Study of 2-D/1-D fusion method with partial current-based coarse mesh finite difference acceleration for 3-D whole-core transport calculation

Abstract

In the conventional nuclear reactor analysis, the whole-core diffusion method is used in which nodal parameters are provided by a single-assembly lattice calculation with the zero net current boundary condition. Thus, the whole-core solution is not transport, because the inter-assembly transport effect is not incorporated. In recent years, three-dimensional (3-D) whole-core neutron transport calculation has been drawing increasing interest due to the demand for an accurate solution, especially since computing power has increased. However, a direct 3-D transport calculation without spatial homogenization still imposes a huge computational burden. To resolve these problems, this dissertation presents a novel approach: nonoverlapping local/global (NLG) iteration. In NLG, a whole-core domain is decomposed into nonoverlapping local problems, with local problem transport solutions then embedded within the partial current-based coarse-mesh finite difference (p-CMFD) methodology in a two-level iterative scheme to provide a whole-core transport solution. On the other hand, two two-dimensional/one-dimensional (2-D/1-D) methods, fusion and hybrid, have been developed and reported in the literature to deal with 3-D heterogeneous reactor problems and to avoid direct 3-D transport calculation. The 2-D/1-D fusion method transforms a 3-D transport problem into 2-D and 1-D transport problems that have a smaller computational burden than the original problem. The hybrid method uses, in the axial direction, an additional diffusion (or $SP_3$) approximation to enhance the efficiency of the calculation. This dissertation performed the stability and accuracy analysis of the two aforementioned methods. The results indicate that the fusion method is stable and it is more accurate than the hybrid method. Hence the 2-D/1-D fusion method is chosen as the local transport solver kernel in NLG iteration. NLG iteration with 2-D/1-D fusion kernel was applied to three configurations of the C5G7 OECD/NEA 3-D benchmark problem and to a modified C5G7 benchmark problem with explicitly modeled cladding in this dissertation. From the numerical results, it can be concluded that NLG iteration with the 2-D/1-D fusion kernel will be a useful computational framework for efficient and accurate reactor core design analysis.