Pin-by-Pin core calculation with NEM-based two-level CMFD algorithm

Abstract

In this thesis, a new global-local hybrid CMFD method has been developed for an efficient parallel calculation of the pin-by-pin heterogeneous core analysis. For the hybrid CMFD method, the one-node CMFD scheme is combined with a NEM-based two-node CMFD method in a non-linear way. One of the major objectives of this research is to parallelize the hCMFD method to solve the steady-state reactor problems comprised of rectangular heterogeneous fuel assemblies in an OpenMP parallel platform. In the global-local hCMFD algorithm, the global problem is a coarse-mesh eigenvalue one, while the local problems are fixed-source one and each local problem is solved with incoming partial current conditions determined by the global solution. The global problem is determined by one-node CMFD formulation, in which two corrections on an interface are introduced to preserve both surface-average net current and flux. For efficient local analysis, the well-established NEM-based CMFD method is utilized to solve the local fixed-source problems. In the case of heterogeneous local problems, each local domain is homogenized using the local solutions and the resulting homogeneous cross sections are utilized in subsequent global problems. In the hCMFD method, all the local problems can be solved in parallel using the incoming partial current boundary conditions. For a consistent approximation of the transverse leakage for the NEM nodal method, the transverse leakage on an interface boundary is constructed with box-averaged flux in the neighboring cell of the adjoining local problem. On a sequential computer, the hCMFD is easily implemented with the sequential sweeping of the partial currents, i.e., outgoing current of a local domain can be used as incoming current for the neighboring local domain. For the verification of accuracy and performance of parallel calculation, four benchmarks problems are analyzed by the hCMFD algorithm. To derive the optimal iteration strategy of the two-level global-local hCMFD algorithm, various sensitivity tests are performed for the test problems. Also, the robustness of the global-local hCMFD is tested in terms of the number of partial current sweepings over the local domains. In this thesis, the parallel performance of the hCMFD algorithm is evaluated on a simple OpenMP platform for the benchmark problems. Based on the numerical results of the study, it is concluded that the new hCMFD method will enable a very efficient parallel calculation of pin-wise heterogeneous two-dimensional core analysis.