deen_ned

7/29/2019 deen_ned

1/11

Nuclear Engineering and Design 241 (2011) 51385148

Contents lists available at SciVerse ScienceDirect

Nuclear Engineering and Design

j o u rn a l h o m ep ag e : www.e l sev i e r. co m/ l o ca t e / n u cen g d es

CFD studies on the phenomena around counter-current ow limitations of gas/liquid two-phase ow in a model of a PWR hot leg

Deendarlianto a,b , , Thomas Hhne a , Dirk Lucas a , Christophe Valle a , Gustavo Adolfo Montoya Zabala ca Helmholtz-Zentrum Dresden-Rossendorf e.V., Institute of Safety Research, P.O. Box 510 119, D-01314 Dresden, Germanyb Department of Mechanical & Industrial Engineering, Faculty of Engineering, Gadjah Mada University, Jalan Graka No. 2, Yogyakarta 55281, Indonesiac Department of Chemical Engineering, Simn Bolvar University, Valle of Sartenejas, Caracas 1080, Venezuela

a r t i c l e i n f o

Article history:Received 6 April 2011Received in revised form 24 August 2011Accepted 24 August 2011

a b s t r a c t

In order to improve the understanding of counter-current two-phase ow and to validate new physicalmodels, CFDsimulationsof a 1/3rdscalemodel ofthe hotlegof a German Konvoi pressurizedwater reactor(PWR) with rectangular cross section were performed. Selected counter-current ow limitation (CCFL)experiments conducted at Helmholtz-Zentrum Dresden-Rossendorf (HZDR) were calculated with ANSYSCFX using the multi-uid EulerEuler modelling approach. The transient calculations were carried outusing a gas/liquid inhomogeneous multiphase ow model coupled with a shear stress transport (SST)turbulence model.

In the simulation, the drag law was approached by a newly developed correlation of the drag coef-cient ( Hhne and Valle, 2010 ) in the Algebraic Interfacial Area Density (AIAD) model. The modelcan distinguish the bubbles, droplets and the free surface using the local liquid phase volume fractionvalue. A comparison with the high-speed video observations shows a good qualitative agreement. Theresults indicate also a quantitative agreement between calculations and experimental data for the CCFL characteristics and the water level inside the hot leg channel.

2011 Elsevier B.V. All rights reserved.

1. Introduction

Two-phase owmay occur in pressurized water reactors (PWR)following a leakage in the primary cooling circuit. In the eventof hypothetical accident scenarios in PWR, emergency strategieshave to be mapped out, in order to guarantee the reliable removalof decay heat from the reactor core, also in case of componentbreakdown. One essential passive heat removal mechanism is thereux-condenser mode. This mode can appear for instance duringa small break loss-of-coolant-accident (LOCA) or because of loss of residual heat removal (RHR) system during mid loop operation atplant outage after the reactor shutdown.

In the hypothetical accident scenario of a loss-of-coolant-accident (LOCA) due to the leakage at any location in the primarycircuit, it is expected that the reactor will be depressurized andvaporization will take place, thereby creating steam in the PWR primary side. Should this lead to reux condensation, whichmaybe a favourable event progression,the generatedsteamwill owtothe steam generator through the hot leg. This steam will condense

Corresponding author at: Helmholtz-Zentrum Dresden-Rossendorf e.V., Insti-tute of Safety Research, P.O. Box 510 119, D-01314 Dresden, Germany.Tel.: +49 351 260 3291; fax: +49 351 260 13291.

E-mail addresses: [email protected] , [email protected] ( Deendarlianto).

in the steam generator and the condensate will ow back throughthe hot leg to the reactor, resulting in counter-current steam/waterow. In some scenarios, the success of core cooling depends onthe behaviour of this counter-current ow ( Deendarlianto et al.,submitted for publication ).

In thereux-condensermode, a partof thecondensatewill owsback to the reactor core in counter-current to the steam ow. Thecounter-current ow of steam and condensate is only stable fora certain range of mass ow rates. If the steam mass ow rateincreases too much, the condensate is clogged in the hot leg. Thecondensate is carried over by the steam and partially entrained inthe opposite direction to the steam generator. This phenomenonis known as the counter-current ow limitation (CCFL) or ooding,and could affect the cooling of the reactor core. Detailed examplesof such LOCA scenarios leading to the reux condenser mode canbe found in Jeong (2002) .

A lot of experiments were carried in the past in order tounderstand the phenomena around CCFL in a model of hotleg PWR. Several experimental correlations were developed topredict the CCFL on them, but they are only valid in spe-cic experimental ranges. Therefore, high resolution experimentaldata at reactor typical boundary conditions is needed. In orderto improve the transient analysis of counter-current two-phaseows, experimental studies were conducted at Helmholtz-ZentrumDresden-Rossendorf (HZDR). A 1/3rd scale model of the hot leg of

0029-5493/$ see front matter 2011 Elsevier B.V. All rights reserved.

doi: 10.1016/j.nucengdes.2011.08.071
http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.nucengdes.2011.08.071http://www.sciencedirect.com/science/journal/00295493http://www.elsevier.com/locate/nucengdesmailto:[email protected]:[email protected]:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.nucengdes.2011.08.071http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.nucengdes.2011.08.071mailto:[email protected]:[email protected]://www.elsevier.com/locate/nucengdeshttp://www.sciencedirect.com/science/journal/00295493http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.nucengdes.2011.08.071

7/29/2019 deen_ned

2/11

Deendarlianto et al. / Nuclear Engineering and Design 241 (2011) 51385148 5139

Fig. 1. Schematic drawing of the hot leg model test section (dimension: in mm) ( Deendarlianto et al., 2008 ).

a German Konvoi PWR with rectangular cross section was used(Deendarlianto et al., 2008; Valle et al., 2009a ).

The nuclear thermal-hydrauliccommunity is facing today inter-esting challenges. These include the development and validation of new computational tools that will be used for improved and moredetailed analysis as well as new generations of reactors. Currenttrends are toward multi-dimensional, -scale, -physics approachesfor such analyses ( Yadigaroglu, 2005 ). For this purpose, the com-putational uid dynamics (CFD) tool is considered to be able tosimulate most of two-phase ow congurations encountered innuclear reactor power plants.

The most widely used analysis to model the CCFL in a PWR hotleg is based on the one dimensional two-uid models as reportedby Ardron and Baneerjee (1986) , Lopez de Bertodano (1994) andWongwises(1996) . In thisapproach, theswitching point, where theow condition changes from sub to critical condition in stratiedow is approached by the experimental correlation in a speciedow direction. Wang and Mayinger (1995) did two-dimensionalanalyses of counter-current model of UPTF Test TRAM A2 & Test11 using a two-uid model. They implemented the interfacial fric-

tion factor proposed by Lee and Bankoff (1983) and Ohnuki (1986)into the code FLOW3D. They reported that satisfactory resultswere obtained, whereas, under the reux condensation condi-tions, numerical computation reveals thatdifferent ow structuresappeared in the region away from the ooding curve and in theregion near the ooding curve.

Murase and his co-workers in Tsuruga -Japan are in disagree-ment with the study of Wang and Mayinger. They claimed that theeffects of wall friction can not be correctly evaluated by using two-dimensional analysis. The given boundary conditions at the inletand outlet of the hot leg in the above work might affect the calcu-lated ow patterns in the hot leg. For this reason they conducted3-DCFD calculations. Their CFDworks can be found in Murase et al.(2009) , Minami et al. (2009, 2010b) , Kinoshita et al. (2009) , and

Utanohara et al. (2009) .Murase et al. (2009) , Minami et al.(2009, 2010b) and Utanoharaet al. (2009) conducted 3-D CFD simulations on counter-currentow in a PWR hot-leg airwater two-phase ow in a 1/15th scalemodel. This calculation model reproduced the size of experimentaltest facility at Kobe University as reported by Minamiet al.(2010a) .Their works included the effects of interfacial friction correlation(Utanohara et al., 2009 ), ow patterns and CCFL ( Murase et al.,2009; Minami et al., 2009, 2010b ). They used the volume of uid(VOF) and EulerEuler two-uid models on the commercial CFDcode FLUENT6.3.26. The required interfacial friction correlations inthe EulerEuler two-uid model were selected from a combinationof available 1-D experimental correlations for the cases of annularand slug ow that gave the best agreement with the experimental

data. They concluded that it is better to use the two-uid model

with suitable interface friction correlation than VOF model. Thepredicted ow patterns, hysteresis behaviours, and CCFL charac-teristics agree well with their experimental data. Meanwhile thosecorrelations were obtained on the basis of one dimensional anal-ysis, which might affect the calculation results. The use of the 1-Dexperimentalcorrelationto the 3-Dproblem might be notaccuratewhen we look into the local physics of the phenomenon.

The development of a general model closer to physics andincluding less empiricism is a long-term objective of the HZDR research programs. Here local geometry independent models formass,momentum,heattransfer, andscalar transportare developedand validated. Such models are an essential precondition for theapplication of CFD codes to the modelling of ow related phenom-ena in nuclear facilities. One of the developed scientic methodsto solve the above problems was the new concept of drag coef-cient in the algebraicinterfacial area density model (AIAD)( Hhne,2009 ).

The aim of this paper is to simulate the phenomena around theCCFL in a PWR hot legwithnewly developedof newconceptof dragcoefcient in the AIAD model to the EulerEuler problem. It allows

the detection of the morphological form of the two phase ow andthe corresponding switching via a blending function of each corre-lation from one object pair to another. The new drag correlation inthis model considers the 3D effects of the simulated phenomenon.

2. Experimental apparatus and procedures

The details of the experimental apparatus and procedure usedin the present study were described in the previous papers(Deendarlianto et al., 2008; Valle et al., 2009a ) and only the mainfeatures are presented here. Fig. 1 shows a schematic drawing of the test section. The tested uids were airwater and saturatedsteamwater. Two vessels simulate the reactor pressure vessel(RPV) simulator and steam generator (SG) separator are connected

by a test section that simulates the 1/3rd scale model of the hotleg PWR of a German Konvoi Pressurized Water Reactor. Both ves-sels are identical vessels with 0.8 m 0.5m 1.55m ( D W H )cubic shape. The water level in both vessels was determined by themeasurement of the differential pressure between the top and thebottomof the vessels with differential pressure transducers. A vor-tex meter was used to measure the injected water mass ow rate.The injected air mass ow rate was measured and controlled usinga thermal mass ow meter, the steam ow rate over the pressuredrop through an ISA nozzle.

The test section is composed of a horizontal rectangular chan-nel, a bend that connects it to an upward inclined and expendedchannel, and a quarter of a circle representing the steam genera-tor inlet chamber. The horizontal part of test section is 2.12 m long

and has a rectangular cross-section of 0.05 m 0.25m. The riser is

7/29/2019 deen_ned

3/11

5140 Deendarlianto et al. / Nuclear Engineering and Design 241 (2011) 51385148

Fig. 2. Schematic diagram of the experimental apparatus ( Deendarlianto et al., 2008 ).

0.23 m long, hasan inclination of 50 to the horizontal plane and anexpansionangle of 7.5 . Theinnerand outer bend radii of curvaturewere 0.25 and0.5 m, respectively. Unfortunately,due to theoveralldimensions of the hot leg model, it was not possible to visualizethe complete test section. Therefore, a region of observation hadto be chosen. Previous investigations (e.g. Ohnuki et al., 1988 ) indi-cate that the most agitatedow regionis located near the bend andthat a recirculation zone forms there. Consequently, it was chosento observe the bendedregion of the hotleg andthe steam generatorinlet chamber as shown in Fig. 1. The ow behaviour was recordedby a high-speed video camera at frequencies of 60100Hz and ashutter speed of 1/500 to 1/1000 s.

This experimental apparatus is put in a pressure tank, where it

was operated in pressure equilibrium with the inner atmosphereof the tank as shown in Fig. 2. The photo of the hot leg model isshown in Fig. 3. Air and nitrogen were injected to increase thepressure in the tank to a maximum operation pressure of 5.0 MPaof the airwater and the steamwater experiments respectively.The detailed principle of the pressure equilibrium technique wasdescribed by Prasser et al. (2006) and Valle et al. (2009b) .

In the experiment, a constant water ow rate was injected atthe bottom of the SG simulator from where it can ow throughthe test section to the RPV simulator. The gas was injected into theRPV simulator from the top and owed through the test section incounter-current to the water owto the SG separator. The increaseof the water level in the RPV simulator was used to determine thewater ow rate streaming over the test section (discharge ow).

The onset of ooding was dened as the limiting point of stability

of the counter-current ow, indicated by the maximum air massow rate at which the down-owing water mass ow rate is equalto the inlet water mass ow rate.

3. Computational modelling

In the present simulation, the ow was treated as transient.The problem is three dimensional (3-D), consequently, it has tobe solved by applying computational uid dynamics (CFD) meth-ods. Such multiphase codes resolve the conservation equations for

Fig. 3. Hot leg model at the TOPFLOW test facility.

Fig. 4. Air velocity near the free surface ( Hhne and Valle, 2010 ).

7/29/2019 deen_ned

4/11


Fig. 5. Calculation model.

mass, momentum and energy and they are distinguished by thedifferent approaches and strategies used in describing the phys-ical closure models. For the solution of the described task, anEulerEuler approach was used.

The EulerEuler approach assumes that at least two uids arecontinuously penetrating each other. The volume fraction of theuids in each cell sums to unity. For each uid, the full set of conservation equations is solved. The mechanisms of interactionbetween the uids are the momentum transfer modelled by theow resistance, the mass transfer modelled by phase change andthe energy transfer modelled by heat conduction. Both phases areassumed to be adiabatic and incompressible; therefore, the twolatter interactions in the present problem are not relevant in thepresent problem.

In the present simulation, we solve the conservation andmomentum equations of the two-uid model, which have the fol-lowing form

( k k)t

+ ( k kU k) = 0 (1)

( k k)t

+ ( k kU kU k) = k pk + k k g + k( v + t k)

+ D,k (2)

where the subscript k denotes phase gas or liquid, is the density,u is the velocity vector, t is the time, p is the pressure, g is thegravitational acceleration, is the volume fraction, is the shearstress ( v is the average viscous shear stress and t is the turbulentshear stress) and D is the interfacial shear stress.

The drag force is derived from the interfacial shear stress(F D = D A), is most conveniently expressed in terms of the dragcoefcient C D

F D = C D A LG| (U L U G)| 2 (3)

where LG is the average density, |( U L U G)| is the relative velocityand Ais the projected area of the body in ow direction (interfacialarea density). In the present work, the capabilities of the C D in AIADmodel and will be examined.

3.1. Free surface and AIAD model

In the previous experimental paper ( Deendarlianto et al., 2008 )it was reported that there are three morphologies at CCFL condi-tion. Those are bubble ow, stratied ow with a free surface andentrainment liquid droplet. Hhne and Valle (2010) noted thatthe CFD simulation of the free surface can be performed by usingthe multi-uid EulerEuler modellingapproachavailable in ANSYSCFX. However it requires a careful treatment of several aspects of

Table 1Calculation runs.

Exp. run Drag coefcient mL [kg/s] Range of mG [kg/s] Pressure [MPa]

30-09 (air/water)C D = 0.44 0.3 0.1830.274 0.15AIAD 0.3 0.1830.274 0.15

11-01 (steam/water) AIAD 0.3 0.4900.669 1.50

7/29/2019 deen_ned

5/11


Fig. 6. Injected gas mass ow rate as a function of time according to experimentalinjected gas ow rates proles.

the model. Those are interfacial area density, turbulence modelnear free surface and inter-phase momentum models. The sepa-ratemodels arealsonecessary fordispersedparticles andseparatedcontinuous phases (interfacial drag, etc.).

The suitable methodology within the EulerEuler approach isto use the momentum exchange coefcient depends on the localmorphology.For thatreason, Yegorov(2004) proposed an AlgebraicInterfacial Area Density (AIAD) model to solve the above problem.The basic conceptions of the proposed model are:

- The interfacial area density allows the detection of the morpho-logical formand the correspondingswitching for eachcorrelationfrom one object pair to another.

- It provides a law for interfacial area density and the drag coef-cient for full range of 0 L 1.

- The interfacial area density in the intermediate range is set to theinterfacial area density for the free surface.

The interfacial area density A also depends on the morphologyof the phases. For bubbles, the interfacial area density is dened asfollows

AB =6 GdB

(4)

Here dB is the bubble diameter and G is the gas void fraction.For a free surface an important requirement for the model is

the normalizing condition: the volume integral of the area densitymust be equal to the real surface area. It means that integration of the area density along a normal to the surface must yield unity:

+

A dn = 1 (5)

For a free surface, the interfacial area density is dened as absolutevalue of the gradients of the liquid fraction in x, y and z directions,and is written as

AFS = | L| = L

x

2

+ Ly

2

+ Lz

2

(6)

Next, the average density LG is dened as

LG = G G + L(1 G) (7)

where L and G are the liquid andgasdensities, respectively. In thebubble regime, where the G is low, the average density accordingto Eq. (7) is close to liquid phase density L. According to the owregime (bubbly ow, droplet ow or stratied ow with a free sur-face), thecorrespondingdrag coefcientsandarea densitieshave tobe applied. This problem can be solved by introducing a blendingfunction f . Introducing void fraction limits, the blending function

Fig. 7. Calculated liquid level in hot leg channel using C D = 0.44 (airwater, mL = 0 .30kg/s and mG = 0 .25 kg/s).

Fig. 8. Interfacial behaviour of airwater in hot leg channel during ooding (airwater, mL = 0 .30 kg/s and mG = 0 .25 kg/s).

7/29/2019 deen_ned

6/11


Region 2

Region 3

0

0.02

0.04

0.06

0.08

100806040200time [s]

W a t e r l e v e l r i s e [ m ]

0.1

0.15

0.2

0.25

0.3

CFD

Experiment A i r m a s s f l o w r a t e [ k g / s ]

Onset of flooding

Water level rise in RPV

Zero liquidpenetration

[Air-water]

(a) Air-water ( HZDR experimental run of 30-09)

0

0.02

0.04

0.06

0.08

50403020100

time [s]

W a t e r

l e v e

l r i s e

[ m ]

0.45

0.475

0.5

0.525

0.55

CFDExperiment

Water level rise in RPV

S t e a m m a s s f l o w r a t e [ k g / s ]Onset of flooding

[Steam-water] Zero liquidpenetration

(b) Steam-water ( HZDR experimental run of 11-01)

Region 1

Region 1 Region 2 Region 3

Fig. 9. Evolution of the water level in the RPV simulator in function of the gas owrate ( mL = 0 .30 kg/s).

and length scales for bubbly and droplet regimes are respectivelydened as

f B = 11 + e AB( G B,limit )(8)

f D =1

1 + e AD( G D,limit )(9)

Next, the blending function for the free surface is dened as

f FS = 1 f B f D (10)

Then, the area density and thedragcoefcient arerespectively welldened in the domain by

A = f FS AFS + f B AB + f D AD (11)

C D = f FSC D,FS + f BC D,B + f DC D,D (12)

After a validation study for this work the void fraction limits of B,limit = 0.3 respectively D,limit = 0.7 and blending coefcients of B = D = 70 were used.

In simulation of free surface ows, Eq. (3) does not represent arealisticphysical model. It is reasonable to expect thatthe velocitiesof both uids in the vicinity of the interface are rather similar. Toachieve this result, it is assumed that the shear stress near the sur-face behaves like a wall shear stress on both sides of the interfacein order to reduce the velocity differences of both phases.

Hhne and Valle (2010) derived a drag coefcient in the AIADmodel for the free surface application. In their proposal, a shearstress like a wall shear stress is assumed near thesurface from bothsides to reduce the velocity differences of both phases as shown inFig.4 . Here, a viscous uid movingalong a solid like boundarywillincur a shear stress, the no-slip condition, the morphology regionfree surface is the boundary layer, the shear stress is impartedonto the boundary as a result of this loss of velocity.

W,i = iu iy

y= 0(13)

Finally, thedrag coefcient of thefree surface can be obtained fromthe substitution of the Eqs. (3), (6) and (13) and is locally depen-dent on the fraction of phases, liquid density, and the slip velocitybetween the phases:

C D =2( L L + G G)

L| (U L U G)| 2(14)

where wall shear stresses of the gas and liquid L and G onto the

free surface are a function of the viscosity of both phases, the areaof free surface and the gradient of void fraction in x, y, z axes.

In the simulation, the drag coefcient of the bubble, a constantvalue of C D,B = 0.44 is taken, based on the drag of rigid spheresat the medium to high Reynolds number regime. For the dragcoefcient of the droplet, the C D,D =0.44 is also taken. On theother hand, the drag coefcient of the free surface, C D,FS, refers toEq. (14) .

Fig. 10. Flow structure of the counter-current gas/liquid two-phase ow near the elbow before ooding (airwater (30-09), mG

= 0 .181 kg/s, and t =5s).

7/29/2019 deen_ned

7/11


Fig. 11. Flow structure of the counter-current gas/liquid two-phase ow near the elbow around ooding ( mG = 0 .268 kg/s) of airwater test (HZDR experimental runningof 30-09); (a) calculated water volume fraction and (b) visual observation obtained from experiment.

3.2. Computational setup and boundary conditions

Numerical errors in CFD simulation are the result of a com-bination of many aspects, which includes the mesh resolution,discretisation method, time step size and convergence errors. Val-idation of the computational models used must be made usingspecic or known conditions basedon gathered experimental data.However, the separation of different numerical effects is difcult;for example, numerical diffusion in the ow direction can smudgeor smooth the estimated values downstream of the perturbation.The effect is similar to the selection of too large a turbulent viscos-ity for some turbulence models. In CFD analysis, demonstration of grid independence is also a basic requirement as indicated in theERCOFTAC Best Practice Guidelines 2001( Menter, 2002 ). However,due to insufcient computer resources, in real technical problemsmostly it cannot be achieved ( Farkas and Tth, 2010 ).

In the present work, the phenomenon of gas/liquid counter-current two-phase ow in model of a PWR hot leg was approachedby using the EulerEuler inhomogeneous model available in thecommercial CFD code of ANSYS CFX 12.0. In our simulations, very

carefully developed structured mesh for most of the ow eld wasadequate, at which the local renement on them was carried out.As a result, the ow domain was modelled with a structured meshconsisting of 248,610 hexahedral elements and 281,076 nodes asshown in Fig. 5.

The following parameters were used in the simulations. Bothphases have treated as isothermal and incompressible. Buoyancyeffects between the two-phase were taken into account by thedirection of gravity term. The turbulence properties at the inletof air and water were set using the turbulence intensity of 5% inboth phases. The air outlet was modelled with an opening bound-ary condition. The inner surface of the channel walls has beendened as hydraulically smooth with a non-slip boundary condi-tion applied to both gas and liquid phases. A time step 10 4 s anda maximum of 15 coefcient loop were taken to model the ow. Aconvergence in terms of the RMS values of the residuals to be lessthan 10 4 could be assured most of the time.

The SST buoyancy turbulence model and upwind advectionscheme were used in the simulation. The SST model works bysolving a turbulence/frequency-based model ( k ) at the wall

7/29/2019 deen_ned

8/11


Fig. 12. Flow structure of the counter-current gas/liquid two-phase ow near the elbow around ooding ( mG = 0 .530 kg/s) of steamwater ( HZDR experimental run of 11-01); (a) calculated water volume fraction and (b) visual observation obtained from experiment.

and standard k in the bulk ow. A blending function ensures asmooth transition between those two models ( Menter, 1993 ). Inthe calculations, newly developed C D in AIAD model (Eq. (14) ) wasimplemented into ANSYS CFX12.0 via the command language CEL

(CFX expression language).Three (3) calculations of two HZDR experimental runs (30-09and 11-01), and they are summarized in Table 1 have been per-formed. In the calculation, the injected gas ow rates were variedintimeaccordingto the measured gas ow rates as shown in Fig.6 .The calculations were performed in parallel of 4 processors of aLinux cluster (operating system: Linux Scientic 64 bit, 32 AMDOpteron Computer Nodes, node conguration: 2 AMD Opteron285(2.6 GHz,dual-core),16 GB memory).Typicalcomputationtimefor each case was about 4 months.

4. Results and discussions

A rst transient simulation was carried out by using a constant

of C D = 0.44 for the experimental running of 30-09. This value is

applicable to the drag coefcient of a spherical particle in a liquidunder turbulent regime. It was proposed by Wallis (1969) and isalso a default value from ANSYS CFX. Theresults areshown in Fig.7 .In the gure, the calculated ow structures of the counter-current

airwater two-phase ownear the elbow at high airmass owrate(mG = 0.25kg/s) i.e. during ooding are shown. From the transientow simulation the predicted ow was characterized by the thinliquid lm in the bend region. A hydraulic jump as the transitionfrom supercritical to subcritical ow is observed near the bendedregion, which is in contradiction to the experimental observationshown in Fig.8. In fact, the formationof liquidslugs and the entrain-ment of liquid droplets during ooding as observed in experimentcould not be reproduced here. This result indicates that interfacialfriction seems to be important at the wave crest during ooding,whereas C D = 0.44 is not big enough to promote a liquid slug andbreak it in the hot leg channel.

For the above reason the newly developed drag coefcient inAIAD model as described in Section 3.1 was used to simulate the

CCFL in a model of PWR hot leg in the following. In contrast to

7/29/2019 deen_ned

9/11


the rst simulation with constant C D, the onset of CCFL can besimulated by using this model. Fig. 9 shows the calculated andexperimental results of the average water level rise in the RPVsimulator in function of time. In the gures, those are plottedagainst the corresponding experimental data of the injected gasmass ow rate. In the gure, (a) and (b) correspond to the casesof airwater and steamwater, respectively. The calculated waterlevel rise and injected gas mass ow rate are shown by the blueand pink curves, respectively. The points correspond to measuredvalues of the water level. From the gures, the phenomena can beexplained as follows.

1. In both cases, the water level rise is similar and can be dividedinto three regions. In the rst region, the water level increaseswith a constant slope independently of the injected gas massow rate. In the second region, the slope of the curve of waterlevel in the RPVsimulator beginsto decrease.It means that a partof thewater injectedin the SG separatordoes notow to the sideof the RPV simulator anymore. This point is known as the onsetof ooding (CCFL). With further increase of the gas mass owrate, the calculated water level rise shows a plateau (region 3).This means that none of the injected water in the SG separatorows to the RPV simulator side.

2. The characteristics of the water level rise described above seemto conrm the experimental observations of Deendarlianto et al.(2008) . They dened the region 1, region 2, and region 3 as thestable counter-current ow, partial delivery region and zero liq-uid penetration respectively. Next, it is also noticed that thecalculated results are generally in agreement with the exper-imental data as shown in Fig. 9(a) and (b). However a minordifference of the water level between the experiment and thecalculation of the steamwater case in the regions of the partialdelivery and zero liquid penetration was found. In the experi-ment, the water level rise in the region 3 tends to increase withthe time, meanwhile it is almost constant in the calculation. Thedifference is believed to be due to the turbulence behaviour atthe interface of steam and water, such as turbulence dampingeffect, was not accounted in the calculation. Therefore it shouldbe considered in the future.

Fig. 10 illustrates the ow structure of the counter-currentgas/liquid two-phase ow near the elbow at low gas mass owrate ( mG = 0 .181 kg/s) i.e. before the onset of ooding for theairwater case. Here new drag coefcient in AIAD model was alsoapplied. In the gure, (a) and (b) correspond to the calculatedwater volume fraction and the visual observation obtained fromthe experiment respectively. From Fig. 10(a), a thin liquid lm wasfound in the bend region, and a hydraulic jump is clearly visiblenear the bended region, which agrees well with the experimentalobservation shown in Fig. 10(b).

Figs. 11 and 12 illustrate the ow structure of the counter-

current gas/liquid two-phase ow near the elbow at high gasmass ow rate i.e. ooding condition for the airwater and thesteamwater cases respectively. In the gure, (a) and (b) corre-spond respectively to the time variation of the calculated watervolume fraction ( t = 0.1s) and the visual observation obtainedfrom the experiment. From the gures, it is clearly shown that abigger wave is generated by the merging of small waves due to theinterfacial drag. The entrainment of liquid droplet observed in theexperiment can also be reproduced in the simulation. Here a highmigration of the crests of the waves toward the steam generator inwhich the down-owing liquid lmis disturbedplays an importantrole on the backwards liquid transport during ooding.

A comparison of the ooding curves between the CFD cal-culation and experiment is shown in Fig. 13 . For a meaningful

comparison, the non-dimensional supercial velocity J k , named as

Fig. 13. CCFL characteristics.

Wallis parameter is used. Here the Wallis parameter in Fig. 13 isdened as

J k = J k 1 gH . kL G (15)where the subscript k indicates gas and liquid phases, J the super-cial velocity, and H the height of the channel (cf. Valle et al.,2011 ). Close inspection of Fig. 13 reveals that the calculated CCFL pass through the range of HZDR experimental data, indicating agood agreement between the calculation and experimental data. A

comparison between both simulation results indicates that the gasvelocities at zero liquid penetration (( J L)

1/ 2 = 0.0) of both cases isalmost the same. This means that there is a minor effect of the uidviscosities on the zero liquid penetration point, whereas it is notaccounted in the Wallis parameter. The same behaviour was alsofound by Nariai et al. (2010) experimentally. This behaviour canbe explained by the fact that at this point the net-ow in the owpath is equal to zero. Therefore, liquidviscosity whichmainly inu-ences the wall and interfacial frictions does not play an importantrole here. Next it noticed also that the difference on the oodingcurves between both cases increases with ( J L)

1/ 2 . It indicates thatthe liquid viscosity has an increasing effect with the liquid massow.

In order to make a quantitative comparison of the water level

inside the hot leg channel between the experiment and calculation,an interface capture method was developed (see Montoya et al.,2011 ). To capture the gasliquid interface in the camera frames, animage processing algorithm was developed. This technique allowsthe representation of the interface by a water level as a functionof locus in the channel x and the time t . Nevertheless, for a com-parison between the CFD calculation and experimental result, asurface similar to the interface in the camera pictures has beendened. Therefore an isosurfac e with a void fraction of 50% waschosen andthe coordinates of its intersectionwiththe verticalmid-plane was exported from ANSYS CFX. With this simplication, thethree-dimensional shape of the isosurface was not taken into theaccount.

The time-averaged water level prole for both of experiments

and CFD calculations is shown in Fig. 14 . In the gure, (a) and (b)

7/29/2019 deen_ned

10/11


Fig. 14. The comparison on the water level inside the hot leg channel between the calculations and experiments.

correspond to the cases of airwater and steamwater, respec-tively. The water level data are presented as a function of locationand of the supercial gas velocity. Qualitatively, Fig. 14 shows thatthe trend obtained for the simulation is similar to the measure-ment. In the experiment with airwater ( Fig. 14(a)), before theonset of ooding the average water level prole in the horizon-

tal part increases at a distance far from the inclined part. However,a detailed comparison shows a minor deviation between simula-tions and measurements. Possible reasons for the discrepancy arethethree dimensionaleffects found in thewater level measurementor the using inlet boundary conditions in the simulation. Thereforefuture works on those problems should be considered.

5. Summary

The new concept of the drag coefcient in the AIAD model wasimplemented to describe the whole phenomena of CCFL in a modelof a PWR hot leg by applying CFD methods. The presented resultsshowed a clear progress in the simulation of the relevant phe-nomena, in which only AIAD model allows the correct simulation

of CCFL physically. The developed approach will also enable the

answering of many practical questions relating to the CCFL phe-nomenon. Moreover, further improvement of the model should becarried out. Here the usage of the morphology detection algorithmshould also be possible also in vertical ow regimes. Therefore, itis necessary to include the modelling of non-drag forces (lift force,wall lubrication force, virtual mass force,etc.)in the AIAD model as

well as the available for polydispersed ows. Next the turbulencedamping procedures should include the existence of small surfaceinstabilities in the macroscopic model. In addition the numericalapproach of the AIAD model should be improved to further reducethe calculation time.

Acknowledgements

This work is carried out within the frame work of a currentresearch project funded by the German Federal Ministry of Eco-nomics and Technology, project number 150 1329. The authorswould like to thank also the TOPFLOW team for their work on thetest facility and the preparation of the experiments.

Dr. Deendarlianto is an Alexander von Humboldt Fellow in

the Institute of Safety Research, Helmholtz-Zentrum Dresden-

7/29/2019 deen_ned

11/11


Rossendorf e.V., Dresden, Germany. The present research is alsosupportedby theAlexander vonHumboldtFoundation in Germany.

References

Ardron, K.H., Baneerjee, S., 1986. Flooding in an elbow between a vertical and ahorizontal or near horizontal pipe; part II: theory. International Journal of Mul-tiphase Flow 12 (4), 543558.

Deendarlianto, Valle, C., Lucas, D., Beyer, M., Pietruske, H., Carl, H., 2008. Experi-mentalstudyon theair/water counter-current owlimitation in a model of thehotleg ofa pressurized waterreactor.Nuclear Engineering andDesign 238(12),33893402.

Deendarlianto,Hhne,T., Lucas,D., Vierow, K. Gasliquidcountercurrent two-phaseow in a PWR hot leg: a comprehensive research review. Nuclear Engineeringand Design, submitted for publication.

Farkas, T.,Tth, I.,2010.Fluent analysis ofa ROSA cold legstraticationtest.NuclearEngineering and Design 240 (9), 21692175.

Hhne, T., 2009. Experiments and numerical simulations of horizontal two-phaseow regimes. In: Proceeding of the Seventh International Conference on CFD inthe Minerals and Process Industries, Melbourne, Australia.

Hhne, T., Valle, C., 2010. Experimental and numerical simulations of horizontaltwo-phase ow regimes using an interfacial area density model. The Journal of Computational Multiphase Flow 2 (3), 131143.

Jeong, H.Y., 2002. Predictionof counter-currentow limitation at hot leg pipe duringa small break LOCA. Annals of Nuclear Energy 29 (5), 571583.

Kinoshita, I., Utanohara, Y., Murase, M., Minami, N., Tomiyama, A., 2009. Numericalcalculations oncountercurrent gasliquidow in a PWRhot leg(2) steamwaterow under PWR plant conditions. In: Proceeding of the 13th International Top-

ical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-13), KanazawaCity, Japan, September 2009.Lee, S.C., Bankoff, S.G., 1983. Stability of steamwater counter-current ow in an

inclined channel: ooding. Journal of Heat Transfer 105, 713718.Lopez de Bertodano, M., 1994. Counter-current gasliquid ow in a pressurized

water reactor hot leg. Nuclear Science and Engineering 117, 126133.Menter, F., 1993. Zonal two-equation k turbulence models for aerodynamic

ows. AIAA Journal 93, 2906.Menter, F., 2002. CFD best practice guidelines for CFD code validation for reactor

safety applications. ECORA FIKS-CT-2001-00154.Minami, N., Utanohara, Y., Kinoshita, I., Murase, M., Tomiyama, A., 2009. Numerical

calculations on countercurrent gas-liquid ow in a PWR hot leg (1) airwaterow in a 1/15-scale model. In: Proceeding of the 13th International TopicalMeeting on Nuclear Reactor Thermal Hydraulics (NURETH-13), Kanazawa City, Japan, September 2009.

Minami, N.,Nishiwaki, Nariai, T.D.,Tomiyama, A.,Murase,M., 2010a.Countercurrentgasliquid owin a PWRhotlegunderreux cooling (I)airwatertests for1/15-scale model of a PWR hot leg. Journal of Nuclear Science and Technology 47 (2),142148.

Minami, N., Murase, M., Tomiyama, A., 2010b. Countercurrent gasliquid ow ina PWR hot leg under reux cooling (II) numerical simulation of 1/15-scaleairwater tests. Journal of Nuclear Science and Technology 47 (2), 149155.

Montoya, G.A., Deendarlianto, Lucas, D., Hhne, T., Valle, C., 2011. Time depen-dent interfacial behavior during counter-current gasliquid two-phase ow ina modelof thehotleg ofpressurizedwaterreactor(PWR). In:19thInternationalConference on Nuclear Engineering (ICONE), Chiba, Japan, 1619 May 2011.

Murase, M., Utanohara, Y., Kinoshita, I., Minami, N., Tomiyama, A., 2009. Numericalcalculations on countercurrent airwater ow in small-scale models of a PWR hot leg using a VOF model. In: Proceeding of the 17th International Conferenceon Nuclear Engineering (ICONE 17), Brussels, Belgium, July 2009.

Nariai, T., Tomiyama, A., Vallee, C., Lucas, D., Murase, M., 2010. Countercurrent owlimitation in a scale-down model of a PWR hot leg. In: Proceeding of the 8thInternational Topical meeting on Nuclear Thermal-Hydraulics, Operation and

Safety (NUTHOS-8), Shanghai, China, October 2010.Ohnuki, A., 1986.Experimentalstudy of counter-current two-phaseow in horizon-

tal tube connected to inclined riser. Journal of Nuclear Science and Technology23 (3), 219232.

Ohnuki, A., Adachi, H., Murao, Y., 1988. Scale effects on countercurrent gasliquidow in a horizontaltube connectedto an inclined riser. Nuclear Engineering andDesign 107, 283294.

Prasser, H.M., Beyer, M., Carl, H., Manera, A., Pietruske, H., Schutz, H., Weiss,F.P., 2006. The multipurpose thermal hydraulic test facility TOPFLOW: anoverview on experimental capabilities, instrumentation and result. Kerntechnik71, 163173.

Utanohara, Y., Kinoshita, I., Murase, M., Minami, N., Tomiyama, A., 2009. Effectsof interfacial friction correlations on numerical calculations for countercurrentgasliquidow in a PWRhot leg.In: Proceedingof the13th InternationalTopicalMeeting on Nuclear Reactor Thermal Hydraulics (NURETH-13), Kanazawa City, Japan, September 2009.

Valle, C., Deendarlianto, Beyer, M., Lucas, D., Carl, H., 2009a. Air/water counter-currentow experiments in a model of thehot legof a pressurized waterreactor. Journal of Engineering for Gas Turbines and Power Transactions of the ASME131, 022905.

Valle, C.,Seidel, T.,Lucas, D.,Beyer, M.,Prasser,H.-M., Pietruske, H.,Schtz,P., Carl,H., 2009b. Inuence of the uid properties on co-current two-phase ows in ahorizontal channel connected to a riser. In: Proceedings of the 7th World Con-ference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics(ExHFT-7), Krakow, Poland, 28 June03 July 2009, pp. 443452.

Valle, C., Seidel, T., Lucas, D., Tomiyama, A., Murase, M., 2011. Comparison of counter-current ow limitation experiments performed in two different mod-els of the hot leg of a pressurised water reactor with rectangular cross-section. Journal of Engineering for Gas Turbines and Power Transactions of the ASME133 (5) (article 052917).

Wallis, G.B., 1969. One-dimensional Two-phase Flow. McGraw Hill, New York.Wang, M.J., Mayinger, F., 1995. Simulation and analysis of thermal-hydraulic phe-

nomena in a PWR hot leg related to SBLOCA. Nuclear Engineering and Design155, 643652.

Wongwises, S., 1996. Two-phase countercurrent ow in a model of a pressurizedwater reactor hot leg. Nuclear Engineering and Design 166 (2), 121133.

Yadigaroglu, G., 2005. Computational uid dynamics for nuclear applications: from

CFD to multi-scale CFMD. Nuclear Engineering and Design 235, 153164.Yegorov, Y., 2004. Contact condensation in stratied steamwater ow. EVOL-ECORA-D07.

deen_ned

Documents