Assessment of RELAP5 matrix solvers for a two-phase natural circulation loop

The heat transfer mechanism of natural convection has been extensively studied as a passive heat removal system of new nuclear power plants. Considering this aspect, the main objective of this study is to present an assessment of RELAP5 linear-equation solver under a transient two-fluid model for a two-phase natural circulation loop (NCL). For this assessment, three different approaches of linear-equation solvers for the hydrodynamic model are presented: the sparse matrix solver based on the Lower-Upper (LU) decomposition, the Border-Profile Lower Upper (BPLU) solver and the iterative method named Generalized Minimal Residual Method (GMRES). For comparison purposes, an experimental natural circulation loop made of glass tubes and using water as working fluid is analyzed. The onset of nucleate boiling observed during the experiment was predicted by all RELAP5 solvers as well as the representation of flow oscillations along the loop. Furthermore, it was noticed that the choice of the solver algorithm has a strong influence on the prediction of the two-phase natural circulation phenomena, since different wavelengths and amplitudes of flow instabilities were obtained for each approach.