Estimativas de dose devido a acidentes nucleares baseadas em medidas de campo e otimização por enxame de partículas

• Ano de defesa: 2017
• Autor(a) principal: Przewodowski Filho, André
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Nuclear; UFRJ
• Palavras-chave em Português: Engenharia Nuclear; Problema inverso; Correção distribuição de dose; CNPQ::ENGENHARIAS::ENGENHARIA NUCLEAR
• Link de acesso: http://hdl.handle.net/11422/11119

It is well-known that under severe nuclear power plant (NPP) accident conditions, the accident progression becomes unpredictable and the source term estimation may be imprecise by orders of magnitude, leading to dose predictions very far from reality. This work presents a novel approach to improve radiation dose predictions based on field measurements and particle swarm optimization (PSO). The idea is to find a correction matrix, which applied to the originally estimated (incorrect) dose distribution map, generates a corrected one, which better fits to the actual field measurements. The proposed correction matrix is a concatenation of geometric transformations, weighted by a multiplication factor, aimed to fit the shape and dose intensities of the original to the actual dose distribution map. In order to find optimum transformations, a PSO has been used. Results demonstrate that PSO was able to find good correction matrix, which can be used to better project future dose distributions.