resultados encontrados apontam a grande aplicabilidade das RNA’s para modelar a frente de carbonatação do concreto, e que o CARBEM se apresenta como uma ferramenta potencial para a determinação da profundidade de carbonatação, vida útil de projeto, cobrimento da armadura e captura de CO2, contribuindo para o estudo da durabilidade, sustentabilidade e vida útil das estruturas de concreto.
The seek for mathematical models capable of estimating and predicting the front of concrete carbonation has been increased due to the carbonation deepness, which is directly linked to the life cycle of the structures. However, some factors, such as the complexity, the number of involved variables, and the lack of linearity of the problem are contributors to make the determination of these models difficult. In order to soften these difficulties, computational tools have been developed, enabling the optimization of problems modeling. One of these tools are the Artificial Neural Networks (ANN’s) that have similar characteristics to biological neural networks, efficiently helping in the recognition of patterns and behaviors, making these networks ideal for the modeling of the carbonatation deepness. Thinking about developing a software to study the durability and the life cycle of the concrete structures that are subjected to the carbonatation process, the degradation of the concrete was modeled utilizing Multilayer Perceptron ANN’s, implemented in C++ object-oriented language. Altogether, 1200 neural networks with distinct characteristics were implemented, which was selected the one with best behavior. The chosen network presented a correlation of 0.98, maximum error of 3.8 mm, and medium error below 1.0 mm. Comparisons with literature models show the potential of the developed network, that is ready to be used to generate secondary models, estimate the life cycle of projects, predict the minimum covering of the reinforcement, and estimate the CO 2 capture due to concrete carbonation. These functions are all found in a coupled software called CARBEM. Besides this, the software has a neat graphic interface that facilitate its utilization, being totally developed in JAVA language. It is worth noting that the results show the great applicability of the RNAs in the modeling of the front of concrete carbonation, and also the utilization to determine the carbonation deepness, the project life cycle, and the CO 2 capture. Therefore, the CARBEM software contributes to the study of the durability of concrete structures, as well as the sustainability of the projects.
The seek for mathematical models capable of estimating and predicting the front of concrete carbonation has been increased due to the carbonation deepness, which is directly linked to the life cycle of the structures. However, some factors, such as the complexity, the number of involved variables, and the lack of linearity of the problem are contributors to make the determination of these models difficult. In order to soften these difficulties, computational tools have been developed, enabling the optimization of problems modeling. One of these tools are the Artificial Neural Networks (ANN’s) that have similar characteristics to biological neural networks, efficiently helping in the recognition of patterns and behaviors, making these networks ideal for the modeling of the carbonatation deepness. Thinking about developing a software to study the durability and the life cycle of the concrete structures that are subjected to the carbonatation process, the degradation of the concrete was modeled utilizing Multilayer Perceptron ANN’s, implemented in C++ object-oriented language. Altogether, 1200 neural networks with distinct characteristics were implemented, which was selected the one with best behavior. The chosen network presented a correlation of 0.98, maximum error of 3.8 mm, and medium error below 1.0 mm. Comparisons with literature models show the potential of the developed network, that is ready to be used to generate secondary models, estimate the life cycle of projects, predict the minimum covering of the reinforcement, and estimate the CO 2 capture due to concrete carbonation. These functions are all found in a coupled software called CARBEM. Besides this, the software has a neat graphic interface that facilitate its utilization, being totally developed in JAVA language. It is worth noting that the results show the great applicability of the RNAs in the modeling of the front of concrete carbonation, and also the utilization to determine the carbonation deepness, the project life cycle, and the CO 2 capture. Therefore, the CARBEM software contributes to the study of the durability of concrete structures, as well as the sustainability of the projects.
Descrição
Trabalho de Conclusão de Curso apresentado a Banca Examinadora do Curso de Engenharia Civil
de Infraestrutura da UNILA, como parte dos requisitos para obtenção do Grau de Bacharel em
Engenharia Civil. Orientadora: Profa. Dra. Edna Possan. Coorientadora: Prof a . Dr a . Gláucia Maria Dalfré.
Palavras-chave
Modelagem da carbonatação do concreto, Redes Neurais Artificiais, Software, Captura de CO 2, Vida útil de projeto
Citação
FELIX, Emerson Felipe. Desenvolvimento de software para a estimativa da profundidade de carbonatação, vida útil e captura de CO2 de estruturas de concreto empregando RNA’s. 75 p . Trabalho de conclusão de curso ( Graduação em Engenharia Civil de Infraestrutura) – Universidade Federal da Integração Latino-Americana, Foz do Iguaçu, 2015.