Any device for mitigation of coremelt accidents must be suitable for a wide range of possible scenarios. Frequently the design of mitigative measures is based on a few enveloping scenarios. This work presents two alternative methods with respect to characteristic in-vessel scenarios. The first one is comprehensive MAAP3.0B study, where the selection of the scenarios is based on the results of a probabilistic safety analysis. In this way on the one hand more than 80% of all coremelt scenarios could be analyzed by a limited number of calculations and on the other hand likelihood distributions of some key results (e.g. begin of melting, slumping, RPV-failure, hydrogen mass, mass and temperature of melt at slumping) could be generated. The second method uses the accident progression event tree (APET) method to generate characteristic plant damage states and their estimated likelihoods.

The MAAP study showed that core degradation begins before 3 hours in about 35% and after 19 hours in 65% of all coremelt accidents. The duration of core degradation (begin melting til slumping) amounts to 40-80 minutes. At the time of slumping 50-70% of UO₂ relocate to the lower plenum in 25% and 70-76% in 75% of all scenarios. The total hydrogen mass was in 75% less than 600 kg (38% Zr-oxidation), the maximum mass was 820 kg (51% Zr-oxidation).

With the APET method a more detailed melt mass - likelihood distribution at the time of vessel failure has been obtained. The most likely (72%) melt group CM3 contains 70-90% of total UO₂ followed by group CM2 (21%) with 50-70% UO₂ and group CM4 (3%) with 90-95% UO₂. The likelihood of scenarios where vessel failure already occurs when the fraction of relocated melt is low (<50% UO₂) has been estimated to 2%.

The test LOFT-FP-2 was conducted under the auspices of the Organization for Economic Cooperation and Development at the LOFT (Loss-of-Fluid) facility, a 1:50 volumetrically scaled pressurized water test reactor, of the Idaho National Engineering Laboratory on July 9, 1985. The objective of the test was the examination of fission product release and transport phenomena during a small break loss of coolant accident. The temperature in a special designed fuel element in the center of the nuclear core exceeded 2100 K for several minutes.

The high temperature caused the degradation of the fuel element and the release of fission products and aerosols. The test was terminated by flooding the reactor pressure vessel with water. At the beginning of the flooding phase the temperature in the core increased again due to the exothermal metal-water-reaction of the rod cladding and caused further damage of the core with extensive hydrogen generation and fission product release.

The test was calculated from scram until the end of the quenching of the hot core with the system code ATHLET-CD in order to verify the code for reactor case. All important phenomena which occurred during the transient, like loss of coolant, core uncovery, heat-up and degradation, release and transport of fission products and aerosols and the final flooding were calculated well. The approach of coupling the modules which describe core degradation, fission product release and transport with the modules which describe thermalhydraulics and neutron kinetics turned out to be successful.

The code ICARE2[1] has been developed in IPSN (France) since 1987. The code was intended for simulation of thermal hydraulic processes during severe accidents of PWR NPP type core. A number of modules for different physical phenomena were included in the code: mechanical destruction of fuel rods, radiation heat transfer, main chemical reactions, etc. However the majority of models used in the code was based on the simplified approaches, which can be applied only for definite conditions. Complexity of described object required generalisation the code in a part of development more physically reasonable models. Such modernisation was conducted in 1992-1994 in co-operation NSI RAS-IPSN. Following modules were implemented in the code ICARE2 during this period (models were developed in NSI RAS in the frames of a detailed mechanistic code package SVECHA[2]):

1. UZRO (chemical behaviour during UO₂ZrH₂O interactions) - the solution of system of equations for oxygen diffusion in multilayered structure.
2. DROP (melt relocation) - the solution of one dimensional equations for heat and mass transfer for liquid masses on the surface of the cladding.
3. CROX (fuel-cladding deformation) - the analysis of mechanical behaviour of multilayer structure ().

Finally the code with new modules were tested on international experiments PHEBUS-B9+, CORA-13, PBF1.4, CORA-W1 and CORA-W2 (ISP-36 - international standard problem as result of common efforts of KfK Germany and NSI "Kurchatov Institute" Russia).

In this report the results of ICARE2 code validation on CORA-W2 experimental data are presented.

The main objectives of CORA-W2 test were investigation of physical processes (onset of temperature escalation, H2-generation, cladding failure, melting, oxidation, relocation, absorber rod behaviour) as separate effects, evaluation of their integral contribution and examination the reliability of severe accident computer codes for experiment with WWER-1000 bundle type.

The code ICARE2 with new physical models was applied to that experiment and it was shown:

1. The new modules being implemented in ICARE2 codes show adequate operationallity in frame work of integral SFD code (non significant increase of computing time, well correlation with other set of models and modules of standard ICARE2 code).
2. Validation results are in good agreement with the integral test CORA-W2 for following key quantities: hydrogen generation, temperature behaviour of the bundle on different elevations, destruction of fuel rod cladding, dynamic of melt relocation and oxidation.

FPT0 was the first test of the program and went much farther into the so called late phase than any other integral test performed so far; it fulfilled all the following objectives: high cladding oxidation without starvation conditions, fuel dissolution, FP release and significant bundle degradation with loss a rod-like geometry using a bundle of 20 fresh fuel rods plus one SIC rod pre-irradiated for 9 days.

In this paper are presented the comparisons between the calculated evolution of the bundle using ICARE2 code and the experimental results. In spite of the fact that the current V2 mod1 version of the code is only validated for the prediction of the early phase of core degradation, this code is currently used as a tool for the understanding of the late phase degradation behaviour of the bundle.

It is shown in a first time that the calculations performed with ICARE2 code, after adjustment of the Zr-rich melt relocation criteria (cladding embrittlement), simulate the experimental fuel temperatures at all measurement levels, up to the end of the oxidation escalation. During this phase, relocations of small amounts of materials occured, inducing partial blockages in the bundle between the two space grids, and causing locally limitation of the clad oxidation.

During the heat up phase of the test it is shown how these local material relocations are one of possible cause that initiated the loss of rode-like geometry in the bundle (calculated temperatures in the shroud near mid-phase during this final phase are overestimated); other possible processes are currently investigated.

These phenomena led to a formation of a molten pool in the lower part of the bundle.

In these conditions it is presently difficult to reproduce the degradation scenario of the bundle during the heat-up phase and the final state of the bundle. Nevertheless the flexibility of the code and the possibility to use different modelling options enabled some observed or expected degradation processes to be imposed: upper grid catcher effect, candling of upper rod plugs, inner rod slumping...

Finally phenomena which are modelled in ICARE2 code, and which are important to calculate FPT0 test have been identified. Main deficiencies concern the non prediction of the loss of rod-like geometry and the lack of transition toward a core debris configuration. In these areas, new development efforts are promated.

The programme is divided into eight areas including the Source Term Project. The objective of this project, which involves eighteen European organizations, is to establish a consensus on the state of knowledge and the key areas of uncertainty affecting the source term under severe accidents. Discussions to define the Source Term Project concluded that the optimum use of resources would be gained by dividing the programme into two phases. The first year (1993) was based on a series of assessments, sensitivity studies and state-of-the-art reviews to identify uncertainties in the database and areas of consensus. Work in the second phase (1994/95) is benchmark studies. The work encompasses fission product release from fuel, transport in the reactor coolant circuit and behaviour in the containment. Benchmark activities in support of the Phebus-FP project are also undertaken.

The main results of the assessment activities are described in this paper, together with preliminary results from the experimental work conducted in the second phase.

Three components are described. The liquid metal phase corresponds to the metals issued from the molten core and the possible ablation of a metallic wall (pressure vessel). The liquid oxide phase is divided into heavy oxides issued from the core and ligth oxides generated by the wall ablation. Finally the gas phase represents the gas volume above the molten core, and those possibly generated by the wall ablation.

Four mass balance equations (metals, oxides, light oxides, gas), two energy balance equations (metals and oxides), and three momentum balance equations (metals, oxides and gases) are written. The discretization uses either a rectangular or a cylindrical 3D meshing. The equations are solved with a two-step quasi-implicit method.

The constitutive laws of TOLBIAC (interfacial heat and mass transfer, wall heat transfers) are taken from the literature, or very simple models with current trends. The molten core configuration is mainly a stratified one: heavy oxides at the bottom, above it a metal layer and finally a layer of the light oxides generated in the liquid metal in contact with the wall. Other configurations may also occur, depending on the wall position (vertical or horizontal) and nature (metal, oxide with or without gas generation). A more or less high mixing of the phases then takes place near the walls, and the constitutive laws have to represent it.

However, the physical properties of the components (mixing of several materials) are not well known, and the constitutive laws may therefore be not very accurate.

The crusts formation at the wall, as well as at the upper surface, is taken into account, as a thermal resistance and a variation of the liquid volume.

The ablation of the solid structure is calculated: the geometry of the cells in contact with a solid wall changes with time. Liquid and gas are generated, depending on the nature of the wall material. The ablation velocity depends on the wall heat transfer coefficient and on the ratio between the conduction heat flux in the wall and the total wall heat flux. It is calculated by solving the thermal equation in the wall with a fine meshing near the ablation front.

A 1D conduction calculation is performed in order to determine the wall temperature profile. A wall typically consists in a concrete layer in contact with the molten core, followed by a steel layer, with imposed external temperature and heat transfer coefficient.

The fission energy source is modeled and also the heat losses at the upper surface (radiation or exchange with a water layer).

The qualification of the code is under progress. For a better physical representation, the constitutive laws have to be improved.

Some results are presented in order to illustrate the code potentialities: the behavior of a molten core in a vessel lower head on one hand, and in an external core catcher on the other hand.

Apart from this, modelling of the propagation phase -the explosion itself- is underway and will be presented.

• The results from experimental series with Sn, Al₂O₃ and UO₂-ZrO₂) melts demonstrate a strongly different behavior: only mild, barely sustained propagation for Sn (main reference experiment: KROTOS-21), strong escalations to high, supercritical pressures of ~100 MPa with Al₂O₃ (main reference experiment: KROTOS-28), applying the same trigger, and strong steam production and fine fragmentation already during premixing with UO₂-ZrO₂, currently preventing triggering and energetic interactions (KROTOS 32 to 37).
• The experiments can be taken as essentially one dimensional (test tube of 9.5 cm inner diameter and height of 1.25 m, upwards wave propagation after triggering from below).
• Indications on the integral premixture conditions are at least given by the level-meter measuring the level swell from which the total steam content can be estimated, the thermocouples indicating the arrival of melt at certain locations.
• A series of pressure transducers along the test tube allow detailed measurements of the behavior after trigerring. These serve as a basis for validating thermal detonation codes. From comparisons with resulting shock wave velocities furhter indications can directly be obtained on the composition of the premixture. In addition pressurization of the expansion vessel is also measured.

Since in KROTOS-21 and KROTOS-28 the triggering and escalation behavior is best characterized, these experiments give the basis for checking thermal detonation codes. At present, calculations of KROTOS-21 with IDEMO/1,5/, ESPROSE/4,6,7/ and TEXAS-IV/8,9/, calculations on KROTOS-28 with IDEMO/3,11/ and ESPROSE/4,7/ have been published. Calculations on another Al2O3-experiment (KROTOS-26) have also been presented in /4,8/; although early triggering makes comparisons difficult.

The pressure development in the tests can be predicted with the different codes reasonably well. Essential features of the processes can be reproduced and thus the codes appear to be in general appropriate to describe the phenomena, although details of the experimental results have not yet been reproduced. However, different assumptions on the premixture and the dominant processes have been used, yielding different interpretations of the experimental results. In /1,5/ it has been shown with IDEMO that hydrodynamic fragmentation is at all not sufficient to explain the sustained pressure waves and short-time peaks with the Sn-experiments. Thus, an additional, strong contribution of thermal fragmentation has been assumed as a possible explanation suggesting also other possibilities, such as heterogeneous water heating. This view is in principle supported by TEXAS-IV calculations in /8/, however with a much smaller thermal fragmentation which was sufficient due to emphasizing steam production. On the other hand, calculations from ESPROSE/6/ emphasize heterogeneous water heating as well as steam production showing hydrodynamic fragmentation to be sufficient under these assumptions.

Nevertheless, the task remains to understand the importance of the respective contributions and effects and, finally, to come to a general description which can be extrapolated to a wide range of conditions. This can also be supported by attempts to get more information on the premixture conditions, e.g., by calculations with premixing codes. Such calculations have already been performed with TEXAS/8/, however further confirmation is needed as well.

The situation with the Al₂O₂ experiments can be characterized similarly. Here, agreement between the calculations with IDEMO/5,11/ and ESPROSE/4,7/ exists that the hydrodynamic fragmentation is strong enough to explain the rapid escalation to very high pressures obtained experimentally. A stronger escalation from ESPROSE may be attributed again to the assumption of nonhomogeneous heating of water in contrast to the IDEMO calculation. Again, there remains a need to clarify the details of behavior and thus to come to a generalized description of the exchange processes.

In the present contribution the above status of analysis of the Sn and Al₂O₂ experiments, especially KROTOS-21 and KROTOS-28 will be discussed, considering and evaluating in detail the models applied in the different codes and the main questions posed from this investigations and the experimental results. Based on this information, new calculations with IDEMO and TEXAS IV will be performed and discussed in direct, detailed comparison. Concerning the different behavior in the UO₂-ZrO₂ experiments, the main questions will also be discussed and elaborated. Since these are obviously related to the premixing behavior and more specifically to the breakup of the melt stream, first attempts to clarify the differences will be undertaken based on film boiling and jet breakup considerations/8,10,12/.

The final intention of this study is to determine the various features which characterize the mixing, pressure escalation and propagation phases which occur during melt/coolant interactions the thus discover why these characteristics may be substantially different for Sn, Al₂O₂ and UO₂-ZrO₂ melts. This raises the concept of scaling of the explosion phenomena and what must be preserved between various tests.

1. Burger, M. et al., (IKE), Schins, H., et al., (JRC Ispra): Examination of Thermal Detonation Codes and Induced Fragmentation Models by Means of Triggered Propagation Experiments in a Tin/Water Mixture, Nucl Eng Design 131 (1991) 61-70.
2. Hohmann, H., Hegallon, D., Schins, H., Yerkess, A.: FCI Experiments in Aluminum Oxide/Water System. CSNI Specialists Mtg. on Fuel-Coolant Interactions, Santa Barbara CA, Jan 5-8, 1993.
3. Hohmann, H., et al.: Advance in the FARO/KROTOS Melt Quenching Test Series, 22nd Water Reactor Safety Information Mtg., Bethesda, Maryland, October 23-26, 1994.
4. Yuen, W. W., Theofanous, T. G.: The Prediction of 2D Thermal Detonations and Resulting Damage Potential, CSNI Specialist Mtg on Fuel-Coolant Interactions, Santa Barbara CA, Jan 5-8, 1993.
5. Burger, M., Buck, M., Muller, K., Scahtz, A.: Stepwise Verification of Thermal Detonation Models: Experimentation by Means of the KROTOS Experiments. CSNI Specialist Mtg. on Fuel-Coolant Interactions, Santa Barbara CA, Jan. 5-8, 1993.
6. Yuen, W. W., Chen X., Theofanous, T. G.: On the Fundamental Microinteractions that Support the Propagation of Steam Explosions, Proc. NURETH-5, Salt Lake City UT, Sept. 22-24, 1992, Vol II, 627-636.
7. Theofanous, T. G., Yuen, W. W.: The Prediction of Dynamic Loads from Ex-vessel Steam Explosions, Int. Conf. on “New Trends in Nuclear System Thermohydraulics,” May 30-June 2, 1994, Pisa Italy.
8. Tang, J., Corradini, M. L.: Modeling of the Complete Process of One-Dimensional Vapor Explosions. CSNI Specialist Mtg. on Fuel-Coolant Interactions, Santa Barbara CA, Jan 5-8, 1993.
9. Bang, K. H.: The Role of Fragmentation Rate in Vapor Explosion Propagation: Comparison of Models, Proc. Int. Seminar on the Physics of Vapor Explosions, Tomakomai, Japan, Oct. 25-29, 1993, pp 229-233.
10. Burger, M., Cho, S. H., v Berg, E., Schatz, A.: Breakup of Melt Jets as Precondition for Premixing, Modeling and Experimental Verification. CSNI Specialist Mtg. on Fuel-Coolant Interactions, Santa Barbara CA, Jan 5-8, 1993.
11. Burger, M., et al.: Analysis of the Thermal Detonation Experiment KROTOS-28 with the IDEMO Code. Intl “Energic-Technic-Unmelt,” U. Brochmeier (Ed), Ruhu-Univ Bochum, Inst. fur Energietechnik, Buchum, Germany, 1994.
12. Burger, M., v Berg, E., Cho, S. H., Schatz, A.: Modeling of Jet Breakup on a Key Process in Premixing Proc. Int. Seminar on the Physics of Vapor Explosions, Tomakomai, Japan, Oct. 25-29, 1993, pp 229-233.

To cope with the consequences of a severe accident means to deal with different phenomena which may threaten the integrity of the containment or may lead to an enhanced fission product release into the environment:

• high pressure reactor pressure vessel failure
• energetic molten fuel-coolant interaction (in-vessel and ex-vessel)
• direct containment heating
• molten core concrete interaction (basemat meltthrough)
• hydrogen combustion
• long term pressure and temperature increase in the containment.

• for primary side depressurization
• for hydrogen control
• for stabilization and cooling of the melt
• for containment heat removal.

To show the adequacy and feasibility of these new features quite a lot of R&D work has to be relied on and still to be performed, starting with generic questions which help to get more insight into the phenomena involved going to more design related questions to proof and demonstrate the adequacy of chosen design measures.

Research and development cooperations have recently been intensified to achieve this ambitious goal. The German Forschungszentrum Karlsruhe and the French research institutions of the CEA are working closely together with the industry having between themselves a strong cooperation in performing experimental and analytical investigations, complementing and supporting each other. Generic investigations of different institutes, institutions and universities partly sponsored by the federal government are necessary and helpful in understanding, calculating and verifying the phenomena involved. Within a greater frame parties from the European Countries have found together withing the nuclear fission safety programmes of the European Commission to reach a common understanding and to find solutions to open questions. Agreements and an open information exchange between the worldwide parties involved in nuclear energy are rounding up the picture.

Several events which are postulated in the licensing and safety assessment of CANDU reactors result in a degradation of the normal heat removal mechanisms from the fuel. In the more severe of these incidents, the pressure-tube temperatures may increase as a result of convective heat transfer from the hot coolant or steam in the channel, and due to thermal radiation from the overheated fuel bundles. If the pressure tube becomes hot enough, it will deform and contact its surrounding calandria tube. Upon contact, stored heat in the pressure tube is transferred across the interface to the calandria tube, conducted through its wall, and transferred from its outside surface into the surrounding moderator. This process significantly improves the transfer of heat from the fuel bundle to the moderator, helping to reduce fuel temperature and fission product releases.

Verification of the effectiveness of the moderator as a heat sink during a LOCA has been the focus of an extensive and ongoing research program at AECL-Research Whiteshell Laboratories. The general methodology used in the research program has been to perform small-scale separate-effects experiments to develop and validate mathematical models which describe the phenomena. These validated models are then integrated into a code (CATHENA) linking the various phenomena to characterize the fuel channel response to a LOCA. Full-scale integrated experiments are then performed to validate the code.

This paper describes the operation of the moderator as a passive heat sink in a CANDU reactor and recent experiments that verify its operation in certain severe accident situations. Recent code development and validation work, aimed at modelling this phenomenon, are also reviewed.

Nevertheless, there is a probability, however low, that the moderator heat sink could fail in such an accident. The pioneering work of Rogers (1984) for such a severe accident using simplified models show that the fuel channels would fail and a bed of dry, solid debris would be formed at the bottom of the calandria which would heat up and eventually melt. However, the molten pool of core material would be retained in the calandria vessel, cooled by the independently cooled shield-tank water, and would eventually re-solidify. Thus, the calandria vessel would act as an inherent "core-cathcer".

The present paper reviews subsequent work on the damage to a CANDU core under severe accident conditions and describes an empirically based mechanistic model of this process. It is shown that, for such severe accident sequences in a CANDU reactor, the end state following core disintegration consists of a porous bed of dry solid, coarse debris, irrespective of the initiating event and the core disassembly process.

The paper describes an improved model for the subsequent heat-up of the debris bed which includes all the significant mechanisms that may affect the thermal behaviour of the debris. The paper also describes a detailed model for the thermal behaviour of the molten pool formed by eventual melting of the solid debris bed which takes into account internal heat generation and buoyancy-driven circulation in the pool, the formation of solidified crusts on the upper and lower surfaces of the pool, heat generation in the crusts, radiation from the upper crust and radiation absorption, emmision and transmission in the steam over the pool, as well as the natural convection and nucleate boiling processes of the shield-tank water on the outer surfaces of the calandria vessel.

Application of these models to a dominant-frequency severe accident sequence in a CANDU-6 reactor is described. Results for reference conditions for the thermal transient of the debris bed show that its heat-up is relatively slow, with melting in the interior of the bed beginning about two hours after heat-up begins. However, the upper and lower surfaces of the debris remain well below the melting point and heat fluxes to the shield-tank water are well below critical heat fluxes for the existing conditions. The calandria vessel remains well-cooled and retains its integrity throughout the transient. Sensitivity studies in which important parameters are varied over wide ranges yield the same conclusions, with the results indicating that, for larger pore sizes, melting of the debris may not even occur.

Results for the thermal behavior of the subsequent molten pool show that the calandria vessel wall will be protected by a thick solid crust below the pool which will grow with time. Calandria wall temperatures and heat fluxes to the shield-tank water ensure that the calandria vessel will maintain its integrity in this phase of the transient also. Again, sensitivity studies varying the important parameters confirm the general validity of this result. Should the shield-tank water eventually boil off, the calandria vessel will fail, but thsi will not occur in less than a day, giving the operators adequate time to provide water from emergency supplies to the shield tank, which will then retain the re-soldifying core indefinitely.

Contributors to the predicted effectiveness of cooling of a degraded core in a CANDU, in addition to the inherent heat sink provided by the separate moderator and the shield tank water are the low power density, about 15.6 MW/Mg of fuel in a CANDU-6 based on the design power, and the extensive dispersion of the debris bed in the calandria resulting in a shallow molten pool depth of about 1 metre maximum and about 0.65 metre average for the reference case for a CANDU-6. These factors help to explain the different predicted behaviour of a degraded CANDU core relative to that of an LWR in typical melt-progression scenarios.

The results of this work confirm and provide more confidence in the conclusions of the early studies that the calandria vessel will retain its integrity in severe accidents in a CANDU reactor and will contain a disintegrated or molten core for a long period without operator intervention, thus acting as an effective core-catcher.