Annals of Nuclear Energy 37 (2010) 1036–1047

Contents lists available at ScienceDirect

Annals of Nuclear Energy

journal homepage: www.elsevier .com/locate /anucene

Experimental and calculation results of the integral reflood test QUENCH-14with M5� cladding tubes

J. Stuckert a,*, J. Birchley b, M. Grosse a, B. Jaeckel b, M. Steinbrück a

a Karlsruhe Institute of Technology (KIT), Hermann-von-Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germanyb Paul Scherrer Institut (PSI), CH-5232 Villigen PSI, Switzerland

a r t i c l e i n f o

Article history:Received 4 February 2010Received in revised form 14 April 2010Accepted 22 April 2010Available online 14 May 2010

Keywords:Core refloodHydrogen source termNiobium-bearing cladding material

0306-4549/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.anucene.2010.04.015

* Corresponding author. Tel.: +49 7247 82 2558; faE-mail address: juri.stuckert@kit.edu (J. Stuckert).

a b s t r a c t

The QUENCH-14 experiment investigated the effect of M5� cladding material on bundle oxidation andcore reflood, in comparison with tests QUENCH-06 (ISP-45) that used standard Zircaloy-4 andQUENCH-12 that used VVER E110-claddings. The PWR bundle configuration of QUENCH-14 with a singleunheated rod, 20 heated rods, and four corner rods was otherwise identical to QUENCH-06. The test wasconducted in principle with the same protocol as QUENCH-06, so that the effects of the change of clad-ding material could be observed more easily. Pre-test calculations were performed by the Paul ScherrerInstitut (Switzerland) using the SCDAPSIM, SCDAP/RELAP5 and MELCOR codes. Follow-on post-test anal-yses were performed using SCDAP/RELAP5 and MELCOR as part of an ongoing programme of model val-idation and code assessment. Alternative oxidation correlations were used to examine the possibleinfluence of the M5� cladding material on hydrogen generation, in comparison with Zircaloy-4.

The experiment started with a pre-oxidation phase in steam, lasting �3000 s at �1500 K peak bundletemperature. After a further temperature increase to maximum bundle temperature of 2073 K the bundlewas flooded with 2 g/s/rod water from the bottom. The peak temperature of �2300 K was measured onthe bundle shroud, shortly after quench initiation. The electrical power was reduced to average value of2 W/cm during the reflood phase to simulate effective decay heat level. Complete bundle cooling wasreached in 300 s after reflood initiation.

The development of the oxide layer growth during the test was essentially defined by measurementsperformed on the three Zircaloy-4 corner rods withdrawn successively from the bundle. The withdrawalof Zircaloy-4 and E110 corner rods after the test allowed a comparison of the different alloys in one test.One heated rod with M5 cladding was withdrawn after the test for a detailed analysis of oxidation degreeand measurement of absorbed hydrogen.

Post-test examinations showed neither breakaway cladding oxidation nor noticeable melt relocationbetween rods. Different from the QUENCH-14 (M5�) findings, the QUENCH-12 test with the E110 clad-dings performed under similar conditions had resulted in intensive breakaway effect at cladding andshroud surfaces during the pre-oxidation phase and local melt relocation on reflood initiation.

The hydrogen production in QUENCH-14 up to reflood was similar to QUENCH-06 and QUENCH-12bundle tests. During reflood 6 g hydrogen were released which is similar to QUENCH-06 (4 g) but muchless than during quench phase of QUENCH-12 (24 g). The reason for the different behaviour of the Zr1%Nbcladding alloys is the different oxide scale properties of the materials.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The most important accident management measure to termi-nate a severe accident transient in a Light Water Reactor (LWR)is the injection of water to cool the uncovered degraded core.The main objective of the QUENCH programme is investigation ofhydrogen production that results from the water or steam interac-tion with overheated elements of fuel assembly. Other ultimate

ll rights reserved.

x: +49 7247 82 2095.

goals of programme are to identify the limits (temperature, injec-tion rate, etc.) for which successful reflood and quench can beachieved (Hering and Homann, 2007b; Steinbrück et al., 2010)and to compare recently used cladding materials.

Zircaloy-4 was used as standard rod cladding material in twelveof fifteen QUENCH experiments performed to date (Sepold et al.,2007; Sepold et al., 2009). One bundle experiment, i.e. QUENCH-12, was performed with Nb-bearing E110 cladding material inVVER geometry (Stuckert et al., 2009). QUENCH-14 was the firstexperiment in the frame of the advanced cladding materials(ACM) test series and investigated the effect of M5� cladding

Fig. 1. QUENCH facility: containment and test section.

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1037

material (industrial product made by AREVA) on bundle oxidationand core reflood. Comparison is made with test QUENCH-06 (ISP-45) that used standard Zircaloy-4 (Sepold et al., 2004; Heringet al., 2002) and is a reference case for the ACM series. Informationon the chemical composition of Nb-bearing M5� cladding materialis given e.g. in eferences (Peng et al., 2005; Duriez et al., 2008).Since then a second ACM experiment, QUENCH-15 was conductedusing Zirlo™ cladding.

The QUENCH-14 experiment was successfully conducted at theForschungszentrum Karlsruhe on 2 July 2008 with the help of pre-test calculations performed by the Paul Scherrer Institute (PSI)using SCDAPSIM, SCDAP/RELAP5 and MELCOR, modified locally asnecessary for M5� oxidation kinetics based on separate-effectsdata from the QUENCH program (Steinbrück, 2008; Grosse, 2008).

With respect to the oxide layer thickness it has to be noted thatthe M5� cladding material has been optimized for usual opera-tional temperatures (maximum of �620 K for outer surface and�720 K for inner surface). In this temperature range the oxide layerthickness of M5� cladding is significantly smaller when comparedto Zircaloy-4 (Garner et al., 2006). The tendency of lower oxidationfor M5� remains unaffected up to �1300 K (Steinbrück, 2008). Thisdifference between those two materials is even increased near1300 K for thick oxides because spalling of oxide scales (breakawayeffect) is typical for Zircaloy-4 due to a phase transition in theoxide. At temperatures higher than 1300 K the oxidation rate ofboth alloys changes to similar values.

The breakaway oxidation with formation of oxide micro-crackscould have developed in the QUENCH-06 test during the pre-oxida-tion phase at those bundle elevations extensively exposed to tem-peratures �1300 K. However, Zircaloy-4 claddings of this referencebundle did not show evidence of oxide scale spalling because of amoderate degree of oxidation during pre-oxidation. Contrary toQUENCH-06, the counterpart test QUENCH-12 with E110 (Zr1%Nb)claddings and E125 (Zr2.5%Nb) shroud, showed very intensiveoxide scale spalling consequently with increased cladding hydrid-ing and intensive hydrogen release during flooding. Hydrideembrittlement of bundle elements can cause severe bundle dam-ages under particular conditions e.g. in a spent storage pool, whichwas observed in an accident with a special cleaning tank in Paks(OECD, 2008). The performance of the QUENCH-14 bundle test en-abled to check the advantages of M5� material under specified se-vere accident conditions.

Fig. 2. Cross section of the QUENCH-14 bundle.

2. QUENCH facility

The main component of the QUENCH test facility is the test sec-tion with the test bundle (Fig. 1). The facility can be operated intwo modes: a forced-convection mode (typical for most QUENCHexperiments) and a boil-off mode. QUENCH-14 was conducted inforced-convection mode, in which, superheated steam from thesteam generator and superheater together with argon as a carriergas for off-gas measurements enter the test bundle at the bottom.The system pressure in the test section is around 0.2 MPa absolute.The test section has a separate inlet at the bottom to inject waterfor reflood (bottom quenching). The argon, the steam not con-sumed, and the hydrogen produced in the zirconium-steam reac-tion flow from the bundle outlet at the top through a water-cooled off-gas pipe to the condenser where the steam is separatedfrom the non-condensable gases. The water cooling circuits forbundle head and off-gas pipe are temperature-controlled to guar-antee that the steam/gas temperature is high enough so that con-densation at the test section outlet and inside the off-gas pipe isavoided.

The test bundle is approximately 2.5 m long and is made up of21 fuel rod simulators (Fig. 2). The fuel rod simulators are held in

position by five grid spacers, four are made of Zircaloy-4 and theone at the bottom of Inconel 718. Except the central one all rodsare heated. Heating is electric by 6 mm diameter tungsten heatersof length 1024 mm installed in the rod centre (lower edge of heat-ers corresponds to bundle elevation 0 mm). Electrodes of molybde-num (length 300 + 576 mm; Ø 8.6 mm) and copper (length390 + 190 mm; Ø 8.6 mm) are connected to the tungsten heatersat one end and to the cable leading to the DC electrical power sup-ply at the other end. The heating power is distributed between twogroups of heated rods. The distribution of the electric power withinthe two groups is as follows: about 40% of the power is releasedinto the inner rod circuit consisting of eight fuel rod simulators

1038 J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047

(in parallel connection) and 60% in the outer rod circuit (12 fuel rodsimulators in parallel connection). The tungsten heaters are sur-rounded by annular ZrO2 pellets. The rod cladding of the heatedand unheated fuel rod simulators is M5� with 10.75 mm outsidediameter and 0.725 mm wall thickness. All test rods are filled withKr at a pressure of approx. 0.22 MPa absolute. The rods were con-nected to a controlled feeding system that compensated minor gaslosses and allowed observation of the first cladding failure as wellas a failure progression.

There are four corner rods installed in the bundle. Two of them,i.e. rods ‘‘A” and ‘‘C”, are made of a solid rod at the top and a tube atthe bottom and are used for thermocouple instrumentation. Cornerrod A is a Zircaloy-4 rod whereas corner rod C is made of E110(Zr1%Nb). The other two rods, i.e. rods ‘‘B” and ‘‘D” (solid Zirca-loy-4 rods of 6 mm diameter) are particularly designed to be with-drawn from the bundle to check the amount of ZrO2 oxidation andhydrogen uptake at specific times. In QUENCH-14, for the firsttime, all four corner rods were used for analysis of the oxide layerthickness. Rod B was pulled out of the bundle after the pre-oxida-tion phase of the experiment and rod D was pulled before quench-ing; rods A and C were removed after test termination. One heatedrod (#16) with M5� cladding was withdrawn from the bundle afterthe test, too, for axial scanning of oxide and absorbed hydrogendistributions.

The test bundle is surrounded by a 3.25 mm shroud of Zircaloy-4 with a 37 mm thick ZrO2 fiber insulation extending from the bot-tom to the upper end of the heated zone and a double-walled cool-ing jacket of stainless steel over the entire length. The annulusbetween shroud and cooling jacket is purged (after several cyclesof evacuation) and then filled with stagnant argon at 0.22 MPaabsolute. The annulus is connected to a flow- and pressure-con-trolled argon feeding system in order to keep the pressure constantat the target of 0.22 MPa and to prevent an access of steam to theannulus after shroud failure. The 6.7-mm annulus of the coolingjacket is cooled by argon flow from the upper end of the heatedzone to the bottom of the bundle and by water in the upper elec-trode zone. Both the absence of a ZrO2 insulation above the heatedregion and the water cooling are to avoid too high temperatures ofthe bundle in that region.

The off-gas including Ar, H2 and H2O is analyzed by a state-of-the-art mass spectrometer Balzers ‘‘GAM300” located at the off-gas pipe �2.66 m downstream the test section. The mass spec-trometer allows also to indicate the failure of rod simulators bydetection of Kr release.

The test bundle, shroud, and cooling jacket are extensivelyequipped with sheathed thermocouples at different elevationswith an axial step of 100 mm. There are 40 high-temperature(W/Re) thermocouples in the upper hot bundle region (eleva-tions between 650 and 1350 mm) and 35 low-temperature(NiCr/Ni) thermocouples in the lower ‘‘cold” bundle region (be-tween �250 and 550 mm). At elevations 950 and 850 mm thereare two centreline high-temperature thermocouples in the cen-tral rod, which are protected from oxidising influence of thesteam. Two thermocouples isolated from steam are installedat the same elevations inside the corner rods A and C. Otherbundle thermocouples are attached to the outer surface of therod cladding. The shroud thermocouples are mounted at theouter surface of Zircaloy-4 shroud. Additionally the test sectionincorporates pressure gauges, flow meters, and a water leveldetector.

Fig. 3. Test scenario of bundle tests QUENCH-06, -12, -14.

3. Test conduct and results of on-line measurements

The QUENCH-14 test phases were as follows (the temperaturegiven is the one at hottest elevation):

PhaseI

Stabilisation at �873 K

PhaseII

Heatup with �0.3–0.6 K/s to �1500 K

PhaseIII

Pre-oxidation in a flow of 3 g/s of superheated steamand 3 g/s argon for �3000 s at relatively constanttemperature of �1500 K. Withdrawal of corner rod Bat the end

PhaseIV

Transient heatup with 0.3. . .2.0 K/s from �1500 to�2050 K in a flow of 3 g/s of superheated steam and3 g/s argon. Withdrawal of corner rod D �30 s beforequench initiation

PhaseV

Quenching of the bundle by a flow of �41 g/s ofwater

The test was conducted in principle with the same protocol asQUENCH-06 and QUENCH-12 (Fig. 3). The hottest zone during thewhole test was the bundle region between 950 and 1050 mm (Figs.4 and 5). The axial temperature profile in these figures shows an in-crease from �250 to 1000 mm and a decrease for the upper part ofthe bundle which is caused by increased radial heat radiation lossesin region due to the uninsulated shroud from 1000 mm upwards.

A moderate temperature excursion was observed following thebeginning of reflood initiation in all three tests, as compared witheach other in Fig. 6. The peak temperature reached during the con-duct of QUENCH-14, �2300 K, was measured on the shroud at ele-vation of 950 mm, shortly after the quench initiation. Similarlybrief shroud temperature escalations were observed in QUENCH-06 and QUENCH-12.

The quench criterion and reflood rate were identical to those inQUENCH-06. The flooding with 41 g/s water at room temperaturewas initiated together with the fast water injection to fill the lowerplenum. The electrical power was reduced to 3.9 kW during the re-flood phase, approximating effective decay heat levels. The propa-gation of quenching by the two-phase fluid was determined fromthe times of rapid decrease to saturation temperature measuredby the surface thermocouples at the different bundle elevations.The duration of complete bundle cooling was about 300 s (Fig. 7)and thus similar to those in QUENCH-06 and QUENCH-12.

On-line measurement of hydrogen production during theQUENCH-14 test revealed 34 g in the pre-oxidation and transientphases, and 6 g in the quench phase. The total amounts releasedare similar to those in QUENCH-06, i.e. 32 g and 4 g, respectively,but less compared to QUENCH-12 with 34 and 24 g, respectively(Fig. 8).

Fig. 4. Axial temperature profiles before transient.

Fig. 5. Axial temperature profiles before flooding.

Fig. 6. Moderate temperature escalation at the end of transient.

Fig. 7. QUENCH-14 bundle cooling progress.

Fig. 8. Progress of hydrogen production during tests QUENCH-06, -12, and -14.

Fig. 9. QUENCH-14 corner rods withdrawn from the bundle.

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1039

4. Post-test appearance

Inspection of four corner rods withdrawn from the QUENCH-14bundle at the end of (1) pre-oxidation (rod B), (2) transient (rod D)and (3) test (rods A and C) clearly demonstrate severe breakaway

oxidation at corner rod C made of E110 (Zr1%Nb) which is not seenat the surfaces of the other three corner rods (A, B and D) made ofZircaloy-4 (Fig. 9). The peak oxide layer thicknesses for these Zirca-loy-4 rods were measured at elevation �950 mm with the follow-ing values: 180 lm at the end of the pre-oxidation phase (rod B),360 lm before reflood (rod D) and about 700 lm after the test(rod A).

Before disassembly of the QUENCH-14 shroud/test bundle unitthe empty channels of four withdrawn corner rods were visuallyinspected. The post-test endoscopy of the bundle was performedwith an OLYMPUS IPLEX videoscope. The endoscopy showed nei-ther breakaway cladding oxidation nor noticeable melt formation.Some light deposits on the cladding surface were observed, whichprobably originated from interaction with parts of relocated downoxidised grid spacer GS4 from elevation of 1050 mm (Fig. 10).

The post-test appearance of the QUENCH-14 bundle supportsthe findings of the separate-effects tests with short specimens(Steinbrück, 2008; Grosse, 2008) that even at elevated tempera-

Fig. 11. Elevation 50 mm: homogeneous oxide layer for QUENCH-14 (top) andspalling of outer oxide scale for QUENCH-12 (bottom).

Fig. 12. Comparison of axial oxide layer profiles for bundles QUENCH-14, QUENCH-12 and QUENCH-06.

1040 J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047

tures there is less spalling of the oxide scale for M5�, evident forrod No. 16 which was removed from the bundle prior to encapsu-lation of the test bundle, compared to standard Zircaloy-4 used inthe QUENCH-06 bundle. Spalling was strongest in the QUENCH-12test bundle with E110 cladding demonstrating extensive break-away oxidation (Stuckert et al., 2009). This behaviour was notfound in the inspection of rod No. 16 from the QUENCH-14 testbundle (Fig. 11).

The final oxide layer thickness of the QUENCH-14 claddings wasevaluated from post-test examination. Below 850 mm theQUENCH-14 cladding oxidation degree is less than the correspond-ing cladding oxidation for the QUENCH-06 bundle (Fig. 12). Above900 mm the oxide layer thickness of the QUENCH-14 cladding isbetween the corresponding values for the QUENCH-12 andQUENCH-06 bundles. The reason for the increased oxide layerthicknesses of the QUENCH-12 cladding at hot bundle elevationswas breakaway damage of the oxide layer during the previouspre-oxidation phase, whereas thicker oxide layers of the hotter re-gions of the QUENCH-14 bundle develop, as illustrated in Fig. 5,due to higher temperatures.

Development and relocation of the metal melt into the spacesbetween rods is a very important factor for possible increasedhydrogen production due to intensive interaction between meltand steam. Melting of the cladding internal metallic layer was ob-served for some fuel rod simulators at the QUENCH-14 bundle ele-vations between 900 and 1100 mm. The QUENCH-14 bundle cross-section at the hottest elevation of 950 mm shows practically nomolten pool formation in the space between rod simulators(Fig. 13). Contrary to the QUENCH-12 test and similar to theQUENCH-06 test (Fig. 14), the melt was localised between the pel-let and the cladding external layer. There are also no shroud meltpenetrations into the QUENCH-14 bundle interspace. Correspond-ingly, no sharp increase of hydrogen release was observed on re-flood initiation for QUENCH-14 and QUENCH-06, contrary toQUENCH-12 (Fig. 8).

The axial distributions of the hydrogen absorbed in the four cor-ner rods and cladding tube of rod 16 were determined by quanti-tative analysis of neutron radiographs (Grosse et al., 2006). Onlya small amount of hydrogen was absorbed in the corner rod Bwithdrawn after pre-oxidation phase. A maximal value of about1.3 at.% was found at elevation 950 mm. Rod D withdrawn beforequenching shows significantly higher hydrogen concentrations;at elevation 970 mm and above 1100 mm maximum values of 3at.% were determined. For rod A withdrawn after the test a maxi-mum value of about 7.5 at.% was determined at elevation880 mm. Large differences in the hydrogen absorption during the

Fig. 10. QUENCH-14 endoscopic view at elevation 1000 mm showing interactionbetween cladding and relocated spacer grid.

test were found amongst materials M5�, Zry-4 and E110(Fig. 15). In corner rod C (E110 alloy) hydrogen was absorbed overa wide axial range. The maximum value in this rod (�32 at.%) isabout four times higher than in the Zry-4 rod (7.5 at.%) and aboutone order of magnitude higher than in the M5� cladding tube. Thereason for the different hydrogen uptake of the three materials isthat the breakaway of the oxide scale open pathways by whichhydrogen can readily penetrate through to the unoxidised metal.Breakaway occurred strongly on corner rod C (E110 alloy) butnot significantly on rod A (Zircaloy-4 alloy) or on the M5� claddingtube. A detailed account of cladding hydriding due to breakaway(Grosse et al., 2009) and the influence of (1) cladding outer-surfaceprocessing and (2) different bulk impurities (Ca, Al, Mg, F) onbreakaway oxidation is presented in (Billone et al., 2008).

5. Computer analyses of QUENCH-14

5.1. Analytical tools used

Versions of the MELCOR code package (Gauntt et al., 2005) havebeen developed by Sandia National Laboratories for nuclear plant

Fig. 13. Cross-sections from tests QUENCH-14, QUENCH-12 and QUENCH-06 at thehottest bundle elevation of 950 mm.

Fig. 14. Comparison of melt localisation in bundle tests QUENCH-14, QUENCH-12and QUENCH-06 at 950 mm.

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1041

safety analysis under the sponsorship of the US Nuclear RegulatoryCommission (US NRC). MELCOR is used extensively worldwide andversion 1.8.5 has been used by PSI as the primary system-levelcode for plant application in Switzerland. Version 1.8.6 was re-cently designated as the current production version and during

the last few years has been assessed in readiness for application.The simplified treatment of physical processes by many of theMELCOR models makes it necessary to perform back-to-back com-parison with empirical data and other code systems. For example, apartial two-fluid formulation in which the phases are essentiallyseparated is used to model the two-phase thermal–hydraulics. In

Fig. 15. Hydrogen uptake determined for different rods withdrawn after theQUENCH-14 test.

Fig. 16. Mass gain measurements for different cladding materials.

1042 J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047

contrast SCDAP/RELAP5 (Siefken et al., 1997) uses a full two-fluidtreatment together with a more complete treatment of the early-phase core degradation processes than MELCOR; it has been widelyused in planning and post-test analyses of the QUENCH experi-ments, and the models are extensively benchmarked (Sepoldet al., 2004; Birchley et al., 2006; Birchley et al., 2008). For the pres-ent analyses, MELCOR 1.8.6 and SCDAP/RELAP5 were used in tan-dem, with the latter used as the lead tool in order to providefurther comparison for assessment of MELCOR. A variant ofSCDAP/RELAP5/MOD3.2 is used here, which includes dedicatedmodels for QUENCH developed by Hering (Hering and Homann,2007a). Extensions to the MELCOR code to enable QUENCH simula-tion were developed by Cole (Cole, 2001). Previous MELCOR anal-yses of QUENCH performed at PSI have been carried out usingversion 1.8.5 (Birchley et al., 2005).

The SCDAP/RELAP5 input model for QUENCH-14 is unchangedfrom the one used for analyses of previous QUENCH experiments.The model comprises a single radial and 16 axial hydraulic nodesfor the test section with ten nodes for the 1 m tungsten heatedlength, and extends to include the upper region and off-gas pipe.The cooling jackets are represented by separate hydraulic systemsand containment is represented as a boundary condition. Althoughthe number of axial nodes comparable with many plant models,the node length is shorter, as dictated by the need to resolve thesteep temperature profile in QUENCH. The MELCOR 1.8.6 inputwas adapted from the 1.8.5 model by means of the supplied con-verter followed by some manual modification to achieve compati-bility of the lower volume with the input specification of the 1.8.6version. A simpler hydraulic noding is used with MELCOR thanwith SCDAP/RELAP5, with just four axial nodes for the tungstenheated length but with ten core component nodes to resolve theaxial profile. The comparative levels of detail reflect customarypractice in many applications.

A point about QUENCH is the significant fraction of total electri-cal power that is dissipated by the electrical resistance of the exter-nal circuitry and its contact with the copper electrodes. This isrepresented (with all codes) by a user-chosen constant. Since theresistance is not known and can vary from test to test, perhaps alsoduring a test, sensitivity studies were performed also with respectto its value.

It is known that the hydrogen generation and likelihood of anexcursion during reflood can depend strongly on the choice of oxi-dation correlation. As well as the default of MELCOR (Urbanic andHeidrick, 1978), several alternatives were used, including a recentprovisional correlation for M5� (Grosse, 2008) (further referred toas GM5) to characterise better the oxidation in QUENCH-14 and to

assess any potential sensitivity. As discussed earlier, one of thepurposes of the QUENCH-ACM experiments is to examine the ef-fect of different cladding materials on the oxidation behaviour.For this reason three of the corner rods (A, B, D) were made fromZry-4 and one from E-100 (C). Unavailability of M5� in suitableform precluded its use for the corner rods and the shroud whichwas also made from Zry-4. Since the shroud presents a significantfraction of the oxidisable surface, the overall oxidation behaviourin QUENCH-14 is a combination of the two materials. Neither theSCDAP nor MELCOR codes can represent different cladding materi-als within the same model. The best that can be done is to performseparate calculations with different oxidation criteria.

SCDAP/RELAP5 uses the pairing of Cathcart-Pawel (CP) (Pawelet al., 1979) and Urbanic–Heidrick (UH) correlations for oxidationin the temperature regimes below 1853 K (low/intermediate) andabove 1873 K (high), with linear interpolation in between. Thecombination is part of the MATPRO material property library (Sief-ken et al., 1997), volume 4. Since model changes cannot be madevia input, local code versions were developed with the Leistikow(L)/Prater-Courtright (PC) correlation (Schanz et al., 2004) and withthe GM5/UH pair of correlations applied in the low/intermediate(T < 1853 K) and high (T > 1873 K) regimes respectively. It is notedthat the L correlation (Schanz et al., 2004) yields similar kinetics toCP, although PC kinetics are much faster than UH in the high re-gime. The L/PC model was therefore used to examine the kineticsabove 1873 K. The GM5 correlation was obtained from data inthe range 1073–1673 K and formulated as two branches, with a‘‘step” at 1323 K which is approximately the transition tempera-ture between the monoclinic and tetragonal phases of the oxide,as shown in Fig. 16. Similarly reduced kinetics were also observedat the lower temperatures in experiments with Zry-4, Duplex andE110. The Grosse data do not extend into the high temperature re-gime above 1853 K. In the SCDAP implementation both brancheswere implemented and the correlation merged with MATPRO (i.e.UH) at the higher temperature. A variant of MATPRO was alsoimplemented with the Zry-4 low temperature branch from (Gros-se, 2008) replacing CP at T < 1300 K.

MELCOR allows just two branches to be input, so it was not pos-sible to fully represent all three regimes: GM5(low), GM5(high),and T > 1853 K. Therefore the model was applied in two differentways: (i) the two branches of the GM5 model with smoothing be-tween 1300 and 1350 K; (ii) the upper branch at T < 1800 K andeither PC or UH at T > 1900 K (with interpolation in between) inthe attempt to cover the full range of temperatures. Case (i) coversthe entire period of pre-oxidation and in particular the slower oxi-dation rate at T < 1300, while case (ii) captures the faster kinetics atT > 1800 K but not the slower oxidation at lower temperatures.

Fig. 17. MELCOR 1.8.6 calculation of QUENCH-06 bundle temperature at 950 mm, with alternative oxidation models.

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1043

The present analysis is the first application by PSI of the MEL-COR 1.8.6 code version to QUENCH. An input model was developedby converting the MELCOR 1.8.5 QUENCH model according to theversion 1.8.6 input specifications and benchmarked against theQUENCH-06 data. Good agreement was obtained for the importantsignatures without significant changes to the input, as shown inFig. 17. The same input model was then used in the planning anal-ysis for QUENCH-14.

5.2. Analyses of the QUENCH-14 results

As part of the of planning support for conduct of QUENCH-14,and also as part of the preparation for analyses of the experimentalresults, pre-test analyses were performed using both the MELCORand SCDAP/RELAP5 codes (Birchley et al., 2009). The planning ef-fort involved best-estimate calculations to determine a nominalpower likely to provide the target conditions, and sensitivity stud-ies with respect to uncertainties such as oxidation kinetics andexternal resistance to identify important thresholds and indicatemargins.

Fig. 18. MELCOR 1.8.6 calculation of QUENCH-14 shroud temperat

Post-test calculations were performed with MELCOR 1.8.6 andSCDAP/RELAP5, using the same best-estimate input models as usedin the pre-test planning analyses. The experimental boundary con-ditions for electrical power and inlet flows were used in the basecase calculations, imposed as functions of time except that refloodinitiation was specified according to the same maximum tempera-ture conditions as in the experiment. This was done to provide con-ditions at the start of reflood as close as possible to the experiment.

In QUENCH-14 the highest temperature was measured on theshroud, and reflood was initiated at 7213 s at which time the tem-perature at the 950 mm elevation was 2120 K. Since the shroudthermocouples are the most reliable, that location was used forthe comparison. Figs. 18 and 19 compare the measured tempera-tures and hydrogen generation with MELCOR using three applica-tions of GM5: low/high, GM5(high)/UH and GM5(high)/PC. Ineach case the injection was initiated when the peak temperatureof 2120 K was reached at 7230 s, i.e. only very slightly later thanin the experiment.

There was little difference between the calculated temperaturehistories since the structure is assumed to collapse upon melting

ure at 950 mm, with alternative forms of the GM5 correlation.

Fig. 19. MELCOR 1.8.6 calculation of QUENCH-14 total hydrogen generation, with alternative forms of the GM5 correlation.

1044 J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047

and no longer participates in the oxidation, while the calculatedtemperature of the ZrO2 insulation continues to increase. The firstoption gave the best agreement for hydrogen generation, while theother options gave (similar) moderate overestimates, withGM5(high)/PC showing a faster generation rate during the hightemperature phase.

Comparisons of the same measured temperatures with SCDAP/RELAP5 are shown in Fig. 20. Alternative oxidation correlationswere used, namely GM5(low/high)/UH, identified as the best esti-mate, and L/PC. The best estimate case, using GM5/UH, generallyfollowed the data closely. Remarkably, and perhaps rather fortu-itously, the temperature of 2120 K at which reflood was initiatedwas reached at the same time in the calculations as in the experi-ment. There was no strong sensitivity to choice of oxidation modeluntil temperatures exceeded 1800 K, since GM5(high) and L kinet-ics are similar. However, above 1800 K the PC correlation calcu-lated a much stronger escalation than was observed in theexperiment. Unlike MELCOR, SCDAP/RELAP5 calculates continuedoxidation until temperatures well above the metallic meltingpoint.

Fig. 20. SCDAP/RELAP5 calculation of QUENCH-14 shroud tem

The effect can be seen more clearly in Fig. 21 which concen-trates on the transient and reflood phases, when SCDAP/RELAP5in conjunction with GM5/UH gave remarkably close agreementwith the data. The strong excursion using L/PC was still calculatedeven when the reflood initiation time brought forward to 7170 s,corresponding to the same temperature of 2120 K as applied inthe experiment. Further sensitivity studies on time of reflood initi-ation did not affect the results. As can be seen from Fig. 22, GM5/UH also gave the best overall agreement for hydrogen generation.

SCDAP/RELAP5 correctly captured the bundle refilling (Fig. 23)apart from the initial temporary ingress of water, which was dueto the rapid injection into the lower volume and resulted in someearly cooling of the bundle and quenching of the lower part. Incomparison MELCOR showed a more rapid fill. (The rise in levelstalled at 1000 mm in the experiment, probably due to the occur-rence of shroud failure near the top.) The thermal response is com-pared in Fig. 24, also showing good agreement for quenchprogression except for the cooling induced by the initial ingressof water. The MELCOR input model uses a coarser hydraulic meshthan SCDAP/RELAP5, while the code itself provides a simpler and

perature at 950 mm, with alternative oxidation models.

Fig. 21. SCDAP/RELAP5 calculation of QUENCH-14 shroud temperature at 950 mm, with alternative oxidation models (reflood).

Fig. 22. SCDAP/RELAP5 calculation of QUENCH-14 total hydrogen generation, with alternative oxidation models.

Fig. 23. Comparison between experimental data, the SCDAP/RELAP5 and MELCORresults for collapsed level in bundle during QUENCH-14 reflood.

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1045

less complete treatment of the transient two-phase hydrodynamicinteractions that are particularly strong during reflood. As seenfrom Fig. 25, the MELCOR calculation showed a lower heat removalrate and later quenching in the upper part of the bundle. Thiswould have resulted in less steam generation and liquid entrain-ment, and hence the faster filling of the bundle. (It is noted thatthe shroud failure was actually calculated by MELCOR, and for thatreason the temperatures at the uppermost nodes are not availableduring reflood. Comparison is therefore made with the associatedheat structure temperature.)

Concerning oxidation modelling, the best overall agreementamong the calculations was obtained by the GM5(low/high) model,whether used with MELCOR or SCDAP/RELAP5. Comparison of theMELCOR calculations for the pre-oxidation phase (T < 1500 K) alsoindicates a clear improvement with GM5(low/high), again indicat-ing the effect of slower kinetics in the low temperature regime. TheSCDAP/RELAP5 calculations show a slight underestimate whenusing GM5 and a slight overestimate when using L/PC (only

Fig. 24. Comparison between experimental data and the SCDAP/RELAP5 results for shroud temperatures during QUENCH-14 reflood.

Fig. 25. Comparison between experimental data and the MELCOR results for shroud temperatures during QUENCH-14 reflood.

1046 J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047

Leistikow applies at these temperatures), but the differences areprobably within the influence domain of other uncertainties.Assessment of the models is more complicated during the transientshortly before and just after the start of reflood, because differentelevations of the bundle span all three temperature regimes. How-ever, good agreement was obtained during the transient and re-flood phases with UH in the high temperature regime. The threeregime model of GM5(low/high) and UH provided the best agree-ment of those used, which was possible to use in a special versionof SCDAP/RELAP5. Cases which included PC overestimated thetemperature escalation and hydrogen generation during thesephases of the experiment.

6. Conclusions

The QUENCH-14 experiment investigated the effect of niobium-bearing M5� cladding material on bundle oxidation and corereflood, in comparison with test QUENCH-06 (ISP-45) that usedstandard Zircaloy-4, and QUENCH-12 (ISTC 1648.2) that used nio-bium-bearing E110 alloy. The test was conducted with an electricalpower history very similar to QUENCH-06. After the pre-oxidationat peak bundle temperature of �1500 K during �3000 s, the elec-tric power was ramped from 540 to 940 W/rod at a rate of0.3 W/s/rod. The desired maximum bundle temperature of 2073K was reached after about 1200 s. Following a fast water injection

J. Stuckert et al. / Annals of Nuclear Energy 37 (2010) 1036–1047 1047

to quickly fill the lower volume, reflooding of the bundle by waterinjection at 2 g/s/rod was initiated, and the electrical power wasreduced to an average decay heat level of 2 W/cm. The cooling to400 K took about 300 s.

The evolution of axial peak of oxide layer thickness during thetest was evaluated with three withdrawn Zircaloy-4 rods with fol-lowing results: 180 lm at the end of the pre-oxidation phase,360 lm before reflood and about 700 lm after the test. Post-testinvestigations of the bundle showed neither breakaway oxidationof M5� cladding nor melt relocation; some melting of the claddingoccurred but the molten cladding material was localised betweenpellets and cladding oxide layer. The M5� cladding absorbed signif-icantly less hydrogen than the E110 corner rod, which showedintensive breakaway oxidation.

Measured hydrogen production during the QUENCH-14 testwas 34 g in the pre-oxidation and transient phases and 6 g in thequench phase being similar to those in QUENCH-06, i.e. 32 g and4 g, respectively. Corresponding values for QUENCH-12 were 32and 24 g. The reasons for such very different hydrogen productionduring the quench phase of tests QUENCH-12 and QUENCH-14were: (1) a degradation of oxide layers caused by the breakawayeffect in case of the QUENCH-12 bundle and absence of this effectfor the QUENCH-14 bundle, and (2) oxidation of the melt relocatedbetween rods during quenching of the QUENCH-12 bundle.

The MELCOR and SCDAP/RELAP5 models both correctly repro-duced the transient prior to reflood, and both provided satisfactoryagreement for the fairly mild oxidation during reflood in QUENCH-14. The reflood progress was captured more effectively by SCDAP/RELAP5, due to the more detailed hydraulic noding and more com-plete hydrodynamic modelling. Two-phase flow has been identi-fied by the MELCOR developers for model improvement in futurecode versions. Until then SCDAP/RELAP5 may be preferred forassessment of reflood quenching, particularly in situations wherecoolant injection is limited, as has been exemplified by a recentplant assessment study (Haste et al., 2010).

Input models for MELCOR 1.8.6 and SCDAP/RELAP5 were devel-oped and benchmarked against QUENCH-06 data. A range ofoxidation correlations was used in conjunction with both codes,among them the trial correlation for M5�. The models were usedin the QUENCH-14 analyses, the results of which confirm theinterpretations based on direct comparison between QUENCH-06and -14. The Urbanic–Heidrick correlation correctly captured thehydrogen generation and thermal response due to rapid oxidationat temperatures above 1800 K for this case without significantdegradation, and where the oxide scale remained unbroken andprevented release of any molten metallic. Close examinationreveals that the low temperature branch of the M5� trial correla-tion obtained by Grosse provides better agreement than the Urbanic-Heidrick correlation at temperatures below 1300 K, although theimprovement compared with the Leistikow is rather marginaland further study is necessary. In any case, the Grosse correlationis highly promising, and continuation of the experiments under awider range of conditions and for the different cladding materialsis most desirable. Inclusion of a low temperature oxidation modelfor M5� in the reactor accident analysis codes is strongly recom-mended, to enable the kinetics to be captured over the whole tem-perature range of interest. An analogous extension of the MATPROcorrelation might be considered for Zry-4 and other claddings,pending further data acquisition and model development.

Acknowledgements

The KIT work is sponsored by the HGF Programme NUKLEAR.The PSI authors thank Swissnuclear for significant financial supportto conduct these research activities.

References

Billone, M., Yan, Y., Burtseva, T., Daum, R., 2008. Cladding Embrittlement DuringPostulated Loss-of-Coolant Accidents. Report NUREG/CR-6967, ArgonneNational Laboratory.

Birchley, J., Haste, T., Jäckel, B., 2005. Calculational Support for the QUENCH-10 andQUENCH-11 experiments. In: Proceedings of the 11th International QUENCHWorkshop, Karlsruhe (CD-ROM (ISBN 3-923704-51-8)).

Birchley, J., Haste, T., Hering, W., Homann, C., 2006. Pre-test analytical support forexperiments QUENCH-10, -11 and -12. In: Proceedings of the InternationalConference Nuclear Energy for New Europe (NENE 2006), Portoroz, Slovenia.

Birchley, J., Haste, T., Dubourg, R., Dutheillet, Y., Lamy, J.-S., Maliverney, B.,Austregesilo, H., Bals, C., Trambauer, K., Steinbrueck, M., Stuckert, J., 2008.Post-test analysis of the QUENCH-13 experiment. In: Proceedings of theInternational Conference on Nuclear Energy for New Europe (NENE 2008),Portoroz, Slovenia.

Birchley, J., Jaeckel, B., Haste, T., Steinbrueck, M., Stuckert, J., 2009. Analysis of theQUENCH -14 bundle test with M5� cladding. In: Proceedings of the 17thInternational Conference on Nuclear Energy (ICONE-17), Brussels, Belgium (CD-ROM, Paper 75099).

Cole, R., 2001. MELCOR Core reflood modelling and applications to QUENCHExperiments. CSARP Meeting, Bethesda, Maryland, USA, May.

Duriez, C., Dupont, T., Schmet, B., Enoch, F., 2008. Zircaloy-4 and M5� hightemperature oxidation and nitriding in air. Journal of Nuclear Materials 380(2008), 30–45.

Garner, G.L., Mensah, I.Q., Mardon, J.P., 2006. Performance of alloy M5� claddingand structure at burnups beyond the current licensing limit in US reactors.Transactions of International Meeting Top Fuel, Salamanca, Spain, pp. 173–177.

Gauntt, R.O., Cash, J.E., Cole, R.K., Erickson, C.M., Humphries, L.L., Rodriguez, S.B.,Young, M.F., 2005. MELCOR Code Manuals – Version 1.8.6. USNRC NUREG/CR6119 Rev. 3, SAND2005-5713, Sandia National Laboratories.

Grosse, M., 2008. Comparison of the high temperature steam oxidation kinetics ofadvanced cladding materials. In: Proceedings of International Congress onAdvances in Nuclear Power Plants (ICAPP ‘08), CD-ROM, Paper 8069, Anaheim,USA.

Grosse, M., Lehmann, E., Vontobel, P., Steinbrueck, M., 2006. Quantitativedetermination of absorbed hydrogen in oxidised Zircaloy by means ofneutron radiography. Nuclear Instruments and Methods in Physics Research,A 566, 739–745.

Grosse, M., Lehmann, E., Steinbrück, M., Kühne, G., Stuckert, J., 2009. Influence ofoxide layer morphology on hydrogen concentration in tin and niobiumcontaining zirconium alloys after high temperature steam oxidation. Journalof Nuclear Materials 385 (2009), 339–345.

Haste, T., Birchley, J., Richner, M., 2010. Accident Management following loss-of-coolant accidents during cooldown in a westinghouse 2-loop PWR. NuclearEngineering and Design 240, 1599–1605.

Hering, W., Homann, Ch., 2007a. Improvement of the SCDAP/RELAP5 Code withRespect to FZK Experimental Facilities. Report FZKA 6566, Karlsruhe.

Hering, W., Homann, C., 2007b. Degraded core reflood: Present understanding andimpact on LWRs. Nuclear Engineering and Design 237, 2315–2321.

Hering, W., Homann, C., Lamy, J.S., Miassoedov, A., Schanz, G., Sepold, L., Steinbrück,M., 2002. Comparison and Interpretation Report of the OECD InternationalStandard Problem No. 45 Exercises (QUENCH-06). Report FZKA 6722, Karlsruhe.

OECD Report, 2008. Final Report of the OECD–IAEA Paks Fuel Project, NEA/CSNI/R 2.Pawel, R.E., Cathcart, J.V., McKee, R.A., 1979. The kinetics of oxidation of Zircaloy-4

in steam at high temperatures. Journal of the Electrochemical Society 126,1105–1111.

Peng, D.Q., Bai, X.D., Chen, B.S., 2005. Corrosion behavior of carbon-implanted M5alloy in 1 M H2SO4. Applied Surface Science 245, 215–222.

Schanz, G., Adroguer, B., Volchek, A., 2004. Advanced treatment of Zircaloy claddinghigh-temperature oxidation in severe accident code calculations. Part I.Experimental database and basic modelling. Nuclear Engineering and Design232, 75–84.

Sepold, L., Hering, W., Homann, C., Miassoedov, A., Schanz, G., Stegmaier, U.,Steinbrück, M., Steiner, H., Stuckert, J., 2004. Experimental and ComputationalResults of the QUENCH-06 Test (OECD ISP-45). Report FZKA 6664, Karlsruhe.

Sepold, L., Hering, W., Schanz, G., Scholtyssek, W., Steinbrück, M., Stuckert, J., 2007.Severe fuel damage experiments performed in the QUENCH facility with 21-rodbundles of LWR-type. Nuclear Engineering and Design 237, 2157–2164.

Sepold, L., Lind, T., Pintér Csordás, A., Stegmaier, U., Steinbrück, M., Stuckert, J., 2009.AgInCd control rod failure in the QUENCH-13 bundle test. Annals of NuclearEnergy 36, 1349–1359.

Siefken, L.J., Coryell, E.W., Harvego, E.A., Hohorst, J.K., 1997. SCDAP/RELAP5/MOD3.2Code Manual. USNRC NUREG/CR-6150 Rev. 1, INEL-96/0422, Idaho NationalEngineering Laboratories.

Steinbrück, M., 2008. Oxidation of zirconium alloys in oxygen at high temperaturesup to 1600 �C. Oxidation of Metals 70, 317–329.

Steinbrück, M., Grosse, M., Sepold, L., Stuckert, J., 2010. Synopsis and outcome of theQUENCH experimental program. Nuclear Engineering and Design 240, 1714–1727.

Stuckert, J., Birchley, J., Grosse, M., Haste, T., Sepold, L., Steinbrück, M., 2009.Experimental and post-test calculation results of the integral reflood testQUENCH-12 with a VVER-type bundle. Annals of Nuclear Energy 36, 183–192.

Urbanic, V.F., Heidrick, T.R., 1978. High temperature oxidation of Zirclaoy-2 andZircaloy-4 in steam. Journal of Nuclear Materials 75, 251–261.