Identificação de danos via metamodelos, análise modal e Inferência Bayesiana em estruturas de engenharia

• Autor(a) principal: Rangel, Isabela Cristina da Silveira e Silva
• Data de Publicação: 2019
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13630
• topic: Identification of Structural Damage; Response Surface Model; Kriging Surrogate Model; Bayesian Inference; Identificação de danos estruturais; Modelo de Superfície de Resposta; Metamodelo Kriging; Inferência Bayesiana; CNPQ::CIENCIAS EXATAS E DA TERRA::MATEMATICA::MATEMATICA APLICADA

The present work is concerned with the identification of structural damage in a simply supported Euler-Bernoulli beam and in a simply supported Kirchhoff plate. The damage state of the structure is continuously described by a cohesion field, spatially discretized by the finite element method (FEM), and a response surface model (RSM) and a Kriging surrogate model are considered to provide a polynomial relation between the nodal cohesion parameters and elements of the flexibility matrix of the system. The problem of identifying damages is formulated as an inverse problem, whose objective is to estimate the cohesion parameter of the structure. Inverse problems in parameter estimation are formulated and solved by Bayesian approach, such that uncertainties inherent to the experimental data were included in the likelihood function. The inverse problem solution consists of marginal density probability functions for each parameter, obtained by sampling the distribution of interest. A numerical sampling was performed with the Monte Carlo method via Markov Chains, through the Delayed Rejection Adaptive Metropolis (DRAM) algorithm. The proposed approach, built on a RSM and a Kriging surrogate model of the flexibility matrix, presented results comparable with those obtained with the approach built on the FEM of the structure and it also presented an extremely higher computational performance.