Coolant mixing in a PWR - deboration transients, steam line breaks andemergency core cooling injection - experiments and analyses

H.-M. Prasser, G. Grunwald, Th. Höhne, S. Kliem, U. Rohde, F.-P. WeißForschungszentrum Rossendorf, Institute of Safety Research

P.O.B. 510119 D-01314 Dresden, GermanyTel: (+49) 351 260 3460, Fax: (+49) 351 260 3460, Email: prasser@fz-rossendorf.de

Abstract - The reactor transient caused by a perturbation of boron concentration orcoolant temperature at the inlet of a Pressurized Water Reactor (PWR) depends on themixing inside the reactor pressure vessel. Initial steep gradients are partially lessened byturbulent mixing with coolant from the unaffected loops and with the water inventory of thereactor pressure vessel. Nevertheless the assumption of an ideal mixing in the downcomerand the lower plenum of the reactor leads to unrealistically small reactivity inserts. Theuncertainties between ideal mixing and total absence of mixing are too large to beacceptable for safety analyses. In reality, a partial mixing takes place. For realisticpredictions it is necessary to study the mixing within the three-dimensional flow field in thecomplicated geometry of a PWR. For this purpose a 1:5 scaled model (the ROCOMfacility) of the German PWR KONVOI was built. Compared to other experiments, theemphasis was put on extensive measuring instrumentation and a maximum of flexibility ofthe facility to cover as much as possible different test scenarios. The use of specialelectrode-mesh sensors together with a salt tracer technique provided distributions of thedisturbance within downcomer and core entrance with a high resolution in space and time.Especially the instrumentation of the downcomer gained valuable information about themixing phenomena in detail. The obtained data was used to support code development andvalidation. Scenarios investigated are: (1) Steady-state flow in multiple coolant loops with atemperature or boron concentration perturbation in one of the running loops. (2) Transientflow situations with flow rates changing with time in one or more loops, such as pumpstart-up scenarios with deborated plugs in one of the loops or onset of natural circulationafter boiling-condenser-mode operation. (3) Gravity driven flow caused by large densitygradients, e.g. mixing of cold emergency core cooling water entering the RPV through theECC injection into the cold leg. The experimental results show an incomplete mixing withtypical concentration and temperature distributions at the core inlet. which strongly dependon the boundary conditions. CFD calculations were found to be in good agreement with theexperiments.

I. INTRODUCTION

A perturbation of boron concentration or coolanttemperature at the inlet of a Pressurized Water Reactor(PWR) may lead to a reactivity excursion. The inducedpower peak and the released fission energy depend onthe inserted reactivity. As a matter of rule theperturbation has to be expected not in all loops of theprimary circuit in the same time – the probability of anappearance in one of the loops only is much higher. Onthe way from the reactor inlet to the core entrance thiswater comes into contact with coolant arriving from theunaffected loops and with the water inventory of thereactor pressure vessel. The boron concentration andtemperature distributions at the core inlet are thereforethe result of a turbulent mixing process in the flow pathof the reactor downcomer and the lower plenum.

3D neutron kinetic computer simulations haveshown quite early, that a uniform distribution of theperturbation at the core inlet, i.e. the assumption of anideal mixing, does not deliver conservative results. In

fact, the predicted power peak was about five timeslower compared to analyses where a total absence ofmixing was assumed [1]. In this work, attempts wereinitially made to implement stationary mixing matrixesin a 3D neutron kinetics code (DYN3D). The remaininguncertainties were too large to be acceptable for safetyanalyses. There are numerous postulated accidentscenarios, where in case of uniform mixing the coresurvives without consequences or even without reachingcriticality, while neglect of the mixing leads to theprediction of severe fuel rod failures.

In reality, a partial mixing takes place. In theresult, the real distribution of fluid condition at the coreentrance lies in-between the two extreme cases bothconcerning shape and amplitude of the perturbation. Tomake realistic predictions about the consequences it isnecessary to study the coolant mixing, which is acomplicated three-dimensional fluid dynamic problem.Modern CFD codes running on powerful computersbecame available to model the coolant flow in thecomplex geometry of a PWR a couple of years ago.

Nevertheless the use of CFD for this complicated task isa challenge still today. Due to the high safety relevanceof the coolant mixing phenomenon it is necessary tovalidate computer codes and to verify computationalresults using experimental data. This explains the vividscientific interest in coolant mixing experiments. In ourown experiments presented in this paper, we putemphasis on flexibility of the test facility to cover asmuch as possible different test scenarios, and, above all,on measuring instrumentation in order to meet therequirements of code development and validation asmuch as possible by providing parameter distributionswith a high resolution in space and time.

Since the characteristic Reynolds numbers for thelarge dimensions of the reactor pressure vessel (RPV)are very high, turbulent mixing dominates overmolecular diffusion. For this reason in a given velocityfield a temperature perturbation and a boronconcentration perturbation behave in the same way,however in case of high temperature differences densitygradients may cause a feedback to the velocity field bygravity effects. From the point of view of fluid dynamicboundary conditions, the mixing scenarios relevant forthe safety of PWRs can be subdivided into three largegroups:

1. Steady flow in multiple coolant loops with atemperature or boron concentration perturbation in apart of the running loops. The flow may be driveneither by running main coolant pumps or by steadynatural circulation. A typical case is the mixing ofover-cooled water arriving from one of the steamgenerators in case of a main steam-line break.

2. Transient flow situations with flow rates changingwith time in one or more loops. Also here the flowmay be driven either by coolant pumps or by naturalcirculation. Prominent examples are: (a) inadvertentstart-up of the main coolant pump in a situationwhere boron-free coolant was unintentionallyaccumulated in the corresponding loop, e.g. by amalfunction of the auxiliary feed water system duringa shut-down of the reactor or due to heat exchangertube leakage in a steam generator, (b) onset of naturalcirculation in a late phase of a small-break loss-of-coolant accident after an accumulation of boron freecondensate by boiling-condenser mode operation.

3. Gravity driven flow caused by large densitygradients. This phenomenon is observed only if themain coolant pumps are not working, otherwise themomentum driven flow induced by the pumps isalways dominating and density effects can beneglected. Gravity effects are characteristic, forexample, for the mixing of cold emergency corecooling water entering the RPV through the ECCinjection into the cold leg. Due to the higher densityof the ECC water, the formation of a streak of coldwater flowing down in the downcomer is observed.This causes thermal loads on the hot reactor pressurevessel. Beside the prediction of the parameters at the

core inlet, the exact knowledge of the spatialdistribution of the ECC water is essential for theassessment of the pressure vessel behavior (Pre-stressed Thermal Shock = PTS).

The literature reports about a number of testfacilities to study coolant mixing, each considering thegeometrical specifics of various reactor types [2-9]. Incase of [2-5], the facilities were solely designed to studythe inadvertent pump start-up, i.e. they dispose of acirculation pump only in one of the loops; the otherloops were just passive ones. Other test facilities werededicated to mixing of cold emergency core cooling(ECC) water injected into the cold leg of a PWR [6-9].In order to study coolant mixing inside the reactorpressure vessel of a German type PWR the test facilityROCOM (Rossendorf Coolant Mixing Model) was builtand operated during the last three years. The facilityrepresents a KONVOI type PWR (1300 MWel)developed by Siemens KWU in a linear scale of 1:5.

II. ROCOM TEST FACILITY

Fig. 1 Reactor model of the test facility ROCOM,German KONVOI type PWR, 1300 MWel,linear scale: 1:5, volumetric scale 1:125

ROCOM consists of a reactor pressure vesselmodel (Fig. 1) with four inlet and four outlet nozzles.Since it was planned to use the facility for a wide rangeof assumed mixing scenarios, ROCOM was equippedwith four fully functioning loops (Fig. 2), i.e. it has four

circulation pumps, which are driven by motors withcomputer controlled frequency transformers. In this way,a wide variety of flow rate regimes, such as four-loopoperation, operation with pumps off, simulated naturalcirculation modes and flow rate ramps can be realized.For natural circulation conditions the correspondingpumps are operated at low rotation speed by means ofthe frequency transformer system. Beginning from thebends in the cold legs which are closest to the reactorinlet, the geometrical similarity between model andoriginal reactor is respected until the core inlet. The core

itself is excluded from the similarity, i.e. cross mixingwithin the core is not investigated. The reactor model ismanufactured of acrylic glass (Perspex, see Fig. 3). Atotal view of the test facility is given in Fig. 4. Thereactor model exactly follows the geometry of theoriginal PWR with respect to the design of the nozzles(diameter, surface curvature, diffuser parts), thecharacteristic extension of the downcomer cross sectionbelow the nozzle zone, the so-called perforated drum inthe lower plenum and the design of the core supportplate with the orifices for the coolant.

Fig. 2 ROCOM test facility with four loops and individually frequency controlled circulation pumps

The flow rate in the loops is scaled according tothe transit time of the coolant through the model. Sincethe geometrical scale is 1:5, the transition time of thecoolant is identical to that of the original reactor, whenthe coolant velocity is scaled down by 1:5, too. Thenominal flow rate in ROCOM is therefore 185 m³/h perloop.

As a result of the scaling and the increasedviscosity of cold water compared to that of the hotcoolant, the Reynolds numbers (Re) in the model arelower than those in the original reactor by factor of 190.

At the maximum flow rate (300 m³/h) delivered by thepumps, a Re ratio of 100 can be achieved. However, theRe numbers (of the order of 105 in the ROCOM facility)correspond to highly turbulent flow conditions, thusboth velocity fields and mixing condition are supposedto be independent of the mass flow rate (if densitygradients are absent). This assumption is supported byown measuring results obtained for flow rates varied inthe feasible range, by scale-up investigations using aCFD code [10], as well as by findings of Dräger [11,12], who took advantage from real-scale experimentscarried out at Russian type VVER-440 reactors.

To vary the flow rate the frequency of the powersupply for the pumps is controlled. Computer controlledfrequency transformers allow to run the pumpsaccording to predefined time dependencies of the flowrate, given as data maps for each loop individually.

Fig. 3 Photo of the reactor pressure vessel model ofROCOM made of acrylic glass (Perspex)

The facility is operated with water at roomtemperature and ambient pressure. Since the coolantmixing is mainly based on turbulent dispersion it ispossible to use a tracer substance to model differencesof either boron concentration or coolant temperature. Incase of ROCOM the coolant in the disturbed loop waslabeled by injecting a sodium chloride solution into themain coolant flow upstream of the affected reactor inletnozzle. For this purpose, loop 1 of ROCOM wasequipped with a mixing device, which distributes thetracer solution uniformly across the cross section of themain circulation pipe. Start and stop of the injectionprocess was controlled by magnetic valves. Theconcentration distribution of the tracer within thereactor model is measured by electrical conductivitysensors.

In case of the experiments on ECC injection, themixing device was replaced by an injection nozzle,which was geometrically similar to the originalKONVOI reactor. There, a horizontal injection line isused, which has a diameter of 0.2 m (at ROCOM ∅ 40mm). It is connected to the cold leg under an angle of

45 deg. The higher density of the cold ECC water wassimulated by adding sugar (glucose), since densitygradients cannot be created by temperature differences,because the facility cannot be heated up. Fortunately,the viscosity of glucose solution becomes large only atconcentrations, where the relative density increase iswell above the 10 % necessary for the experiments. Asugar solution with the corresponding density of 1100kg/m³ has a viscosity which is still just by factor ofabout 3 higher than that of pure water. The tracer cantherefore still be envisaged as a low-viscous fluid.

Fig. 4 General view of the ROCOM test facility

III. INSTRUMENTATION

The tracer distribution in the reactor model wasobserved by so-called electrode-mesh sensors, whichmeasure the distribution of the electrical conductivityover the cross section of the flow duct. Sensors of thiskind were introduced by Johnson [13], who used themto measure the integral gas fraction in the cross section.Reinecke et al. [14] presented a tomographic devicevisualizing sequences of gas fraction distributions witha rate of about 100 frames per second. Our owndevelopment was aimed at a direct conductivitymeasurement between pairs of crossing wires to avoidtomographic reconstruction algorithms [15] and to reacha time resolution of up to 10 000 frames per second[16]. Two crossing grids of electrodes insulated fromeach other are placed across the flow duct. Theelectrodes of the first grid (transmitter electrodes) aresupplied with short voltage pulses in a successive order.The currents arriving at the electrodes of the secondgrid are recorded (receiver electrodes). After a completecycle of transmitter activation a complete matrix oflocal conductivities is obtained. Special methods ofsignal acquisition [15] guarantee that each value of thematrix depends only on the local conductivity in theimmediate vicinity of the corresponding crossing pointbetween transmitter and receiver electrodes.

Measured local conductivities are afterwardsrelated to reference values. The result is a so-calledmixing scalar that characterizes the instantaneous shareof coolant originating from the disturbed loop (i.e.where the tracer is injected) at a given location insidethe flow field. The scalar is dimensionless. Assuming

similarity between tracer field and the temperature andboron concentration fields it can be used to apply theexperimental results to the original reactor. Thereference values correspond to the unaffected coolant(index 0) and the coolant at the disturbed reactor inletnozzle (index 1). The difference between the tworeference values is the magnitude of the perturbation.The mixing scalar Θ is defined as follows:

0,B1,B

0,Bt,z,y,x,B

01

0t,z,y,x

01

0t,z,y,xt,z,y,x CC

CC

TT

TT

−−

≅−

−≅

σ−σσ−σ

=θ

Here, σ is the electrical conductivity, T is thetemperature and CB the boron concentration. Which ofthe two parameters temperature or boron concentrationis represented by the measured mixing scalar dependsonly on the right choice of the reference values and theset-up of the boundary conditions in the experiment. Ifthe tracer is for example injected corresponding totemperature changes in a given scenario, the mixingscalar represents the temperature. Additional measureshave to be taken, if the velocity field would beinfluenced by a feed-back from the temperature-dependent fluid density. This is the case in the testsmodeling the cold ECC water injection described lateron.

Fig. 5 Mesh sensor for measuring tracer distributionsupstream of the reactor inlet nozzle (16x16electrodes)

Mesh sensors are placed at four positions of theflow path. The first sensor (Fig. 5) is flanged to thereactor inlet nozzle (Fig. 1) in loop 1. It is aimed at theobservation of the distribution at the reactor inlet, whichis used as reference distribution for calculating mixingscalars. It has 2 x 16 electrode wires, which are crossingunder an angle of 90 deg, i.e. the measuring matrix hasthe dimension of 16 x 16. The pitch of the wires is 8.9mm, the total amount of measuring points is 216, sincesome of the crossing points are outside the circularcross section of the pipe.

The second and the third sensors are located in thedowncomer (Fig. 1 and Fig. 6). They represent 4 x 64

measuring positions, i.e. the azimuthal distribution ofthe tracer concentration is measured at 64 angularpositions with a pitch of 5.625 deg. Over the radiusthere are 4 measuring positions with a radial pitch of 13mm. The downcomer sensors consist of radial fixingrods with orifices for four circular electrode wires. Rodsand wires are separated electrically by small ceramicinsulation beads. The rods are acting as radialelectrodes, i.e. each rod corresponds to an azimuthalmeasuring position.

a) total view

b) closer view to the electrodes

Fig. 6 Mesh sensor for the downcomer (4x64measuring points)

The fourth sensor is integrated into the coresupport plate (Fig. 7). 2 x 15 electrode wires arearranged in a way that the wires of the two planes crossin the centers of the coolant inlet orifices of each fuelelement. In this way, the tracer concentration ismeasured at each fuel element inlet.

a) total view

b) electrode wires integrated into the core supportplate

Fig. 7 Mesh sensor at the core entrance (15x15measuring points, each fuel element isinstrumented)

All four sensors are connected to a signalacquisition unit disposing of measuring matrixcomprising of a 32 x 32 points. The sensors occupydifferent quadrants of this matrix. In total, about 1000measuring points are recorded, the measuring frequencyis 200 Hz. In most of the cases 10 successivemeasurements were averaged and the result was storedwith a frequency of 20 Hz, because the characteristicfrequency of the observed phenomena were notrequiring a higher sampling frequency.

IV. MIXING UNDER STEADY FLOW CONDITIONS

The tests for steady flow conditions wereperformed with different numbers of operating pumpsincluding the normal operation regime with all fourpumps running [17]. The pumps were working atconstant rotation speed. The perturbation was modeledby injecting the tracer into loop 1, the mixing deviceguaranteed an equal distribution over the cross sectionof the main circulation pipe. In the tests presentedbelow, the tracer was injected over a long period (≈10s), which was sufficient to establish a quasi-stationaryconcentration field. The injection was stopped justbefore the tracer cloud reappeared at the reactor inletafter completing an entire circulation in the primary

circuit. The average conductivity measured by thesensor at the reactor inlet during the quasi-stationaryperiod was used as reference value σ1. The secondreference value σ0 is the conductivity measured beforethe injection started.

Fig. 8 Time averaged tracer concentration distributionat all four sensors during quasi steady-stateflow, scaled nominal flow rate (185 m³/h in allfour loops), perturbation in loop 1 (red arrow)

The distributions measured by the sensors in thedowncomer and at the core entrance were also averagedover the quasi-stationary period. In the result, averagemixing scalars, i.e. so-called mixing coefficients wereobtained, which show the average share of coolantarriving from the disturbed loop at the given measuringlocation (Fig. 8). In the presented case all maincirculation pumps are delivering identical flow rates.This corresponds to an early stage of a main steam-linebreak scenario and the tracer represents over-cooledcoolant coming from the disturbed steam generator. Inthis case the labeled water occupies a sector of 90 deg atthe upper downcomer sensor. The azimuthal position ofthis sector is shifted from the angular position of theinlet nozzle (22.5 deg) towards 45 deg corresponding tothe symmetry conditions. At the lower downcomersensor, the slopes of distribution become more smoothcaused by the action of turbulent mixing at theboundary of the labeled water stream. At the coreentrance, a characteristic distribution is found, whichshows a maximum near the symmetry position of 45deg at the side of the azimuthal position of the affectedinlet nozzle. The maximum value of the mixingcoefficient reaches 91 %, i.e. there is a fuel elementreceiving almost the full amplitude of the perturbation.There is almost no tracer found in large regionsopposite to the affected loop. Nevertheless, the effect ofthe perturbation on the reactivity is much smaller thanin case of assumed absence of mixing due to the smoothslopes of the distribution.

Fig. 9 Time sequence of instantaneous mixing scalars showing the arrival of the tracer slug, scaled nominal flow rate(185 m³/h in all four loops), perturbation in loop 1 (arrow), maximum indicated by bold number [%]

On the first glance, these results seem to prove thatthe flow field is similar to what was found in the verycomprehensive mixing experiments carried out byDräger [11, 12] for the Russian type pressurized waterreactor VVER-440. There, it was found that the flowfield in the downcomer can be approximated by apotential flow with quite straight streamlines in thedowncomer. A closer look to the transient behaviorduring the arrival of the tracer slug shows that the flowconditions in the downcomer are more complicated(Fig. 9). Tracer arrives at first at two separate azimuthalpositions. Only later the two streams fuse together. Thisis caused by secondary vortices in the downcomer,which are induced by the widening of the downcomer inthe region below the inlet nozzles. There, the width ofthe downcomer is increased by the reduction of the wallthickness of the reactor pressure vessel (see Fig. 1). Theresulting diffuser angle of 18 deg is too big to avoidflow separation.

0

0.5

1

1.5

2

2.5

3

0 45 90 135 180 225 270 315 360

azimuthal position / °

c / m

/s

LDA

CFX

outlet nozzleshatched

Fig. 10 Azimuthal distribution of the vertical velocitycomponent at the lower end of the downcomer,LDA measurements compared to CFXcalculations [18]

Secondary vortex flow is the result, causing a non-uniform velocity field at the lower end of the down-comer, which was found also by LDA measurements(Fig. 10). The vortices are rotating in the plane of theunwrapped downcomer, they are intermittently chang-ing their strength, which results in strong fluctuations ofthe tracer concentration at the periphery of the affectedsector of the core entrance (Fig. 11). The assumption of

a potential flow can therefore not be applied for thistype of reactor.

Fig. 11 Fluctuations of the mixing scalar due toturbulence in the maximum and in the slope ofthe distribution at the core entrance

V. CIRCULATION START-UP WITH A SLUG OFDEBORATED COOLANT IN THE LOOP

Most relevant boron dilution transients are initiatedby an initiation of the flow in a loop, where an accumu-lation of boron-free coolant took place. Initially thecoolant is at rest. Only in this state can large slugs ofboron-diluted coolant be accumulated. This coolant cancome into motion either by a pump start-up or an onsetof natural circulation. Both kinds of tests were perform-ed at ROCOM. For example, a pump start-up scenariois presented. The flow rate in the affected loop wasincreased according to Fig. 12. It was assumed that aquantity of 36 m² of boron-free coolant is present in theloop. This is a worst-case condition covering allpossible scenarios of this kind. The deborated coolantpasses around the core barrel instead of flowing directlydownstream. This effect can be seen in Fig. 13, wheresequences of azimuthal distributions in the downcomerare plotted in an unwrapped way against time. At theupper sensor in the downcomer the tracer arrives stillbelow the affected inlet nozzle. With growing time thetracer spreads in the azimuthal direction. Subsequently,at the lower sensor two maxima at azimuthal positions

on the back side of the downcomer are observed. In theresult the tracer appears at the core entrance at the sidenearly opposite to the affected inlet nozzle. A sequenceof instantaneous mixing scalars at the core entrance isshown in Fig. 18a together with theoretical results. Themaximum value reached 59 % in this test. It stronglydepends on the volume of the initial slug of deboratedcoolant, i.e. the amplitude of the deboration of the coredecreases, if the slug is smaller. Similar results wereobtained in tests where the flow rate was close tonatural circulation conditions.

0 5 10 15 20Time [s]

0.0

0.2

0.4

0.6

0.8

1.0

Nor

mal

ized

Val

ues

[-]

Velocity

VelocityLoop 2-4Slug Concentration

Loop 1

Fig. 12 Flow rate in loop 1 and counter flow in loops2-4 during pump start-up, averaged tracerconcentration at the reactor inlet nozzle

Fig. 13 Time history of azimuthal mixing scalardistributions in the downcomer during pumpstart-up, unwrapped downcomer plots, affectedloop: red arrow

VI. SIMPLIFIED MATHEMATICAL MODEL

A simplified mathematical model (SAPR - Semi-Analytical Perturbation Reconstruction) was developedto generalize experimental results using some linearproperties of the transport equation for the temperaturerespectively the boron concentration [19]. By means ofa superposition of data measured at the core entrancefor short tracer pulses (quasi Dirac impulses) generatedat the reactor inlet, the length and the amplitude of thedisturbance can be varied off-line, i.e. without repeatingthe experiments. Since the flow rate in the loop ischanging, the system response is depending from themoment of injection. Therefore, the Dirac impulseshave to be put in different points on the flow rate ramp.

Fig. 14 Dirac impulse-like perturbation at the reactorinlet (upper plot) and maximum mixing scalaras function of time taken from the systemresponse (lower plot) for the pump start-up test

The method was checked using the experimentaldata of the test described in the previous chapter. Theapplied input pulses and the experimentally obtainedresponses are shown in Fig. 14. Additional intermediateDirac-responses were derived by interpolation. The timehistory of the tracer concentration during the longinjection (Fig. 12) was reconstructed by superposingthese Dirac perturbations (Fig. 15). The superposition ofthe corresponding time responses is in good agreementto the directly measured one (Fig. 18b). Among others,the method can be used to correct the decrease of thetracer concentration in the experimental slug. Due to thefact that tracer solution was injected with a constant

flow rate into the main flow which was increasing intime, the concentration in the experimental slug wasdecreasing (corresponding to increasing boronconcentration) during the injection period (Fig. 12). Inpractice the slug is more or less homogeneous. Anotherapplication of the method is the coupling of 3D neutronkinetic codes with thermal hydraulic system codes.

0 5 10 15 20 25Time [s]

0.0

0.2

0.4

0.6

0.8

1.0

mix

ing

scal

ar [-

] ExperimentReconstructionDeviationImpulse functions

Fig. 15 Reconstruction of a long tracer slug by super-position of Dirac pulses at the reactor inlet

VII. APPLICATION OF THE SAPR MODEL TO AHYPOTHETICAL BORON DILUTION EVENT

The SAPR model has been coupled with the 3Dreactor dynamics code DYN3D [20] to provide realisticboron concentration or temperature distributions asboundary conditions for transient analyses. TheDYN3D code has been developed in the Institute ofSafety Research of the Forschungszentrum Rossendorf.DYN3D has an extensive verification and validationbasis [21, 22] and is widely used for the analysis ofreactivity initiated accidents in light water reactors withCartesian and hexagonal fuel assembly cross sectiongeometry [23, 24].

The analysis presented here has been carried outfor the beginning of an equilibrium fuel cycle of ageneric four-loop pressurized water reactor. Themacroscopic cross section library needed for the corecalculations has been generated by the 2D neutrontransport code HELIOS. The library contains crosssection sets dependent from burn-up and the thermo-hydraulic feedback parameters in a range of variationbeing relevant for the transient under consideration. Thereactor is assumed at hot zero power in a subcriticalstate. The Xenon- and Samarium-distributioncorresponds to the full power state. All control rods areinserted, except the most effective, which sticks at fullywithdrawn position. The considered boron dilutionscenario is based on the analysis of the restart of thefirst main coolant (MCP) pump in a PWR after a steamgenerator tube rupture accident [25]. It was assumed inthis analysis, that a slug of boron-free coolant has beencreated in the main circulation loop and is driven intothe core by start-up of the first MCP. The coolant in thelower plenum has a temperature of 192 °C and a boroncontent of 2200 ppm. The temperature of the deborated

slug is 210 °C. The initial subcriticality of this state(before the restart of the circulation) is determinedwith -7787 pcm.

12 14 16 18 20 22 24Time [s]

0

2000

4000

6000

8000

Nuc

lear

Pow

er [M

W]

Fig.16 Nuclear power in the case of the 36 m3

unborated slug

The MCP reaches its full mass flow rate about 15 safter the start-up. The boron front reaches the core entryabout 12 s after switching-on the MCP. The borondilution in the reactor core due to the slug causes asuper prompt critical reactivity insertion leading to avery short power pulse with a magnitude of more than7000 MW (Fig. 16).

00.2

0.40.6

0.81.0

0

0.2

0.4

0.6

0.8

0

100

200

300

400

500

L

inea

r Rod

Pow

er [k

W/m

]

x [-]

y [-]

Fig. 17 Radial distribution of the linear rod powerdensity at the moment of maximum power

It is limited due to the strong, practically promptlyacting Doppler feedback of the fuel temperature. Thehalf width of the power peak was so small that asignificant enhancement of the coolant temperature did

not occur at that moment. Due to the further deboration,the positive reactivity insertion is continued after thefirst power peak. Typical secondary power peaks areobserved. As can be seen from figure 17, the radialpower distribution over the reactor core is very non-uniform. At the location of the power maximum (near tothe position of the sticking out control rod), coolantboiling with a maximum void fraction of up to 70 %occurs for a short time. However, no occurrence of heattransfer crisis was obtained, so that the claddingtemperatures remained below 260 °C. Safety marginsare not violated.

VIII. CFD CALCULATIONS

The code CFX-4 [26] is used for complementarycalculations in test facility and original reactor geo-metry. The coolant flow is assumed to be incom-

pressible. The standard k, ε turbulence model is used.The inlet boundary conditions (velocity, temperature,boron concentration etc.) are set at the inlet nozzles.The boundary conditions at the core outlet are pressurecontrolled. For the description of temperature, boronrespectively salt tracer concentrations, passive scalarfields are used. Special attention was paid to the exactrepresentation of the nozzle region with the curvatureradii at the four inlet nozzles, the passes of the hot legsthrough the downcomer and the diffuser-type extensionof the downcomer below the nozzles. Due to the greatimportance of this region for the coolant mixing, thegrid is very dense in these regions. The generated gridconsists of about 450000 nodes. The core support plate,the core and the perforated drum are modeled as aporous region. The porosity is determined by relatingthe area of orifices to the total area of the sieve.

a) Experimental results

b) Semi-analytical model SAPR

c) Calculation with CFX

Fig. 18 Time sequences of selected instantaneous mixing scalar distributions at the core entrance during pump start-up,36 m³ deborated coolant in loop 1, maximum indicated with bold number at corresponding fuel element position

The body forces taking into account distributedfriction losses in the sieve plate are determined to obtainequal pressure drop over the structures as they arederived from the detailed construction. Steady state

calculations with at least 1000 iterations last a fewhours, transient calculations a few weeks on an SGIOrigin 200 workstation.

0 5 10 15 20 25 30Time [s]

0.00

0.10

0.20

0.30

0.40

0.50

0.60

Mix

ing

Sca

lar

[-]

Experiment

2 Sigma

SAPR

CFX

Maximum

Fig. 19 Time history of the instantaneousely observedmaximum mixing scalar at the core entranceduring the pump start-up test

Fig. 20 Stream lines in the reactor pressure vesselcalculated by CFX-4 for the case of pumpstart-up in one of the loops

The result of a post-test calculation of the pumpstart-up experiment described above is given in figure18c. There are quantitative deviations of few percent

concerning the time history of the instantaneousmaximum mixing scalar (Fig. 19). Larger deviations arefound at a number of fuel element positions mainly inthe central region of the core. Furthermore, thecalculated distributions show a certain delay (≈1 s)compared to the experiment. The calculation results arein good correspondence to the experiment concerningthe main mixing phenomena, such as azimuthal positionof tracer appearance and the shape of the founddistributions. In general it was found that CFX isreproducing most of the measured results with asatisfactory reliability in case of the single-phase flowcharacteristic for the investigated mixing scenarios, aslong as density effects are absent. The latter caserequires additional code development and validationactivities.

CFD calculations are powerful tools to explainfluid dynamic effects by plotting stream lines. This wasdone in figure 20 for the pump start-up scenario. It isclearly visible that the coolant has the tendency to flowaround the core barrel. The flow is divided into twoparts, which meet again at the side opposite to theaffected inlet nozzle. This explains the shape of thetracer distributions presented in figure 18.

IX. ECC WATER INJECTION WITH DENSITYEFFECTS

During emergency core cooling the higher densityof the ECC water leads to a streak of cold water that isflowing downwards in the downcomer. This causesthermal loads on the RPV. Furthermore, in the case ofinadvertent injection of low borated ECC water, a borondilution transient would be initiated. The transient isthen determined by the resulting boron concentrationdistribution at the core inlet.

Due to the fact, that the test facility cannot beheated up, the necessary density differences weresimulated by adding sugar (glucose) to the water that isinjected into the cold leg. To observe the mixing of theECC water, this water was again labeled by smallamounts of sodium chloride, enhancing theconductivity. Generating density differences by highsalt concentrations is not possible, because themeasurement system is very sensitive and would besaturated at high salt concentrations.

The goal of the experiments presented in thecurrent work was the generic investigation of theinfluence of density differences between the primaryloop inventory and the ECC water on the mixing in thedowncomer. To separate the density effects from theinfluence of other parameters, a constant flow in theloop with the ECC injection nozzle was assumed in thisstudy. The mass flow rate was varied between 0 and15 % of the nominal flow rate, i.e. it was kept in themagnitude of natural circulation. The other pumps wereswitched off. The density difference between ECC and

loop water was varied between 0 and 10 %. Figure 21summarizes the boundary conditions of theexperiments. Altogether 20 experiments have beencarried out (dots in figure 21). In all experiments, thevolume flow rate of the ECC injection system was keptconstant at 1.0 l/s. The normalized density is defined asthe ratio between ECC water density and density offluid in the circuit. All other boundary conditions areidentical.

The experiments without density effects serve asreference experiments for the comparison. Fig. 22shows the time evolution of the tracer concentrationmeasured at the two downcomer sensors in anunwrapped view. The downwards directed red arrowindicates the position of the loop with the runningpump, in that case delivering 10 % of the nominal flowrate. At the upper downcomer sensor, the ECC water(injected in each experiment from t = 5 to t = 15 s)appears directly below the inlet nozzle. Due to themomentum created by the pump, the flow entering thedowncomer is divided into two streams flowing rightand left in a downwards directed helix around the corebarrel. At the opposite side of the downcomer, the twostreaks of the flow fuse together and move downthrough the measuring plane of the lower downcomersensor into the lower plenum. Almost the whole

quantity of ECC water passes the measuring plane ofthe lower downcomer sensor at the side nearly oppositeto the azimuthal position of the affected loop.

0.00 0.05 0.10 0.15Loop mass flow rate [-]

1.00

1.02

1.04

1.06

1.08

1.10

1.12

norm

aliz

ed E

CC

den

sity

[-]

Fr = 1.00

Fr = 4.00

Fr = 7.00

Fr = 10.0

Fig. 21 Test matrix of ECC injection experiments,indication of observed flow regime [27]

Fig. 22 Time evolution of the mixing scalar at the twodowncomer sensors in the experiment with 10 %loop flow rate and no density difference

Fig. 23 Time evolution of the mixing scalar at the twodowncomer sensors in the experiment with 10 %loop flow rate and 10 % density difference

The maximum tracer concentration of the ECCwater in the downcomer is 20 % of the injected waterconcentration at the upper sensor and 8 % at the lowersensor. Such a velocity field is typical for single-loop

operation. It has its maximum at the opposite side of thedowncomer and a minimum at the azimuthal position ofthe running loop, which has been found in velocity

measurements by means of a laser-Doppler anemometerat the ROCOM test facility [18], too.

Figure 23 shows the experiment, carried out at thesame flow conditions, but the density differencebetween the injected ECC water and the primary loopcoolant is now 10 %. In that case a streak formation ofthe water with higher density is observed. At the uppersensor, the ECC water covers a much smaller azimuthalsector. The density difference partly suppresses thepropagation of the ECC water in circumferencialdirection. The ECC water falls down in an almoststraight streamline and reaches the lower downcomersensor directly below the affected inlet nozzle. Onlylater, coolant containing ECC water appears at theopposite side of the downcomer. The maximumconcentration values observed at the two downcomersensors are significantly higher than in the case withoutdensity differences, i.e. 37.5 % and 20.0 % from theinitial concentration in the ECC water tank. Thevisualizations of the behavior of the ECC water in thedowncomer reveals that in case of momentum drivenflow, the ECC water covers nearly the whole perimeterof the upper sensor and passes the measuring plane ofthe lower sensor mainly at the opposite side of thedowncomer. When, on the other hand, density effectsare dominating, the sector at the upper measuringdevice covered by the ECC water is very small. TheECC water falls down straightly and passes the sensorin the lower part of the downcomer below the inletnozzle of the working loop.

Subsequently, variations of the density werecarried out to identify the transition region betweenmomentum driven and density driven flow. A densitydifference of 4 % was found to belong to anintermediate region. In this case, the ECC water reachesthe opposite side of the downcomer and the regionbelow the inlet nozzle position near the lower sensoralmost at the same time. That means, that one part ofthe ECC water follows the stream lines of the externalmomentum driven flow field and another part directlyfalls down due to the internal momentum created bydensity differences.

Based on these observations, the set of experimentsconducted according to the matrix in figure 21, wasdivided into three groups: density dominated flow (���momentum dominated flow (��� ���� �� ��� ����region (*). The conditions at the inlet into thedowncomer were used to calculate Froude-numbers ofthe experiments according to the following formula [6]:

inain

in

sg

vFrρ

ρ−ρ⋅⋅=

where vin is the velocity at the reactor inlet(combined loop and ECC flow), g is the gravitationalacceleration, s is the width of the downcomer, ρin thedensity of the incoming flow, calculated with the

assumption of homogeneous mixing between ECC andloop flow, and ρa the density of the ambient water in thedowncomer. Lines of constant Froude-numberscalculated by means of this formula are shown infigure 21. All experiments, identified as densitydominated are located in the region left of the iso-lineFr = 4.0 and all momentum dominated points are foundright of the iso-line Fr = 7.0. These two numbers arecritical Froude numbers separating the two flowregimes for the ROCOM test facility. Between the twocriterial values a transition region is located.

Fig. 24 Time evolution of the mixing scalar at the twodowncomer sensors in the experiment with 0 %loop flow rate and 10 % density difference

Density effects are extremely developed in anexperiment with stagnant coolant in the primary loop(Fig. 24), where the fluid circulation is initiated only bystarting the ECC injection (injection time was 40 s). Atthe upper sensor, the ECC water appears unmixed andcovers a sector of only about 15 degree. The data fromthe lower downcomer sensor show clearly buoyancyinduced turbulent structures. As can be concluded fromthese data, the water with higher density accumulates inthe lower plenum.

X. SUMMARY

ROCOM is a test facility for experimental coolantmixing studies over a wide variety of possiblescenarios. It is equipped with advanced instrumentation,which delivers high-resolution information character-izing either temperature or boron concentration fields in

the investigated pressurized water reactor. The obtainedhuge data basis and the fact that the coolant mixing wasstudied not only by characterizing the resultingdistributions at the core entrance, but also by highresolution measurements at two axial positions in thedowncomer, represent unique contributions to under-standing of coolant mixing in pressurized waterreactors.

Experimental results show that the mixing isincomplete, and that for different scenarios, thedistributions of the mixing scalar show large qualitativedifferences. The test results were successfully appliedfor CFD code validation. 3D simulations were carriedout with the code CFX-4.

An approach to generalize experimentally obtainedmixing scalars was developed on the basis of thesuperposition of system responses on Dirac impulse-like perturbations. After coupling the obtainedsimplified model with a three-dimensional neutronkinetics code (DYN3D) a hypothetical transient causedby pump start-up with a deborated slug in thecorresponding loop was analyzed. Despite extremelyconservative assumptions about the quantity of thedeborated coolant (36 m³) and additional unfavorableconditions, like stuck control rods etc., no fuel roddamage was theoretically predicted.

ACKNOWLEDGEMENT

The project this paper is based on is funded by theBMWi (Federal Ministry of Economics and Technologyof Germany) and is registered with No. 150 1216.

NOMENCLATURE

a index ambient

CB g/l Boron concentration

CFD abbr. computational fluid dynamics

ECC abbr. emergency core cooling

Fr 1 Froude number

g m/s² gravity constant

in index inlet

Re 1 Reynolds number

RPV abbr. reactor pressure vessel

s m width of downcomer

t s time

T deg C temperature

x m coordinate

y m coordinate

z m coordinate

θ 1 mixing scalar

ρ kg/m³ density

σ m²/s surface tension

REFERENCES

1. U. Grundmann, U. Rohde: Investigations on a borondilution accident for a VVER-440 type reactor bythe help of the code DYN3D, ANS Topical Meetingon Advances in Reactor Physics, Knoxville, TN,April 11-15, 1994, proc. vol. 3, pp. 464-471.

2. D. Alvarez et al.: Three dimensional calculationsand experimental investigations of the primarycoolant flow in a 900 MW PWR vessel, Proc.NURETH-5, 1992, vol. II, pp. 586-592.

3. Alavyoon, B. Hemström, Andersson, R. Karlsson:Experimental and computational approach toinvestigating rapid boron dilution transients inPWRs, CSNI Specialist Meeting on Boron DilutionReactivity Transients, State College, Pennsylvania,USA, 1995, October 18-20, 1995.

4. O. I. Melikhov, V. I. Melikhov, S. E. Yakush, Yu.A. Bezrukov, S. A. Logvinov: Analysis of borondilution in VVER-1000 reactor, Proc. Ann. Meetingon Nucl. Technology 2001, pp. 117–120,INFORUM GmbH, Bonn, Germany.

5. V. N. Ulyanovsky, Yu. A. Bezrukov et al.: Themixing of coolant with different boron concentrationat the VVER-1000 core inlet during RCP start-up,Proc. Ann. Meeting on Nucl. Technology 2000, pp.115–118, INFORUM GmbH, Bonn, Germany.

6. HDR safety program – thermal mixing in the coldleg and downcomer of the HDR test rig, ReportPHDR 91-89, FZ Karlsruhe, 2000, Germany.

7. K. Umminger, W. Kastner, J. Liebert, T. Mull:Thermal Hydraulics of PWR’s with Respect toBoron Dilution Phenomena: Experimental Resultsfrom the Test Facilities PKL and UPTF, Ninth Int.Topical Meeting on Nuclear Reactor ThermalHydraulics (NURETH–9), San Francisco,California, USA, 1999.

8. M. Gavrilas and K. Kiger: ISP-43: Rapid BoronDilution Transient Experiment, Comparison Report,NEA/CSNI/R(2000)22.

9. B. Woods: UM 2x4 Loop Experimental Findings onthe Effect of Inertial and Buoyancy forces onAnnular Flow Mixing for Rapid Boron DilutionTransients, Ph.D. Thesis, University of Maryland,USA, 2001.

10. Höhne, T.: Vergleich von Kühlmittelströmung und -vermischung in einem skalierten Modell des DWRKonvoi mit den Vorgängen im Originalreaktor,FZR-Report FZR-210, 1998.

11. Dräger, P., Ackermann, G., Siklossy, P.:Theoretische und experimentelle Ermittlung derTemperaturverteilung in der Spaltzone des WWER-440, Kernenergie 28, 7, 293-297, 1985

12. Dräger, P.: Makroskopische Kühlmittelvermischungin Druckwasserreaktoren, PhD Thesis, TH Zittau,1987.

13. I. D. Johnson: Method and Apparatus for MeasuringWater in Crude Oil, United States Patent, No4,644,263, Date of Patent: Feb. 17, 1987.

14. N. Reinecke, M. Boddem, G. Petritsch, D. Mewes:Tomographisches Messen der relativen Phasen-anteile in zweiphasigen Strömungen fluider Phasen,Chem. Ing. Tech. 68 (1996) 11, S. 1404-1412.

15. H.-M. Prasser, A. Böttger, J. Zschau (1998), A newelectrode-mesh tomograph for gas-liquid flows,Flow Measurement and Instrumentation 9, 111-119.

16. H.-M. Prasser, J. Zschau, D. Peters, G. Pietzsch, W.Taubert, M. Trepte: Fast wire-mesh sensors for gas-liquid flows - visualisation with up to 10 000 framesper second, International Congress on AdvancedNuclear Power Plants (ICAPP), June 9-13, 2002 -Hollywood Florida, USA, paper 1055.

17. G. Grunwald, Th. Höhne, H.-M. Prasser, F.-P. Weiß(2000), Coolant mixing in PWR at constant coolantflow in the primary circuit, Proc. Annual Meeting onNuclear Technology ‘00, pp. 109 – 114, INFORUMGmbH, Bonn.

18. G. Grunwald, Th. Höhne and H.-M. Prasser (2001),Laser-Doppler-velocity measurements in thedowncomer of the ROCOM test facility incomparison with numerical results, Report FZR-334, ISSN 1437-322X, Rossendorf (in German)

19. S. Kliem and H.-M. Prasser (2001), A semi-analyt-ical model for the description of coolant mixing in-

side the KONVOI reactor pressure vessel duringpump start-up, Proc. Annual Meeting on NuclearTechnology ‘01, pp. 107 – 112, INFORUM GmbH,Bonn.

20. U. Grundmann, U. Rohde (1996), DYN3D – A 3-dimensional core model for steady-state andtransient analysis in thermal reactors , Proc. Int.Conf. on the Physics of Reactors PHYSOR’96 (pp.J70 – J79).

21. U. Grundmann, U. Rohde, “Verification of the CodeDYN3D/R with the Help of InternationalBenchmarks,” Forschungszentrum Rossendorf,Report FZR-195 (1997).

22. R. Kyrki-Rajamäki, U. Grundmann, A. Kereszturi,“Results of Three-Dimensional Hexagonal DynamicBenchmark Problems for VVER Type Reactors.,”Proc. Int. Conf. on the Physics of ReactorsPHYSOR96, pp. J229 - J238 (1996).

23. K.N. Ivanov, U. Grundmann, S. Mittag, U. Rohde,Comparative Study of a Boron Dilution Scenario inVVER Reactors. Ann. Nucl. Eng. 26, 1331-1339(1999).

24. W. Tietsch, J. Tischer, “Analysis of Boron Dilutionduring Small Break LOCAs in a 2x4 Loop B&WType PWR,” In: K.G. Bauer (Ed.), Proc. Ann.Meeting on Nuclear Technology, Inforum GmbH,Bonn, Germany, pp. 143-147, (1999).

25. R. Reinders, “Three-Dimensional Calculation of theBoron Concentration in the Reactor Pressure Vesselafter a SG Tube Rupture Accident,” (in German),Proc. Annual Meeting on Nuclear Technology ’98.107 – 110, INFORUM GmbH, Bonn, Germany(1998).

26. CFX-4.3 Flow Solver User Guide, AEA Technology(1999).

27. S. Kliem, H.-M. Prasser, G. Grunwald, U. Rohde etal. (2002), ROCOM Experiments on the influence ofdensity differences on the coolant mixing inside thereactor pressure vessel of a PWR, Annual Meetingon Nuclear Technology ’02, accepted.