Abstract
(Englisch)
|
Melt through experiments CORVIS
One of the objectives of the research on nuclear safety is to evaluate the behaviour of the reactor pres-sure vessel (RPV) under thermal and mechanical loading in case of a severe accident. The re-search pro-ject CORVIS (Corium Reactor Vessel Interaction Studies) conducted at PSI is an investi-gation of a poten-tial failure of the RPV lower head when attacked by molten corium during a core melt accident. Melting through ex-periments are performed, at atmospheric pressure, using models of RPV lower heads pene-tration tubes of reactor-typical designs. The corium substitute is an alumina/iron thermite melt of which the oxidic and/or the metallic part can be applied. The decay heat is simulated by a su-stained heating of the corium substitute with a submerged electric arc. Heating up of the steel structu-res, structural deformations and thermal effects of oxide crusts are observed. To give an order of the scale of the experiments, the test vessels have a diameter of up 740 mm, the models of the RPV wall (vessel bottom) are plates manufac-tured of reactor steel with a thickness of 100 mm, tube penetrations inserted in the plate have scale 1:1 compared to reactor designs, the oxide melt used has a mass of 450 kg and the specific power of the arc heater is ³250 W/kg of melt. Computational modell³ing supple-ments the measurements in order to provide a deeper understanding of the physical processes under-lying the observations, in order to apply the knowledge to a wider range of conditions, for example high pressure.
A long series of tests has been performed since February 1993 through September 1996:
· 16 experiments to develop and qualify the experimental technique (93/2-96/3),
· 1large scale test on a lower head without penetrations (93/6),
· 2 large scale tests on a lower head with a BWR drain line according to the specifications by General Electric (94/12-96/5),
·1 large scale test on a PWR lower head with instrument tube penetrations according to specifi- cations by Westinghouse (96/9).
Three experiments were carried out in 1996. Including the experience from earlier experiments, se-veral conclusions can now be made about the RPV response under melt attack.
Experiment 01/11 (March 28,1996)
This experiment was the first test of a sustained heating in a melt of pure aluminium oxide. The test plate did not carry penetra-tions. The experiment demonstrated that a ceramic melt tends to solidify and form a crust; no melting of the test plate occurred.
Experiment 03/2 (May 30,1996)
Experiment 03/2 was to completed the investigation of the drain line by in-vestiga-ting its behaviour under attack by an almost pure oxidic melt. The oxide melt with an initial temperature of 2245 °C - 2270 °C flowed into the drain line and filled it com-pletely along its 7 m length. The melt flow was stopped by the steel plug at the end of the tube extension. Otherwise, the melt would have penetrated further. For a real core melt accident it was con-cluded that the melt would penetrate the entire length of that section of the drain line which does not contain any residual water. Despite small displacements the test section did not fail. It was clear however, that the drain line would have been torn off at higher inter-nal pressure, due to accumulation of creep damage. Abla-tion of the drain line by the flowing oxide melt was negligible.
Experiment 04/1 (September 19,1996)
The test section was a model of a PWR lower head according to a reactor of Westinghouse design. The test plate carried 5 penetrations with long tube extensions at the lower side, representing the guide tubes for the movable neutron flux detectors. The construction of the penetrations were the same as those of the Swiss PWR Beznau. A pure oxide melt of 450 kg was used. A small amount of oxide melt entered the 4 open tubes and penetrated to distances up to 4 m. In particular, the penetrations had not slipped out of their seats in the test plate. Again, the measured temperatures and the other observations pro-vide data to support develop-ment and qualification for analyses of the lower head behaviour.
Validation of cumputational models
PSI performed analyses of the CORVIS experiments in order to interpret the data in terms of the processes which govern the thermal attack and possible melt-through of the lower head, and to vali-date models to be applied to plant accident situations. The analyses focussed mainly on the following aspects of the behaviour: the thermal transport within the molten pool and to the structure, the free-zing of the molten material as it cools, the formation and remelting of a solid crust, the thermal re-sistance between the solid crust and the lower head structure, melt behaviour in penetration tubes and the ther-mal and me-chanical response of the heated, ablated structure. The main tool was the ADINA finite-element set of computer codes which provide for computational fluid dynamic (CFD), heat con-duction, and structural analyses.
Detailed CFD calculations with ADINA demonstrated strong convection effects in the region above the bottom of the heating element, whereas, the fluid below the heating element was comparatively stag-nant, and thermally stratified. This result is corroborated by the CORVIS experimental data and results of small scale CORVIS-counterpart experiments performed at FZK. Post-test analyses show that the material froze readily on the lower plate. The insights gained from the detailed calculations make it pos-sible to define a simple engineering model for the thermal transport, based on enhanced heat conduc-tion in the molten regions.
The crust behaviour was found to be crucial to whether or not the structures are melted. In none of the ex-periments involving ceramic melt did any structural melting occur. In tests involving instrumention tubes, ceramic melt penetrated some distance into the tubes before freezing, but did not ablate the tube wall. The penetration distance was typically larger than predicted by the current models for flowing melts, indi-cating that additional processes, not yet understood, control the heat transport.
In contrast with ceramic melt, a highly superheated metallic melt resulted in heat fluxes from the melt that were much higher than could be conducted away through the wall, and significant structural mel-ting oc-curred. The melting was augmented by the erosive effect of the flowing melt, which removed the newly melted (and not superheated) structural steel and brought additional superheated material to the melting location. In the case of metallic melt flowing through a BWR drainline, CORVIS 3/1, the initial contact temperature was below the melting point, hence a metallic crust formed at first. After a very short period, the large heat flux from the flowing melt quickly remelted the crust and then melted the inside of drain line itself, leading to failure. Calculations correctly reproduced the temperatures and time of failure. Thus the applied calculational method could be validated.
The buildup of a crust strongly inhibited the heat flux from a ceramic melt. Not only does the lower con-ductivity of the solid crust reduce limit the heat transfer, but the imperfect contact or gap between the crust and the structure presents an additional thermal barrier. Analysis of CORVIS 1/11 indicates a time-varying gap resistance, apparently due to thermal deformations. The effect of (thermally or mecha-ni-cally) induced deformation on opening of a gap has important implications for the heat flux to the vessel wall, particularly if liquid water is present in which case additional coo-ling of the structure is possible.
|
