Improvement of quench front modelling for thermohydraulic system codes

Abstract

The accident of Fukushima in Japan in March 2011 highlighted the need of further research on severe accidents and on the prediction capabilities of current integral system codes. The present work deals with simulation of quenching of hot particles bed, which may form in the core from the melting and fragmentation of core components during a severe accident. The coolability of hot debris bed, still generating decay heat and threatening to re-melt, is a key-issue in terms of severe accident management. Codes that are supposed to be used as decision-making tools have to be able to calculate accident sequences quickly and accurately enough. For this reason, the computation domain is typically coarsely discretized, yielding large mesh cells (mesh size > 20 cm). The capabilities of COCOMO-3D regarding simulations of quenching of hot debris bed is first assessed against experiments such as DEBRIS and PEARL, or by simulating the quenching of a reactor-scale debris bed, on which larger mesh cells can be generated like for integral codes. The latter simulation with large cells yields strong computation instabilities, due to the fact that two-phase cells, represented as a homogeneous water-steam mixture, are no longer representative of the real topology. Therefore, a new method is developed in order to track and reconstruct the quench front in an unstructured meshing evolving with time. Additionally, the mass, momentum and energy conservation equations for water and steam have to be locally adapted in order to take into account the moving quench front. Moreover, the present work proposes a method to reproduce the geometry of the debris bed domain, since a coarse meshing cannot reproduce pre-defined bed geometries properly, i.e. without smearing of the boundaries. The new modelling is verified against base cases that are analytically solvable. Finally, the capability of simulating with coarse meshing (and large cells), quickly and without any instabilities, was assessed by repeating the reactor-scale simulations.