Determinação de propriedades térmicas, radiativas e coeficientes de convecção com a otimização global topográfica

• Autor(a) principal: Jardim, Lucas Correia da Silva
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13731
• topic: Topographical global optimization; Heat transfer; Conduction; Radiation; Nelder-mead; Levenberg-Marquardt; Mersenne Twister; Sobol sequence; Otimização global topográfica; Transferência de calor; Condução; Radiação; Nelder-Mead; Levenberg-Marquardt; Mersenne-Twister; Sequência de Sobol; Otimização matemática; Problemas inversos (Equações diferenciais); Calor -Transmissão; Estimativa de parâmetro; Calor - Condução; CNPQ::CIENCIAS EXATAS E DA TERRA::MATEMATICA::MATEMATICA APLICADA

To estimate unknown parameters from available experimental data is a problem commonly encountered in several areas of engineering. A very used technique to obtain this estimate is the formulation and solution of an inverse problem. When implicitly formulated, through the maximum likelihood hypothesis, this problem is treated as a global optimization problem where it is necessary to obtain the minimum of a function composed by the sum of the quadratic residuals between experimental values and those obtained from a model. In this work, three heat transfer inverse problems are formulated and solved to obtain estimates of parameters such as thermal conductivity, specific heat at constant pressure, scattering albedo, among others. To do so, the technique known as Topographic Global Optimization (TGO) is employed. This technique distributes points in a search space and, through the topographical information of the objective function, selects points considered topographical minimum. These minima are then used as the initial solution for a local search method. The main objective of this work is to compare the quality of the solution obtained with the TGO when it is implemented with random point generators and different local search methods. For the generation of initial points, the pseudo-random routine known as Mersenne Twister is compared to the Sobol Sequence, which generates quasi-random points. For the local search, the well-known Nelder-Mead and Levenberg-Marquadt methods are used. Therefore, four different combinations of TGO are used - the Wilcoxon Test is used at the end for a more reliable comparison between them. The obtained results, besides pointing out the qualities and defects of the different approaches, shows a better final performance when TGO is used with the Sobol Sequence and the Nelder-Mead.