Professional Documents
Culture Documents
h i g h l i g h t s
• We present the status modeling of the fuel–coolant interaction premixing stage in the computer code MC3D.
• We also propose a general state of the art, highlighting recent improvements in understanding and modeling, remaining difficulties, controversies and
needs.
• We highlight the need for improving the understanding of the melt fragmentation and oxidation.
• The verification basis is presented.
a r t i c l e i n f o a b s t r a c t
Article history: Fuel–coolant interaction is a complex mixing process that can occur during the course of a severe acci-
Received 5 April 2014 dent in a nuclear power plant involving core melting and relocation. Under certain circumstances, a steam
Received in revised form 5 August 2014 explosion might develop during the mixing of the melt and the water and induce a loss of integrity of
Accepted 13 August 2014
the containment. Even in the absence of an explosion, studying the mixing phenomenon is also of high
interest due to its strong impact on the progression of the accident (debris bed formation, hydrogen pro-
duction). This article is the first of two aiming at presenting both a status of research and understanding of
fuel–coolant interaction and the main characteristics of the model developed in the 3-dimensional com-
puter code MC3D. It is devoted to the premixing phase whereas the second is related to the explosion
phase. A special attention is given to major difficulties, uncertainties and needs for further improve-
ments in knowledge and modeling. We discuss more particularly the major phenomena that are melt
fragmentation and film boiling heat transfer and the challenges related to modeling melt solidification
and oxidation. Some highlights related to the code verification are finally given.
© 2014 Elsevier B.V. All rights reserved.
1. Introduction be of the order of some millimeters. The melt drops will rapidly
cool down and settle on the vessel bottom or on the pit floor. If the
During a severe accident in a nuclear power plant (PWR, BWR), fragmentation is complete and if melt drops are sufficiently cooled
the core may melt and flow down into the reactor vessel lower to solidify, the melt will form a so-called debris bed. This debris
plenum. In case of failure of the vessel, the melt will then flow bed might further be cooled down again by the water, which might
down in the reactor pit. If the core melt, called “corium”, relocates stop the melt progression. This stage is called the premixing of the
in a region containing liquid water, either in the vessel (in-vessel fuel–coolant interaction. However, the flow during the premixing
situation), or in the pit (ex-vessel), then melt and water will enter in might also enter a critical phase and further destabilize to produce
a complex thermohydrodynamical interaction called fuel–coolant an energetic event called steam explosion which may endanger
interaction. The most straightforward result is the fragmentation the containment integrity. The triggering of a steam explosion may
and dispersion of the melt in the water. Depending on the con- occur under several mechanisms and cannot yet be predicted for
ditions, the melt length scale of the fragmented part (drops) will practical situations. Then, particularly in reactor accident analysis,
it still has to be considered as a stochastic event.
Whatever the situation, it is now clear that the sole study of
∗ Corresponding author. Tel.: +33 442199114. the explosion phenomenon is not sufficient to eliminate the risk of
E-mail address: renaud.meignen@irsn.fr (R. Meignen). loss of containment integrity in common situations. First realized
http://dx.doi.org/10.1016/j.nucengdes.2014.08.029
0029-5493/© 2014 Elsevier B.V. All rights reserved.
512 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
(premixing stage) and the resulting melt droplets (of the order of
some millimeters) exchange heat with the coolant, leading to a
more or less intense boiling. This mixing stage has several impacts.
First, it may provide an initial state to a possible steam explosion,
following some triggering event. However, if no steam explosion
occurs, the mixing itself may endanger directly the containment
through a global pressurization resulting from the boiling process,
or even the internal structures1 themselves by local pressurization
of the pit, although this later case seems potentially dangerous only
in extreme situations with high vessel pressures and large breaks.
The third physical issue of the mixing phase is the production of a
debris bed whose characteristics will be a function of fragmentation
and solidification of the melt. Depending on the state of this bed
(porosity, particle size, geometrical configuration), the melt might
be cooled definitely, which would terminate the melt progression.
The coolability issue is also obviously dependent on the amount
of water remaining in the cavity after the mixing process. Last, all
experiments show that oxidation should very easily occur during
this fast transient. This should lead to a fast and massive production
of hydrogen that might lead to a potentially explosive atmosphere.
It is also to be noticed that a steam explosion with little mechan-
ical effects would nevertheless modify the following of the mixing
phenomenon in a way that cannot be handled at the moment.
Fig. 1. The general issue of ex-vessel fuel–coolant interaction. In contrast with several other areas in nuclear severe accident
studies (e.g. MCCI), it can be highlighted that the FCI problem was
rapidly tackled through the development of complex multiphase
studies hypothesized premixing configurations where important CFD (CMFD) tools. The motivation for the development of such
mitigating effects (void and solidification) could not be a priori tools is probably the recognition of the numerous highly non-linear
postulated and were not considered. This led to unrealistic high interactions that should hardly be described with simple 0-D mod-
potential pressure loads on containment structures. It was then els without an in-depth understanding of the phenomena. Several
found that a more precise characterization of the premixing, i.e. codes have thus been developed with various levels of complexity
the initial state of the explosion, was necessary to assess the risk. and hypotheses of modeling (Meignen et al., 2005), among which
FCI studies, in the frame of NPP severe accident analyses, started we will cite the 1-dimensional TEXAS (Kim and Corradini, 1978),
with the investigation of the so-called ␣-mode which was pre- the 2-dimensional JASMINE (Moriyama et al., 2006) and the 3-
sumed to be one of the major risks of containment failure (NRC, dimensional PM-ALPHA (Theofanous et al., 1999a,b) and JEMI codes
1975). An important step was done by Theofanous and Yuen (1995) (Schröder et al., 2009).
who applied the so-called ROAAM (risk oriented accident analysis The computer code MC3D is devoted to multiphase thermal-
methodology, Theofanous, 1996), using PM-ALPHA and ESPROSE hydraulic flow studies and evaluations in the field of nuclear safety
codes (Theofanous et al., 1999a,b) to characterize the risk and con- (Berthoud and Valette, 1994). Its field of applications has evolved
cluded to the low probability of failure of the containment through several times and its major use now concerns the evaluation of
this mode. Theofanous even questioned its physical feasibility. (molten) fuel–coolant interaction (FCI) in power and experimental
Nowadays, this particular risk is still estimated as very limited. reactors. Two models describing the premixing and the explosion
However, the possibility of projection of objects of any kind on the stages were developed. Note that the premixing model is also being
containment wall after the upper vessel failure is certainly not to used for the evaluation of the melt dispersion during a DCH event
be totally discarded, particularly if the upper part of the vessel is (Meignen and Janin, 2010) and for debris bed cooling (Pohlner et al.,
hot. Further, the probability of failure of the lower part of the ves- 2014).
sel itself was considered and the analyses made in particular in the The objective of this paper is double: presenting the current sta-
frame of the SERENA-1 project (OECD, 2006) led to the conclusion tus of development of the premixing models in MC3D (version 3.7),
that this risk is also weak, provided that the vessel strength is not and, through this presentation, highlighting the existing difficulties
affected by aging or heating. The focus regarding the risk of steam and challenges for a sufficient resolution of the problem,2 conclud-
explosion is then now put on potential explosive interactions in the ing by the perspectives of improvements for the near future. This
reactor pit, if the vessel comes to fail and the corium to flow into a article is the first of two aiming at presenting the general status
flooded pit. It is nonetheless important to stress that non-explosive of FCI modeling through a presentation of MC3D code. The second
interactions in the vessel is still an important issue if one wants to paper is devoted to the explosion modeling. The reading of these
be able to capture adequately the subsequent global pressurization, papers might be complemented by a short review of recent pro-
hydrogen production and formation of the debris bed. gresses made in this area and ranking of priorities by Meignen et al.
Restricting our attention to the ex-vessel situation, Fig. 1 (2014b).
provides a sketch of the general issue of fuel–coolant interaction. After a short highlight on premixing phenomena, a general
This figure gives links between each individual process with a description of the mathematical model is provided. Then the key
rough classification that might be subject to discussion: simply
not addressed for various reasons, weakly addressed, in general
due to starting studies, or fully addressed, i.e. taken in serious
1
consideration, whatever the level of accuracy. FCI is seen to have By “internal structures”, we mean the reactor pit, floors and various walls.
2
“Resolution” is to be understood here in the sense of understanding and ade-
numerous impacts, steam explosion being one of them. The story quate modeling. The term “sufficient” is employed to mean that the “resolution”
then starts with the failure of the vessel and flowing of the liquid level must be considered regarding adequate objectives in terms of modeling and
corium into the flooded pit. The melt starts to mix with the water confidence to extrapolations to reactor situations.
R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527 513
Fig. 3. The four steps for achievement of determination of fuel–coolant interaction for application to steam explosion and debris bed formation. The main topics and of the
two phases of SERENA program are indicated.
514 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
since the risks are now better assessed. Nevertheless, as already Indeed, being able to propose such double approach in the same
mentioned, experiments themselves were performed with simpli- tool is also a particular challenge. It will be important to have this
fying assumptions. The remaining uncertainties are then mostly in mind while reading the rest of the paper, other way the pro-
related to aspects which were not or only weakly addressed. Fur- posal for several models at different levels of complexity for the
ther improvements then require also additional experimental data same phenomenon might be quite confusing, whereas they are
addressing different configurations. complementary.
Melt fragmentation was then historically the first aspect to be
investigated but, despite important progresses and in view of the 3. MC3D’s mathematical model for premixing
above discussion, it still needs further work to address simplifying
assumptions. Indeed, jet fragmentation in isothermal (non-boiling) The computer code MC3D is primarily to be considered as a tool-
situations is already quite poorly understood and modeled, except box allowing the resolution of physical models for multiphase flow
in some very specific configurations widely studied in some indus- problems. The mathematical model is sufficiently robust to cover
tries. The difficulty here is amplified by the fact that the ambient a very large spectrum of physical problems, from flows in geologic
medium configuration and composition is dependent on the frag- media (waste management context, see Berthoud, 2006) to strong
mentation itself (Fig. 2). detonation waves. The code structure is very flexible and coding
Void in premixture can probably be adequately characterized of new models is in general very fast. Developers have therefore
only with the use of CFD models. Its impact on melt fragmenta- mainly to derive physical models, which in the case of FCI, is the
tion (Fig. 2) is still uncertain and numerous models use simplifying core of the problem. For the numerical model of a physical prob-
assumptions (e.g. no impact in the “thin film” model). The eval- lem, the developer has to define mass components which might
uation of the amount of void is obviously difficult due to the be grouped in volume mixtures, i.e. with components having the
very specific conditions out of the scope of the large majority of same volume (e.g. gases), then in momentum mixtures, i.e. with
experimental research. Nevertheless, it is now quite clear that in components having all the same velocity (dynamical equilibrium),
conditions with large melt jet release, this one embedded in a large and then grouped in energy mixtures, i.e. with same temperature
vapor film and thus fragmentation occurs primarily in a gaseous (thermal equilibrium) (Fig. 4).
environment. The simulations also indicate that this vapor film is We use the classical Eulerian “two-phase” method (e.g. Ishii and
not stable but subject to large scale instabilities. Hibili, 2006) where each component/mixture is described by mass,
Solidification is now suspected to be one of the leading effects momentum and energy equations. These balance equations are
in the mitigation of steam explosion with UO2 /ZrO2 melts as com- solved with a semi-implicit method derived from the ICE method
pared to other melt materials (e.g. alumina which was used in some (Harlow and Amsden, 1971). A staggered grid is used (Cartesian or
KROTOS experiments, see Huhtiniemi et al., 2001). This effect has cylindrical) where velocities are expressed at the faces and other
always been accounted for in the MC3D models but it has received variables at the center of meshes.
recently renewed attention, and is now introduced in all available The main variables for resolution in the premixing model are:
(and still developed) models.
Oxidation has also been introduced in the MC3D model since • the volume fractions ˛;
a long time but verification and improvements are limited by the • the mass fractions of the gas components;
small number of experimental data. More, the few experimental • the pressure P (one single pressure);
results are quite confusing from apparent full mitigation of the • the velocities, U, V, W;
explosion to multiplication of loads by several orders of magnitude • the temperatures T.
(Cho et al., 1998). Today, from our point of view, this aspect is the
one on which efforts should be put primarily since it is very likely
The process starts by considering the momentum equation
that in the reactor situation the melt will contain large amount of
which is written with the finite difference method. For a momen-
un-oxidized metals (e.g. vessel failure from the “focusing effect”).
tum mixture m, the balance is written in non-conservative and
The problem is seen to be in itself very challenging. How-
mostly implicit form regarding the velocities and pressures:
ever, it is made even more complex by the fact that the initial
and boundary conditions, i.e. accident scenarios, are themselves −
→n+1
very uncertain, which obliges to investigate various conditions that Vm −V m
n −
→
• ˛nm m
n
t
+ V m n)
n ∇ (V
m = −˛nm ∇ (P)n+1 +
might be later found unrealistic. The ex-vessel situation in the
PWR/BWR involves:
nmel
(V lN+1 − V m
n+1/n
˛nm m
ng+ Klm n+1
)+
likely, non-axisymmetric conditions (side break) i.e. needing in
principle 3D evaluations;
l=1,l =
/ m
ncons
ncons
n+1/n n+1
potentially, very large breaks (some tens of centimeters); cc (Vl − V m
n+1
) + Tadd
potentially, pressurized ejection of the melt. c=1,c ∈ l c =1,c ∈ m
Fig. 4. Generic arrangement of grouping and relations between component in MC3D framework.
(˛n+1 n+1 n n
ncons
• v v −˛v v ) + ˜ nv
˛ ˜ vn ·
Sf n+1
· Vm =
n+1/n
cc
dt ˝
faces c,c
(˛n+1 en+1 een+1 −˛n n n Sf
• e e ee ) n+1
e
dt
+ ˜ ne
˛ ˜ en ẽen · ˝
· Vm + Fig. 5. The two configurations for describing the melt and transfers between both
fields.
⎡ faces ⎤
˛n+1
nc
• one describing a continuous configuration, e.g. jets, pools or liquid
P n+1 ⎣ n+1 ⎦
−˛n S n+1/n
e e
+ ˜ ne . ˝f .Vm
˛ = ˚n+1/n + cc Hc +
dt films, which is called the JET field;
faces c,c • one describing the drops, supposed to be in a discrete state.
Tadd
Both fields are interconnected though the fragmentation of the
In these equations, the subscripts v and e represents the con- JET field and the coalescence of the DROPS.
sidered volume and energy mixtures, the tilde (e.g. ˛) ˜ stands for The JET field is modeled with a VoF-PLIC (volume of fluid, Piece-
“convected” quantities at the faces (e.g. quantity in upwind cell for wise linear interface construction) that allows a precise description
O(1) convection scheme), represents the mass transfers, the of the flow convection with negligible numerical diffusion. The
heat fluxes and H the enthalpy of donor phase. Sf is the area of method is well-known and we will here give only the major lines of
the faces and is the volume of the cell. Heat fluxes are under modeling. The interface is supposed to be approximated locally by a
the general form (except for radiation), ˚ = hn An+1 Tn+1 , with h straight line (or plan in 3D). Thus the melt is precisely localized into
the heat transfer coefficient, written explicitly, A the interfacial the cell. The orientation of the interface is obtained by inspection
area between two fluids and T the temperature difference, both of the surrounding cells and assuming the functional form:
expressed implicitly. In the energy equation, Tadd stands for vari-
˛j = ai xi + c
ous additional terms that cannot be detailed here as conduction,
i
chemical energy, volumetric power sources. The kinetic energy is
neglected. where ˛j is the continuous fuel volume fraction. The ai = ∂˛j /∂xi are
The numerical resolution method is shortly described in the the coordinates of the normal vector, and are calculated by a least
appendix. It consists in rearranging the balance equations in order square method. The interface is then placed according to the real
to obtain a matrix containing only the pressure. This matrix is volume of the JET in the cell. In principle, the factor c should give
solved and other variables are deduced. To do so, the momentum the position of the interface but this is not sufficiently precise.
equations are treated separately from the mass and energy balances The convection scheme is depicted in Fig. 6 for the simple case
in order to express the velocities as functions of local surrounding of two fluids (jet and gas) and 2D configuration. The convected vol-
pressures. Then, these expressions for the velocities are introduced umes of each fluid during the time step at a face are pictured with
in the mass and energy equations so that they are only functions of dashed surfaces, the height being the product between velocity and
the pressure, volume fractions and temperatures. These balances time step and the base being the common surface between the jet
are solved with an iterative Newton–Raphson procedure. and the cell face. A similar method has been sought for the 3D cylin-
drical case with interfaces with conical forms but the complexity
of modeling has been found too high to be incorporated in the code
4. Constituents and flow configuration with sufficient confidence on robustness. Thus, for 3D cases, the
model is available only in Cartesian geometry. The model is nev-
In the PREMIXING model of MC3D, the melt is described with 2 ertheless partly available in cylindrical 3D calculation for the R–Z
different fields (Fig. 5): directions. The direction is described with a simple 2nd order
516 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
hfilm
e=
kfilm
The film void fraction is transported with the drops. The purpose
of this specificity will be addressed further when looking at heat
and mass transfers.
Fig. 6. Principles of the VOF-PLIC convection scheme for a JET-gas configuration. Vj :
jet velocity through the lower face, Vg : gas velocity, dt: time step. The nature of the continuous phase (liquid or gas) depends on
the relative void fraction defined by:
˛g − ˛film
˛g,rel =
˛l + ˛g − ˛film
Fig. 8. Example of visualizations of experiments of instability and fragmentation of water jet in air (left, reproduction from Marmottant and Villermaux, 2004) and mercury
jet in water (right, reproduction from Hall and Fletcher, 1995).
CE ≈ 7.5 ± 2.5. This correlation is based on the assumption that the - an adequate spatial averaging of the properties;
fragmentation rate is proportional to the jet velocity and then the - an averaging of the fluid properties.
breakup length becomes independent of it. This assumption comes
from the Kelvin–Helmholtz model when the vapor film is not taken Since it is believed that the jet rate of distortion and fragmen-
into consideration (thin film assumption). There is obviously a large tation is dictated by large-scale effects, we need an averaging and
uncertainty but the Epstein correlation does not seem to be applica- filtering process over a given distance R. Furthermore, this will limit
ble for corium jets in water. Originally, the Meignen (CONST model the effect of mesh size. For a given property V, the averaging is done
in MC3D) correlation was adjusted to fit the FARO L11 and L14 data as:
(Magallon and Huhtiniemi, 2001), the only ones available at that
time, which are at saturated conditions, thus with large films sur- (x − y)˛(y)Vdy
rounding the jet. A saturated CCM test is seen to be largely out of V (x) =
(x − y)˛(y)dy
the linear tendency. It was suspected to be due to large scale distur-
bances originating from the injection system. Indeed, several tests
have shown a very fast apparent breakup when the jet is already 1 if ||x − y|| < R
turbulent before penetrating the water. The correlation is seen to (x − y) =
0 else
be also quite consistent with sub-cooled data, although a better
specific adjustment can be found. Since, the mixing zone width is of the order of the jet radius, we
If the fragmentation rate can be approached with relatively use it as the characteristic length R. Then, since the KHF model uses
simple arguments, much more difficulty regards the size of the one simple ambient fluid, we need to provide an averaging process
drops. Indeed, the size seems to be dependent on very complex to define this ambient fluid. There are several possibilities to take
processes which are not yet fully understood. Fig. 10 provides the into account the multiphase environment. In the MC3D model, it
mean Sauter diameter obtained from the sieving of the debris in is chosen to consider a “mixed” fluid with averaged properties. To
various tests with corium melts. The injection velocity in the exper- define them, again, several ways are again possible. We use the
iments is in general quite uncertain. However, these uncertainties property of the Kelvin–Helmholtz model that the leading term of
cannot explain the large scatter of the experimental data. It will the instability is the kinetic energy V2 . Denoting spatial averages
be remarked that the experiments involving quite large melt jets with a bar, we get for the ambient fluid:
(5 cm) in large water pool (∼70 cm) tend to give particles with
diameters in between 1.5 and 3 mm. In contrast, there is a group of 2 2
experimental results with small particles, between 0.4 and 1 mm, ˛g g + ˛l l ˛g g (Vg ) + ˛l l (Vl )
a = , Va = + Vj
which were all obtained for experiments with smaller jets (2 cm) in ˛g + ˛l ˛g g + ˛l l
narrower water pools (20 cm). Such difference due to the geometry
was discussed, in particular in the frame of the SERENA-2 project, where Vg = Vg − Vj and Vl = Vl − Vj .
but could not find any explanation up to now. The standard CONST A first validation and calibration of the model is obtained by
model in MC3D does not provide a law for the drop size and it is comparison with data from experiments with low melting point
up to the user to fix it. It is recommended to choose a value slightly melts (e.g. Wood’s metal) which can be considered as isothermal,
above the requested Sauter diameter in order to better take into i.e. without boiling. The solidification is also expected to have a
account the thermal aspects attached to a real population of drops: weak impact.
the smaller drops are cooled earlier and then have a lower weight The drop diameter is computed from the wavelength of the KH
for the global energy transfers. Based on the experimental results, instability model, as being proportional of the most unstable wave
a value of 3 mm5 is then recommended as being a reasonable con- .
servative choice regarding the explosion (the large drops solidify
less rapidly and produce less void). Dd = Nd
This situation is expected to yield sufficiently conservative
results in the common situations with vertical central jets. How- The corresponding wavenumber is given by
ever, there is certainly a need for a more elaborate model. In MC3D, 2
2 (Vj − Va ) j a
kmax =
3
j + a
5
Latest IRSN reactor applications use in fact 2.5 mm.
R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527 519
Fig. 12. Comparison of the evaluation of the breakup lengths of low melting point
melt jet dispersion in water with computations with MC3D. sNd =
Nd . See also
legend of Fig. 11. Data: same as Fig. 11 except KTH (Haraldsson, 2000): Cerrobend-70,
d = 25 mm;
Fig. 11. Comparison of the evaluation of drop size from simulations of low melt-
ing point melt jet dispersion in water. sNd =
Nd . IKE (Cho et al., 1991): Wood’s
metal, d = 4 mm; Matsuo (Matsuo et al., 2008): U-Alloy 78, d = 7, 10, 15, 20 mm; of the actual values of this crucial property. Thus, for the evalu-
KMU (Bang et al., 2003): Wood’s metal, d = 10, 20 mm, ANL (Spencer et al., 1986): ations, the actual parameter to take into account is
Nd . For the
Wood’s metal, d = 22 mm. ISPRA (see Bürger et al., 1995): Wood’s metal, d = 28 mm.
computation, we impose a surface tension of 1 N/m, likely to be
MC3D calculation, density = 9000 kg/m3 , surface tension = 1 m/s.
representative.
The jet diameters in the experiments were ranging from 4 to
A rational of the model can be obtained if we simplify a bit the
50 mm. The model should be in principle less accurate at small jet
equations with the assumption a « j which is correct for most of
diameter, i.e. not far larger than the drop size. The reported exper-
our applications. Denoting, ˇ = a /j we have then imental diameters are the median values of size distribution. All in
all, a very large scatter is observed in the experimental data, nearly
kmax
ci = ˇ(Vj − Va )2 − one order of magnitude. The uncertainty on surface tension can-
j not explain such large discrepancy and further efforts are clearly
needed in this direction. The jet size and duration of the injection
It is then worth to notice that the velocity can be further approx-
should also influence the results as they influence the entrainment
imated for most of the wave numbers to
of the water along the jet. The situation is nevertheless quite confus-
ci = ˇ(Vj − Va ) ing and for the comparison, we choose to compute a representative
situation, without trying to reproduce a specific experiment. Apply-
Then the fragmentation rate is f = Nf ˇ(Vj − Va ) and, as ing the model with a value
Nd = 2 leads, as shown in Fig. 11, to a
L/D = (Vj − Va )/2 f , we recover the Epstein correlation by identi- rather satisfactory result. Tendency curves of the form d = a·vb , are
fying Nf with 1/CE . A typical value close to 0.12 is then expected for also shown for the calculation results. The dependency with the
Nf . velocity is approximately at the power b = 4/3. A parameter
Nd = 2
The surface tension acts as a low-pass cut-off regarding the wave is also correct to conform to FARO results, if we use a surface ten-
number, this cut-off being close to the maximum in fact and then, sion of about 0.5 N/m. Nevertheless, results with
Nd = 4 are also
ˇ(Vj − Va )2 ≈ (kmax /a )
. With Dd = Nd max we are then led to the shown to check for the sensitivity. On the average, the ratio of the
following expression: obtained diameter for the two cases is not 2 as expected, but 1.5. It
is likely to be due to a stronger entrainment of the water with the
a (Vj − Va )2 Dd
≈ 2
Nd drops when they are smaller, which reduces the velocity difference,
and thus increases the drop size.
which is in fact nothing else than the classical Weber criterion for In order to conform to the Epstein correlation, the fragmenta-
drop stability. This criterion is also often used to characterize the tion parameter Nf should be of the order of 0.12. The Epstein model
mean drop size in a mixture. From this expression, a typical value does not take into account the jet acceleration which in the case
of 2 for Nd can be expected if the critical Weber number is taken as of metals (or corium) is important. In fact, taking this effect into
12. Inversely, a typical value of 0.5 for Nd (half a wavelength) would account leads to a non-constant breakup length, tending to zero
yield an equivalent Weber number of about 3. at zero velocity. This is shown in Fig. 12, together with exper-
We firstly check the consistency with experimental data in imental results and evaluations with MC3D. The theoretical KH
quasi-isothermal conditions or with low boiling conditions. It is model results are shown with two values of Nf . The breakup lengths
noticed that, as the drop size is estimated from the debris, solidifi- are also in reasonable agreement with the experimental data and
cation occurs during the flow. Here, the impact of the solidification follow the trends expected from the model, i.e. an augmentation
on the final size is not known and it is postulated that it is small. with the velocity due to an effect of the acceleration of the jet. The
The result of the computation of the mean drop size compared to lengths with suggested value Nf = 0.15 are in the upper range of the
the available experimental data is shown in Fig. 11. In these exper- experimental values and a value of 0.2 should be more accurate
iments, several low melting point metals are used (Wood’s metal, for these particular tests. It is noticed that at high velocity, the jet
Cerrobend-70, U-alloy 78 . . .) and, unfortunately, the surface ten- becomes quite unstable and the breakup length oscillates around
sion seems to be very uncertain for most of them. The value for an average value given in the graph. The breakup length seems nev-
the Wood’s metal ranges from 0.38 N/m (Narayanan et al., 2006) to ertheless to become more or less constant, in agreement with the
“∼1” in Bang et al. (2003). Note that the same uncertainty exists for Epstein model.
corium melts. As the drop size should be (according to the model), Also in Fig. 12 is shown the result of the Saito correlation, which
in the absence of solidification effect, proportional to the surface is often used to characterize the breakup length (e.g. in JASMINE
tension, it is clear that there is a need for the proper determination code).
520 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
L j V
= 2.1 j
D l gDj
5.2. Perspective
To do so in MC3D, a multi-field Eulerian description (MUSIG) was criterion:ddaughter = 2 which can be obtained easily also
c (Vda )
developed in the code. In this method, the particles are distributed from the Kelvin–Helmholtz theory is a characteristic Weber num-
in several classes with fixed particle size. Each class evolves as the ber. In principle, the energy criterion of Weber should impose a
drop field in the standard description except that: characteristic value below 12. However, the fragment production is
a complex process and the fluid might be locally entrained before
- all classes have the same velocity (homogenous MUSIG descrip- being fragmented, which will result in a higher characteristic
tion); Weber number.
- the energy balance is evaluated explicitly. However, it is to be noticed that, for the model to give reasonable
agreement with experiments, it is necessary to jump to a hetero-
A first step was then to impose a spectrum shape, e.g. log-normal geneous MUSIG description with several velocity fields. This is due
or gamma distribution, while keeping the current models. As these to strong entrainment of the smallest fragments which leads to a
distributions need two parameters, one needs to fix an additional premature stop of the fragmentation process with one single veloc-
parameter. This was done by imposing a ratio between the mass ity field. After some tests, it comes that a reasonable agreement is
median diameter and the mean Sauter diameter (1.4 seems a good obtained with 3 associated velocity fields: one for the very large
compromise). Such method was also chosen for the recent devel- drops, one for the medium-size drops, one for the smallest frag-
opments of the JEMI code. ments. Currently, the smallest are in fact placed in the same velocity
We are however also currently seeking for an alternative and field as the water, i.e. they are in hydrodynamical equilibrium with
improvement where the distribution would be directly evaluated. water.
This can be done through the use of a secondary fragmentation Note that, in the considered conditions with fully liquid drops,
mechanism. It is recalled that it is believed that the drop size is the viscosity is too low to have an impact (the Ohnesorge number
affected by local characteristics whereas the jet fragmentation itself is very small). However, in case of partial solidification, an effective
is affected by more global characteristics thus with large scales viscosity might be considered, leading to a different criterion. It is
(the jet scale). With this description, we are able to disconnect the also noticed that, for an accurate computing, one needs accurate
primary fragmentation rate from the final drop size. In the model frictions since fragmentation is partly stopped due to entrainment
under development (Castrillon Escobar et al., 2013), it is considered of the drops with water. It is not possible to discuss this aspect in
that the large scale deformation is equivalent to a primary fragmen- detail in the present article. It is nevertheless important to highlight
tation, still calculated with previous methods, but imposing large the quite large uncertainty on this issue when the drops are deform-
initial deformation (Fig. 13). The resulting large drops (or deforma- ing and fragmenting. With the development of high precision CFD
tion) are then subject to finer fragmentation until a criterion for calculations, this issue has received some attention recently (see
stability is reached. e.g. Kahre and Yang, 2013) and some efforts might have to be under-
Fragmentation of the drops is computed with the rather classical taken in this area for the computation of FCI. Also, currently, each
Ranger and Nicholls time scale, which leads to the following mass fragmentation step is viewed as producing daughter drops with a
1 ˛d Vda √ 1 Vda a single size, whereas it is likely that each individual process actually
transfer rate:frag = t ∗ d a = ˛d d t ∗
frag
Dd
frag
Dd d produces its own size spectrum.
R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527 521
h_mod / h_exp
mass transfer processes and we will restrict to the most spe-
1,5
cific and important one, i.e. the film boiling configuration. It is
indeed easy to verify that, despite the high melt temperatures
1
considered here, radiation is dominant only when the melt is
in a gaseous environment. This is accentuated by the fact that
0,5
the melt surface temperature rapidly decreases. Nevertheless, in
general, and particularly in reactor situations, the melt spends
0
an important time in the vapor produced by the vaporization 0 50 100 150 200 250 300 350 400
of the water and then radiation is an important heat transfer Subcooling (K)
mode.
Film boiling has been and is still a topic of intensive research. Fig. 14. Ratio between the computed heat transfers (h mod) and those measured
Several models and correlations have been proposed, all with limi- (h exp) in the TREPAM experiments (Berthoud and D’Aillon, 2009, including those
at pressures up to 240 bars) and in the experiments from Inoue et al. (1997). It
tations. MC3D proposes several correlations but the default one is a
is noticed that in the case of filaments, the Epstein–Hauser model provides the
composition of the forced convection film boiling from Epstein and necessary corrections to the case with spheres.
Hauser correlation (1980), slightly modified, and a pool film boiling
correlation from Sakurai et al. (1990). However, the correlation is In such conditions, film boiling heat transfer is weakly depend-
also quite complex and is not detailed here. In most of the cases, ent on the melt temperature.
the regime is forced convection and we will focus on it. It will be noticed that the recent TREPAM experiments
There have been a number of debates on the problem of film (Berthoud and D’Aillon, 2009) have confirmed, for the case of
boiling and the accuracy of the existing correlations. The choice small filaments (lower than 0.5 mm) the accuracy of the model,
made for MC3D was dictated by the fact that the correlation of and then the accuracy of the Epstein–Hauser correlation for high
Epstein–Hauser was obtained from an analytical model and is more temperatures (up to 3000 K), high pressures (up to 22 Mpa) and
easily tractable and understandable. However, results are similar to high velocities (up to 40 m/s). Fig. 14 shows a comparison with
those with the Liu and Theofanous correlation (1996, also available the calculated and the measured heat transfer coefficients. Also
in MC3D). are included in this figure the comparison with the data from
The Epstein–Hauser model makes the hypothesis of a constant Inoue et al. (1997). The level of accuracy is the same as those in
film thickness at the front of the ball/drop and is in principle limited previous experiments and is also comparable to the experimental
regarding the thickness of the film that might appear in specific uncertainty (the scatter of data for similar conditions yields an
situations with low velocity and low subcooling. The analysis con- uncertainty of a factor of 2 in TREPAM). It is remarkable to note
ducted by Epstein and Hauser yields a complex relation that has no that, even at pressure slightly beyond the critical pressures (strong
explicit solution: only an iterative procedure can give the result. To subcooling), the model is still sufficiently accurate regarding
overcome the difficulty, they proposed a correlation with a combi- all general uncertainties in the FCI problem (the melt is not a
nation of the results for the two extreme cases of high subcooling composition of solid spheres (or filaments)).
(thin film) and saturated cases (thick film). Epstein and Hauser Interestingly, it is noticed that, despite the fact that both factors
paper does not give information of the accuracy of the procedure in the brackets can strongly differ, the influence of the sub-cooling is
but it is easy to verify that the error is bounded within 20% for finally rather mild. At one bar, the difference between saturated and
the cases of interest here. The modified Epstein–Hauser correlation strongly sub-cooled cases is about 25% so that, as a first approxima-
used in MC3D reads: tion, the simplified case of strong sub-cooling Nusub may be used.
ϕEH, mod = hEH, mod (Td − Tsat ), hEH, mod = However, in a global FCI situation, more vapor is produced with
2 B 4 1/4
Cc 3 g Re1/2 1
+
2 saturated conditions so that, all in all, the cooling is much more
2 D ˇ 24ACc 4 A
Cpg ·(Td −Ts ) ˇ.l ·Cpg ·(Ts −Tl )
effective in sub-cooled conditions.
A = Pr ·(h −h ) , B = ·Pr ·(h −h ) . Prl , ˇ = Then, into our point of view, the film boiling heat transfer is not
g vsat g g v,sat
1/2
lsat l,sat
to be considered as an important challenge and we estimate that
g 1/2 g
l , Cc = 2 researches on this issue are of low priority for the FCI problem. More
In the above correlation, stands for the thermal conductiv- problematic is in fact the result of the heat transfer: part of the heat
ity, for the density, for the kinematic viscosity, Cp for the heat goes directly to the water and part goes to evaporation to produce
capacity, h for the enthalpy, Re and Pr for respectively Reynolds and void (that might further re-condense). In fact, in steady conditions,
Prandtl numbers, the subscript g refers to the gas, d to the drop and it can be considered that all the heat goes into water since a steady
l to the liquid. void volume is formed (film and gas pocket in the rear). The film
The first term in the bracket gives the heat transfer close to boiling models cannot yield an answer to the problem of the volume
saturation, whereas the second term corresponds to the strongly of vapor formed on the rear of the drop since the heat transfer
sub-cooled cases. Compared to the original correlation, the fitting occurs for most at the front. It is foreseen to handle this difficulty in
factor Cc has also been introduced in the first term in bracket. How- the near future and currently a simple parametric approach is used
ever, considering the general uncertainties (see discussion below), where the heat flux leaving the drop and going to water is adjusted
this modification is not decisive. in order to have the following situations:
It can be easily verified that, when the second term in the bracket - in saturated cases: all the heat is used for vaporization; vapor
is dominant, i.e. at sub-cooling larger than about 50 K, the model is might further condense;
reduced to a heat transfer correlation around a sphere at saturation - in sub-cooled cases, if DT > 10 K, all the heat goes to water, pro-
temperature in water: vided that a minimum vapor fraction exists. This minimum is
given by the film thickness from the film boiling model;
ϕEH, highsub ∝ Re1/2 Pr 1/2 (Tsat − Tl ) - linear interpolation in between.
522 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
Fig. 18. The general issue of oxidation during premixing. The colors for the final
effect indicate whether it is positive for safety (green) or negative (red). Increasing
void is considered as positive, although this needs further clarifications. (For inter-
pretation of the references to color in this figure legend, the reader is referred to the
web version of this article.)
Fig. 19. Visualization of a 2D calculation of the FARO L-28 experiment (Magallon and Huhtiniemi, 2001). The jet and hot drops are represented with red colors. The solidified
drops are represented as black dots. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
the water out of the tank. This ejection makes in fact the calculation
of the experiment particularly difficult but gives useful information
on the void in the mixture. It is seen that the void evolves in fact as
large bulges traveling along the jet. This is a typical behavior also
visible in reactor evaluations.
A similar calculation of the (subcooled) FARO L-31 test shows
a similar trend except that the void is far smaller. In the experi-
ment, no overflow of the water occurred. In both cases the pressure
rises steadily. However, the rise is about ten times larger in L28.
Indeed, in our computations, the pressure rise in L31 is essentially
due to the production of hydrogen due to oxidation of UO2 . The
net pressurization tends to nearly 0 if no hydrogen generation is
accounted in L31. The experimental pressurization is recovered
when the hydrogen mass is close to 40 g, i.e. close to the exper-
imental finding (about 50 g). Such effect is responsible for only
Fig. 20. Penetration of the melt for the calculations of the FARO L-31 and L-33. The
about 1 bar in L-28 (estimated in the experiment from the mass
time and height zeros are adjusted for the comparison (the water height in L-33 is
of hydrogen generated), compared to the about 10 bar from pure 1.6 m, whereas it is 1.45 m in L31).
vaporization (overall experimental pressurization is 11.5 bar), and
oxidation can then be omitted at first order for the interpretation
of this experiment. In Fig. 21 is shown what we call the “explosivity criterion” (mass
Finally, it is noticed that we also compute the debris bed for- of melt that should be subject to explosion, arbitrarily defined here
mation. This issue has not been discussed previously. It was the as the mass of liquid melt in cells with an amount of water larger
object of particular attention in the frame of the SARNET-2 project than 30%) for three calculations with three different drop size (the
(WP 5.3, Kudinov et al., 2012). This activity showed that the evalu-
ation of the debris bed when it is made partly from agglomerated
or un-fragmented melt is also a challenge that needs further effort.
In the present L-28 calculation, the jet found on the ground on the
water tank is obtained from a re-coalescence process of the drops
insufficiently solidified. The usual criterion that is used for this is
the mean temperature between solidus and liquidus (this criterion
should not be confused with the one for explosion discussed in the
section relative to solidification). This criterion was chosen from a
rough analysis of the forces exerted on the drops when they set-
tle on the ground with the same hypothesis as for the criterion for
explosion.
The FARO L-33 was devoted to steam explosion with a triggering
after about 1 s of flow. The test was similar to L-31 but with a higher
pressure (5 bar versus 2) and a slightly higher water level (1.6 m
versus 1.45). The global behavior of our calculation is very similar
to the one for L31. The penetration for our reference calculation
with Ddrop = 2.5 mm follows the same trends (Fig. 20). Due to the
larger water height, the contact with the bottom is slightly delayed.
However, whereas the penetration for L-31 is rather satisfactory, Fig. 21. Explosivity criteria for three calculations of FARO L-33 with three different
drop size. Red: D = 2.2 mm, green: D = 2.5 mm, blue: D = 3 mm. Abscissa: real calcu-
the drops contact the bottom at t ∼ 1.3 s in the L-33 calculation
lation time. The explosivity criterion here is the mass of liquid melt in cells with an
whereas the experimental report gives a value of 1 s. This would amount of water larger than 30%. (For interpretation of the references to color in
mean a faster penetration than L-31 that cannot be explained. this figure legend, the reader is referred to the web version of this article.)
R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527 525
thicker the line, the larger the drops). We estimate that the amount
of melt drops in contact with water is small in any case, smaller
than 7 kg. At the time of melt bottom contact, when the trigger is
applied this mass ranges from about 2 kg to about 5 kg. We notice
strong variations of the criterion with time, due to variations of void
in the mixture, due to pulsations of the vapor film around the jet.
Indeed, the explosion calculations are found also very sensitive to
the chosen time, in coherence with the explosivity criterion (see
companion paper in this issue). This highlights a strong depend-
ency of the explosion strength on the void and then on the exact
time of triggering. Considering this criterion, the most challenging
explosion would have been at about 1 s, when the melt reaches
approximately 40 cm from the bottom.
The assessment of the jet fragmentation model, already dis-
cussed previously, is now investigated in more realistic conditions
using again the FARO L-28 and L-31 tests. It is recalled that L-28 is
with saturated water and thus with a considerable void production Fig. 23. Example of 3D ex-vessel premixing calculation in a simplified French
whereas L-31 is strongly sub-cooled, with a small void produc- 1300 MWe reactor geometry and assumed central break. The presence of the lat-
tion. Then these two cases are expected to embed the impact of eral corridor induces a rupture of the axisymmetric behavior, deflation of the jet
two-phase flow configuration along the jet for a wide range of con- and subsequent breakup.
Fig. 22. Correlations between the breakup length and parameter Nf on the left and mean drop diameter with parameter Nd on the right for all L28 and L31 calculations. Plain
symbols are related to the sensitivity study calculations and empty symbols are for reference calculation with parameters chosen to yield the best result.
526 R. Meignen et al. / Nuclear Engineering and Design 280 (2014) 511–527
lateral breaks for safety studies. This is a very serious limitation Appendix: resolution procedure.
and it is necessary investigating rapidly such configurations.
To do so, additional experiments would be, of course, of great Once, from the momentum equations, the formal relationships
help. [U, V, W ]n+1
m = Q n (P n+1 ) + Rn are introduced in the mass and
energy balances, one gets a set of balances M with the volume frac-
tions, pressures and temperatures as main variables Y. At a given
9. Conclusions, perspectives for improvements time step n + 1, one has to solve M · Yn+1 = 0 by the iterative Newton
method. A Taylor decomposition leads us to solve at each iteration
This paper synthetizes the main developments of the FCI pre- k:
mixing models in the computer code MC3D and highlights the ∂M
major difficulties and challenges for a sufficient resolution of the ıY + M · Y k = 0
∂Y
problem. It has been shown that the models are at a devel-
To do so, the system is again simplified by decomposing for each
opment state that they can be used for both interpretations of
cell Yi = Xi + Pneighbours and this leads formally to a system of the
most of the existing experiments and for reactor applications.
form:
However, there still exist serious limitations to cover a sufficient
range of conditions for nuclear safety and this limits the over- A · ıX + BıPneighbours = M · Y k
all confidence in the evaluations. Currently, model verifications
are limited to cases with UO2 /ZrO2 melts and axisymmetric sit- The matrix A can be decomposed in a product of two triangular
uations with central break. The limitations not only concern the inferior and superior matrixes (LU decomposition). Multiplication
models: there is a real lack of data for verifications of further mod- by L-1 of the previous system leads to:
els. U · ıX + L−1 BıPneighbours = L−1 · M · Y k
Inoue, A., Fujii, Y., Lee, S., 1997. Studies on transient film boiling heat transfer from Narayanan, K.S., Das, S.K., Jasmin Sudha, A., Rao, E.V.H.M., Lydia, G., Murthy, S.S.,
thin wires penetrating through liquid interface. In: Proceeding of the Int. Semi- Kumaresan, M., Harvey, J., Kasinathan, N., Rajan, M., 2006. Assessment of thermal
nar on Vapour Explosion and Explosive Eruptions, Sendaï, Japan, p. 123. and hydrodynamic fragmentation in molten fuel–coolant interaction with sim-
Ishii, M., Hibili, T., 2006. Thermo-fluid Dynamics of Two-phase Flow. Springer, ulant system. In: Proceedings of ICONE14 International Conference on Nuclear
New-York. Engineering, Miami, FL, USA, July 17–20.
Kahre, P., Yang, V., 2013. Drag coefficient of deforming and fragmenting liquid NRC, 1975. Reactor Safety Study. An Assessment of Accident Risks in U.
droplets. In: ILASS Americas, 25th Annual Conference on Liquid Atomization S. Commercial Nuclear Power Plants, WASH-1400-MR; NUREG-75/014-MR.
and Spray Systems, Pittsburgh, PA, May 2013. http://www.osti.gov/scitech/servlets/purl/7134131
Kim, B., Corradini, M.L., 1978. Modeling of small-scale single droplet fuel/coolant OECD, 2001. In-Vessel and Ex-Vessel Hydrogen Sources, NEA/CSNI/R(2001)15.
interactions. Nucl. Sci. Eng. 98 (1), 16–28. https://www.oecd-nea.org/nsd/docs/2001/csni-r2001-15.pdf
Kudinov, P., Davydov, M., Pohlner, G., Bürger, M., Buck, M., Meignen, R., 2012. OECD, 2006. Research Programme on Fuel–Coolant Interaction Steam Explo-
Validation of the FCI codes against DEFOR-A data on the mass fraction of agglom- sion – SERENA, Final Report, NEA/CSNI/R(2007)11. http://www.oecd-nea.org/
erated debris. In: 5th European Review Meeting on Severe Accident Research globalsearch/download.php?doc=6599
(ERMSAR-2012), Cologne, Germany, March 21–23, 2012. Pohlner, G., Buck, M., Meignen, R., Kudinov, P., Ma, W., Polidoro, F., Takasuo, E., 2014,
Liu, C., Theofanous, T.G., 1996, September. Film Boiling on Spheres in Single and Two December. Analyses on ex-vessel debris formation and coolability in SARNET
Phase Flows, DOE/ER/12933.3. frame. Annu. Nucl. Energy 74, 50–57.
Magallon, D., Huhtiniemi, I., 2001. Corium melt quenching tests at low pressure and Sakurai, A., Shiotsu, M., Hata, K., 1990. A general correlation for pool film boiling
subcooled water in FARO. Nucl. Eng. Des. 204 (1–3), 369–376. heat transfer from a horizontal cylinder to subcooled liquid. J. Heat Transf. 112,
Manara, D., Ronchi, C., Sheindlin, M., Lewis, M., Brykin, M., 2005. Melting of stoi- 430–450.
chiometric and hyperstoichiometric uranium dioxide. J. Nucl. Mater. 342 (2005), Schröder, M., Pohlner, G., Buck, M., Bürger, M., Lohnert, G., Vusic, Z., 2009. Inves-
148–163. tigation of main limiting effects to strong steam explosions in 3D geometry
Marmottant, P., Villermaux, E., 2004. On spray formation. J. Fluid Mech. 498, 73–111. considering real accident scenarios. In: Nuclear Energy for New Europe (NENE
Matsuo, E., Abe, Y., Chitose, K., Koyama, K., Itoh, K., 2008. Study on jet breakup 2009), Bled, Slovenia, September 14–17.
behavior at core disruptive accident for fast breeder reactor. Nucl. Eng. Des. Spencer, B.W., Gabor, J.D., Cassulo, J.C., 1986. Effect of boiling regime on melt stream
238 (August (8)), 1996–2004. breakup in water. In: 4th Miami Int. Symp. on Multiphase Transport and Partic-
Meignen, R., 2005. Status of the qualification program of the multiphase flow code ular Phenomena, December 15–17, 1986, pp. 269–296.
MC3D. In: Proceedings of ICAPP’05, Seoul, Korea, May 15–19, 2005, Paper 5081. Spencer, B.W., Wang, K., Blomquist, C.A., McUmber, L.M., Schneider, J.P., 1994.
Meignen, R., Berthoud, G., 1997. Fragmentation of molten fuel jets. In: Proceeding Fragmentation and Quench Behavior of Corium Melt Streams in Water,
of the Int. Seminar on Vapour Explosion and Explosive Eruptions, Sendaï, Japan, NUREG/CR-6133, ANL-93/32.
p. 83. Theofanous, T.G., 1996. On the proper formulation of safety goals and assessment
Meignen, R., Janin, T., 2010. On the analysis and evaluation of direct containment of safety margins for rare and high-consequence hazards. Reliab. Eng. Syst. Saf.
heating with the multidimensional multiphase flow code MC3D. Sci. Technol. 54, 243–257.
Nucl. Install., http://dx.doi.org/10.1155/2010/289792. Theofanous, T.G., Li, G.J., 2008. On the physics of aerobreakup. Phys. Fluids 20,
Meignen, R., Dupas, J., Chaumont, B., 2003. First evaluations of ex-vessel fuel–coolant 052103.
interaction with MC3D. In: 10th International Topical Meeting on Nuclear Reac- Theofanous, T.G., Yuen, W.W., 1995. The probability of alpha-mode containment
tor Thermal Hydraulics (NURETH-10), Seoul, Korea, October 5–9, 2003. failure. Nucl. Eng. Des. 155 (April (1–2)), 459–473.
Meignen, R., et al., 2005. Comparative review of FCI computer models used in the Theofanous, T.G., Yuen, W.W., Angelini, S.S., 1999a. The verification basis of the PM-
OECD-SERENA program. In: Proceedings of ICAPP’05, Seoul, Korea, May 15–19, ALPHA code. Nucl. Eng. Des. 189, 59–102.
2005, Paper 5087. Theofanous, T.G., Yuen, W.W., Freeman, K., Chen, X., 1999b. Verification basis of the
Meignen, R., Raverdy, B., Picchi, S., Lamome, J., 2014a. The challenge of mod- ESPROSE.m code. Nucl. Eng. Des. 189 (1), 103–138.
eling fuel–coolant interaction: Part II – Steam explosion. Nucl. Eng. Des., Theofanous, T.G., et al., 2012. On the physics of aerobreakup, II. Viscous liquids. Phys.
http://dx.doi.org/10.1016/j.nucengdes.2014.08.028. Fluids 24, 022104.
Meignen, R., Raverdy, B., Buck, M., Pohlner, G., Kudinov, P., Ma, W., Brayer, C., Piluso, Theofanous, T.G., et al., 2013. On the physics of aerobreakup, III. Viscoelastic liquids.
P., Hong, S.-W., Leskovar, M., Uršič, M., Albrecht, G., Lindholm, I., Ivanov, I., 2014, Phys. Fluids 25, 032101.
December. Status of steam explosion understanding and modelling. Ann. Nucl. Uršič, M., Leskovar, M., Mavko, B., 2011. Improved solidification influence modelling
Energy 74, 125–133. for Eulerian fuel–coolant interaction. Nucl. Eng. Des. 241 (April (4)), 1206–1216.
Moriyama, K., Takagi, S., Muramatsu, K., Nakamura, H., Maruyama, H., 2006. Evalua- Yang, J.W., Bankoff, S.G., 1987. Solidification effects on the fragmentation of molten
tion of containment failure probability by ex-vessel steam explosion in Japanese metal drops behind a pressure shock wave. ASME J. Heat Transf. 109 (February),
LWR plants. J. Nucl. Sci. Technol. 43 (7), 774–784. 226.