A

apgtPf©

1

pbihitt2Todicttas(cS

0d

Available online at www.sciencedirect.com

Nuclear Engineering and Design 238 (2008) 590–599

Validation of the CFX4 CFD code for containment thermal-hydraulics

M. Houkema ∗, N.B. Siccama, J.A. Lycklama a Nijeholt, E.M.J. KomenNuclear Research and Consultancy Group (NRG), PO Box 25, 1755 ZG Petten, The Netherlands

Received 16 October 2006; received in revised form 23 January 2007; accepted 12 February 2007

bstract

In order to determine the risk associated with the presence of hydrogen in a nuclear power plant containment during a hypothetical severeccident, thermal hydraulic codes are used. Amongst other codes, NRG uses the commercial computational fluid dynamics code CFX4 for this

urpose. Models to describe condensation have been implemented by user coding. This paper describes these models. In addition, an overview isiven of validation activities with the CFX4 model. Experimental results from the following sources have been used: the Kuhn condensation model;he PHEBUS test facility; the PANDA test facility; and the TOSQAN, MISTRA, and THAI test facilities within the OECD International Standardroblem 47. The CFX4 model predictions are fairly good. Deviations originate primarily from the applied wall treatment. Several recommendationsor further development are therefore proposed.

htamhwaweanirfaioAo

2007 Elsevier B.V. All rights reserved.

. Introduction

During a hypothetical severe accident in a nuclear powerlant, which involves significant core damage, hydrogen cane generated and released into the containment. This may resultn local hydrogen concentrations sufficient to pose a threat ofydrogen deflagration or detonation. Such events could causencreases in containment pressure that could affect the con-ainment integrity. Substantial international attention is paid tohis hydrogen control issue (Babic et al., 2006; Royl et al.,006; Forgione and Paci, 2005; Houkema and Willemsen, 2005;katschenko et al., 2005; Siccama et al., 2002), which is onef the key safety issues for nuclear power plants. In order toetermine the risk associated with the hydrogen control issue,t is required to predict the local hydrogen concentration in theontainment during a severe accident. For this prediction, twohermal-hydraulic approaches can be used: the lumped parame-er (LP) approach and the computational fluid dynamics (CFD)pproach. Lumped parameter codes are often referred to asystem codes. NRG has developed the LP code SPECTRA

Stempniewicz, 2000). Though the results with this SPECTRAode were considered as best overall results in Internationaltandard Problem 42 (Lubbesmeyer and Aksan, 2003), LP codes

∗ Corresponding author.E-mail address: houkema@nrg-nl.com (M. Houkema).

•

•

•

029-5493/$ – see front matter © 2007 Elsevier B.V. All rights reserved.oi:10.1016/j.nucengdes.2007.02.033

ave inherent limitations concerning conservation of momen-um. Because CFD codes conserve momentum correctly, NRGlso uses the commercial CFD code CFX4, for which imple-entation of condensation models was needed. Recently, efforts

ave been undertaken to determine the local mass transfer ratesith commercial CFD codes by solving additional species bal-

nce equations with appropriate boundary conditions, that is,ithout the use of engineering correlations (see e.g., Ambrosini

t al., 2001, 2002; Siccama, 2001; Smith et al., 2002; Forgionend Paci, 2005). These commercial CFD codes do generallyot have built-in steam condensations models. Consequently, its necessary to implement such models via user-defined sub-outines. Calculations with CFX4 have been performed forull-scale NPPs (Houkema et al., 2003). The present paper isbout validation of the NRG condensation model in CFX4. Themplemented condensation model will be described and resultsf validation calculations are compared to experimental data.

brief overview will be given on the comparison of resultsbtained with CFX4 versus experimental results from:

The condensation in a vertical tube in the experiment of Kuhnet al. (1997).The PHEBUS test facility within the PHEBUS project, and

PHEBEN-2 fifth framework EU project (Siccama, 2001).The PANDA test facility within the TEMPEST fifth frame-work EU project (Lycklama a Nijeholt and Komen, 2003;Komen et al., 2003).

M. Houkema et al. / Nuclear Engineerin

Nomenclature

A, B, C constants in Antoine equationAwall wall surface area (m2)BD driving force for steam condensationcP specific heat at constant pressure (J/kg K)D diffusion coefficient (m2/s)gm mass transfer coefficient (kg/m2 s)k turbulent kinetic energy (m2/s2)M molecular mass (kg/mol)ni condensation flux (kg/m2 s)P pressure (Pa)Q enthalpy sink (W)R gas constant (8.314 J/mol K)S sinkT temperature (K)x steam mass fractiony distance from the cell centre to the wall (m)y+ dimensionless distance to the wall

Greek symbolsε turbulent dissipation rate (m2/s3)ρ density (kg/m3)ω turbulent frequency (1/s)

Subscriptsb in the bulkbg of the background gascell in the centre of the cell contiguous to the walli at the interfaceref referencewall at the wall

•

hcbl“CnaroTbhbC

2

CAc

2

gflflbtma

c

c

ab

s

IIf

d

It

n

n

i

B

2

ssinks are applied to grid cells contiguous to the walls. The mag-nitude of the sinks depends on the steam condensation rate. The

The TOSQAN, MISTRA, and THAI test facilities within theOECD International Standard Problem 47 (Malet et al., 2002;Tkatschenko et al., 2005; Fischer et al., 2005).

These experiments addressed several important thermal-ydraulic phenomena: wall condensation, atmospheric stratifi-ation, natural circulation, turbulent diffusion, and interactionsetween these phenomena. Within International Standard Prob-em 47, the following has been concluded (Fischer et al., 2005):NRG has demonstrated that results from using an industrialFD-code are of the same quality as the ones obtained withuclear field codes.” NRG has used the commercially avail-ble CFD code CFX4, which is, like STAR-CD and FLUENT,eferred to as an “industrial CFD-code”. These codes are devel-ped for a wide range of non-nuclear and nuclear applications.he so-called “nuclear field codes” are CFD-codes that haveeen specifically developed for nuclear containment thermal-ydraulics (like GASFLOW or GOTHIC). Before ISP47, it was

elieved that “nuclear field codes” were superior to “industrialFD-codes”.

NF

g and Design 238 (2008) 590–599 591

. Model description

The used CFD code is the commercially available programFX4, developed by AEA Technology, nowadays owned byNSYS. CFX4 is a general purpose CFD code. Below, the NRG

ondensation model is described.

.1. Driving force for wall condensation

If steam condensation occurs, there is a net movement ofas towards the wall. This gas movement results in a convectiveow towards the wall. This section describes how the convectiveux is taken into account in steam condensation models. Moreackground information can be found in chapters 16 and 17 ofhe book of Bird, Stewart and Lightfoot (Bird et al., 1960). Theass balance of steam at the interface (subscript i) between gas

nd condensate film at the tube wall is

ondensation flux = steam transported with bulk motion

+ diffusive transport (1)

Suppose, the condensation flux at the wall is

ondensation flux = ni (2)

Consequently, the bulk motion of the gas to the interface islso equal to ni. The flux of steam that is transported with theulk motion is therefore equal to:

team transported with bulk motion = nixi (3)

n this equation, xi is the steam mass fraction at the interface.n a stagnant medium, the diffusive mass transfer between massractions x and xi is equal to:

iffusive transport = gm(x − xi) (4)

n this equation, gm is the mass transfer coefficient. Substitutinghese equations in the mass balance results in

i = nixi + gm(x − xi) (5)

Rearranging this equation results in

i = gmx − xi

1 − xi(6)

Consequently, the driving force (BD) for steam condensations equal to:

D = x − xi

1 − xi(7)

.2. NRG wall condensation model

The NRG steam condensation model consists of sinks ofteam, total mass, enthalpy and the turbulent quantities. These

RG steam condensation model is implemented in the user-ortran of the CFX4 code. Two user-Fortran subroutines are

5 eerin

uflss(ci

n

t

g

dfut

l

TtBtoteicu

Q

wq

S

S

bofdlvaf

2

h

Bsdw

H

H

Spvtfiaupatf

R

t

R

a

ρ

mscw

3K

tgdaile

92 M. Houkema et al. / Nuclear Engin

sed: subroutine USRWTM is used to calculate the steam massuxes and subroutine USRSRC is used to define all sinks. Inubroutine USRWTM, the steam mass fluxes at the condensingurfaces are calculated by the general formulation given by Eq.6), which is here reformulated as Eq. (8). The mass transferoefficient (gm) is a function of the turbulence near the wall, ands calculated in the CFX4 code internally

i = gmxcell − xwall

1 − xwall(8)

If the cells near the walls are sufficiently narrow, the massransfer coefficient reduces to laminar transport:

m = ρD

y(9)

The mass fraction xwall of steam at the surface of the con-ensate film (or possibly water strokes) at the wall is calculatedrom the vapour pressure at the surface. The Antoine equation issed to describe the vapour pressure as a function of the surfaceemperature:

n

(P

1 Pa

)= A + B

T + C(10)

he coefficients A, B and C are fitted on data from steamables. The result of this fitting process is A = + 23.1512,= −3788.02 K, and C = −47.3018 K. In subroutine USRSRC,

he sinks of mass are defined. The steam sink as well as the sinkf the total mass are modelled according to equation (1). Fur-hermore, an enthalpy sink is required in order to remove thenthalpy associated with the mass of steam that is removed. Its assumed that the condensing wall immediately removes theondensation heat. In CFX4, the enthalpy sink Q (energy pernit time) for each cell is calculated by:

= niAwall(cP,stTcell − cP,bgTref) (11)

In case mass is removed, the turbulent quantities associatedith this mass must also be removed. The sinks of the turbulentuantities are calculated by:

k = niAwallkcell, Sε = niAwallεcell,

ω = niAwallωcell (12)

The diffusion coefficients used are obtained from the hand-ook of Vargaftik et al. (1996). The diffusion coefficient of H2 isbtained from H2/N2 or H2/air systems. The diffusion coefficientor steam is obtained from steam/air or steam/N2 systems. Theiffusion coefficients are determined for a constant temperatureevel and are kept constant during the calculation. For the laminariscosity of the mixture also a constant value is used and is evalu-ted at average conditions. At present, NRG investigates the needor implementation of multi-component diffusion coefficients.

.3. NRG mist model

If the partial steam pressure at a certain location becomesigher than the saturation pressure, bulk condensation will occur.

fhpw

g and Design 238 (2008) 590–599

ulk condensation is the process of converting water vapour intomall water droplets; so-called fog or mist. In CFX4, bulk con-ensation has been modelled by an equilibrium reaction betweenater vapour and mist:

2O (vapour, gas)

→ H2O (mist droplets, gas) + heat of condensation (13)

2O (mist droplets, gas) + heat of condensation

→ H2O (vapour, gas) (14)

The mist mass fraction has been defined as an Algebraiclip Model user scalar. The Algebraic Slip Model is a multi-hase model in which the relative velocity of the phases, the slipelocity, is defined as a prescribed value. The slip velocity ishe terminal velocity that mist droplets reach in a gravitationaleld. The magnitude of the slip velocity is based on the expectedverage mist droplet size. In each cell that contains gas, the sat-ration pressure of water vapour is calculated. The saturationressure is calculated by the Antoine equation, Eq. (10). If thectual water vapour fraction in a cell is higher than the satura-ion fraction, the gas mixture is supersaturated and mist will beormed

condensation = kcondensation xsteam (15)

If the water vapour fraction in a cell is lower than the satura-ion fraction, mist – if present – will evaporate.

evaporation = kevaporation xsteam (16)

The total fluid density is a function of the mist mass fractionnd is calculated by the following equation:

= MP

RT(1 + xmist) (17)

The modelled mist can be transported to surfaces by: bulkotion; diffusion; and slip in gravitational direction. The depo-

ition of mist on surfaces is simulated by defining the mistoncentration on wall to be equal to zero, which means mistill instantly disappear on surfaces.

. Validation of the NRG condensation model with theuhn condensation model

Kuhn et al. (1997) conducted experiments on the local heatransfer from condensation in the presence of non-condensableases inside a vertical tube. The condenser tube has an inneriameter of 47.5 mm and is 3.37 m long. The tube consists of andiabatic entrance of 0.81 m, followed by a 2.4-m long condens-ng section. Local heat fluxes are measured accurately at severalocations in the tube. On the basis of the results obtained (Kuhnt al., 1997), physically based correlations have been developed

or the heat and mass transfer of the vapour-gas side. The totaleat transfer coefficient from the gas to the wall cooling waterredicted by the correlations in their article is in close agreementith experimental values. The Kuhn model has the following

eerin

sfiibfaepNcsbubauiNidbsmc

•••

2o

3c

dmpa

ommtt

3

ta

3

begtmcfip5oao

3

bapoTl3rate is negligible, provided that the y+ values are in the correct

M. Houkema et al. / Nuclear Engin

hortcomings: the steam condensation rate may become unde-ned; entrance effects are not taken into account; free convection

s not taken into account; and the gas temperature may fallelow the saturation temperature. However, the model is use-ul to compare the NRG model with, because the entrance effectnd the effect of free convection are small for the studied geom-try. Unfortunately, the experimental local heat fluxes are notresented in the article of Kuhn et al. (1997). Therefore, theRG condensation model is compared to the physically based

orrelations in this article. In order to obtain a good compari-on, the experimental apparatus is modelled in CFX4 and theoundary conditions that are selected fall in the range that issed in the article. In order to obtain a sensible comparisonetween the models, the physical properties in both modelsre identical and a fixed condensing surface temperature wassed. The inlet was modelled as an ‘Inlet Boundary’, with spec-fied velocity, turbulence intensity and dissipation length scale.ine two-dimensional computational meshes have been used

n order to obtain y+ values (representing the non-dimensionalistance between the wall and the first computational cell)etween 0.5 and 200. The corresponding CPU time for a steady-tate calculation ranges from 1 s up to 20 min for the finestesh. Three turbulence models have been used for the CFD

alculations:

the standard k–ε model.the low-Reynolds k–ε model.the k–ω model.

The first turbulence model requires preferably y+ higher than0, while the latter two work best for y+ around 1. The influencef the y+ value on the results is presented below.

.1. Results of the comparison between the twoondensation models

The results of the NRG CFX4 condensation model with

ifferent turbulence models are compared to the condensationodel of Kuhn. The total condensation rates of both models are

resented as a parity plot in Fig. 1. Different condensation ratesre achieved by variation of: the steam fraction; the temperature

Fig. 1. Comparison with the Kuhn model.

ryt

Fc

g and Design 238 (2008) 590–599 593

f the gas; and the wall temperature. Fig. 1 shows that the maxi-um difference between the CFX4 results and the condensationodel of Kuhn is 11%. The standard deviation between these

wo models is 3%. It is noted that the standard deviation betweenhe experiment and the Kuhn correlations is 6.4%.

.1.1. Influence of the turbulence modelFig. 1 shows that the influence of the turbulence model on

he condensation rate is negligible, provided that the y+ valuesre in the correct range.

.1.2. Influence of the y+ value at the wallIt is well known that the y+ value has a large influence on

oundary layer flow predictions. In order to investigate the influ-nce of the y+ value on the three turbulence models, nine differentrids were used to model the experimental tube. The results ofhe calculations with different grids and the three turbulence

odels are shown in Fig. 2. This figure shows the ratio of theondensation rates calculated by CFD and the Kuhn model. Thegure shows that the low-Reynolds k–ε and k–ω models indeederform best at y+ lower than 2. At y+ values higher than aboutthese low-Reynolds models show a sudden over-prediction

f the condensation rate. The standard k–ε model works bestt y+ higher than 20. At lower values, the condensation rate isver-predicted.

.2. Summary

The NRG condensation model is compared to the physicallyased correlations described in literature, which is referred tos the Kuhn condensation model. The NRG model gives com-arable condensation rates as the Kuhn model for a large rangef conditions, like steam mass fractions between 30 and 99%.he difference between the two condensation models is always

ower than 11%. The standard deviation of all calculations is%. The influence of the turbulence model on the condensation

ange. The low-Reynolds k–ε and k–ω models perform best at+ lower than 2. At higher y+ values – higher than about 5 –hese low-Reynolds models show a sudden over-prediction of

ig. 2. The influence of the y+ value at the wall on the calculation of theondensation rate.

594 M. Houkema et al. / Nuclear Engineering and Design 238 (2008) 590–599

Fig. 3. The structured mesh of the PHEBUS vessel, as used in the CFX4 model.

eering and Design 238 (2008) 590–599 595

thtrCt

4

tfcebmslparnhisdttr

4

iPa(tlmtFc

gpftlptaRtne

Ft

4

tm(goC2PtWncia

g(dicted helium stratification. This stratification, as presented inFig. 5, could explain the discrepancy between the experimentalpressure and the pressure predicted with the lumped parametercodes.

M. Houkema et al. / Nuclear Engin

he condensation rate. The standard k–ε model works best at y+

igher than 20. Generally, it is required to keep the y+ smallerhan 100–300. At values lower than about 20, the condensationate is over-predicted when using the standard k–ε model. TheFD result is independent of the y+ value at y+ values lower

han 2.

. Validation against experiments in test facilities

A shortcoming of the comparison with the Kuhn model ishat only forced convection conditions are considered. In aull scale reactor containment, also free convection flows andombinations of forced/free convection boundary layers can bexpected (Houkema et al., 2003). Therefore, experiments haveeen performed in facilities representing (small-size) contain-ents. NRG validated her condensation models in CFX4 for

everal of these experiments. An attempt has been made to fol-ow the Best Practice Guidelines established within the ECORAroject (Menter, 2002). If these Best Practice Guidelines (BPG)re followed, independency of both mesh and time step areeached and proven. The mesh independency is proven by run-ing the computation on at least three meshes. The first meshas the largest cell size. In each following mesh, the cell sizes halved, until the solution does not change between two con-ecutive meshes. Time step independency is proven by usingifferent time steps until the solution does not change betweenwo consecutive time steps. However, due to the large compu-ational effort that they impose, the scope of the BPG has to beeduced for most applications.

.1. The PHEBUS test facility

The NRG condensation model has been validated with exper-mental data from the PHEBUS test facility, as used within theHEBUS FTP1 project (Lycklama a Nijeholt and Hart, 1999),nd within the PHEBEN-2 fifth framework EU FPT0 projectSiccama, 2001). Fig. 3 presents the PHEBUS vessel, containinghree condenser tubes. The low Reynolds number k–ω turbu-ence model and a structured grid have been applied. In this

odel the boundary layer is fully resolved and the condensa-ion rate is calculated from diffusive flux to the condensers.ig. 4 shows the agreement between the calculated and measuredondensation rates.

In both of the PHEBUS projects, the NRG predictions were inood agreement with the experiment. In addition to the presentedrediction of the condensation rate, the calculated temperaturesollow the experimentally observed trends of the temperatures,hough the experimental variation was lower than the calcu-ated variation. The calculated pressure follows the experimentalressure closely. Summarising, validation of the code withhe PHEBUS experiment showed very good results with the

pplication of a low Reynolds number modification. This loweynolds number k–ω turbulence model with full resolution of

he boundary layers was selected for this validation, becauseear-laminar flow and free convection boundary layers werexpected.

ig. 4. The measured and predicted condensation rate for the PHEBUS FPT0est.

.2. The PANDA test facility

For the majority of applications, NRG used the condensa-ion model in combination with the standard k–ε turbulence

odel with wall functions. In International Standard Problem 42Lubbesmeyer and Aksan, 2003) and in the fifth Framework Pro-ramme project TEMPEST (Testing and Enhanced Modellingf Passive Evolutionary Systems Technology for Containmentooling) (Lycklama a Nijeholt and Komen, 2003; Komen et al.,003; Wichers et al., 2003), experiments were performed in theANDA facility. PANDA is a relatively large thermal-hydraulicest facility, which can be used for investigating Advanced Light

ater Reactor containment system behaviour and related phe-omena. Prototypical fluids such as water and steam are used asoolant at prototypical conditions. The total height of the facilitys 25 m and it is designed for operating conditions up to 10 barnd 200 ◦C.

In ISP42 it was observed that the lumped parameter codesave a systematic error of the containment pressure in PANDALubbesmeyer and Aksan, 2003). The CFX4 results of NRG pre-

Fig. 5. CFX4 prediction of helium stratification in the PANDAfacility.

5 eerin

4I

clITmacbNik

4

srAaaifibFp(

4

nACuotrsa

F

thti

4.3.3. The THAI test facilityThe THAI test facility is a 60 m3 vessel with several

obstructions. In this way, the vessel is intended to represent a

96 M. Houkema et al. / Nuclear Engin

.3. TOSQAN, MISTRA and THAI facilities innternational Standard Problem 47

In order to assess the capabilities of both LP and CFDode approaches for the hydrogen control issue, the OECD hasaunched International Standard Problem 47 (ISP47). WithinSP47, experiments have been performed in the French facilitiesOSQAN (IRSN) and MISTRA (CEA), and in the Ger-an THAI (Becker Technologies) facility. These experiments

ddressed several important thermal-hydraulic phenomena: wallondensation, atmospheric stratification, natural circulation, tur-ulent diffusion, and interactions between these phenomena.RG participated in ISP47 by modelling all of the experiments

n the three facilities, applying the CFX4 model with the standard–ε turbulence models with wall functions.

.3.1. The TOSQAN test facilityThe TOSQAN facility (Malet et al., 2002) is a relatively

mall facility (7 m3). In the experiment, wall temperatures wereegulated and central steam and helium injection took place.

two-dimensional axi-symmetric model could therefore bepplied, as was done by the majority of participants (Forgionend Paci, 2005). Using CFX4, NRG analysed several steamnjection rates and phases were studied, as Fig. 6 reveals. Thisgure presents the experimental and calculated pressures (inars) in the different phases of the ISP47 TOSQAN experiments.rom this and other comparisons it was concluded that the CFX4rediction was in fairly good agreement with the experimentMalet et al., 2005).

.3.2. The MISTRA test facilityThe MISTRA test facility (Tkatschenko et al., 2005) is sig-

ificantly larger than the TOSQAN facility: 100 m3 versus 7 m3.s for TOSQAN, central steam and helium injection took place.ooled condenser plates were present in the facility. However,nintended spurious condensation on the containment wallsccurred in the experiment. Since NRG’s model did not include

his spurious condensation, NRG’s attempt to validate the CFX4esults by the MISTRA experiment was limited. Only the steadytates were modelled. For the steam steady state NRG predictedpressure of 3.61 bar, where the measured value was 3.34 bar. In

ig. 6. Experimental and predicted pressure for the TOSQAN experiment.

g and Design 238 (2008) 590–599

he helium-steam steady state, the predicted pressure was againigher than the experimental one: 5.70 bar versus 5.38 bar. Sincehe spurious condensation was not modelled, an over-predictionn the CFX4 result is expected.

Fig. 7. The THAI vessel.

M. Houkema et al. / Nuclear Engineerin

Fp

coutttsr(

mdwnbatbCHmfitwnebasi

5

etaoffi

scmftcp

6

cg

•••

Ch

6

fladrfntvcm

6

amtmcmprbsto

ig. 8. Blind CFD code prediction of the helium concentration in the upperlenum of the THAI vessel.

ompartmented reactor containment. Fig. 7 gives an impressionf the THAI facility. The ISP47 experiment comprised of anpper helium injection phase, followed by an upper steam injec-ion phase at another location. Then, horizontal steam injectionook place in the lower part of the vessel. There were no injec-ions in the fourth phase of the experiment. The experiment waset-up in such a way that though a three-dimensional model wasequired, symmetry allowed modelling of only half the vessel180◦). This was done with the CFX4 model.

The experimental results showed stratification, which wasaintained in the third phase of the experiment. This gas

ensity stratification involves a mixed laminar/turbulent or aeakly turbulent flow. In the CFD analyses, the high Reynoldsumber k–ε turbulence model was applied. This model haseen developed for fully turbulent flows and over-predicts themount of turbulence in weakly turbulent flows. Consequently,he CFD results consistently showed a much more intense tur-ulent mixing than observed in the experiments. TherewithFX4 predicted a too rapid weakening of the stratification.ence, although the CFD code contains models for the drivingechanisms for both creating as well as for destroying strati-cation, it incorrectly predicted the stratification behaviour in

he THAI experiment, as can be seen in Fig. 8. NRG’s resultsere fairly good for the first two phases, but the results doot maintain the stratification in the third phase. It cannot bexcluded that the experimental conditions that were used asoundary conditions in CFX4 somehow cause this discrep-ncy (Royl et al., 2006). For instance, there appears to beome asymmetry in the experimental solution, which may be ofmportance.

. Conclusions

To model the condensation heat and mass transfer in the pres-nce of non-condensable gases in CFX4, appropriate sourceerms have been implemented into the mass and energy bal-

nce equations by user coding. These source terms are basedn the steam concentrations gradients at the condensing sur-aces. Furthermore, the assumption of an extremely thin liquidlm has been made. Based on overall measured parameters

mefc

g and Design 238 (2008) 590–599 597

uch as system pressure and total condensate masses, it wasoncluded that condensation is predicted fairly well for theajority of validation cases presented in this paper. Deviations

rom the experimental results appear to originate primarily fromhe applied wall treatment, i.e., the treatment that is used toompute the near wall velocity, temperature, and concentrationrofiles.

. Recommendations

For further improvement of the CFD approach forontainment-thermal hydraulic analyses, recommendations areiven with respect to the following issues:

Condensation.Stratification.Specific best practice guidelines (BPG) for CFD containmentthermal-hydraulic analyses.

In addition, recommendations are given for a combinedFD/LP code approach for performing containment thermal-ydraulic analyses.

.1. Condensation

When Stefan’s law is used to calculate the condensationuxes, the application of the automatic wall treatment would ben improvement. However, the application of this wall treatmentoes still not guarantee that the boundary layers are correctlyepresented. Therefore, the automatic wall treatment could beurther improved by for example the application of an adaptiveear wall mesh refinement. Concerning flow and heat transfer,he impinging jet has been widely used for turbulence modellingalidation. It is recommended to determine what improvementan be made for thermal-hydraulics, especially in relation toass transfer (i.e., condensation) modelling.

.2. Stratification

Sub-models for turbulence damping due to stratification havelready been developed for standard two-equation turbulenceodels. However, some controversy exists in the literature about

he exact model formulations. It is recommended to deter-ine whether the existing formulations are adequate for nuclear

ontainment thermal-hydraulics. Furthermore, it is noted thatore advanced three and four-equation turbulence models are

resented in the literature, which are claimed to be supe-ior to standard two-equation turbulence models for complexuoyancy-driven flows. It is recommended to determine whetheruch more advanced models are really required for nuclear con-ainment thermal-hydraulics. Mixed laminar/turbulent flows areften related to stratification. For such flows, low Reynolds

odifications for turbulence models are available in CFD. How-

ver, they currently have to be used always together withull resolution of the boundary layers. Therefore, the appli-ation of low Reynolds modifications is currently often not

5 eerin

ptt

6t

mtiitotc

6

cwaroLccfioibtimacLccccborta

R

A

A

B

B

F

F

H

H

K

K

L

L

L

M

M

M

R

S

S

S

98 M. Houkema et al. / Nuclear Engin

ractical. It is proposed to implement low Reynolds modifica-ions in combination with the earlier discussed automatic wallreatment.

.3. Specific best practice guidelines for CFD containmenthermal-hydraulic analyses

Present user guidelines and nodalisation rules for CFD codesust be expanded, and they must be applied. For example

he turbulence simulation in the CFD models was not accord-ng to best practice guidelines. It is recommended to start annternational activity to develop specific BPG for containmenthermal-hydraulic CFD analyses. These BPG should amongstthers include guidelines concerning jet modelling for con-ainment thermal-hydraulics as well as guidelines for wallondensation.

.4. Combined CFD/LP code approach

Lumped Parameter codes have proven that valuable resultsan be obtained for containment thermal-hydraulics. LP codesill always require less CPU time than CFD, and thus are suit-

ble for sensitivity calculations, which are required for nucleareactor safety analyses that have to deal with the stochastic naturef initial and boundary conditions. Therefore, it is evident thatP codes will always remain the main workhorse for analysingontainment thermal-hydraulic accident scenarios. However, LPodes have some inherent limitations due to the applied simpli-ed flow model description. This results in a large user influencen LP code results, as evidenced in ISP47 by the large variationn the outcomes of LP code results. Therefore, combined use ofoth LP codes and CFD is recommended, where LP will act ashe main workhorse for the containment analyses, whereas CFDs used to analyse a few selected accident scenarios which require

ore detailed analyses of the (local) phenomena that occur. Inddition, it is recommended to use CFD to analyse a few selectedritical accident scenarios which are difficult to analyse usingP codes due their inherent limitations. Possibly, the use of CFDan guide the LP code user in selecting the proper models andode parameters for bridging these limitations. The proposedombined approach requires some further development of bothodes (Fischer et al., 2005). The overall objectives of the com-ined LP/CFD approach are to improve and guarantee the qualityf the containment thermal-hydraulic analyses, and therewith toeduce the safety margins. In the proposed way, containmenthermal-hydraulic analysis will benefit from the best of both LPnd CFD.

eferences

mbrosini, W., Forgione, N., Mazzini, D., Oriolo, F., He, S., 2001. CFD anal-ysis of evaporation cooling experimental tests. In: Proceedings of the Ninth

International Conference on Nuclear Engineering (ICONE 9), Nice, France,April 8–12, p. 2001.

mbrosini, W., Forgione, N., Mazzini, D., Oriolo, F., 2002. Computational studyof evaporative film cooling in a vertical rectangular channel. Heat Transfer.Eng. 23, 25–35.

S

g and Design 238 (2008) 590–599

abic, M., Kljenak, I., Mavko, B., 2006. Modeling of non-homogeneous con-tainment atmosphere in the ThAI experimental facility using a CFD code.In: Proceedings of 14th International Conference on Nuclear Engineering(ICONE14), Miami, FL, USA, July 17–21.

ird, R.B., Stewart, W.E., Lightfoot, E.N., 1960. Transport Phenomena. JohnWiley, NewYork.

ischer, K., Royl, P., Heitsch, M., Lukianov, A., Sato, N., Babic, M., Houkema,M., Wiles, L.E., George, T.L. Wheeler, C.L., Paci, S., 2005. InternationalStandard Problem ISP-47 on Containment Thermal-Hydraulics. Step 2:ThAI. Volume 4: Comparison Report of Open Phases I–II–III–IV. BeckerTechnologies GmbH, Eschborn, Report Nr. BF-R 70031-4.

orgione, N., Paci, S., 2005. Computational analysis of vapour condensa-tion in presence of air in the TOSQAN facility. In: Proceedings of the11th International Topical Meeting on Nuclear Reactor Thermal-Hydraulics(NURETH-11), Avignon, France, pp. 1–14.

oukema, M., Siccama, N.B., Willemsen, S.W., Komen, E.M.J., 2003. CFDanalyses of steam and hydrogen distribution in a nuclear power plant. In:Proceedings of the 10th International Topical Meeting on Nuclear ReactorThermal Hydraulics (NURETH-10), Seoul, Korea, October 5–9.

oukema, M., Willemsen, S.M., 2005. Assessment of Unintended PAR Ignition,NRG report 22007/04.64050/C, Petten, The Netherlands.

omen, E.M.J., Willemsen, S.M., Roelofs, F., 2003. Modelling SWR-1000Building Condenser Behaviour. CFD Analysis of the PANDA BC4 Contain-ment Phenomena. NRG report 20407/03.53012/C, Petten, The Netherlands.

uhn, S.Z., Schrock, V.E., Peterson, P.F., 1997. An investigation of condensationfrom steam—gas mixtures flowing downward inside a vertical tube. NuclearEng. Design 177, 53–69.

ubbesmeyer, D., Aksan, S.N., 2003. ISP42 (PANDA Tests), Open Phase Com-parison Report. NEA/CSNI/R(2003)7, OECD Nuclear Energy Agency.

ycklama a Nijeholt, J.A., Hart, J., 1999. CFD simulation of thermal-hydraulicsin Phebus FP containment vessel. Nuclear Europe Woldscan (NEW),January–February, 1999.

ycklama a Nijeholt, J.A., Komen, E.M.J., 2003. Modelling ESBWR passivecontainment cooling system performance. CFD Analysis of Condensa-tion and Helium Distribution in the PANDA Drywells. NRG report20407/03.53769/C, Petten, The Netherlands.

alet J., Porcheron E., Cornet P., Brun P., Norvez O., Menet B., Thause L.,2002. ISP-47, International Standard Problem on Containment thermal-hydraulics, Step 1: TOSQAN-MISTRA, Phase A: air-steam mixturesTOSQAN experimental results. IRSN Rapport DPEA/SERAC/LPMAC/02-45, Saclay, France.

alet J., Porcheron E., Cornet P., Brun P., Menet, B., Vendel, J., 2005. ISP-47, International Standard Problem on Containment thermal-hydraulics,Step 1: TOSQAN-MISTRA, TOSQAN Phase B: air-steam-helium mixtures,Comparison Code-Experiments. IRSN Report DSU/SERAC/LEMAC/05-19, Saclay, France.

enter, F., 2002. CFD best practice guidelines for CFD code validation forreactor-safety applications. European Commission report, EVOL-ECORA-D01.

oyl, P., Lee, J.J., Travis, J.R., Breitung, W., 2006. Benchmarking of the 3D CFDCode Gasflow II with Containment Thermal Hydraulic Tests from HDR andThai. In: Conference Proceedings CFD4NRS 2006.

iccama, N.B., 2001. CFD simulation of the flow phenomena in the Phebuscontainment during the FPT0 Release Phase. SAM-PHEBEN2-T05, NRGreport 20397/01.41354/C, Petten, The Netherlands.

iccama, N.B., Houkema, M., Komen, E.M.J., 2002. Determination of the steamand hydrogen distribution in a nuclear power plant. In: NRG Paper Presentedat the Technical Meeting on the Use of CFD Codes for Safety Analysis ofReactor System, including Containment, Pisa, Italy, 11–15 November, p.2002.

mith, B.L., Milelli, M., Shepel, S., 2002. Aspects of nuclear reactorsimulation requiring the use of advanced CFD models. In: TechnicalMeeting on Use of Computational Fluid Dynamics (CFD) Codes for

Safety Analysis in Reactor Systems, Including Containment, Pisa, 11–13November.

tempniewicz, M.M., 2000. Simulation of containment transient response dur-ing accidents in advanced reactor types—the computer code SPECTRA.PhD Thesis. NRG report 21437/00.52167/P, Arnhem, The Netherlands.

eerin

T

V

Wichers, V.A., Malo, J.A., Starflinger, J., Heitsch, M., Preusser, G., Tuomainen,

M. Houkema et al. / Nuclear Engin

katschenko, I., Studer, E., Paillere, H., 2005. Status of the MISTRA programme

for the validation of containment thermal-hydraulic codes. In: Proceedingsof the 11th International Topical Meeting on Reactor Thermal-Hydraulics(NURETH-11), Avignon, France.

argaftik, N.B., Vinogradov, Y.K., Yargin, V.S., 1996. Handbook of PhysicalProperties of Liquids and Gases. Begell House, New York.

g and Design 238 (2008) 590–599 599

M., Huggenberger, M., 2003. Testing and enhanced modelling of passiveevolutionary systems technology for containment cooling (TEMPEST). In:FISA Proceedings November 2003, EU Research in Reactor Safety, Lux-emburg.