A NEW ASSESSMENT OF THE LOFT-WYLE BLOWDOWN TEST … Documents... · the LOFT reactor vessel and broken loop cold leg. The pressure vessel was made from carbon steel, with a volume

2002 RELAP5 International Users SeminarPark City, UtahSeptember 4-6, 2002

n testcility

le ino thewing

downed inor thed. Theng thesuredn, 2)tureal fluid

ctural-flownon-liquidt-fluxfullythree-

s notalente Vea-l is

A NEW ASSESSMENT OF THE LOFT-WYLEBLOWDOWN TEST WSB03R USING RELAP5-3D

B. R. BandiniD.L. AumillerE.T. Tomlinson

Bechtel Bettis, Inc.Bettis Atomic Power Laboratory

P.O. Box 79West Mifflin, PA 15122-0079

Abstract

The RELAP5-3D (version bt03) computer program was used to assess the LOFT-Wyle blowdow(WSB03R). The primary goal of this new assessment is to represent faithfully the experimental faand instrumentation using the latest three-dimensional fluid flow modeling capability availabRELAP5-3D. In addition, since RELAP5-3D represents a relatively new and significant upgrade tcapabilities of the RELAP5 series of computer programs, this study serves to add to its groassessment base.

The LOFT-Wyle Transient Fluid Calibration test facility consisted of an approximately 5.4 m3 pressurevessel with a flow skirt which created an annulus that acted as a downcomer. An instrumented blowloop with an orifice was connected to the downcomer. This facility, built to calibrate the orifices usseveral of the LOFT experiments, simulated the LOFT reactor vessel and broken loop cold leg. Fpresent assessment an existing RELAP5 model developed at INEEL was corrected and upgrademodel corrections included: 1) employing the proper measured downcomer thickness, 2) positioniexperimental instrumentation in its correct location, and 3) setting the fluid conditions to their meainitial values. Model upgrades included: 1) use of more finely-detailed fluid component nodalizatioexplicit modeling of the experimental facility beyond the blowdown orifice, 3) addition of heat struccomponents to represent the heat capacity of structural material, and 4) use of three-dimensioncomponents to model asymmetric portions of the facility.

The new assessment highlights the need to model explicitly the effects of heat storage in strumaterials for slowly evolving transients. The assessment also highlights the sensitivity of chokedlimited calculations to: 1) the model employed, 2) input discharge coefficient values and/or 3) inputequilibrium values. In addition, the assessment demonstrates that an instability in the calculatedfraction at the base of the downcomer obtained using the standard RELAP5-3D Kataoka-Ishii drifcorrelation can be substantially mitigated through the use of the optional Gardner correlation in theone-dimensional model. Finally, the new assessment demonstrates the correct functioning of thedimensional fluid components. For this particular transient, three-dimensional modeling doesignificantly alter or improve agreement with the experimental data in comparison with an equivmodel consisting entirely of one-dimensional fluid components. This assessment shows that thLahey drift-flux correlation in conjunction with the modified LeVeque momentum flux-splitting moderequired to dampen liquid fraction oscillations at the vessel/downcomer interface in the 3-D model.

1


ndtog.

el,

.75ntmt

elylgeheselhek16df

yt

reehyey

mheewmm

hehesm

mK

Introduction

Many of the transients of interest to the thermal-hydraulic safety community (Loss of CoolantAccidents) are characterized by fastdepressurization due to the loss of liquid inventory.This depressurization causes flashing of the liquidas the pressure falls below the saturation pressurefor the fluid temperature. Accurate predictions ofthe time-dependent fluid inventory loss ratethrough a choked orifice in the presence of flashingand varying void distribution profiles in varioussystem components are important for thermal-hydraulic safety programs.

In 1979, Wyle Laboratories in Norco, Californiaconducted a series of experiments in the LOFTTransient Fluid Calibration Facility [1] to measurethe critical flow rate through LOFT break orificesof varying sizes (nozzles L3-1 and L3-2) duringdepressurization transients with varying initialconditions. The objectives of these tests were toobtain orifice calibration data at fluid conditionstypical of small break loss-of-coolant accidentsand to provide a data base for critical flow modeldevelopment. One of these tests (WSB03R,initiated at 14.7MPa (2148 psia) and 557K (542.9oF) with a 16 mm (0.6374 in.) break orifice) hasbecome a standard verification problem for theRELAP5 program [2]. The WSB03R assessmentis updated herein using RELAP5-3D [3]. Theresult is a model which faithfully represents the fulltest facility, provides improved accuracy relative tothe experimental data, and serves as an addition tothe growing assessment base for RELAP5-3D.

Description of the Test

The LOFT-Wyle critical flow experiments weredesigned to measure transient critical flow rate andfluid conditions (pressure, temperature, anddensity) upstream of the orifice within theblowdown line during a depressurization transient.Two different orifice sizes (4 mm and 16 mmnominal diameter) were used in the experimentalprogram. This paper will focus on test WSB03Rperformed with the larger of the two orifices withinitial fluid conditions of 14.7MPa and 557K.

A schematic of the LOFT-Wyle Transient FluidCalibration Facility is shown in Figure 1. Thefacility hardware consisted of a pressure vessel aa blowdown leg. These components are similiarthe LOFT reactor vessel and broken loop cold le

The pressure vessel was made from carbon stewith a volume of approximately 5.4 m3 (190 ft3).The pressure vessel contained a 0.01905 m (0in.) thick carbon steel flow skirt to create aannulus for the LOFT downcomer. The flow skirextended 0.8382 m (33 in.) above and 4.251(167.375 in.) below the centerline of the outleflange. The downcomer fluid was in fullcommunication with fluid in the vessel at thelower extent of the downcomer. Meanwhile thdowncomer/vessel flow path was completeblocked at the top of the downcomer (an axiaelevation corresponding to the vessel head flansurface). The blowdown leg was connected to tvessel outlet flange. This leg consisted of a vesoutlet nozzle, an instrumentation test section, tbreak orifice, a shutoff gate valve, a rupture disassembly, and a discharge pipe and tee. Themm (0.6374 in.) break orifice was axially centerewithin the test section which was constructed o0.3556 m (14 in.) Schedule 160 stainless pipe.

A measurement of the weight of the systemthroughout the blowdown was accomplished bfour load cells which supported the entire weighof the system. These precision transducers, thof which supported the vessel, and one of whicsupported the blowdown leg were the primarmeans for determining the system weight. Thmass flow rate through the orifice was indirectldetermined by electronically differentiating thetime-dependent weight of the system. This sytehad been tested and was determined from tengineering judgement of Wyle personnel to baccurate to within 0.5 kg/s of the actual mass florate at all times during the experiment. A six beagamma densitometer, comprised of two three-beadensitometers mounted on opposite sides of tpipe, was used the determine the density of texiting liquid/vapor mixture at three radial crossections at a position 0.87 m (34.26 in.) upstreaof the break orifice. The fluid temperature 0.87 mupstream of the break orifice and near the bottoof the vessel were measured by ISA Type

2


g4gctweaalltsdn

nte

ite

,snt.e

asere5-

ido

thelllyIneK

s

as.seof

thermocouples. The pressure 0.87 m upstream ofthe break orifice was measured by a pressuretransducer.

The initial conditions for LOFT-Wyle testWSB03R were a system completely filled withdemineralized water at pressure of 14.7MPa (2148psia), a fluid temperature of 557K (542.9oF) at thebottom of the vessel, and a fluid temperature of520K (476.3oF) measured 0.87 m upstream of thebreak orifice. Before initiating the blowdown, thesystem was allowed to ‘soak’ for three hours toequalize the temperature in the fluid and structuralmaterial. The blowdown through the 16 mmorifice was then initiated by venting a cavitybetween two 0.1524 m (6 in.) diameter rupturedisks connected to the downstream flange of theblowdown line shutoff gate valve. As a result of alack of detail in the description of this portion ofthe LOFT-Wyle facility, the rupture disks areestimated, from scaling of various facility sketchesprovided in Reference 1, to be 1.72 m (5.64 ft)downstream of the break orifice. The blowdownwhich begins with the venting and subsequentyielding of the rupture disks was simulated for1500 s.

Original Assessment Model

The input description for the original assessment isdescribed in Volume III of the RELAP5/MOD2Code Manual [2]. An electronic copy of thecorresponding input deck was obtained fromRELAP5-3D program developers, Idaho NationalEngineering and Environmental Laboratory(INEEL). In this model, the 5.4 m3 (190 ft3)pressure vessel, with downcomer in place wasrepresented using 10 one-dimensional (1-D)volumes, 7 above the bottom of the downcomerand 3 below. The downcomer was modeled using6 1-D volumes, 4 below the blowdown pipe, one atthe elevation of the pipe, and one representing theflow skirt extension above the blowdown pipe. Alljunctions at which an area change occurred weremodeled using the RELAP5 smooth area changeoption except for the two junctions connecting theupper portion of the vessel to the lower portion,and the lower portion of the vessel to thedowncomer. Here the RELAP5 abrupt area change

option was invoked. The blowdown pipe leadinup to the break orifice was modeled usinghorizontal 1-D volumes. The volumes representinthe blowdown pipe were connected to the correelevation of the downcomer using a cross flojunction. The 16 mm (0.6374 in.) diameter orificitself was modeled using the RELAP5 abrupt arechange option with an area equal to the actuorifice area. In the original model the defauRELAP5 Ransom-Trapp critical flow model waused [4] with unity subcooled, two-phase, ansuperheated break flow multipliers. The blowdowline beyond the break orifice was not explicitlymodeled. The attributes of the original assessmemodel are summarized in Table 1 under thheading Case 0.

Revised Assessment Model

In previous assessments of RELAP5-3D [5,6],was concluded that faithful representations of thexperimental facility including instrumentationthe boundary conditions and the initial conditionwere required to obtain an undistorted assessmeThis philosophy was used in the creation of threvised assessment model.

A slight error in the representation of thedowncomer thickness in a previous assessment wfound and corrected. As a result, the downcomflow area was reduced 1.6% relative to thdevelopmental assessment model. The RELAP3D fluid volume containing the instrumentationwhich records temperature, pressure, and fludensity upstream of the break orifice was alsadjusted such that its cell-center corresponds tolocation of the instruments. The initial vessepressure was set to the correct experimentameasured value of 14.7MPa (2148 psia).addition, the initial sytem fluid temperatures werset to the experimentally measured values of 557(542.9oF) in the vessel and 520K (476.3oF)upstream of the blowdown orifice.

The revised assessment model which initially waconstructed from one-dimensional fluidcomponents was also revised in several other areFirst, the experimental facility fluid componentdownstream of the 16 mm break orifice wermodeled explicitly. These include the remainder

3


ehee-ricndtheetem

lye,e,ein

ater,een

e.d

eefatalse

tualadt

ein

thd

.ete1e 2.inss

blowdown spool piece #2, the shutoff gate valve,the rupture disk assembly, and the discharge pipingfrom the rupture disks to the ‘tee’ outlet to theatmosphere. Secondly, the heat capacity andthermal conductivity associated with the structuralmaterial of the vessel downcomer and blowdownloop piping was modeled explicitly through the useof passive heat structures. In addition, in order toobtain better steady-state initial conditions for thetransient blowdown calculation, the model wasinitialized for 10 seconds with the rupture disksintact.

In the new 1-D model the pressure vessel, withdowncomer in place, was represented using 16volumes, 14 above the bottom of the downcomerand 2 below. The downcomer was modeled using11 volumes, 8 below the blowdown pipe, one at theelevation of the pipe, and two representing the flowskirt extension above the blowdown pipe. Alljunctions within the vessel and the downcomerregions were modeled as smooth. Form losscoefficients of zero were employed at all junctionsexcept for those joining the lower vessel head tothe downcomer and to the upper vessel regions. Atthese locations small form loss coefficients of 0.1were employed, each representing one-half ofthose associated with a 180 degree piping bend.The blowdown pipe leading up to the break orifice(which includes spool piece #1 and the upstreamportion of instrumented spool piece #2) wasmodeled using 6 horizontal volumes connected toone another with smooth junctions. The volumesrepresenting the blowdown pipe were connected tothe correct elevation of the downcomer using aside exiting junction with non-zero form losscoefficients calculated for the existing physicalgeometry (1.13 in the forward direction and 1.0 inreverse). The 16 mm (0.6374 in.) diameter orificeitself was modeled as a single abrupt junction withan area equal to the actual orifice area.

Volumetric fluid components downstream of theorifice included: 1) a single volume representingthe first (smaller diameter) downstream portion ofspool piece #2, 2) two volumes representing thefinal (larger diameter) downstream portion of spoolpiece #2, 3) two volumes to separately representthe inlet and outlet sections of the shutoff gatevalve, 4) 8 volumes to represent the ~ 2.4 m long

straight run of downstream piping, 5) a singlvolume which represents the ‘tee’ connected to tend of the downstream piping run, and 6) a timdependent volume to represent atmospheconditions beyond the ‘tee’. All junctions betweevolumes downstream of the orifice were modeleas smooth, except for an abrupt area change atrupture disk location (between the volumrepresenting the downstream portion of the gavalve and the first volume representing the ~ 2.4long straight run of downstream piping). Inaddition, non-zero form loss coefficients were onapplied to the junction representing the gate valvthe junction representing exit to the atmospherand the junctions at which an area changoccurred. Passive heat structures were includedthe revised model in order account for the hestorage capacity of the system vessel, downcomand blowdown piping run. Heat structures wernot included in the piping downstream of thrupture disks because their effect on the evolutioof the transient was considered to be negligiblThe majority of the heat structures were modeleas cylindrical regions with best-estimatdimensions and compositions. In contrast, thpartially hemispherical upper and lower regions othe vessel structure were modeled as two hestructure components, a slab and a cylindricregion. The thickness and surface area of theheat structures were adjusted to preserve the acsurface area and volume of the true vessel hestructures. A schematic of the fluid componennodalization used in the revised 1-D LOFT-WylWSB03R test assessment model is presentedFigure 2.

Comparison of Model Results

In the initial revised assessment calculations bothe default RELAP5 Ransom-Trapp [4] anoptional Henry-Fauske [7] critical flow modelswere employed with unity break flow multipliersIn addition, the non-equilibrium parameter for thHenry-Fauske critical flow model was retained aits default value of 0.14. The attributes of thesinitial revised models are summarized in Tableunder the respective headings, Case 1 and CasAs a consequence of high vapor velocitiesexplicitly-modeled test section componentdownstream of the orifice, Courant limitation

4


rly0y

erntethhet

atlye

g

00ofeehehedstehl,es.wwel

e

eld

ict

eredcalntl,at

caused the calculational time steps in the revisedmodel to always be small. Thus, the calculationaltime step sizes were determined by the Courantlimit and ranged from an initial value of 0.00156 sat the beginning of the transient graduallyincreasing to a value of 0.0125 s at the conclusionof the transient.

Figures 3 through 6 compare three important fluidparameters obtained from these calculations(designated as Case 1 and Case 2) with thoseobtained experimentally and with those calculatedwith the original INEEL assessment model(designated as Case 0). These parameters include:1) system pressure upstream of the break orifice, 2)break flowrate, 3) integrated break orifice massflow, and 4) fluid density upstream of the breakorifice. The experimental results for three of thefour parameters include error bars. The error barson the measured mass flow rate represent WyleLaboratory’s engineering judgement, while theerror bars on pressure and density represent thestated uncertainty in the measurement equipment.The experimental value for integrated break orificemass flow does not include error bars since this isan indirectly derived quantity.

Of the parameters being compared, the transientpressure measurement upstream of the orifice wasjudged to be the measured parameter with the leastuncertainty. Figure 3 shows that the originalassessment model overpredicts upstream pressurefor the first ~950 seconds before falling in line withmeasurement at later times. The revised 1-Dassessment model using the default Ransom-Trappcritical flow model slightly underpredicts pressurefor the first 200 seconds of the transient whileoverpredicting pressure, sometimes significantly,from that time onward. This alternate prediction ofpressure versus that obtained in the original modelis mainly the result of the addition of heatstructures, which tend to keep pressure higher for alonger period of time by releasing stored heatenergy into the fluid later in the transient. Theinclusion of heat structures, which more faithfullyrepresent the LOFT-Wyle experimental facility,bring to light fortuitous compensating errors in thetime-dependent pressure calculated by the originalassessment model. Meanwhile, the revisedassessment model with the optional Henry-Fauske

critical flow model also underpredicts pressure fothe first ~250 seconds while subsequentoverpredicting pressure from ~250 to 100seconds, beyond which time satisfactoragreement is acheived.

Figure 4 depicts the most significant parametcalculated by this experiment, the time-dependerate of mass flow from the system. Here thoriginal assessment model compares well wiexperimental data out to about 100 seconds, thigh mass flow rate portion of the transient awhich the fluid upstream of the orifice exists insubcooled state. This good agreement is mosfortuitous. This was proven to be a result of thincorrect initial fluid conditions in the originalmodel through a sensitivity calculation run usinthe correct initial conditions in the originalassessment model. From 100 seconds to ~4seconds, the time period during which the levelthe stratified two phase fluid upstream of the orificis believed to remain above the centerline orificposition, the initial assessment underpredicts tmass flow rate. Beyond 400 seconds, where tlevel of the fluid upstream of the orifice is believeto fall below the orifice, the original model predictthe experimentally-observed drop in mass flow rareasonably well. The revised 1-D model whicuses the default Ransom-Trapp critical flow modeunderpredicts the mass flow rate from thbeginning of the transient out to ~400 secondBeyond this time the model overpredicts mass florate, especially the timing of the sharp drop in florate as the two-phase mixture drops below the levof the orifice. In the LOFT-Wyle experiment thebreak orifice is centered in the test spool piecoriented in the primary direction of fluid flow. Insuch case, if the horizontally-stratified water levis above or below the break orifice, the donerenumerical scheme may underpredict or overpredthe junction void fraction. The horizontalstratification entrainment off-take model inRELAP5-3D [3] cannot be applied in this cassince the orifice is not in an upward, downward, oside-centered orientation with respect to thupstream volume. Meanwhile, the reviseassessment model using the Henry-Fauske critiflow model generally predicts the time-dependemass flow rate from the system rather weldeviating from the experimental data somewhat

5


-e

ct

s-alell

yis

atly00,istalofly

eere

-erntgetoerse

af

yer

wor’ota,ae

times less than 100 seconds and between 400 and500 seconds.

Figure 5 depicts the time-integrated mass flowfrom the system. The curves shown in this figureare merely the integrals over time of the quantitiesshown in Figure 4. Here the original assessmentmodel compares rather well with experimental dataexcept for a slight underprediction between ~100and ~700 seconds, which corresponds to flow rateunderprediction from ~100 to ~400 seconds and acompensating overprediction from ~400 to ~700seconds. Meanwhile, the revised 1-D model whichuses the default Ransom-Trapp critical flow model,significantly underpredicts integrated mass flowfrom the beginning of the transient out to ~700seconds. Beyond this time the model overpredictsintegrated mass flow, eventually exhausting a totalof ~100 kg (220 lb) more fluid from the system by1500 seconds than was observed experimentally.Finally, the revised assessment model using theHenry-Fauske critical flow model slightlyunderpredicts integrated mass flow from thebeginning of the transient out to ~500 seconds.Beyond this time this model also overpredictsintegrated mass flow, again exhausting a total of~100 kg (220 lb) more fluid from the system by1500 seconds than was observed experimentally.

The amount of fluid which was exhausted in thesimulations is determined by the entrainment ofliquid in the lower head of the vessel. The impactof interfacial drag on this phenomena will bediscussed in a later section of this paper.

Figure 6 depicts the time-dependent average fluiddensity upstream of the orifice. Here the originalassessment model mispredicts the timing of thelarge drop in fluid density which occurs at ~100seconds, the subcooled to saturated transitionpoint. This is again the result of incorrect initialfluid conditions. At ~500 seconds the originalmodel does, however, predict the sharp drop inaverage fluid density due to the changing of fluidconditions upstream of the orifice from two-phasefluid to vapor rather well. The revised 1-D modelwhich uses the default Ransom-Trapp critical flowmodel does a much better job of predicting theinitial drop in fluid density at the subcooled tosaturated transition point. However, the prediction

of the drop in average fluid density during the twophase to vapor transition occurs later than in thoriginal assessment model, highlighting the fathat, in this case, more faithful facility modeling(i.e. inclusion of heat structures) uncoverfortuitous compensating errors in the timedependent fluid density calculated by the originassessment model. Finally, the revised 1-D modwhich uses the Henry-Fauske critical flow modeaccurately predicts the initial drop in fluid densitat the subcooled to saturated transition point. Thmodel also predicts the timing of the drop in fluiddensity at ~500 seconds rather well. The fact ththis model does not predict the experimentalmeasured density plateau between ~100 and ~4seconds is most likely not a liability in the modelbut is instead related to the manner in which thdata was obtained. The reported experimendensity is the average value from the three setsdensitometers. Since each densitometer onviews a single chord cutting across thinstrumentation spool, any given densitometreading will exhibit a sharp density change as thtwo-phase mixture level falls through its narrowlyfocused field of view. Since three densitometreadings (which pass through chords at differeelevations) are combined to determine the averaupstream mixture density, this average will tendexhibit a sharply defined density change, raththan a gradual change as the level of the two-phahorizontally-stratified mixture level falls below theelevation of the highest densitometer beam.

Sensitivity to Choking ModelParameters

The orifice used in LOFT-Wyle test WSB03R isunique design, with a length to diameter ratio o~3.4. This orifice design does not directlcorrespond to any that was used to develop eiththe Ransom-Trapp or Henry-Fauske critical flomodels. As such, input discharge coefficients fboth models, as well as the ‘non-equilibiumcoefficient in the Henry-Fauske were ‘tuned’ tobtain better agreement with experimental daprimarily orifice mass flow or blowdown rate asfunction of time. Figures 7 through 10 detail thbest results obtained from this ‘tuning’ of thecritical flow correlations in the break orifice of the

6


ntl,

sseDwtss

elw,dasd

f0gll.otaly.

idDwtdyelll

y

heeenyntr

Dt’-

revised assessment model. The attributes of theserevised models with adjusted critical-flowparameters are summarized in Table 1 under therespective headings, Case 3 and Case 4.

Figure 7 shows that the revised 1-D assessmentmodel using the Ransom-Trapp critical flow modelwith non-physical subcooled, two-phase, andsuperheated critical flow discharge coefficients of1.2, still slightly underpredicts pressure for the first200 seconds of the transient. However, from 200to 800 seconds, while pressure is stilloverpredicted, the agreement with experiment ismuch improved over that obtained with unityRansom-Trapp discharge coefficients in either theoriginal or the revised LOFT-Wyle RELAP5models. The use of critical flow dischargecoefficients greater than 1.0 represents a non-physical situation, indicating that the Ransom-Trapp model is being used outside its range ofapplicability. This situation may be a commonoccurance, as alluded to in the section of theRELAP5-3D User’s Guidelines which providesrecommendations for break modeling [3].Meanwhile, the revised assessment model was runusing the Henry-Fauske critical flow model with acritial flow discharge coefficient of 0.84 and a non-equilibrium parameter of 1000, indicating a‘frozen’ model, or one in which the liquid isassumed not to vaporize. This prevents the liquidand vapor temperatures from being the same (i.e.non-equilibrium) through the orifice. This modelyields the best overall prediction of theexperimentally-measured pressure upstream of theorifice throughout the transient.

Figure 8 again depicts the time-dependent ratemass flow from the system. The revised 1-D modelwhich uses the Ransom-Trapp critical flow modelwith critical flow discharge coefficients set to anon-physical value of 1.2 predicts the mass flowfrom the beginning of the transient out to ~400seconds very well. Beyond this time the modelstill somewhat overpredicts mass flow rate. But,the time at which the mass flow falls sharply as thetwo-phase mixture drops below the level of theorifice is now much better predicted. Meanwhile,the revised assessment model using the Henry-Fauske critical flow model with a critial flowdischarge coefficient of 0.84 and a non-equilibrium

parameter of 1000, predicts the time-dependemass flow rate from the system very welincluding the timing of the drop in mass flow rateat ~400 seconds.

Figure 9 again depicts the time-integrated maflow from the system (the time integration of thquantities shown in Figure 8). The revised 1-model which uses the Ransom-Trapp critical flomodel with critical flow discharge coefficients seto 1.2, somewhat underpredicts integrated maflow from the beginning of the transient out to~550 seconds. Beyond this time the modcontinues to overpredict integrated mass floexhausting a total of ~120 kg (264 lb) more fluifrom the system by 1500 seconds than wobserved experimentally. Meanwhile, the revise1-D model using the Henry-Fauske critical flowmodel with a critial flow discharge coefficient o0.84 and a non-equilibrium parameter of 100predicts integrated mass flow from the beginninof the transient out to ~450 seconds very weBeyond this time this model also continues toverpredict integrated mass flow, exhausting a toof ~100 kg (220 lb) more fluid from the system b1500 seconds than was observed experimentally

Figure 10 depicts the time-dependent average fludensity upstream of the orifice. The revised 1-model which uses the Ransom-Trapp critical flomodel with critical flow discharge coefficients seto 1.2 does a good job of predicting the timing anmagnitude of the drop in average fluid densitwhen the two-phase mixture drops below the levof the orifice. Finally, the revised 1-D modewhich uses the Henry-Fauske critical flow modewith a critial flow discharge coefficient of 0.84 anda non-equilibrium parameter of 1000, veraccurately predicts the timing of the drop in fluiddensity at ~500 seconds. As stated previously, tfact that this model does not predict thexperimentally measured density plateau betwe~100 and ~400 seconds is most likely not a liabilitin the model, but an experimental measuremeanomaly resulting from the use of a limited numbeof densitometers with a very narrow field of view.

Based upon the above observations from 1-RELAP5-3D models it appears that the ‘beschoice of orifice choking parameters for LOFT

7


t-)

sy-te

n,

sln

s

gtheenss

.2ss

-D

ef0d

eese

-eroe

drinrto

Wyle blowdown test WSB03R are the Henry-Fauske correlation with a critial flow dischargecoefficient of 0.84 and a non-equilibriumparameter of 1000 (e.g. Case 4).

One common observation can be made uponreviewing the revised 1-D RELAP5-3D resultsobtained with various choked-flow models (bothwith and without optimized discharge/non-equilibrium coefficients). That is, all revisedcalculations tend to overpredict the total fluid masslost from the system over the course of thetransient by ~100 kg. Based upon the results ofprevious studies [5] it is believed that thismisprediction may be a direct consequence of alimitation in the ability of RELAP5-3D toaccurately predict liquid droplet carryover involumes with relatively high void fractions. Thisdeficiency in calculating the magnitude of dropletcarryover may be traced to inaccuracies in theprediction of drag between the liquid and vaporphases. RELAP5-3D uses the drift-flux velocity todetermine the interfacial drag. The drift-fluxvelocity in turn depends on the flow regime. Anindication of RELAP5-3D having difficultypredicting drift-flux velocity and thus interfacialdrag in the LOFT-Wyle blowdown experiment maybe obtained by observing Figure 11, the time-dependent liquid fraction at the junction betweenthe lower vessel and the downcomer forcalculational Case 4. Here large high-frequencyoscillations are observed in the liquid fraction atthe entrance to the downcomer. This is a directconsequence of the interdependencies inherent inthe default Kataoka-Ishii drift-flux correlation [8]employed in RELAP5-3D. In this model the drift-flux velocity is dependent upon the flow regimewhich in turn is dependent upon the void fraction.Thus oscillations in the flow regime produceoscillations in the drift-flux velocity and interfacialdrag which in turn cause the void fraction tooscillate, each inter-dependent component drivingthe next.

Interfacial Drag Study

In order to reduce these oscillations and, ifpossible, obtain a better prediction of the liquidcarryover and thus the total fluid mass loss from

the system during the transient, two optional drifflux correlations were tested. These include: Athe Gardner correlation [9] (implemented withCard 1 option 82) with modified bubbly and slugflow regime interfacial heat transfer coefficient(Card 1 option 61) and B) the Vea-Lahecorrelation [10]. The use of either of these driftflux correlations requires the use of an alternaformulation for the drift-flux distribution parameter(Card 1 option 78). As discussed in detail iReference 5, the Gardner drift-flux correlationappropriate for large pipes (D>0.24 m), iindependent of flow regime. But, its optionaimplementation in RELAP5-3D is dependent upomass flux. The correlation is only used for lowmass flux situations. For high mass flux situationthe default Kataoka-Ishii drift-flux correlation isused. In addition, the modified bubbly and sluflow regime interfacial heat transfer coefficien(Card 1 option 61) used here in conjunction witthe Gardner drift-flux correlation, uses a Laplacnumber formulation (independent of relative phasvelocity) rather the default Weber number criterioto calculate bubble size. This formulation washown to reduce oscillatory behavior in a previouanalysis [5]. Finally, the Vea-Lahey drift-fluxcorrelation, also appropriate for large pipes (D>0m), is independent of both flow regime and maflux.

Both the Gardner and Vea-Lahey drift-fluxcorrelations were investigated separately in the 1RELAP5-3D model of the LOFT-Wyle WSB03Rblowdown experiment with the Henry-Fauskcritical flow model using a discharge coefficient o0.84 and a non-equilibrium parameter of 100(frozen model). The attributes of these revisemodels with alternate drift-flux correlations arsummarized in Table 1 under the respectivheadings, Case 5 and Case 6. The effect of thealternate drift-flux correlations on the timedependent liquid fraction at the vessel/downcominterface can be obtained by viewing the twcurves on Figure 12. This figure shows that thGardner drift-flux model (as implemented inRELAP5-3D) and associated modified bubbly anslug flow regime interfacial heat transfecoefficients dampen the calculated oscillationthe liquid fraction at the vessel/downcomeinterface quite significantly after ~550 seconds in

8


e

lnelreda

sall),eechrale-t.lgheeralo-enre

ellf

eofhedreatee-reasse-

of

tns

the blowdown. During time frame from ~500 to~550 seconds liquid fraction oscillations persist,since RELAP5-3D uses the default Kataoka-Ishiidrift-flux model at higher mass fluxes. In addition,Figure 12 shows that the Vea-Lahey drift-fluxmodel tends to predict distinct regions ofoscillatory behavior, each ~50 to ~100 secondsapart, finally culminating in a continuousoscillation beyond ~1350 seconds. This behavior,although somewhat less oscillatory than thatresulting from the use of the default Kataoka-Ishiidrift-flux model, is not satisfactory.

In spite of the large effect on predicted liquidfraction at the vessel/downcomer interface, the useof the three alternative drift-flux models only havea small effect on comparison to measured testparameters. The most relevant product of thisstudy of alternative drift-flux formulations is theeffect on the total fluid mass lost from the systemover the course of the transient. This effect isdepicted in Figure 13. The Vea-Lahey drift-fluxcorrelation calculates system fluid mass losses verysimiliar to those obtained with the default Kataoka-Ishii correlation (within 10 kg at all times).Meanwhile, one also notices that the Gardner drift-flux correlation slightly underpredicts (with respectto both experimental data and the Kataoka-Ishiicorrelation) the total fluid mass loss from thesystem out to ~500 seconds. After that point theGardner correlation overpredicts system fluid massloss. But, the overall agreement with the end-stateexperimental data has been improved. That is, thecalculation employing the Gardner drift-flux modeltend to overpredict the total fluid mass lost fromthe system over the course of the transient by only~70 kg (155 lb) rather than ~100 kg (220 lb).

Three-Dimensional Modeling

At this time it was determined that the use of thenew RELAP5-3D hydrodynamic modelingcapability might further reduce the differencebetween the computational simulation and theLOFT-Wyle experimental results. This conclusionwas based upon observations of the asymmetry ofthe experimental facility. The most significantasymmetry was the single blowdown legcontaining the 16 mm (0.6374 in.) orifice

connecting to only a small azimuthal sector of thdowncomer.

In order to better model this experimentaasymmetry, the eleven-volume downcomer regiocomponent as well as the two-volume lower vesshead component (as depicted in Figure 2) weconverted from one- to three-dimensional fluicomponents. The conversion was performed instraight-forward manner. The relatively thin0.08731 m (3 7/16 in.) thick downcomer wamodeled as a cylindrical annulus with one radinode, 11 axial nodes (identical to the 1-D modeand 4 azimuthal nodes (each one a 90 degrsector). The lower vessel head component, whiconnects to both the downcomer and the centvessel region was also converted from a ondimensional to a three-dimensional componenThis component was modeled in cylindricageometry with two radial mesh, with the outer rinof mesh set to be the same thickness as tdowncomer. As was the case for the downcomregion, this 3-D component employed 4 aximuthnodes, each a 90 degree sector. Axially this twvolume tall 3-D component employed the samheights as the original 1-D component. But, iorder to model correctly the hemispherical lowevessel head (which rests entirely in the lower of thtwo axial volumes), the outer ring of volumes inthe lower axial volume set was eliminated from thsystem by setting its junction connections to aother volumes to zero. In addition, the volumes othe four lower inner ring volumes were modifiedsuch as to maintain the true volume of thhemispherical lower vessel. Finally, the numberpassive heat structures adjacent to either tdowncomer or lower vessel head were multiplieby a factor of four, with a concurrent factor-of-foureduction in volume and surface area. Thadditional structures were necessary to accomodthe four azimuthal quadrants of the new thredimensional fluid components. This heat structurepresentation ensured that the total structural mand surface area would be equivalent in the threand one-dimensional fluid component modelsthe LOFT-Wyle experimental facility.

Otherwise, the remainder of the RELAP5-3Dmodel of LOFT-Wyle test WSB03R was noaltered. This includes the use of smooth junctio

9


d.dsetensdhele

h

disiseg

ly,re,dm

tal1).

mere-d

he

x,l

le.

seicr,

with no form-loss coefficients (except for a smallform loss of 0.1 which represents the 180 degreebend at the lower vessel/downcomer and lowervessel/upper vessel junctions) in thevessel/downcomer modeling. In addition, the‘best” critical flow model from the 1-D analysiswas used in the system simulation employing 3-Dfluid components. This model was the Henry-Fauske critical flow model with a dischargecoefficient of 0.84 and a non-equilibriumparameter of 1000.

At first a preliminary RELAP5-3D calculation wasrun with the Gardner drift-flux model (withmodified bubbly and slug flow regime interfacialheat transfer coefficients) to see if oscillations inthe liquid fraction in any azimuthal quadrant of thevessel/downcomer interface persisted. Figure 14shows that, unlike the case of the fully 1-D model,the Gardner drift-flux correlation does not reduceoscillations in the liquid fraction at thevessel/downcomer interface in the 3-D model.This calculation is designated as Case 7 in both thefigure and in Table 1. Similiar oscillations wereobtained with both the default Kataoka-Ishii andoptional Vea-Lahey drift-flux models. An optionin RELAP5-3D known as the modified LeVequemomentum flux splitting option for the semi-implicit scheme (Card 1 option 93) [11] was thenactivated in an attempt to dampen the oscillationsin the liquid fraction at the 3-D vessel/downcomerinterface. The calculation with these modelingattributes is designated as Case 8 in Table 1 andFigure 14.

As implemented in three-dimensional fluidcomponents in RELAP5-3D, the LeVeque methodpermits varying proportions of the accurate butnumerically unstable central-differencing schemeto be used in combination with the standardupwind differencing scheme. This technique isused to compute cell-centered fluid velocities fromthose directly calculated in adjacent junctions. TheLeVeque method allows proportionately morecentral-differencing to be employed depending onhow closely the spatially-varying velocity profilecan be approximated as a linear function. Inpractice the LeVeque method produces moreaccurate results than the purely upwinddifferencing scheme, especially in situations where

a stratified vapor/liquid interface is encountereIn addition, a modification of the LeVeque methoallows the R-direction fluid momentum equationin cylindrical geometry to be solved much moraccurately in volumes abutting the coordinacenterline. Both of the above-mentioned situatioin which the modified LeVeque method shoulproduce increased solution accuracy exist in tthree-dimensional components of the LOFT-Wymodel.

The modified LeVeque method in conjunction witthe Vea-Lahey drift-flux model was found toproduce the least amount of oscillation in the liquifraction at the vessel/downcomer interface. Thgeneral lessening of the oscillatory behaviordepicted in Figure 14. The modified LeVequmethod was not found to be as effective in reducinliquid fraction oscillations if either the defaultKataoka-Ishii or Gardner drift-flux models wereemployed.

Figures 15, 16, 17, and 18 compare, respectivethe time-dependent orifice upstream pressuorifice mass flow rate, the time-integrated fluimass lost from the system, and the orifice upstreadensity for the newly developed RELAP5-3Dmodel of the LOFT-Wyle blowdown test withthree-dimensional fluid components to thapreviously obtained using only one-dimensioncomponents (e.g. Case 4 and Case 8 in TableThis 3-D model employed the Vea-Lahey drift-fluxcorrelation, and the modified LeVeque momentuflux splitting option. Upon observation, thesfigures show that the salient calculation results aonly marginally affected by: 1) the use of threedimensional fluid components in the vessel andowncomer, 2) the use of the Vea-Lahey versus tdefault Kataoka-Ishii drift-flux correlation, or 3)the use of the modified LeVeque momentum flusplitting option. In addition, although not shownthe use of the two-phase stratified flow levetracking option in either the 1-D or 3-D fluidcomponent RELAP5-3D models had no noticabeffect on the pertinent results of either calculation

The fact that the 1-D and 3-D results are so clowas not expected, given the radially asymmetrconnection of the blowdown leg to the downcomein the experimental facility. These results do

10


,d

tosaaltt,te’se,hee

alee2.ee

mh

alel-t

dsrn.isct-

of

te

dso],e

however, confirm the proper operation of the multi-dimensional fluid component in RELAP5-3D.This was confirmed by detailed investigation of theRELAP5-3D output. A reasonable distribution ofazimuthal fluid velocities was observed in thethree-dimensional downcomer component. Thatis, the azimuthal velocities were observed to behighest, azimuthally symmetric, and physicallyreasonable in comparison to blowdown legvelocities at the elevation of the blowdown leg.These azimuthal downcomer velocities wereobserved to retain symmetry but becomeprogressively lower at other elevations.

It is believed that radial uniformity of the flowdistribution at the entrance to the downcomer in theLOFT-Wyle blowdown experiment is the majorreason why three-dimensional fluid modeling haslittle effect on calculated results.

Conclusions

The LOFT-Wyle test WSB03R was used toperform an assessment of RELAP5-3D. Faithfulrepresentation of the experimental facilityincluding instrumentation, the boundary conditionsand the initial conditions is required in order toobtain an undistorted assessment. With thisphilosophy in mind, a new 1-D RELAP5 model ofthe LOFT-Wyle facility was created. The majordifferences/improvements between this model andthe original 1-D RELAP5 assessment model were:1) more accurate modeling of the dimensions ofthe downcomer, 2) more accurate modeling of theinstrumentation locations upstream of theblowdown orifice, 3) more realistic modeling ofexperimental system initial fluid conditions, 4)modeling of the portion of the experimental facilitybeyond the break orifice, 5) the use of heatstructures to model the heat capacity of thestructural material immediately adjacent to thefluid, and 6) a larger number of fluid volumes inthe representation of the experimental facility.

The net result of these modelmodifications/improvements is generally improvedagreement with experimental measurement, withthe RELAP5 default Ransom-Trapp critical flowmodel with unity discharge coefficients. Furtherimprovements were observed if the optional

Henry-Fauske critical flow model is employedagain with default values for the discharge annon-equilibrium coefficients.

Since the orifice being modeled has a lengthdiameter ratio of ~3.4 it is not well described aeither a sharp or a long-tube break orifice. Asresult, the independent parameters in both criticflow models were ‘tuned’ in order to achieve bescomparison with experimental measuremenespecially the time-dependent mass flow rathrough the blowdown orifice. These ‘tunableparameters included, the subcooled, two-phaand superheated discharge coefficients for tRELAP5-3D default Ransom-Trapp model and thdischarge coefficient and non-equilibriumparameter for the optional Henry-Fauske criticflow model. Best results were obtained with thRansom-Trapp model when all dischargcoefficients were set to non-physical values of 1.Meanwhile, best results were obtained with thHenry-Fauske model when the dischargcoefficient was set to 0.84 and the non-equilibriuparameter was set to 1000 (a ‘frozen’ model witno evaporation or condensation in the orifice).

The overall best comparison with experimentdata was obtained using the Henry-Fauske modwith a discharge coefficient of 0.84 and a nonequilibrium parameter of 1000. But, the fact thaall revised 1-D RELAP5-3D models overpredicte(by ~100 kg) the total fluid system mass losduring the transient remained a cause for conceBased upon the results of previous studies [5], itbelieved that this misprediction may be a direconsequence of a flaw in the ability of RELAP53D to accurately predict liquid droplet carryover involumes with relatively high void fractions. Thiscan be traced to inaccuracies in the predictioninterfacial drag, which in RELAP5-3D is afunction of the flow-regime-dependent drift-fluxvelocity.

Upon further investigation, it was discovered thathe 1-D RELAP5-3D model suffered from a severoscillation in the liquid fraction at thevessel/downcomer interface beyond ~450 seconinto the transient. This behavior, very similar tthat seen in a previous RELAP5-3D analysis [5was determined to be a result of the flow regim

11


rs

”

dependence of the default Kataoka-Ishii calculateddrift-flux velocity and the resulting interfacial drag.It was believed that by reducing these liquidfraction oscillations, a more accurate prediction ofdroplet carryover and thus time-integrated systemfluid mass loss could be obtained. Towards thisend two other drift-flux velocity formulations, theGardner and Vea-Lahey correlations, which are notflow regime dependent were tested. The Gardnercorrelation (with modified bubbly and slug flowregime interfacial heat transfer coefficients) wasfound to significantly reduce the oscillations in theliquid fraction at the vessel/downcomer interface.However, the elimination of liquid fractionoscillations only provided a slight improvement inthe calculational prediction of total fluid systemmass loss during the transient (an end-stateoverprediction of ~70 kg rather than ~100 kg).

In response to observed asymmetries in theexperimental facility, the downcomer and lowervessel head regions of the RELAP5-3D model ofLOFT-Wyle blowdown experiment were convertedfrom one- to three-dimensional fluid components.However, in contrast to the fully one-dimensionalmodel, which employed the Gardner drift-fluxcorrelation, the Vea-Lahey drift-flux correlation inconjunction with the modified LeVequemomentum flux-splitting model were required toreduce liquid fraction oscillations at thevessel/downcomer interface in the 3-D model.

The basic level of agreement with experimentalresults remained unchanged with the introductionof three-dimensional fluid modeling. Thisinsensitivity to more explicit flow modeling is mostlikely the result of the radial uniformity of the flowdistribution at the entrance to the downcomer.

References

1. Lin, J.E., G.E. Gruen, and W.J. Quapp,“Critical Flow in Small Nozzles for Saturatedand Subcooled Water at High Pressure,” BasicMechanisms in Two-Phase Flow and HeatTransfer Symposium at the Winter Meeting ofthe American Society of MechanicalEngineers, Chicago, Illinois, pp 9-17, 1980.

2. Ransom, V.H. et al., “RELAP5/MOD2 CodeManual, Volume III: DevelopmentalAssessment Problems,” EGG-TFM-7952,December 1987.

3. “RELAP5-3D Code Manuals, Volumes I, II,IV, and V,” Idaho National Engineering andEnvironmental Laboratory, INEEL-EXT-98-00834, Revision 1.1b, February 1999.

4. Trapp, J.A. and V.H.Ransom, “CalculationalCriterion for Nonhomogeneous,Nonequilibrium Two-Phase Flows,”International Journal of Multiphase Flow, pp669-681, 1982.

5. Aumiller, D.L., E.T. Tomlinson, andW.G.Clarke, “A New Assessment of RELAP5-3DUsing the General Electric Level SwellProblem,” Presented at 2000 RELAP5International Users Seminar, Jackson Hole,Wyoming, available as B-T-3320 from DOEOffice of Scientific and Technical Information,2000.

6. Tomlinson, E.T. and D.L Aumiller, “AnAssessment of RELAP5-3D Using theEdwards-O’Brien Blowdown Problem,”Presented at 1999 RELAP5 International UseSeminar, Park City, Utah, available as B-T-3271 from DOE Office of Scientific andTechnical Information, 1999.

7. Henry, R. E. and H.K. Fauske, “The Two-Phase Critical Flow of One-ComponentMixtures in Nozzles, Orifices, and ShortTubes,”Journal of Heat Transfer,93, pp 179-187, 1971.

8. Kataoka, I. and M. Ishii, “Drift Flux Model forLarge Diameter Pipe and New Correlation ofPool Void Fraction,”International Journal ofHeat and Mass Transfer, 30, pp 1927-1939,1987.

9. Gardner, G., “Fractional Vapour Content of aLiquid Pool Through which Vapour isBubbled,”International Journal of MultiphaseFlow, 6, pp 399-419, 1980.

10. Vea, H.W. and R.T. Lahey, Jr., “An ExactAnalytical Solution of Pool Swell DynamicsDuring Depressurization by the Method ofCharacteristics,”Nuclear Engineering andDesign, 45, pp 101-116, 1978.

11. LeVeque, R.J., “Numerical Methods forConservation Laws, Lectures in Mathematics,Birkhauser Verlag, Basel, 1992.

12


13

Table 1

Attributes of the Critical Flow and Interfacial Drag Models Used in This Assessment

Case Critical FlowModel

Critical FlowDischargeCoefficient

Henry-FauskeNon-

EquilibriumParameter

Interfacial DragCorrelation

MomentumSplitting Option

0(Original)

Ransom-Trapp 1.0 − Kataoka-Ishii Standard

Revised Fully 1-D Models

1 Ransom-Trapp 1.0 − Kataoka-Ishii Standard

2 Henry-Fauske 1.0 0.14 Kataoka-Ishii Standard

3 Ransom-Trapp 1.2 − Kataoka-Ishii Standard

4 Henry-Fauske 0.841000

(frozen)Kataoka-Ishii Standard


(frozen)Vea-Lahey Standard


(frozen)Gardner-61 Standard

Revised Models with 3-D Lower Vessel and Downcomer


(frozen)Gardner-61 Standard


(frozen)Vea-Lahey LeVeque

14


Figure 1Axomnometric Projection of Wyle Test Facility

15


Figure 2Schematic of RELAP5-3D Representation of LOFT-Wyle Experiment WSB03R

Inner Vessel (010)

Lower Vessel (050)

Downcomer (110)

16 mmBreakOrifice (125)

Spool Piece #1 andInlet of Sp#2 (120)

Downstream Piping (170) Atmosphere (210)

Downstream Tee (180)

End of Sp#2 (140)

Middle of Sp#2 (130) Inlet of Gate Valve (150)

Outlet of Gate Valve (160)

RuptureDisks(165)

(020)(060)

(115) (135) (145)

(18001) (18002)

GateValve(155)


0

2e+06

4e+06

6e+06

8e+06

1e+07

1.2e+07

1.4e+07

0 200 400 600 800 1000 1200 14000

500

1000

1500

2000

Pre

ssur

e (P

a)

Pre

ssur

e (p

si)

Time (Sec)

Exp dataCase 0Case 1Case 2

rifice

-2

0

2

4

6

8

10

12

14

0 200 400 600 800 1000 1200 1400

0

5

10

15

20

25

30

Mas

s F

low

Rat

e (K

g/se

c)

Mas

s F

low

Rat

e (lb

/sec

)

Time (Sec)


Figure 3: Initial Comparison of Pressure Upstream of the O

16

Figure 4: Initial Comparison of Rate of Mass Outflow


0

500

1000

1500

2000

2500

0 200 400 600 800 1000 1200 14000

1000

2000

3000

4000

5000

Inte

gra

ted M

ass

Flo

w (

Kg)

Inte

gra

ted M

ass

Flo

w (

lb)

Time (Sec)


Figure 5: Initial Comparison of Integrated Mass Outflow

-100

0

100

200

300

400

500

600

700

800

0 200 400 600 800 1000 1200 1400

0

10

20

30

40

50

Densi

ty (

Kg

/m3)

Densi

ty (

lb/ft3

)

Time (Sec)


17

Figure 6: Initial Comparison of Density Upstream of the Orifice


0

2e+06

4e+06

6e+06

8e+06

1e+07

1.2e+07

1.4e+07

0 200 400 600 800 1000 1200 14000

500

1000

1500

2000

Pre

ssure

(P

a)

Pre

ssure

(psi

)

Time (Sec)

Exp dataCase 3Case 4

Figure 7: Comparison of Pressure Upstream of the Orifice Using Modified Discharge Coefficients

-2

0

2

4

6

8

10

12

14

0 200 400 600 800 1000 1200 1400

0

5

10

15

20

25

30

Mas

s F

low

Rat

e (K

g/se

c)

Mas

s F

low

Rat

e (lb

/sec

)

Time (Sec)


18

Figure 8: Comparison of Rate of Mass Outflow Using Modified Discharge Coefficients


0

500

1000

1500

2000

2500

0 200 400 600 800 1000 1200 14000

1000

2000

3000

4000

5000

Inte

gra

ted M

ass

Flo

w (

Kg)

Inte

gra

ted M

ass

Flo

w (

lb)

Time (Sec)


Figure 9: Comparison of Integrated Mass Outflow Using Modified Discharge Coefficients

-100

0

100

200

300

400

500

600

700

800

0 200 400 600 800 1000 1200 1400

0

10

20

30

40

50

Densi

ty (

Kg

/m3)

Densi

ty (

lb/ft3

)

Time (Sec)


19

Figure 10: Comparison of Density Upstream of the Orifice Using Modified Discharge Coefficients


0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

0 200 400 600 800 1000 1200 1400

Liq

uid

Fact

ion

Time (Sec)

Case 4

Figure 11: Liquid Fraction at the Vessel/Downcomer Interface

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

0 200 400 600 800 1000 1200 1400

Liq

uid

Fa

ctio

n

Time (Sec)

Case 5Case 6

20

Figure 12: Liquid Fraction at the Vessel/Downcomer Interface Using Alternate Interfacial Drag Correlations


0

500

1000

1500

2000

2500

0 200 400 600 800 1000 1200 14000

1000

2000

3000

4000

5000

Mass

Flo

w (

Kg)

Mass

Flo

w (

lb)

Time (Sec)


Figure 13: Comparison of Integrated Mass Outflow Using Various Interfacial Drag Correlations

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

0 200 400 600 800 1000 1200 1400

Liqu

id F

actio

n

Time (Sec)

Case 7Case 8

21

Figure 14: Liquid Fraction at the Vessel/DowncomerInterface in 3-D Calculations

(same azimuthal quadrant as blowdown line)


22

0

2e+06

4e+06

6e+06

8e+06

1e+07

1.2e+07

1.4e+07

0 200 400 600 800 1000 1200 14000

500

1000

1500

2000

Pre

ssure

(P

a)

Pre

ssure

(psi

)

Time (Sec)


-2

0

2

4

6

8

10

12

14

0 200 400 600 800 1000 1200 1400

0

5

10

15

20

25

30

Mass

Flo

w R

ate

(K

g/s

ec)

Mass

Flo

w R

ate

(lb

/se

c)

Time (Sec)


Figure 15: Comparison of Pressure Upstream of the Orifice (3-D Versus 1-D Modeling)

Figure 16: Comparison of Rate of Mass Outflow (3-D Versus 1-D Modeling)


23

0

500

1000

1500

2000

2500

0 200 400 600 800 1000 1200 14000

1000

2000

3000

4000

5000

Inte

gra

ted M

ass

Flo

w (

Kg)

Inte

gra

ted M

ass

Flo

w (

lb)

Time (Sec)


-100

0

100

200

300

400

500

600

700

800

0 200 400 600 800 1000 1200 1400

0

10

20

30

40

50

De

nsi

ty (

Kg

/m3)

De

nsi

ty (

lb/ft3

)

Time (Sec)


Figure 17: Comparison of Integrated Mass Outflow (3-D Versus 1-D Modeling)

Figure 18: Comparison of Density Upstream of the Orifice (3-D Versus 1-D Modeling)