Validation of RELAP5-3D© for liquid metal reactor technologies

The present research work set in an international and national context that includes the efforts of several universities and research centers in a strict collaboration. Within the international framework, the Department of Astronautical, Electrical and Energy Engineering (DIAEE) of “Sapienza” University of Rome (UNIROMA1) has been recently involved in the Horizon 2020 (H2020) SESAME (thermal-hydraulic Simulations and Experiments for the Safety Assessment of MEtal-cooled reactors) project. The project aims to contribute to the liquid metal-cooled fast reactors (LMFRs) development, including the advanced numerical approaches for the design and safety evaluation of the technologies. Regarding the national context, R&D efforts are mainly dedicated on the development of the lead-cooled fast reactor (LFR) technologies, involving three main partners: ENEA (Italian National Agency for New Technologies, Energy and Sustainable Economic Development), CIRTEN (Interuniversity Consortium for Technological Nuclear Research) and the industrial partner ANSALDO NUCLEARE. Strong collaboration and comparing among the three partners are devoted to the development of the reference LFR project: the Advanced Lead Fast Reactor European Demonstrator (ALFRED). In this framework, the purpose of the present research activity has been to contribute to the understanding of relevant thermal-hydraulic phenomena that characterize the operations of the LMFR, and to the fundamental validation process of the innovative numerical tools, adopted for safety analysis and licensing of new Generation IV (GEN IV) technologies. The research activity has dealt with the validation of RELAP5-3D© (R5-3D) for applications on liquid metal-cooled pool-type fast reactors. The thesis is divided in seven chapters. The first one is dedicated to the definition of the general background of the research activity. Chapters 2, 3 and 4 are focused on the validation of R5-3D for application on LMFRs. The merits of the simulations are evaluated comparing the results with experimental data from CIRCE (CIRColazione Eutettico) facility, Phénix reactor and PERSEO (in-Pool Energy Removal System for Emergency Operation) facility. The numerical activities have been used to explore and validate different modelling approaches and the acquired know-how has been used to support the design of ALFRED reactor. This topic is examined in chapter 5 that is basically divided in two subsections: the first one analyzes the reference configuration of ALFRED and the second one deals with the revised configuration of the reactor. Chapter 6 presents a methodology for the uncertainty quantification, based on the coupling approach between RELAP5-3D and RAVEN codes. Experimental data from the loop-type facility, called NACIE (NAtural CIrculation Experiment), have been used for the qualification of the methodology. Finally, the main results and guidelines, as well as the weaknesses and the future perspective coming out during the present research activity, are pointed out in chapter 7. CIRCE is a multipurpose pool-type facility aimed to investigate thermal-hydraulics of innovative heavy liquid metal (HLM) cooled pool-type systems. Two experimental campaigns have been considered in the present work, related to two different configurations of the facility: ICE (Integral Circulation Experiment) and HERO (Heavy liquid mEtal pRessurized water cOoled tubes). The experimental campaign promoted in CIRCE-ICE test facility was aimed to investigate the thermal-hydraulics of a complex HLM system and to provide data for validation of computational tools. Two experimental tests have been analyzed in this activity: Test A, consisting in a transition from no-power to full power steady-state conditions, and Test I, consisting in a transition from gas-enhanced circulation (GEC) to natural circulation (NC), simulating a protected loss of heat sink plus loss of flow accident. The computational activity has been addressed to investigate the capability of RELAP5-3D© to predict thermal stratification phenomenon in an HLM pool. Several examples were found in literature concerning the simulation of large tanks with RELAP5. The state of art on the simulation of the thermal stratification in large pool has been confirmed by the calculations performed on CIRCE-ICE. The mono-dimensional approach, using a single channel for pool modelling, highlights discrepancies with the experimental data, failing on the prediction of the axial temperature profile. A high peak temperature in the middle of the tank, which is not observed by the experiment, is caused by a total absence of the natural flow within the pool. In order to verify how the axial conduction within the fluid can improve the computational results, a thermal conduction model has been implemented in the nodalization, using several heat structures that couple adjacent meshes. As expected, the axial conduction reduces the peak but not enough to match the experimental temperature profile. To reproduce buoyancy within the LM tank, the pool has been divided in three vertical channels, connected with cross junctions. This approach (model #1) has shown good capabilities on the thermal stratification evaluation. The qualitative trend is well reproduced, predicting two relevant stratifications in the upper and in the middle volumes of the pool. In the lower part of the tank, the LBE temperature is very well simulated but, at 4 m, the plateau temperature is under-predicted of about 15 K and this discrepancy is maintained up to the cover gas. R5-3D was improved with a fully integrated multi-dimensional (MULTID) modelling scheme, mainly developed for volumes where the movement of the fluid is preferably 1D. The MULTID component has been used for the simulation of CIRCE pool. In addition, the fuel pin simulator (FPS) modelling has been improved for the subchannel analysis (model #2). Several figures of merit have been selected, assessing the capabilities of the two models to reproduce thermal-hydraulics of an HLM cooled pool-type system in safety-relevant operations. The comparison with experimental data has highlighted excellent capabilities of the two models to predict thermal-hydraulics of the main flow path, managing to evaluate the most important features: LBE mass flow rate in both GEC and NC conditions, heat exchange within FPS, heat exchanger (HX) and decay heat removal system (DHR), and heat losses. In addition, model #2 assesses RELAP5-3D abilities as a subchannel analysis code, in both GEC and NC operations. The effect of the radial conduction has been evaluated: specific heat structures have been implemented to reproduce thermal conduction through adjacent subchannels. This analysis has shown small effects of the radial conduction, even if, in low flow rate regimes, such as NC operation, it provides not negligible improvement on the prediction of the temperature profile. In this case, the simulation is in good agreement with experimental data; the highest discrepancies are observed in the edge of the bundle (4 degrees) where the errors can be justified by the uncertainties related to the thermocouples positions. Focusing on the pool simulation, the MULTID component has introduced relevant improvements on the prediction of the thermal stratification phenomenon. The two relevant stratifications have been observed by the calculation. The temperature in the lower part matches very well the experimental measurements. The lower stratification level is well predicted and the temperature hot plateau underestimation has been reduced to 5 K. After the transition from GEC to NC, both the models are able to predict the upper stratification attenuation and the movement of the lower one below the DHR outlet. In the long term, the two calculations provide the same temperature profile that matches well the experimental trend, limiting the discrepancies below 4 degrees. Evaluation of the axial conduction within the pool has been performed, highlighting limited effects, especially when the natural circulation (NC) inside the pool is considered. HERO test section was employed in CIRCE facility to investigate thermal-hydraulics of a double wall bayonet tube (DWBT) steam generator (SG), in a relevant configuration for ALFRED SG. Moreover, a validation benchmark was proposed within the H2020 SESAME project to asses the capabilities of different computational tools to predict the main thermal-hydraulic phenomena in an HLM-cooled pool-type facility. UNIROMA1 supported the definition and the realization of the experimental campaign, developing a nodalization scheme, based on the validated CIRCE-ICE model, for the pre-test analysis. The calculations have investigated different transient scenarios, highlighting the capabilities of HERO test section to guarantee sufficient NC to remove the decay heat in the short term. Based on these results, a set of three experimental tests has been performed, consisting in three protected loss of flow accidents (PLOFAs), occurred during the normal operation of the facility. The numerical activity has concerned two of the three tests, adopting an improved nodalization scheme. Regarding post-test analysis of Test 3, simulation of the full power conditions is globally in agreement with experimental data for all the primary circuit physical quantities monitored, including the thermal stratification phenomenon; some discrepancies are highlighted on the secondary side, mainly due to the lack of some information which determines large uncertainties on the boundary conditions related to the operation of the secondary loop. Starting from the full power steady-state conditions, two transient calculations have been performed, assuming the same boundary conditions, except for the feedwater (FW) mass flow rate after the transition event. Case 1 assumes the reference value of the secondary flow rate (0.078 kg/s), obtained with the energy balance equation applied to the FW pre-heater. This calculation highlights an overestimation of the power removed by the SG. A second calculation has been performed adjusting the total secondary flow rate to 0.047 kg/s; that value guarantees the correct SG power. Both the simulations are in good agreement with experimental data in the first 200 s, reproducing very well the first minutes after the transition event. After the Ar injection cut off, the first calculation provides a good estimation of the minimum value of the LBE MFR, underpredicted by the second calculation of about 2 kg/s. The long term behavior strongly depends on the feedwater mass flow rate. Case 1 shows an overestimation of the whole system energy unbalance, leading to the overprediction of the natural circulation contribution and of the cooling trend. The SG power balance analysis has highlighted an overestimation of the power removed of about 30% of the experimental value. The large uncertainties related to the measurement of the secondary system quantities have suggested the calibration of the secondary flow rate to obtain the correct SG power removed. This assumption has been justified comparing the experimental flow rate, acquired at the inlet of the tubes 0 and 4, with the simulation results: the experimental data are underestimated but the calculation results remain within the experimental uncertainty bands. Looking at the primary system, the assumption of a lower FW flow rate leads to a better agreement with experimental data, providing a good estimation of the long term behavior. Some discrepancies are still maintained on the secondary side, where the steam outlet temperature is overpredicted by the code. The differences could be due to a not perfect agreement of the powder thermal conductivity, which represents a large source of uncertainties. This opens the possibility to continue the post-test analysis for the secondary side, leading an improved experimental results analysis, despite the good global results obtained. Similar conclusions are obtained from the post-test analysis of Test 1, confirming good capabilities of R5-3D to reproduce thermal-hydraulics of HLM-cooled systems in safety-relevant operations. In addition, the subchannel analysis performed has highlighted a good prediction of the subchannel thermal-hydraulics within the FPS bundle in the postulated transient accident. The Phénix Dissymmetric End-of-Life test, proposed for a benchmark exercise within the H2020 SESAME project, offered useful data for the analysis of more complex systems where asymmetrical effects could play a relevant role in the system thermal-hydraulics. As a participant to the validation benchmark, in collaboration with ENEA, UNIROMA1 has developed a detailed nodalization of the reactor, including a three-dimensional modelling of the pools and an assembly per assembly core modelling in the active region, suitable for a neutron kinetic and thermal-hydraulic (NK-TH) coupling calculation. The full power calculation has highlighted a good capability of the code to reproduce the normal operation of the reactor. Starting from the steady-state results, the transient calculation has been performed assuming the dissymmetric test boundary conditions provided by CEA. The asymmetric distribution of the flow rate through the secondary loops leads to an asymmetric operation within the primary system, which is well predicted by R5-3D. The asymmetrical operation of the two secondary systems leads to a dissymmetric evolution of the thermal-hydraulics within the cold pool. Good agreements have been observed between experiment and simulation. In particular, the movement of the hot sodium within the cold pool is well predicted by the code, which is able to predict the local peak temperature at the PP1 inlet. At this regard, the three-dimensional momentum equation adopted in the MULTID modelling seems to provide a good instrument for the evaluation of the temperature and flow distribution within large volumes. Safety and reliability are relevant aspects of the development of GEN IV reactors. In this framework, LMFRs present peculiar characteristics related to the thermophysical properties of the coolant, basically regarding the possibility of coolant freezing, that can occur in long term DHR operation if the thermal power removed by the DHR system exceeds the decay residual power. For this reason, in LMFRs, the DHR system must ensure an efficient power removal, avoiding to overcome technological limits in terms of maximum temperatures, and must prevent coolant freezing in the grace time period. In addition, according to GIF guidelines, passive DHR systems are needed to prevent unexpected evolution of accidental scenarios following a total loss of the continuous electrical power supply. The solution consisting of an isolation condenser (IC) immersed in a water tank, acting as a final heat sink, could meet the above-mentioned characteristics. The operation of such a system, is based on in-tube condensation under NC condition and pool boiling. For this reason, a validation process is required for STH codes, such as R5-3D. In this framework, PERSEO facility provided useful experimental data. UNIROMA1 developed a 1D model of the facility. In order to reduce the computational cost, a mono-dimensional model of the pools (three parallel pipes with cross junctions) has been included. This modelling choice is justified by the negligible effect of the thermal stratification on the system thermal-hydraulics. The numerical activity has shown satisfactory capabilities of the code to reproduce the safety performance of the passive system. The main limitation observed in R5-3D calculations, and in almost all the STH codes adopted in the benchmark exercise, is the significant underestimation of the power exchanged between the HX and the HX pool. This can be attributed to the underprediction of the heat transfer coefficient (HTC) in both the tube-side and pool-side, where the condensation under natural circulation conditions and the pool boiling are outside of the validity ranges of the correlation fully integrated into the code. For this reason, the main improvement adopted in the nodalization has been the application of a constant multiplicative factor (2.4) to the HTC on both the sides. The main target of the validation activity was to qualify R5-3D STH code to support the design of the GEN IV nuclear power plants. The numerical activity was used to explore and validate different modelling approaches and the acquired know-how has been used to support the design of ALFRED reactor, investigating the reference and the improved configurations of the LFR concept. The first numerical activity has concerned the reference concept of the reactor, developed within the LEADER project. Based on the experience learned during the analysis of the experimental campaign performed in CIRCE and on the references found in literature, a thermal stratification phenomenon was expected into the main pool of the reactor. For this reason, a detailed three-dimensional model of the pool has been developed. In addition, an assembly per assembly core modelling, assuming the approach adopted for Phénix simulations, has been developed, allowing the calculation of the power distribution at the beginning of life (BOL) of the reactor using the NK-TH methodology. It is based on the RELAP5-3D/PHISICS coupling calculation. The full power calculation has highlighted a relevant thermal stratification, of about 70 K, in the upper part of the pool, that is not been involved in the primary flow path. Thermal stratification represents a significant technological issue. This was one of the reasons that encourages the designers to develop a revised concept of ALFRED reactor. In this frame, the solution was to include an internal structure, within the Reactor Vessel, that forces the cold lead, exiting the SG, to move upward and then, passing through specific holes in the upper part of the IS, to move downward towards the core inlet. In this way, zones not involved in the flow path are avoided. A thermal-hydraulic model of the new configuration was developed in order to verify, among other issues, the absence of relevant thermal stratifications in both normal and accidental operations. For this purpose, a detailed MULTID component was developed to reproduce the pools of the reactor. The numerical activity has demonstrated the improved pool thermal-hydraulics: significant thermal stratifications are avoided in both full-power operation and SBO scenario. Another issue is related to the possibility of coolant freezing in all the operative conditions, including accidental scenarios. A solution that limits the maximum temperatures in the first instants after the transition event and modulates the power removed from the primary coolant in the long term, has been proposed, adopting an IC, immersed in water pool, equipped with non-condensable (nitrogen) tank. The non-condensable tank allows to passively modulate the power removed by the DHR system, by injecting nitrogen within the IC bundle. The non-condensable flow rate is passively controlled by the pressure difference between the gas tank and the water system. The numerical activity has been aimed to verify the revised concept of the DHR system under a postulated protected loss of offsite power (PLOOP). The simulation shows the capability of the safety system to restrict the maximum temperatures in the first phases of the transient within the technological limits. In the long term, as the depressurization of the secondary system occurs, nitrogen is injected within the IC bundle, degrading the heat exchange. This is enough to limit the power removed by the DHR system to the decay heat value, limiting the Pb minimum temperature to 630 K, about 30 K higher the Pb freezing point. Finally, the last chapter of the thesis has proposed a best estimate plus uncertainty (BEPU) methodology, based on a statistical exploration of the input space considering the associated uncertainties altogether and the analyses of the responses with several validation metrics. The methodology consists in the RELAP5-3D/RAVEN coupled calculation. The objective has been to verify the coupled methodology for the uncertainty quantification of LM-cooled systems. For this purpose, the experimental campaign performed on NACIE facility and the thermal-hydraulic model developed by UNIROMA1 has been considered. The UQ has been based on the perturbation of the input space following a Monte Carlo sampling, propagating the input uncertainties. The analysis of the main outcomes related to selected FOMs, has been performed with three comparison metrics, fully integrated into RAVEN tool. The application of the comparison metrics has shown the capabilities of the methodology, highlighting the merits and the weaknesses of the thermal-hydraulic model. The future perspectives could be the application of the model to more complex models to increase the validation process of the methodology and to apply it to the verification process of the new NPP concepts. The main outcomes and guidelines coming out during the research activity are summarized in four main sections: 1. Pool modelling: • Several modelling approaches have been explored; • Mono-dimensional approach based on a single equivalent channel fails to reproduce large pool thermal-hydraulics; • Mono-dimensional modelling approach, consisting in more parallel channels connected with cross junction, is preferable when satisfactory prediction of pool thermal-hydraulics are required, even if phenomena such as thermal stratification do not assume relevant role in the behavior of the whole system. This modelling approach allow to reduce computational cost; • MUILTID modelling approach is preferable if relevant thermal stratification phenomena are expected, justifying the computational cost increase; • Calculations have highlighted the importance of specific heat transfer correlations applicable for large pool geometries, that are not implemented in the actual version of RELAP5-3D. In this framework, UNIROMA1 is actually working on a modified version of RELAP5 mod. 3.3 that, among other implementations, will include specific correlation for heat transfer in large pools. 2. Liquid metal system operation: • Calculations have shown satisfactory capabilities of R5-3D to reproduce primary system if LMFR under safety-relevant conditions; • Simulation performed in the frame of ALFRED have highlighted the importance of STH codes in the verification and licensing process of an innovative NPP; • R5-3D has shown satisfactory capabilities as subchannel codes; Several modelling choices are explored. 3. Passive safety system: • Good capabilities of R5-3D to reproduce safety performances of a passive DHR system based on an IC immersed in a water tank; • Main limitations related to the prediction of the power removed by the IC, due to the lack of a specific heat transfer correlation for pool boiling and in-tube condensation under NC; • Simulation performed in the frame of the revised ALFRED concept demonstrates the safety capabilities of the DHR system designed for the reactor. Maximum temperatures are restricted within the technological limits and lead freezing is prevented in the grace time period. 4. BEPU methodology: • Methodology based on a statistical exploration of the input space considering the associated uncertainties altogether; • Three comparison metrics have been used for the analysis of the main outcomes, highlighting the merits and the weakness of the thermal-hydraulic model; • The future perspectives could be the application of the model to more complex models to increase the validation process of the methodology and to apply it to the verification process of the new NPP concepts.