Adjoint Based Uncertainty Quantification and Sensitivity Analysis for Nuclear Thermal-Fluids Codes.

[摘要] In this thesis the various methods of obtaining sensitivity derivatives are explored in order to highlight the significant contributors of uncertainty, and the adjoint approach is shown to be the most efficient. The implementation of theadjoint method into the NRC Advanced Gas REactor Evaluator (AGREE) code is performed. Additionally, a method for obtaining an approximation that has a leading error term of third order is derived by calculating both forward and adjoint sensitivities. The approximation is demonstrated to be accurate locally. A method for combining these surrogates is demonstrated using the DAKOTA code.The utility of the method is demonstrated by performing three different analysesrelated to the High Temperature Test Reactor (HTTR) in Japan. First, the bypass flow experiment of Kaburaki and Takizuka is analyzed to inspect the sensitivity and variability of the bypass flow with respect to the cross-flow gap geometry, loss coefficient and boundary conditions. Next, the HENDEL experiment isanalyzed to investigate the factors impacting peak core temperatures and the initial amount of stored energy in the core. Last, an analysis of the actual HTTR reactor is performed. A key result from these analyses is that the cross-flow gap geometry and loss coefficients were of relatively low significance with regard tothe quantities that pertain to core temperature.

[发布机构] University of Michigan

[效力级别] Uncertainty Quantification

[关键词] Adjoint;Uncertainty Quantification;Sensitivity Analysis;Nuclear Engineering and Radiological Sciences;Engineering;Nuclear Engineering & Radiological Sciences