Dimensionamento de vigas em estruturas metálicas utilizando redes neurais artificiais e o método de elementos finitos
Resumo
Resumo: Devido à grande variedade de perfis disponíveis para serem utilizados em estruturas metálicas, também existe uma grande quantidade de configurações possíveis para os elementos estruturais. Por essa razão, o objetivo deste trabalho é desenvolver uma ferramenta utilizando redes neurais artificiais para auxiliar no dimensionamento de vigas com seção do tipo I e H conectadas às colunas por meio de uma ligação do tipo cantoneira dupla. A ferramenta deve fornecer para o vão da viga e a carga aplicada uma configuração com o perfil para a viga, o perfil cantoneira para a ligação, o diâmetro e a quantidade de parafusos. Durante o desenvolvimento do trabalho foi preparado um conjunto de configurações variando seus parâmetros. As configurações foram analisadas utilizando o método de elementos finitos. Na análise pelo método dos elementos finitos os perfis foram modelados utilizando um elemento quadrilateral plano com efeitos do estado plano e de placa de Kirchhoff-Love. Em seguida, com base nos diagramas de esforços internos obtidos com os resultados da análise pelo método de elementos finitos, as configurações foram verificadas de acordo com os critérios de dimensionamento da norma brasileira NBR 8800:2008. Os resultados das verificações de dimensionamento das configurações foram utilizados durante o treinamento das redes neurais. Os resultados da ferramenta utilizando as redes neurais foram avaliados por sua acurácia em fornecer configurações que atendem os critérios de dimensionamento. Para os conjuntos de teste, 98% das configurações indicadas atenderam os critérios de dimensionamento. Portanto, uma vantagem da ferramenta é fornecer de forma simplificada, com custo computacional reduzido e de acordo com as normas técnicas uma configuração para os elementos estruturais. Abstract: Due to the wide variety of profiles available for use in steel structures, there are also a lot of possible configurations for the structural elements. For this reason, the objective of this work is to develop a tool using artificial neural networks to assist in the design of beams with type I section connected to the columns by a double angle connection. The tool must provide for the length of the beam and the applied load a configuration with the profile for the beam, the double angle profile for the connection, the diameter and the number of bolts. During the development of the work a set of configurations was prepared varying its parameters. The configurations were analyzed using the finite element method. In the analysis by the finite element method, the profiles were modeled using a flat quadrilateral element with effects of the plane stress and Kirchhoff-Love plate. Then, based on the internal force diagrams obtained with the results of the analysis using the finite element method, the configurations were verified according to the design criteria of the Brazilian standard NBR 8800:2008. The results of the configuration design checks were used during the training of the neural networks. The results of the tool using the neural networks were evaluated for its accuracy in providing configurations that meet the design criteria. For the test sets, 98% of the indicated configurations met the design criteria. Therefore, an advantage of the tool is to provide, in a simplified way, with reduced computational cost and according to the technical standards, a configuration for the structural elements.
Collections
- Dissertações [164]