Faire avancer la sûreté nucléaire

La Recherchev2


Air ingress issue in PWRs



Email :

Mot de passe :

QUENCH workshop, 13-15 Octobre 2003, Karlsruhe
L. Cantrel, B. Clément, P. Giordano, M. Kissane, C. Séropian.

Type de document > *Congrès/colloque

Mots clés > sûreté, accident grave, calcul réacteur, intrusion d'air, ruthénium

Unité de recherche > IRSN/DPAM/SEMIC/LEPF

Auteurs > GIORDANO Patrice, SEROPIAN Christophe

Date de publication > 13/10/2003


When air is in contact with UO2 irradiated fuel, the latter and its Fission Products (FP) undergo oxidation reactions, especially under high temperature conditions such as those encountered in a severe accident. Some FPs, which are known to be hardly released during severe accident progression in a reactor core under steam oxidizing conditions, may form volatile oxide species under air oxidizing conditions. This is particularly the case for ruthenium. These volatile oxide species may keep a relatively high stability even when passing through lower temperature areas in the primary circuit, and thus may reach the reactor containment under gaseous forms, and then may potentially be released to the outside environment. Circumstances in which air can come in contact with irradiated fuel have been first investigated. Then, air ingress flows have been evaluated, and a review of knowledge on the potential effects on Ru release, transport in circuit up to the environment has been performed.

Basically, two possible accidental situations leading to air/fuel contact may be considered in a Pressurized Water Reactor (PWR) : first when there is a loss of coolant during the shut-down state, and in second when there is a lower vessel head failure (LHF), after core meltdown accident. Probabilistic Safety Assessment level 1 (PSA level 1) carried out at IRSN evaluated probabilities of first situations to ~1.4 10-6/reactor·year ; this figure represents about 7 % of the total frequency of core meltdown -which is quite high in view of the time spent in this shutdown state. Evaluations for such accidents are being revised, likely leading to significant reduction. Indeed, the time window during which the upper head is removed and the reactor pool is not yet filled is very narrow, and once the reactor pool is filled, operators have much more time to intervene. Probabilities related to the second situations are small, but high enough to need consideration. Moreover, such situations have another implication for the source term: in a very oxidising environment (i.e., in air) : previously deposited FPs (during the so-called in-vessel phase of the accident) would partly be volatilised, possibly inducing a delayed release to the containment. In a priority manner, these second situations have been investigated.

Some previous studies have been carried out at Sandia National Laboratories for evaluating in-vessel air flows, examining the “chimney” effect that develops after the reactor vessel is ruptured. It was first concluded that most of the flow regime was concentrated between 10 to 200mol/s ; more recent studies estimated the air flows range between 2mol/s to 20mol/s. The approach presented here is different from the previous one as it is based on thermal-hydraulic calculations with a lumped-parameter code, namely ASTEC, applied to a 900MWe French PWR containment. For these calculations, a quite detailed nodalization scheme was used (more than 50 zones). The scenario investigated was a loss-of-coolant accident (LOCA) due to a 12’’ break located on a hot leg, water injection safety systems being not available. The reactor core underwent severe damage up to the lower head vessel failure (LHF). This scenario was chosen because it represents conditions where an air ingress may be favoured, as size and location of the break reduce pressure losses along the path way from the reactor cavity to the upper part of the containment : if no air ingress would be observed in such a situation, there is little “chance” to get it in other scenarios. In a first step, the whole severe accident transient has been calculated with the ASTEC integral code; boundary conditions coming from this first calculation were used as an input for a pure thermal-hydraulic calculation, with a detailed containment scheme, in a second step.

Results for flow in the containment can be illustrated with three “snapshots” at different instants:

- before LHF: the steam issued from the RPV blowdown at primary circuit break induces very soon a containment pressurization, leading to the failure of the rupture disk connecting the cavity and the upper part of the containment; then, during the in-vessel phase of the accident, a convective flow develops from the upper parts of the containment down to the ventilation chimney compartment through this connection and exits the cavity through the annulus space around the reactor vessel.
- at LHF and during about 11mn, due to the significant gas mass flows produced by MCCI, all gases exit the cavity; the steel door connecting the ventilation room to adjacent rooms is assumed to leak as well. All the oxygen initially present in the cavity is blown away, so there is no way for any air ingress in RPV.
- after a few minutes, flows is again reversed: a natural circulation between the upper containment and the cavity through the RPV takes form, and lasts for several hours. Such a convective loop brings air into the RPV.

Explanation of this flow behaviour is the following : the cavity pressurization due to MCCI is actually overwhelmed by buoyancy forces between the ventilation chimney and the reactor cavity. The temperature difference between these zones is high enough to induce gas density differences. Considering the height of the ventilation chimney (~ 10m), it is clear that pressure difference between these zones due to gravity drives the flow pattern. This situation lasts for several hours ; a quasi-steady state seems to be established. Only 8% of the total mass flow entering the cavity passes through the RPV, while 92% exits the cavity through the vessel annulus space. This is a particular feature of « French like open cavities », for which there is no major role of cavity pressurization. Air molar flow in the RPV can be estimated to be ~7 mol/s.

From this “base case” calculation, several sensitivity calculations were performed to investigate the influence of different parameters. Results indicate that in most cases, an air ingression in the reactor vessel is likely to occur some time after the beginning of MCCI, with a typical air molar flow of ~ 10 mol/s (note the same order of magnitude as for previous studies). Surprisingly, there is a limited role of the RPV chimney effect itself: there is no significant increase of the up-draught flow rate when the temperature of the remaining fuel is increased in the vessel. On the other hand, temperature in the cavity and in the adjacent zones, namely zones above RPV and ventilation chimney, are of much higher importance.

After evaluating in-vessel air flows, a rapid survey of knowledge on Ru behaviour is then proposed. From a large set of small scale experiments on Ru release from fuel under air environment, release rate is found dependent on oxygen partial pressure and temperature. Total Ru release from bare fuel in air was observed in a few minutes at 1600°C, while at 1200°C, a few hours were needed. A typical delay of ~2000s after volatile FP release is observed for Ru release. Interpretation is that Ru must be oxidised before it is released and that competition for oxygen does not favour Ru oxidation until sample is fully oxidized (UO2 included).
Much less experiments are available for Ru transport within the circuit. From available data, at low temperature, the partial pressure of RuO4 is surprisingly found largely higher than the equilibrium one. Interpretation is that the endothermic step, RuO4(g) -> RuO2(g) +O2, prevents the thermodynamically favoured RuO4(g) -> RuO2(c) +O2 .
There is practically no experiment studying the Ru behaviour in a reactor containment building under severe accident conditions. Ru species exiting the circuit after an in-vessel air ingress are supposed to be mainly RuO2 and RuO4. RuO2 would behave as a settling aerosol, and supposed to be non reactive in sump (insoluble aerosol), except if polymeric forms can be formed. RuO2 aerosol deposited on walls are supposed to be able to be oxidized (i.e. RuO2(s) + 2/3 O3(g) -> RuO4(g) ). From preliminary review, it can’t be excluded that a significant fraction of volatile Ru can be found in the reactor containment building (partial pressure ranging form 10-8 to 10-6 bar). Concerning RuO4 species, its behaviour is quite unknown up to now.

Finally, beside this studies, several outstanding points remain and should be addressed:

- what is the remaining fuel state in the vessel when the air ingress occurs?
- what is the amount of Ru which can be released from this remaining fuel ?
- what is the amount of released Ru which can reach the reactor containment?
- and finally, how does Ru behave when entering in the reactor containment?