Identificação de falhas em compósitos laminados via transformações integrais com domínio único e inferência bayesiana

• Ano de defesa: 2019
• Autor(a) principal: Mascouto, Fabricio da Silva
• Orientador(a): Knupp, Diego Campos
• Banca de defesa: Silva Neto, Antônio José da , Zotin, Jose Luiz Zanon
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade do Estado do Rio de Janeiro
• Programa de Pós-Graduação: Programa de Pós-Graduação em Modelagem Computacional
• Departamento: Centro de Tecnologia e Ciências::Instituto Politécnico
• País: BR
• Palavras-chave em Português: Técnica da transformada integral clássica; Domínio único; Condutividade térmica; Compósito laminado; Compositos poliméricos; Materiais laminados; Calor - Transmissão; Localização de falhas (Engenharia)
• Palavras-chave em Inglês: Classical integral transformation technique; Single domain; Thermal conductivity; Laminated composite; Identification of failures
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA MECANICA::FENOMENOS DE TRANSPORTE
• Link de acesso: http://www.bdtd.uerj.br/handle/1/13728

This work presents a physico-mathematical model of the heat transfer in a laminated composite. From this model, two main themes were addressed: i) the first one was the verification of the hybrid solution (numerical-analytical) of a heat transfer problem through the Classical Integral Transform Technique with Single Domain formulation, as well as the Eigenvalue Problem s solution via Finite Differences Method. The great geometries diversity presented in this work was possible due to the use of the single domain formulation, allowing the approach with different shapes and number of layers. The verification was performed by comparing the results with the solution obtained through the software COMSOL Multiphysics ®. The COMSOL s solution was used as a reference in all direct problems. ii) the second one was the identification of adhesion failures in the bonding adhesive between two layers of the composite material. The regions of adhesion failures are characterized by the absence of the adhesive material and the presence of air in its place. Thus, the function of thermal conductivity coefficient in the bonding adhesive was estimated. Since the air has much lower thermal conductivity than the adhesive, the presence of these failures generates an abrupt variation in the estimated function. The function is estimated by solving an inverse heat transfer problem, using a Bayesian Inference approach with the Monte Carlo Markov Chains method. For the generation of the Markov Chain s states, the Metropolis-Hastings algorithm was used. It was decided to use the Total Variation of Parameters as a non-informative priori information and also the constraint to positive values for the thermal conductivity coefficient. The thermal conductivity function was determined for several cases, increasing the associated difficulty degree. The major changes in the geometry of the studied problems relate to faults characteristics, varying their length, the adhesive thickness, the number of failures and the distance between them. In addition, it was studied cases with and without roughness in the material s layers. The experimental thermal data were obtained by at the external surface (upper) of the laminated composite. As no real experiment was performed, the experimental data are synthetic, simulated by adding noise to the respective direct problem solution.