The TMI-2 accident and recently Fukushima accident demonstrated that the nuclear safety philosophy has to cover accident sequences involving massive core melt in order to develop reliable mitigation strategies for both, existing and advanced reactors. Although severe accidents are low likelihood and might be caused only by multiple failures, accident management is implemented for controlling their course and mitigating their consequences. In case of severe accident, the fuel rods may be severely damaged and oxidized. Finally, they collapse and form a debris bed on core support plate. Removal of decay heat from a damaged core is a challenging issue because of the difficulty for water to penetrate inside a porous medium. The reflooding (injection of water into core) may be applied only if the availability of safety injection is recovered during accident. If the injection becomes available only in the late phase of accident, water will enter a core configuration that will differ from original rod-bundle geometry and will resemble to the severe damaged core observed in TMI-2. The higher temperatures and smaller hydraulic diameters in a porous medium make the coolability more difficult than for intact fuel rods under typical loss of coolant accident conditions. The modeling of this kind of hydraulic and heat transfer is a one of key objectives of this. At IRSN, part of the studies is realized using an European thermo-hydraulic computer code for severe accident analysis ICARE-CATHARE. The objective of this thesis is to develop a 3D reflood model (implemented into ICARE-CATHARE) that is able to treat different configurations of degraded core in a case of severe accident. The proposed model is characterized by treating of non-equilibrium thermal between the solid, liquid and gas phase. It includes also two momentum balance equations. The model is based on a previously developed model but is improved in order to take into account intense boiling regimes (in particular nucleate boiling). Moreover, the criteria characterizing the transition between different flow regimes were completed. Currently, the French IRSN sets up two experimental facilities, PEARL and PRELUDE. The aim is to predict the consequences of the reflooding of a severely damaged reactor core where a large part of the core has collapsed and formed a debris bed e.g. particles with characteristic length-scale: 1 to 5mm. This means the prediction of debris coolability, front propagation and steam production during the quenching after the water injection. A series of experiments performed in 2010-2012 at the PRELUDE facility has provided a large amount of new data that are summarized. On the basis of those experimental results, the thermal hydraulic features of the quench front have been analyzed and the intensity of heat transfer regimes is estimated. A three-equation model for the two-phase flow in a heat-generating porous medium was validated. The quantitative validation of model with experimental results was realized and showed that the model provides satisfactory results. The model is able to predict the quench front velocity in the core, steam production (instantaneous and cumulated) as well as the pressure increase during reflood for different particle diameters and different injection liquid flows.