A variant of the ADI method for two-phase flow calculations (Q1327264)

For the purpose of safety analysis of fast-breeder nuclear reactors, computer codes have been developed to describe the three-dimensional single and two-phase flows within the reactor. The author of this paper developed such a code (BACCHUS-3D/TP) in the Argonne National Laboratory in the United States, and the purpose of this paper is to document recent improvements on the numerical schemes for the two-phase flow calculations. Since the pressure distribution within the coolant of the reactor is usually governed by a Poisson equation, the author introduces initially a model Poisson equation with the known exact solution. Solutions are obtained and compared with schemes of matrix inversion, vectorized Gaussian elimination, classical Peaceman and Rachford alternating direction implicit scheme, the ADI-variant of Press-Flannery scheme, and the ADI-variant of Messina-Londrillo (ML-ADI) scheme. It was observed that the schemes of ML-ADI and Gaussian elimination are superior in savings of computing time. However, with the cell number equal or larger than 3000, the ML-ADI scheme becomes faster without significantly reducing the accuracy of the numerical results. The author then reports the results of applications of the ML-ADI and the Gaussian elimination schemes to three-dimensional two-phase flow problems where the effect of compressibility must be accounted for. It can be observed that the ML-ADI scheme is better with proper choice of the number of cycles of the convergence-accelerating parameter over the allowed range and the number of steps per cycle. Future plans of adapting the method of preconditioned conjugate gradients and multi-grid methods are also mentioned.