Estimativa de propriedades termofísicas espacialmente variáveis com seleção de classes de modelos e transformações integrais

Abstract

The identification of spatially varying thermophysical properties is crucial in heat transfer problems in composite materials and Functionally Graded Materials (FGMs), with various applications in engineering, ranging from heat sinks in microprocessors to aerospace engineering. This dissertation addresses the formulation and solution of an inverse heat conduction problem using a Bayesian inference approach with model class selection. The focus is on a heat transfer problem in a thermally thin flat plate. Synthetic experimental data were employed to verify the efficiency of the Transitional Markov Chain Monte Carlo method, in conjunction with integral transforms, in parameter estimation and selecting the most plausible model. The use of integral transforms allowed for analytical solutions of the inverse problem involving the parameterized models chosen to describe the thermophysical properties. The results, using synthetic temperature measurements, demonstrate that the model selection methodology is capable of identifying both qualitatively and quantitatively the probable models in heat conduction problems, especially in contexts involving thermally thin flat plates.