Utilização do método da amostragem para propagação de incertezas de parâmetros físicos em sistemas com material físsil

• Ano de defesa: 2015
• Autor(a) principal: Daniel de Almeida Magalhaes Campolina
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: MCNPX; Método da Amostragem Aleatória; Quantificação de incertezas; Engenharia nuclear
• Link de acesso: http://hdl.handle.net/1843/BUBD-A4BLH6

There is an uncertainty for all the components that comprise the model of a nuclear system. Assessing the impact of uncertainties in the simulation of fissionable material systems is essential for a realistic calculation that has been replacing conservative model calculations as the computational power increases. The propagation of uncertainty in a simulation using a Monte Carlo code by sampling the input parameters is recent because of the huge computational effort required. By analyzing the propagated uncertainty to the effective neutron multiplication factor (keff ), the effects of the sample size, computational uncertainty and efficiency of a random number generator to represent the distributions that characterize physical uncertainty in a light water reactor was investigated. A program entitled GB_sample was implemented to enable the application of the random sampling method, which requires an automated process and robust statistical tools. The program was based on the black box model and the MCNPX code was used in and parallel processing for the calculation of particle transport. The uncertainties considered were taken from a benchmark experiment in which the effects in keff due to physical uncertainties is done through a conservative method. In this work a script called GB_sample was implemented to automate the sampling based method, use multiprocessing and assure the necessary robustness. It has been found the possibility of improving the efficiency of the random sampling method by selecting distributions obtained from a random number generator in order to obtain a better representation of uncertainty figures. After the convergence of the method is achieved, in order to reduce the variance of the uncertainty propagated without increase in computational time, it was found the best number o components to be sampled. It was also observed that if the sampling method is used to calculate the effect on keff due to physical uncertainties reported by manufacturers, there will be a reduction in the value compared to the conservative model. The results made it possible to verify the correct functioning of the program developed and show the potential of the sampling method for propagation of uncertainties, especially when many uncertainties are evaluated together in the same input.