Development of a Drift-Flux Model Based Pinwise Whole Core Thermal-Hydraulics Analysis Code Aiming at Highly Parallelized Execution

Abstract

In order to improve the parallel computation efficiency of neutronics/thermal-hydraulics (T/H) coupled reactor core calculations, a core T/H analysis code ESCOT, that can handle pinwise flow channels in the whole core calculation, is developed based on the drift-flux model and a SIMPLE-like numerical solution scheme. The governing equations are formulated and discretized from a three-dimensional 4-equation model to derive the pressure equation coupled with the equations of scalar variables. The initial verification and validation are performed for single-phase flow conditions to assure the accuracy of the code. The calculated results are comparing with the analytic solutions, experiments, and the results of other codes such as CUPID, CTF and MATRA. The selected problems deal with the following phenomena: pressure drop by gravity acceleration and spacer grids, turbulent mixing, crossflow by friction-flow-split and asymmetric flow inlet, reverse flow by recirculation, and simplified main steam line break (MSLB) accident. It turns out that ESCOT is about 3 times faster than CTF while retaining comparable accuracy. In order to establish an effective linear solver for the pressure equation on parallel computing platforms, the efficiency of various linear solvers is examined. The selected linear solvers are a direct solver, SuperLU, and two Krylov subspace algorithms, GMRES and BiCGSTAB. The BILU3D preconditioner is applied to accelerate the Krylov subspace algorithms, and the Krylov subspace calculation modules are parallelized with OpenMP. The incomplete domain decomposition is applied to forward and backward substitutions to solve the preconditioner equation in parallel. Parallel performance tests are carried out with sample problems, and it is shown that the unpreconditioned BiCGSTAB yields the best performance in terms of computing time and speedup.