Transcription

CFD Evaluation of OECD PSBT Geometry Effects Based on FluidTemperature MeasurementsY. Xu, Y. Sung and E. TatliWestinghouse Electric Company, 1000 Westinghouse Drive, Cranberry Township, PA 16066,[email protected]; [email protected]; [email protected] mixing, void and Departure from Nucleate Boiling (DNB) experimental data from theOrganization for Economic Co-operation and Development / Nuclear Regulatory Commission(OECD/NRC) Pressurized Water Reactor (PWR) Sub-channel and Bundle Tests (PSBT)international benchmark exercises are available for benchmarking system, sub-channel andComputational Fluid Dynamics (CFD) codes simulating fuel thermal-hydraulic behavior. In orderto better understand the mixing test data and the uncertainties in the rod bundle and spacer gridgeometric parameters, several CFD models with different configurations and numbers of spacergrids were built in STAR-CCM (CD-adapco) based on previous benchmarking experience forsimilar geometries to the PSBT rod bundle test section. Geometry effects on fluid temperaturedistribution at sub-channel exits were explored by comparing the CFD predicted results to themeasured test data in the PSBT Phase II/Exercise 1. The CFD evaluation results show that spacergrids with mixing vanes in the test bundle were most likely installed with alternating 90ºrotations, consistent with the actual fuel design practices at the time. The results from the CFDsimulations incorporating the spacer grid rotation are in better agreement with the measured testdata. The CFD evaluation also indicates that fully developed thermal mixing is associated withthe number of spacer grids in the test bundle. CFD modeling and simulation of the full length testbundle and radial geometry is recommended for future benchmark exercises.KEYWORDSCFD, PSBT Rod Bundle, Mixing Vanes, Orientation of Spacer Grids, Turbulent Mixing1. INTRODUCTIONPWR Sub-Channel and Bundle Tests (PSBT) are a series of thermal hydraulic tests conducted by NUPECin typical PWR fuel bundles and operated under prototypical PWR conditions with a wide range ofoperating conditions of system pressures, heating powers, inlet fluid temperatures and flow rates. Thebenchmark database [1] composed by the Pennsylvania State University sponsored by OECD and NRCincludes high quality and high resolution experimental data of void fraction distribution, sub-channelfluid temperatures and DNB powers of single and two-phase flows in single sub-channel and rod bundleNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201541964196

geometries under steady state and transient conditions. PSBT database has been used for many codebenchmarks and assessments in pressure drop, void fraction distribution and steady state or transient DNBpowers as summarized in [2].Phase II/Exercise 1, which is the focus herein, includes fluid temperature measurement test data (thermalmixing). In thermal mixing tests, in the test section named as PSBT A1 [1], the fluid temperatures at thecenter of sub-channels downstream of the end of heated length were measured with thermocouples understeady state conditions. PSBT A1 section consisted of a 5x5 rod bundle with 25 heated rods, three types ofspacer grids and Type C radial power distribution. Type C radial power distribution, an aggressivedistribution (four times between hot and cold rods) compared to others [5], was applied to the rod bundle,creating strong thermal gradients across the cross section which may serve well as a candidate for codeassessments and benchmarking, such as thermal diffusion coefficients benchmarking conducted in [3, 4].Since the test series was primarily designed for assessments and benchmarking of system and/or subchannel codes, detailed geometry information was neither disclosed nor available, particularly theinformation on the spacer grids. The spacer grid design information was obtained from one of thebenchmark participants as stated in [2]. Insufficient information is a big challenge for conducting accurateCFD analyses. On the other hand, CFD codes are capable of capturing geometric effects whencomputational mesh and numerical algorithms are appropriately selected. This test data may provide anopportunity for using CFD codes to determine further geometry details, such as mixing vane orientationsin actual operated tests.Due to the periodic and/or repeatable features in rod bundle designs, computational models with partiallengths are generally used in CFD benchmarking practices [5, 6, 7, and 8] without loss in accuracy inorder to reduce computing costs. This may be appropriate for cases in which the independence of the flowfield to number of spacer grids is demonstrated, but it may not be the case for thermal mixing cases sinceheating power is added continuously along the length of the test section.Several CFD models with different lengths and spacer grid orientations were created to investigate thegeometric effects on thermal mixing in rod bundles. A possible grid orientation was identified and anappropriate CFD modeling approach for thermal mixing simulations in PWR rod bundles wasrecommended.2. PSBT Thermal Mixing Tests and the Development of the CFD Modeling Approach2.1. PSBT Fluid Temperature Tests and DataAmong PSBT series, Phase II/ Exercise 1 was devoted to fluid temperature measurements (thermalmixing). The test section used in this exercise is schematically shown in Figure 1 - a 5x5 rod bundlewithout any unheated thimble tubes. It consisted of two (2) spacer grids with no mixing vanes (NMV) atthe beginning and at the end of the heated length (BOHL and EOHL). Between these two NMV grids,seven (7) spacer grids with mixing vanes (MVG) and eight (8) simple support grids (SSG) weredistributed alternatively along the heated section. The axial locations of the grids (defined at their bottomedges) are tabulated in Figure 2. The table also includes the locations of fluid temperature measurements(FTM) and the test section configuration with rotated MVGs (referred to as MVR). The 3D images ofthese grids are copied from [1] as shown in Figure 1. An axially uniform power profile was appliedbetween BOHL and EOHL in Type C distribution in the radial direction (shown in Figure 2). Coolanttemperatures were measured by thermocouples placed at the centers of sub-channels located 457 mmfrom the EOHL. Sixty tests were conducted with various mass fluxes, system pressures, coolant inlettemperatures, and heating powers. Several tests were marked and recommended by the databasecomposer for assessments, and data set 01-5343 was selected for the CFD simulation. Since the CFDNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201541974197

models use different lengths, it is difficult (if not impossible) to have appropriate inlet temperatures sothat fluid temperatures at the FTM are the same for different CFD simulations. Furthermore, it is thetemperature gradients, not the absolute temperature values that indicate thermal mixing behavior. Ofcourse, the latter plays a secondary role in estimating fluid properties. Therefore, it is more convenient touse temperature differences relative to cross section averaged values for comparison among different CFDsimulations. The operating conditions and the temperature difference data of this test are listed in Figure2.2.2. Development of the CFD Modeling ApproachIt was assumed that the far away upstream geometry features had little effect on flow and heat transferdownstream in rod bundles, such as PWR fuel assemblies [5]. Generally, CFD models are built for partiallength in order to save computing time. However, it is not clear how many spacer grids are sufficient inorder to be able to validate this assumption for this particular case, or whether or not fluid flow andthermal mixing behave the same.In this exercise, partial length and full length (including all grids) CFD models were developed. The bestpractices learned from previous benchmarking exercises [9] were adopted in the selection of meshgeneration settings, turbulence models, and numerical algorithms.Since there was no information on MVG orientation and their relationship to radial power distribution inthe test specifications [1], many combinations of MVG orientations, and the power distribution and MVGorientations were possible. In reality, the most possible variation is that some of MVGs may be rotatedfrom others, consistent with the grid rotations in actual fuel assemblies: i.e., 3 MVGs or 4 MVGs rotated.The 4 MVGs rotation can be realized using radial power rotation. Therefore, radial power rotation can beused to evaluate its relationship with the MVG orientation without building a new geometry.In order to address the above issues, a total of six CFD models were built in STAR-CCM [10]. Theirdetails are listed in Table I, including domain lengths, number of grids and their orientations. The lastcolumn in the table shows number of computational cells in each model. The number in the parenthesiswas obtained from a smaller base size for mesh independence check. All models have the same outletlocation, about L/D 35 from the FTM. At the inlets, constant mass flow rate and uniform temperatureprofiles were applied. All solid surfaces, including the side casing, were treated as no-slip wallboundaries. Typical meshes at the outlet and on MVG region are shown in Figure 3. Very finecomputational meshes were built to capture the small geometric features of MVGs, which could havesignificant effects on the mixing performance. The prism layers were created near solid surfaces andcorners with reasonably good quality. The mesh was extended axially in order to save mesh cells since theflow velocity is dominantly large in the axial direction.Fluid temperatures were extracted for each sub-channel as numbered in Figure 2. These points are definedat the center for whole sub-channels while minor adjustments are needed for side and corner subchannels. The plane of these points is located at the FTM (4115 mm from the BOHL). A straight linealong x-axis (red dotted line in Figure 2) is defined on this plane for line plots of velocity andtemperature. Similarly, another parallel line is created on the plane of 2856 mm where mixing effectsinvolving different number of MVGs are demonstrated.NURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201541984198

Figure 1 Test Section and Spacer Grids [1]NURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201541994199

Figure 2 Test Conditions, Power Distribution, Test Data, and Component ConfigurationsTable I. Details of the CFD DomainsCFD# of CellsLength (mm)SSGNMVMVGMVRDomain(*106)3086 to 450011106.35 (13.71*)MV-C12629 to 4500212012.92MV-C2-25 to 4500827040.25MV-A12629 to 450021119.86MVR-C22172 to 4500312113.17MVR-C3-25 to 4500824330.11MVR-A1(*mesh generation base size equal to 0.6 mm vs. 0.8 mm in original mesh)NURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542004200

Figure 3 Typical mesh scenes3. Evaluation of CFD Simulation ResultsCFD models of the PSBT test section A1 [1] were built and run using STAR-CCM v8.06. Steady stateRANS modeling was adopted with standard k- turbulence model and two-layer all y wall treatments. Adefault quadratic relationship between strain and stress was chosen. Fluid density, dynamic viscosity andthermal conductivity were tabulated versus temperature while specific heat was fitted with a 4th orderpolynomial as a function of temperature. The results were judged converged when monitored variables(local axial velocity, turbulent kinetic energy and temperature) and the residuals stabilized reachingconstant values, special attention paid to the energy balance in the domain. The simulation results wereextracted from the last iteration. Although temperature gradients exist across each sub-channel, it wasverified that variations between averaging across each sub-channel and the values at center points are.Therefore, only the values at the center points were used in following evaluations.3.1. CFD Simulation without MVG RotationThe first set of CFD models was built based on the information disclosed in [1]. Since there is nospecification regarding MVG orientations and the symbols of the MVGs in Figure 1 are identical for allMVGs, it was assumed that the MVGs were aligned identically in the test section. Three different CFDdomain lengths were used to investigate the upstream condition effects on thermal mixing. A typicaltemperature contour plot (MV-C2) is shown in Figure 4a. A high temperature region and a lowNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542014201

temperature region appear at the upper left and lower right corners, respectively, due to swirling flowsproduced by the MVGs. Thermal mixing is clearly seen from the hot region to the cold region through thetemperature gradients. The fluid temperatures at the FTM from all three MV models are compared withthe test data in Figure 4b. Overlapping the two results from MV-C1 shows that mesh independence isachieved with the selected mesh settings. The full length model significantly over-predicts temperaturedifferences between the hot and the cold regions, indicating under-prediction of thermal mixing. Thisindicates that there may be some feature missing in the model. MV-C2 produced the best match with thedata, further proving that inappropriate boundary conditions and geometry features could producemisleading information. 01 534340.00MV C1MV C1 0.630.00MV C2MV A1Temperature Difference (K)20.0010.000.00 10.00 20.00 30.00 40.0005101520Sub Channel IDs25303540a Temperatures at FTMb MV results vs. test dataFigure 4 Temperatures at FTM of MV Cases3.2. CFD Simulation with MVG RotationThe results in Section 3.1 suggested possible incorrect geometry features existed in the models. One ofthe most likely situations was the orientations of the MVGs. It is common practice that MVGs are rotatedalternatively, i.e. the downstream MVG is rotated by 90o of the upstream one, and the next one is rotatedby -90o, and so on. The second set of the CFD models was built based on this assumption (MVR-C2,MVR-C3, and MVR-A1) with other parameters identical to the first set. Similar to the MVG withoutrotation cases, a hot region and a cold region still co-exist in the MVR cases as seen in Figure 5a.However, the hot spots appear at quite different locations, indicating better mixing from the MVRdesigns. Furthermore, comparison of temperatures at the FTM between the CFD and the test data showthe significant improvements in CFD predictions. Especially, predictive errors are almost reduced to halfin the full length model. This indicates that the misalignment of the MVGs was one of the major errorsources in the first set of CFD models. In other words, CFD tools are capable of finding possiblesimulation errors when appropriate modeling approaches are pre-determined. This also shows thatappropriately assessed CFD tools can be utilized to perform design optimization for appropriate physicalprocesses.A closer look may reveal mismatches from the CFD simulations. Temperature trends of the same columns(indicated by the vertical solid lines in Figures 4b and 5b) are inconsistent or opposite between the testand the simulations, gradually decreasing from a lower number sub-channel to a higher one, particularlyNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542024202

in Columns 5 and 6 (i.e. cold region) in the test. Sub-Channel 24 has a lower temperature for both MVand MVR models, while it was higher than the neighboring sub-channels in the test data. Larger errorfrom the full length model indicates that turbulent dissipation may be over-predicted in these models.40.0030.00 01 5343MVR C2Temperature Difference (K)20.00MVR C3MVR A110.000.00 10.00 20.00 30.00 40.0005101520Sub Channel25303540a Temperatures at FTMb MVR results vs. test dataFigure 5 Temperatures at FTM of MVR CasesThe prediction performance of the above models for thermal mixing may be better viewed graphically. Aσయల ȁ்ି்ȁchannel error (݄݈ܿܽ݊݊݁ ݁ ݎ ݎݎ ൌ ೖసభ ೌ ೌ ಷವ ) is defined for an average error between the test dataଷ and the CFD results over all of the sub-channels. Figure 6 shows a plot of the channel errors of theaforementioned models.1412Channel Error (K)1086420MVR C2MVR A1MVR C3MV A1MV C2MV C1Figure 6 Channel Errors (Performance Indicator) of CFD ModelsNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542034203

3.3. Flow Fields due to Inlet Boundary ConditionsLateral velocity is a key indicator of thermal mixing within sub-channels. Vector plots of lateralvelocities at the FTM of the MV models are depicted in Figures 7a, 7b, and 7c. The velocityscale is from 0 to 0.02 m/s for all of the plots. It can be seen that the velocity distribution is quitesimilar for all three cases, but their magnitudes become smaller as the CFD domain becomeslonger. This also indicates an over-prediction of turbulent dissipation. Axial velocity wasextracted from a line at the FTM plane as shown in Figure 7a. The results are presented in Figure7d. Axial velocity is larger in the hot region due to lower density. The differences of axialvelocities between hot and cold regions appear closely correlated with temperature gradients (orthermal mixing); MV-A1 has the largest differences with most under-prediction of thermalmixing; MV-C1 has the smallest with over-prediction of thermal mixing.0 - 0.02 m/s0 - 0.02 m/s0 - 0.02 m/sFigure 7 Flow Fields at FTM of MV ModelsAs a comparison of MV models, the flow fields of MVR models are plotted in Figure 8. The MVat the highest elevation is shown in Figure 2 for illustrating the relationship between lateralNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542044204

velocity direction and the MV orientation. A swirling flow pattern around rods was created bythe vanes. It can also be observed that high lateral flow appears in the channels away from thebundle center, indicating better mixing flow in the MVR models than in the MV models. Thedifferences of the axial velocity magnitudes (seen in Figures 7d and 8d) are mainly due to thedifferent inlet temperatures used.Figure 8 Flow Fields at FTM of MVR ModelsIt was pointed out that many researchers used one or two spans of rod bundles for the PSBTbenchmarking exercises [5, 6, 7 and 8]. This may be appropriate for hydraulic variables such asvelocity and turbulent quantities. It is questionable for predicting the fluid temperatureNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542054205

distribution since heating power is added in the axial direction continuously. Figure 9 shows thevelocity components and fluid temperatures along the dotted line shown in Figure 8 at Elevation2856 mm. This location has one MVG for MVR-C2, two MVGs for MVR-C3 and six MVGs forMVR-A1 from the inlet, where uniform temperature and constant mass flow were applied. It canbe seen that the velocity profiles are virtually independent of the number of MVGs at upstream,particularly the lateral velocity components. Visible differences in axial velocity profiles aremainly from the density differences caused by local fluid temperatures. On the other hand, thedistributions of temperature differences (calculated based on its average across the cross sectionof the elevation) are quite different. The more MVGs are involved, the larger temperaturedifferences. This clearly indicates that a single MVG cannot achieve complete thermal mixing.Figure 9 Velocity and Temperature Fields at 2856 mm of MVR Models4. CONCLUSIONSThe PSBT fluid temperature measurement test data (A1) was used for CFD modeling andsimulations using STAR-CCM with different mixing vane orientations and domain lengths. TheCFD model was able to identify missing geometry features with adoption of the best practiceguidelines and lessons learned from similar prior benchmark practices. It can be concluded thatthe mixing vane spacer grids in the rod bundle tests were rotated, consistent with the actual fuelassembly design at the time. Different from flow fields, which were not sensitive to number ofupstream mixing vanes involved, thermal mixing (temperature gradients) demonstrated muchlonger upstream effects. A CFD model for the whole heated length is recommended for thermalmixing benchmark or assessment; otherwise, additional justification is needed on defining theinlet boundary conditions, particularly the inlet temperature distribution.NURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542064206

CDPSBTSSGBeginning Of Heated LengthComputational Fluid DynamicsDeparture from Nucleate BoilingEnd Of Heated LengthFluid Temperature MeasurementMixing Vane ModelMixing Vane GridRotated Mixing Vane ModelNon-Mixing VaneNuclear Regulatory CommissionNUclear Power Engineering CorporationOrganization for Economic Co-operation and DevelopmentPWR Sub-Channel and Bundle TestSimple Support GridREFERENCES1. A. Rubin, M. Avramova, and H. Utsuno, “OECD/NRC benchmark based on NUPEC PWRsubchannel and bundle tests (PSBT) volume I: experimental database and final problemspecifications,” Tech. Rep. NEA/NSC/DOC(2010)1, USA, NRC/OECD Nuclear Energy Agency.2. M. Avramova, A. Rubin, and H. Utsuno, “Overview and Discussion of the OECD/NRC BenchmarkBased on NUPEC PWR Subchannel and Bundle Tests”, Science and Technology of NuclearInstallations Volume 2013, Article ID 946173, 20 pages, http://dx.doi.org/10.1155/2013/9461733. Y. Sung, R. L. Oelrich Jr., C. C. Lee, N. Ruiz-Esquide, M. Gambetta, C. M. Mazufri, “Benchmark ofSubchannel Code VIPRE-W with PSBT Void and Temperature Test Data”, Science and Technologyof Nuclear Installations, Volume 2012, Article ID 757498, 11 pages doi:10.1155/2012/7574984. M. Valette, “Subchannel and Rod Bundle PSBT Simulation With CATHARE-3”, The 14thInternational Topical Meeting on Nuclear Reactor Thermalhydraulics, NURETH-14-132, Toronto,Ontario, Canada, September 25-30, 20115. K. Ikeda, M. Hoshi, “Development of Mitsubishi High Thermal Performance Grid (CFDApplicability for Thermal Hydraulic Design,” JSME International Journal Series B, Vol. 45, No.3,2002.6. M. E. Conner, E. Baglietto, A. M. Elmahdi, “CFD Methodology and Validation for Single-phase flowin PWR fuel assemblies”, Nuclear Engineering and Design, 240, pp. 2088-2095, 20107. C. Pena-Monferrer, J. L. Munoz-Cobo, S. Chiva, “CFD Turbulence Study of PWR Spacer-Grids in aRod Bundle”, Science and Technology of Nuclear Installations, Volume 2014, Article ID 635651, 15pages, http://dx.doi.org/10.1155/2014/6356518. C. P. Tzanos, “Computational Fluid Dynamics for the analysis of light water reactor flows”, NuclearTechnology, Volume 147, No. 2, August 2004, pages: 181-1909. M. E. Conner, Z. E. Karoutas, Y. Xu, “Westinghouse CFD Modeling and Results for EPRI NESTORCFD Round Robin Exercises of PWR Rod Bundle Testing”, NURETH-16 Paper-13601.10. STAR-CCM 8.06, www.cd-adapco.comNURETH-16, Chicago,NURETH-16,Chicago, IL,IL, AugustAugust 30-September30-September 4,4, 2015201542074207

Figure 2 Test Conditions, Power Distribution, Test Data, and Component Configurations Table I. Details of the CFD Domains CFD Domain Length (mm) SSG NMV MVG MVR # of Cells (*106) MV-C1 3086 to 4500 1 1 1 0 6.35 (13.71*) MV-C2 2629 to 4500 2 1 2 0 12.92 MV-A1 -25 to 4500 8 2 7 0 40.25 MVR-C2 2629 to 4500 2 1 1 1 9.86