A SARNET Benchmark on two VULCANO Molten Core Concrete Interaction Tests

Transcription

1 A SARNET Benchmark on two VULCANO Molten Core Concrete Interaction Tests C. JOURNEAU 1, J.F HAQUET 1, B. LETEXIER 1, A. GRECO 1, B. SPINDLER 2, R. GENCHEVA 3, P. GROUDEV 3, D. DIMOV 4, A. FARGETTE 5, J. FOIT 6, B. MICHEL 7, C. MUN 7, T. SEVON 8, C. SPENGLER 9, F. POLIDORO 10 1 CEA, Cadarache (FR) 4 Energy Inst., Sofia (BG) 7 IRSN, Cadarache (FR) 10 RSE, Milan (IT) 2 CEA, Grenoble (FR) 5 AREVA NP, Erlangen (DE) 8 VTT, Espoo (FI) 3 INRNE, Sofia (BG) 6 KIT, Karlsruhe (DE) 9 GRS, Cologne (DE) ABSTRACT A Molten Core-Concrete Interaction experimental program is underway at the VULCANO facility. It is devoted to the study of 2D long-term dry MCCI with either purely oxidic prototypic corium or with oxide and metal layers. It has been decided to open the results of two oxidic tests VB-U5 (silica-rich concrete) and VB-U6 (limestone-rich concrete) to the SARNET WP6 as a part of the Joint Program of Activities on this topic in the European Severe Accident Research Network of Excellence. In this scope this benchmark was performed. Ten participants from different countries (CEA Cadarache France, CEA Grenoble France, IRSN France, GRS Germany, KIT Germany, VTT Finland, AREVA Germany, EI Bulgaria, INRNE Bulgaria, RSE Italy) took part in this work. Five computer codes were used to perform two independent calculations, each one representing the main phenomena arising during the interaction between prototypic oxidic corium and siliceous or limestone concretes. The purpose of these analyses is to compare code results with those obtained by the experiments and to compare the best-estimate assumptions in the models used in the available MCCI codes. Conclusions on the major uncertainties on models used in these codes can then be drawn. Major differences lie with the estimation of the radiative losses (which control the power remaining available for ablation), the pool temperature evolution and the pool composition. Also some types of modelling rely on explicit assumptions on ablation anisotropy while for others it is only a consequence of the chosen set of correlations. 1 INTRODUCTION A Molten Core-Concrete Interaction (MCCI) experimental program is underway at the VULCANO facility [1]. It is devoted to the study of 2D long-term dry MCCI with either purely oxidic prototypic corium or with oxide and metal layers. It has been decided to use the results of two oxidic tests VB-U5 (silica-rich concrete) and VB-U6 (limestone-rich concrete) for a benchmark on MCCI codes as a part of the Joint Program of Activities on this topic in the European Severe Accident Research Network of Excellence (SARNET). Ten participants from different European countries (CEA Cadarache France, CEA Grenoble France, IRSN France, GRS Germany, KIT Germany, VTT Finland, AREVA Germany, EI Bulgaria, INRNE Bulgaria, RSE Italy) took part in this work. Five computer codes were used to perform two independent calculations, each one representing the main phenomena arising during the interaction between prototypic oxidic corium and siliceous or limestone concretes. The purpose of these analyses is to compare code results with those obtained by the experiments and to compare the best-estimate assumptions in the models used in the available MCCI codes. 1/9 pages

2 In a first section, the tests are described. Then, in a subsequent section, the calculations results will be presented and discussed. A synthesis will then conclude this paper. 2 TEST DESCRIPTION The concrete test sections (Figure 1) are 600 x 300 x 400 mm concrete blocks with a 300 x 250 mm hemicylindrical cavity in which corium is poured [2]. Two types of concretes, representative of nuclear reactor basemats, have been used. They were both made with CEM I 52.5 N cement (from Holcim) and the aggregate granulometry was the same for the two concretes. But, the nature of the aggregates were different: in the first one (Concrete F), the aggregates were mostly siliceous (they came from the Rhine river GSM Rumersheim gravel pit), while, for the second one (Concrete G), the sand and the gravel were a mixture of limestone-siliceous and limestone aggregates from the Durance Granulats Peyrolles ballast pit and the La Nerthe Quarry (Jean Lefebvre Méditérannée), leading to a higher fraction of limestone (Table I). More than 100 K-type thermocouples (Figure 1 left) have been installed in the concrete to monitor its ablation as well as some high temperature C-type thermocouples. Refractory ceramic wall In-corium TCs Concrete to be studied Figure 1: The concrete crucible Left: before corium pouring Right: before concrete pouring TABLE I: Overall chemical composition of the concretes used in VULCANO MCCI experiments (wt.%) Sum may be below 100% due to the presence of minor species Concrete F Concrete G SiO CaO Al 2 O CO H 2 O Fe 2 O /9 pages

3 Four parallel induction coils surround the section and provide sustained heating. The test sections are installed in the VULCANO containment and operate in air at atmospheric pressure. During the experiments a high-temperature oxidic corium melt has been obtained with the VULCANO plasma-arc furnace and poured in concrete test sections. Table IV summarizes their main characteristics. Pseudobinary phase diagrams between concrete and corium have been computed for VB-U5 and VB-U6 (Figure 2) by GEMINI2 correlations from NUCLEA08 database. They show the different physicochemical behaviour of the two studied concretes. TABLE II: Characteristic of the oxidic corium experiments Corium composition (wt.%) Test VB-U5 VB-U6 initial 54% UO 2, 37.4 % ZrO 2, 3.7 % SiO 2, 2.5% Fe 2 O 3, 2.4% CaO 57 % UO 2, 33.1 % ZrO 2, 5 % SiO 2, 2.6 % CaO, 2.3% Fe 2 O 3 Corium mass 28 kg 31 kg Concrete Concrete F Silica Concrete G Limestone Initial Temperature ~2350 K ~2600 K Benchmark Duration time 1 hr 20 2 hr Total injected power 25 kw (first 1h20) 27 kw Net power kw 9 1 kw Commentaire [m1]: It is more relevant to present the benchmark considered time and not the total test duration Figure 2: Pseudo binary diagrams between concrete and corium Left: VB-U5 Right: VB-U6 black: Liquidus line red: solidus line 3/9 pages

4 H (cm) 5th European Review meeting on Severe Accident Research (ERMSAR-2011) 3 COMPARATIVE ANALYSIS OF CALCULATIONS Different codes like TOLBIAC-ICB v3.2 [3], [4], ASTECv2/MEDICIS [5], COSACO [6], CORQUENCH 3.03 [7], WECHSL [8] and CORIUM-2D [9] have been used to compute the main phenomena during the interaction between real concrete material (siliceous-rich and limestone-rich concretes) and molten prototypical corium in VB-U5 and VB-U6 tests. Many research organizations as CEA-Cadarache, CEA-Grenoble, GRS, IRSN, VTT, AREVA, EI, KIT, INRNE and RSE have been involved to recalculate and explain pool/concrete interface behaviour. Some of the participants, like CEA-Cadarache and CEA-Grenoble have presented more than one calculation per test to compare different assumptions available in code modelling. The main purpose of this benchmark is to compare code results with the experimental results, to compare the best-estimated assumptions and to define major uncertainties in the models used in these codes. The other purpose is to point out the key points in the physical phenomena that will allow future improvement of the available MCCI codes. An example of the obtained final cavity shapes in case of the VB-U5 test is displayed on next Figure 3. Clearly, codes imposing the anisotropy (by fitting a multiplicative factor applied to the convective heat transfer coefficients or imposing values of thermal resistances at pool/concrete interfaces) or using heat transfer models leading intrinsically to anisotropy permit to get a better agreement for the final cavity shape. Crucible shapes at the end of calculations of VB-U5 test R (cm) VTT GRS AREVA IRSN EI INRNE CEA_gre_base CEA_gre_case2 CEA_gre_case3 KIT CEA_cad_2336K CEA_cad_2625K RSE EXPERIMENT INITIAL Figure 3: Crucible shapes at the end of calculation for VB-U5 Pool temperatures calculated for VB-U5 are shown in Figure 4 below. Most curves indicate a trend of monotonous decrease. In most of calculations the pool temperature decreases faster during the first transient phase and this decrease slows down at later times, only exceptions are TOLBIAC (CEA) and CORIUM-2D (RSE) calculations with the highest pool/crust interface temperature. It must be noted that pool temperature has been measured only during a small period around 1200 s (2403 K) is somehow overestimated as it 4/9 pages

5 Mass fraction (%) Temperature (K) Exp. values 5th European Review meeting on Severe Accident Research (ERMSAR-2011) has arisen from recent analysis of subsequent VULCANO experiments (the real temperature can be 30 to 200K lower). Only TOLBIAC calculations which follow the liquidus temperatures (Figure 2) give fairly good estimates of the pool temperature Pool temperatures versus time for VB-U5 test GRS Time (s) VTT AREVA IRSN EI INRNE CEA_gre_base CEA_gre_case2 CEA_gre_case3 KIT CEA_cad_2625K CEA_cad_2336K RSE EXPERIMENT Figure 4: Pool temperatures versus time for VB-U5 45% 40% 35% 30% UO2 final mass fractions for VB-U5 VTT; 39,35% GRS; 37,29% IRSN; 39,39% AREVA; 35,65% EI; 36,96% INRNE; 39,03% KIT; 35,55% 25% 20% 15% 10% 5% CEA_gre_ base case; 3,52% CEA_gre_ case2; 14,71% CEA_gre_ case3; 5,81% CEA_cad_ 2625K; 10,70% CEA_cad_ 2336K; 9,32% 0% Figure 5: UO2 final mass fractions for VB-U5 (Experimental range 34-53%) The computed final composition has been compared from the Post Test Analyses results. Figure 5 presents for instance the final UO 2 fraction. Most of the calculations consider a mixing of all the corium with all the ablated concrete and provide final values between 35 and 40 %. On the opposite, TOLBIAC-ICB considers that crust remain at the same composition from the time they have been formed (i.e. a time where there was less concrete in the pool) and in some case even consider that the crust correspond to the most 5/9 pages

6 H (cm) 5th European Review meeting on Severe Accident Research (ERMSAR-2011) refractory species [3] (i.e. mainly (U,Zr)O 2 ). Therefore the melt is depleted in UO 2 and results between 3.5 and 15% have been computed. Experimentally, corium-rich zones (with up to 53 wt% UO 2 ) and corium-poor zones (down to 35 wt% UO 2 ) have been observed. Contrary to TOLBIAC-ICB calculations, all other calculations fit between these two bounds and are thus deemed acceptable. Therefore, it is suggested that if crusts are formed, they must be frequently renewed so that their composition does not vary significantly to that of the rest of the pool. If one consider the experimental final (U,Zr)O 2 composition (57-91 %), the pool liquidus at the end would be between 2320 and 2600 K. Concerning the VB-U6 experiment with limestone rich concrete, similar fits have been observed. For instance, Figure 6 presents the experimental and computed cavity shapes. Crucible shapes at the end of calculations of VB-U6 test VTT GRS AREVA CEA_gre_base CEA_gre_case2 CEA_gre_case3 IRSN INRNE CEA_cad_9kW CEA_cad_9kW-Pcond KIT EI R (cm) RSE EXPERIMENT INITIAL Figure 6: Final cavity shape for VB-U6: Experimental and calculated profiles. 6/9 pages

7 Void fraction % 5th European Review meeting on Severe Accident Research (ERMSAR-2011) 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Void fractions for VB-U6 test Time (s) GRS VTT CEA_gre_base CEA_gre_casa2 CEA_gre_case3 AREVA IRSN INRNE CEA_cad_9kw CEA_cad_9kW-Pcond KIT Figure 7: Estimated void fractions in VB-U6 Figure 7 presents the computed void fractions in VB-U6. In the initial stage of calculation the greater values are calculated by GRS, IRSN, INRNE and AREVA (around 30%) whereas TOLBIAC (CEA) gives lower values around 10 % at the initial transient decreasing very quickly and reaching around 3 % at the end of the test. One of the causes of discrepancy may reside by the different assumptions made to adapt the superficial gas velocity - void fraction drift correlations to a 2D gas injection geometry. TOLBIAC averages the volumetric gas flow rate over the total concrete area, whereas the other codes consider the volumetric gas flow rate through the pool horizontal section. After one hour, even for this test with a 66 % CaCO 3 concrete, all calculations leadto void fractions lower than 7 %. Post Test Analyses indicate a higher porosity at the end of the test, around 15 %. 4 CONCLUSIONS From this benchmark exercise, it must be noted that there are many similarities in the predicted trends, in the ranges accounted. Nevertheless some major differences between modelling approaches can be observed. The ablated volume is controlled by the ablative energy, thus by the amount of energy radiated through the upper surface which depends on the code heat transfer models (heat convection distribution and interface structure) and also on the upper crust interface temperature; moreover most codes overestimate the ablated concrete volume if significant conduction heat losses through the concrete are not taken into account especially in case of VB-U6 (It has been estimated to account for 1 kw for CEA Cadarache with TOLBIAC-ICB to 2.5 KW for IRSN with ASTECv2/MEDICIS on a net injected power of 9 kw, depending on the modelling hypotheses). Impact of the ablation occurring during the initial transient phase can also be important, in particular in experiments. Cavity shape are rather well predicted: the VB-U5 with siliceous concrete required taking into account anisotropy, either explicitly (with ASTECv2/MEDICIS, TOLBIAC- 7/9 pages

8 ICB) or implicitly (with CORQUENCH and WECHSL); axial ablation is generally overestimated. TOLBIAC-ICB calculations provide good estimates of the pool temperature, even if uncertainties on experimental measurements remain, whereas the other models give some discrepancies of several hundreds kelvins at least in the initial MCCI phase of the VBU5 experiment; however, the calculated temperatures cannot be compared with the experimental one in the longer term since the overall pool temperature evolution was not measured. This has been corrected in subsequent tests [10]. All codes predict a small final void fraction (even with limestone test VBU6); the reason is probably that the drift flux model used to calculate the void fraction, even if different approaches are used in codes to adapt 1D experimental data to a 2D configuration with gas sparging from sides, is not well suited to 2D configurations, in particular in small scale experiments in which sides area is greater than bottom area, contrary to the reactor case. Crusts, if they exist, shall have a composition close to the current pool composition. Therefore it infirms the assumption that crust could keep the same composition from the beginning (i.e. a time where there was less concrete in the pool), and/or assuming that crust contains a higher fraction of most refractory species, like TOLBIAC-ICB code hypotheses. This shows that crust formation, if any, is renewed all along the experiment and without any significant segregation. Up to now, it is still not possible to propose a comprehensive modelling of MCCI that could predict the observed anisotropy and all the parameters of the experiment. But it must be reminded that we are using multi 0D quasi-steady state modelling to model an intermittent ablation process with a complex geometry both at the interface and a complex convection pattern in the pool because of combined effects of gas bubbling and solutal convection. Considering this benchmark exercise, it must be stressed that VB-U5 and VB-U6 were among the first successful VULCANO MCCI experiments. They used Gen 2 plant concretes and prototypical corium. Recent tests (VBES-U2 [10] e.g.) would have better measurements (better estimation of radiated and ablated powers, valid measurements of temperature every ~10 min). So a new benchmark with this experiment or a new test represents a significant opportunity to enrich our MCCI scientific knowledge. Indeed the general ablation pattern can be rather well predicted by most codes but no single calculation has been able to compute all the parameters of the experiments, i.e. mainly 2D ablation, evolutions of bulk pool temperature and upper pool interface temperature. REFERENCES [1] C. Journeau, P. Piluso, J.F. Haquet, E. Boccaccio, V. Saldo, J.M. Bonnet, S. Malaval, L. Carénini, L. Brissonneau, Two dimensional interactions of corium with concretes: The VULCANO VB Test series, Ann. Nucl. Ener. 36, pp (2009). [2] C. Journeau, J.M. Bonnet, E. Boccaccio, P. Piluso, J. Monerris, M. Breton, G. Fritz, T. Sevón, P. H. Pankakoski, S. Holmström, J. Virta, European Experiments on 2D Corium-Concrete Interaction: HECLA and VULCANO, Nuclear Technology, 170, (2010). [3] B. Spindler, B. Tourniaire, J.M.Seiler, MCCI Analysis and Application with the TOLBIAC-ICB Code based on Phase Segregation Model, ICAPP 05, Seoul Korea, May /9 pages

9 [4] B.Spindler, E. Dufour, TOLBIAC-ICB V3.2: Code description, Note technique DEN/GRE/DTN/SE2T/LITA/NT, /0. [5] M. Cranga, R. Fabianelli, F. Jacq, M. Barrachin, F. Duval, The MEDICIS code, a versatile tool for MCCI modelling, Proceedings of ICAPP05, Seoul, Korea, May th 2005 [6] G. Lohnert, R. Gadow, Temporary Melt Retention in the Reactor Pit of the European Pressurized Water Reactor (EPR), IKE [7] M. T. Farmer, The CORQUENCH Code for Modeling of Ex-Vessel Corium Coolability under Top Flooding Conditions Code Manual Version3.03, OECD/MCCI-2010-TR03. [8] J. J. Foit, M. Reimann, B. Adroguer, G. Cenerino, S. Stiefel, The WECHSL-Mod3 Code: a computer program for the interaction of a core melt with concrete including the long term behavior; Model descriptions and User s manual,. FZKA 5416 (1995). [9] F. Parozzi, R. Fontana, CORIUM-2D Advanced Code, Enel Struttura Ricerca, PAM , June [10] C. Journeau, L. Ferry, P. Piluso, J. Monerris, M. Breton, G. Fritz, T. Sevon, «Two EU-funded tests in VULCANO to assess the effects of concrete nature on its ablation by molten corium, ERMSAR 2010, Bologna, Italy, May /9 pages