The In-Vessel Retention (IVR) strategy for Light Water Reactors (LWR) intends to stabilize and isolate corium and fission products in the reactor pressure vessel and in the primary circuit. This type of Severe Accident Management (SAM) strategy has already been incorporated in the design and SAM guidances (SAMGs) of several operating small and medium capacity LWRs (reactors below 500 MWe, e.g. VVER440) and is part of the SAMG strategies for some Gen III+ PWRs of higher power such as the AP1000 or the APR1400. However, the demonstration of IVR feasibility for high power reactors requires using less conservative models as the safety margins are reduced.
In Europe, the IVMR project aims at providing new experimental data and a harmonized methodology for IVR. A synthesis of the methodology applied to demonstrate the efficiency of IVR strategy for VVER-440 in Europe (Finland, Slovakia, Hungary and Czech Republic) was made. It showed very consistent results, following quite comparable methodologies. The main weakness was identified in the evaluation of the heat flux that could be reached in transient situations, e.g. under the “3-layers” configuration, where the “focusing effect” may cause higher heat fluxes than in steady-state (due to transient “thin” metal layer on top). Analyses of various designs of reactors with a power between 900 and 1300 MWe were also made. Different models for the description of the molten pool were used: homogeneous, stratified with fixed configuration, stratified with evolving configuration. The last type of model provides the highest heat fluxes (above 3 W/m2) whereas the first type provides the lowest heat fluxes (around 500 kW/m2) but this model is not realistic due to the immiscibility of molten steel with oxide melt. Obviously, there is a need to reach a consensus about best estimate practices for IVR assessment to be used in the major codes used for safety analysis, such as ASTEC, MELCOR, SOCRAT, MAAP, ATHLET-CD, SCDAP/RELAP, etc. Despite the model discrepancies, and leaving aside the unrealistic case of homogeneous pool, the average calculated heat fluxes can reach, in many cases, values which are well above 1 MW/m2. This could reduce the residual thickness of the vessel considerably and threaten its strength and integrity. Therefore, it is clear that the safety demonstration of IVR in high power reactors requires a more careful evaluation of the situations which can lead to formation of either a very thin top metal layer provoking the focusing effect or significantly overheated metal, e.g. after oxide and metal layer inversion. Both situations are illustrated in this paper. The demonstration also requires an accurate thermo-mechanical analysis of the ablated vessel. The standard approach based on “yield stress” (plastic behaviour) is compared with more detailed calculations made on realistic profiles of ablated vessels. The validity of the standard approach is discussed.
The current approach followed by many experts for IVR is a compromise between a deterministic analysis using the significant knowledge gained during the last two decades and a probabilistic analysis to take into account large uncertainties due to the lack of data for some physical phenomena, e.g. associated with molten pool transient behaviour, and due to excessive simplifications of models. A harmonization of the positions of safety authorities on the IVR strategy is necessary to allow decision making based on shared scientific knowledge. Some elements that might help to reach such harmonization are proposed in this paper, with a preliminary revision of the methodology that could be used to address the IVR issue. In the proposed revised methodology, the safety criterion is not based on a comparison of the heat flux and the Critical Heat Flux (CHF) profiles as in the current approaches but on the minimum vessel thickness reached after ablation and the maximum pressure load that is applied to the vessel during the transient. The main advantage of this revised criterion is in consideration of both steady-state and transient loads on the RPV. Another advantage is that this criterion is more straightforward to be used in a deterministic approach.