advanced numerical simulation and modelling for reactor...
TRANSCRIPT
EPJ Nuclear Sci. Technol. 6, 42 (2020)© C. Demazière et al., published by EDP Sciences, 2020https://doi.org/10.1051/epjn/2019006
NuclearSciences& Technologies
Available online at:https://www.epj-n.org
REVIEW ARTICLE
Advanced numerical simulation and modelling for reactorsafety � contributions from the CORTEX, HPMC, McSAFEand NURESAFE projectsChristophe Demazière1,*, Victor Hugo Sanchez-Espinoza2, and Bruno Chanaron3
1 Department of Physic, Division of Subatomic and Plasma Physics, Chalmers University of Technology, 412 96 Gothenburg,Sweden
2 Institute for Neutron Physics and Reactor Technology (INR), Karlsruhe Institute of Technology (KIT),Hermann-vom-Helmholtz-Platz-1, 76344 Eggenstein-Leopoldshafen, Germany
3 Commissariat à l’Energie Atomique et aux Energies Alternatives, Centre de Saclay, 91191 Gif-sur-Yvette Cedex, France
* e-mail: d
This is an O
Received: 12 March 2019 / Accepted: 4 June 2019
Abstract. Predictive modelling capabilities have long represented one of the pillars of reactor safety. In thispaper, an account of some projects funded by the European Commission within the seventh FrameworkProgram (HPMC and NURESAFE projects) and Horizon 2020 Program (CORTEX and McSAFE) is given.Such projects aim at, among others, developing improved solution strategies for the modelling of neutronics,thermal-hydraulics, and/or thermo-mechanics during normal operation, reactor transients and/or situationsinvolving stationary perturbations. Although the different projects have different focus areas, they all capitalizeon the most recent advancements in deterministic and probabilistic neutron transport, as well as in DNS, LES,CFD and macroscopic thermal-hydraulics modelling. The goal of the simulation strategies is to model complexmulti-physics and multi-scale phenomena specific to nuclear reactors. The use of machine learning combinedwith such advanced simulation tools is also demonstrated to be capable of providing useful information for thedetection of anomalies during operation.
1 Introduction
The safe and reliable operation of nuclear power plantsrelies on many intertwined aspects involving technologicaland human factors, as well as the relation between those.On the technological side, the pillars of reactor safety arebased on the demonstration that a reactor can withstandthe effect of disturbances or anomalies. This includes theprevention of incidents and should an accident occur, itsmitigation.
Predictive simulations have always been one of thebackbones of nuclear reactor safety. Due to the extensiveefforts the Verification and Validation (V&V) of thecorresponding modelling software these represent, most ofthe tools used by the industry are based on coarse mesh inspace and low order in time approaches developed whencomputing resources and capabilities were limited. Becauseof the progress recently made in computer architectures,high performance computing techniques can be used for
pen Access article distributed under the terms of the Creative Comwhich permits unrestricted use, distribution, and reproduction
modelling nuclear reactor systems, thus replacing thelegacy approaches by truly high-fidelity methods.
In parallel with the more faithful modelling of suchsystems, the monitoring of their instantaneous state isbecoming increasingly important, so that possible anom-alies can be detected early on and proper actions can bepromptly taken. On one hand, over 60% of the currentfleet of nuclear reactors is composed of units more than30 years old, therefore, operational problems are expectedto be more frequent. On the other hand, the conservatismin design previously applied to the evaluation of safetyparameters has been greatly reduced, thanks to theincreased level of fidelity achieved by the currentmodelling tools. As a result, nuclear reactors are nowoperating more closely to their safety limits. Operationalproblems may be also accentuated by other factors (e.g.use of advanced high-burnup fuel designs and heteroge-neous core loadings).
In this paper, a brief account of four projects previouslyor currently funded by the European Commission in thearea of the simulation and themonitoring of nuclear reactorsystems is given. Despite the differences in nature betweenthose projects, the key objectives and achievements with
mons Attribution License (http://creativecommons.org/licenses/by/4.0),in any medium, provided the original work is properly cited.
2 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
respect to advanced numerical simulation and modellingfor reactor safety will be given particular emphasis. Thepaper will conclude with some recommendations for thefuture.
A glossary defining all the used abbreviations can befound at the end of the paper.
2 Short description of the respective projects
2.1 CORTEX
The CORTEX project (with CORTEX standing for CORemonitoring Techniques and EXperimental validation anddemonstration) is a research and innovation actionfinanced by the European Commission. The projectformally started on September 1, 2017 for a duration offour years. The overall objective of CORTEX is to developa core monitoring technique allowing the early detection,localization and characterization of anomalies in nuclearreactors while operating.
Being able to monitor the state of reactors while theyare running at nominal conditions is extremely advanta-geous. The early detection of anomalies gives the possibilityfor the utilities to take proper actions before such problemslead to safety concerns or impact plant availability. Theanalysis of measured fluctuations of process parameters(primarily the neutron flux) around their mean values hasthe potential to provide non-intrusive online core monitor-ing capabilities. These fluctuations, often referred to asnoise, primarily arise either from the turbulent character ofthe flow in the core, from coolant boiling (in the case of two-phase systems), or from mechanical vibrations of reactorinternals. Because such fluctuations carry valuable infor-mation concerning the dynamics of the reactor core, onecan infer some information about the system state undercertain conditions.
A promising but challenging application of corediagnostics thus consists in using the readings of the(usually very few) detectors (out-of-core neutron counters,in-core power/flux monitors, thermocouples, pressuretransducers, etc.), located inside the core and/or at itsperiphery, to backtrack the nature and spatial distributionof the anomaly that gives rise to the recorded fluctuations.
Although intelligent signal processing techniques couldalso be of help for such a purpose, they would generally notbe sufficient by themselves. Therefore, a more comprehen-sive solution strategy is adopted in CORTEX and relies onthe determination of the reactor transfer function orGreen’s function, and on its subsequent inversion.
TheGreen’s function establishes a relationship betweenany local perturbation and the corresponding space-dependent response of the neutron flux throughout thecore. In CORTEX, state-of-the-art modelling techniquesrelying on both deterministic and probabilistic methods arebeing developed for estimating the reactor transferfunction. Such techniques are also being validated inspecifically designed experiments carried out in tworesearch reactors.
Once the reactor transfer is known, artificial intelli-gence methods relying on machine learning techniquesare used to recover from the measured detector signals
the driving anomaly, its characteristic features andlocation.
More information about the CORTEX project can befound in [1].
2.2 HPMC and McSAFE
The projects HPMC (High Performance Monte CarloMethods for Core Analysis) and McSAFE (High Perfor-manceMonte CarloMethods for SAFEty Analysis) are twocollaborative research projects funded by the EuropeanCommission in the seventh Framework Program (2011–2013) and Horizon 2020 Program (2017–2020) with themain goal of developing high fidelity multi-physicssimulation tools for the improved design and safetyevaluation of reactor cores. The peculiarity of HPMCand McSAFE is the focus on Monte Carlo neutronicssolvers instead of deterministic ones, in order to take profitof the huge and cheap available computer power currentlyavailable.
The scientific goal of the HPMC was the “proof ofconcept“ of newly developed multi-physics codes fordepletion analysis taking into account thermal hydraulicfeedbacks, static pin-by-pin full LWR core analysisconsidering local feedback, and the development of time-dependent Monte Carlo codes including the behaviour ofprompt and delayed neutrons for accident analysis.
Based on the success and promising results of theHPMC project, the goal of the McSAFE project thatstarted in September 2017 is to become a powerfulnumerical tool for realistic core design, safety analysisand industry-like applications of LWRs of Generation IIand III [2,3]. For this purpose, the envisaged developmentswill permit to predict important core safety parameterswith less conservatism than current state-of-the-artmethods and they will make it possible to increase theperformance and operational flexibility of nuclear reactors.Moreover, the multi-physics coupling developments arecarried out within the European Simulation platformNURESIM developed during different projects in theseventh Framework Program such as NURESIM, NURISPand NURESAFE [4], heavily relying on the open-sourceSALOME-software platform. In this context, the EuropeanMonte Carlo solvers MONK, SERPENT, and TRIPOLIare coupled with the subchannel thermal-hydraulic codeSUBCHANFLOW and with the thermo-mechanic solversTRANSURANUS using the ICoCo-methodology [5]. Atpresent, the application and demonstration are done forLWRs and SMRs. However, the peculiarity of the codesand methods make their application possible to the Gen-IIIand Gen-IV reactors as well as to research reactors, forwhich the complicated geometry and physics of the corecan only be adequately simulated by Monte Carlo codes.
Finally, all developed methods and codes are validatedagainst plant data of European VVER and PWR plants aswell as using test data of the SPERT Series IV E REA.
2.3 NURESAFE
NURESAFE (NUclear REactor SAFEty simulation plat-form) is a collaborative research project funded by the
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 3
European Commission in the seventh Framework Program[5,6]. The project started early 2013 for a duration of threeyears. The main objective of NURESAFE was to develop aEuropean reference tool for higher fidelity simulation ofLWR cores for design and safety assessment.
The simulation tool developed by the NURESAFEproject includes deterministic core physics codes, thermal-hydraulics and fuel thermo-mechanics codes, all integratedin a software platform whose name is NURESIM. Thisplatform provides a capability for code coupling, capabilityof paramount importance as the main phenomena occur-ring in reactors involve an interaction between theabovementioned physics. The NURESIM platform alsooffers an uncertainty quantification, which is necessary forvalidation and safety evaluation.
The scope of the NURESIM platform includes thesimulation of steady states of LWRs and design basisaccidents of LWRs. This platform was initially created inthe framework of former collaborative projects within thesixth and seventh Framework Programs (NURESIM andNURISP), during which core physics and thermal-hydraulics codes were first integrated. In NURESAFE,the platform was extended to more codes, particularly fuelthermo-mechanics codes. An important part of theNURESAFE work was also dedicated to:
– the demonstration of the multi-physics capability of theplatform;–
advanced CFD modelling; – uncertainty quantification and validation.3 Key objectives with respect to advancednumerical simulation and modellingfor reactor safety
3.1 Introduction
As earlier mentioned, most of the modelling tools used bythe nuclear industry were developed when computingresources and capabilities were limited. Although nuclearreactors are by essence multi-physics and multi-scalesystems, the techniques that were then favoured relied onmodelling the different fields of physics and sometimes thedifferent scales by different codes that were only thereaftercoupled between each other. In the current best-estimateapproaches, the modelling of neutron transport, fluiddynamics and heat transfer is thus based on a multi-stagecomputational procedure involving many approximations.
On the neutronic side, deterministic approaches havebeen used primarily, due to their lower computational costcompared to probabilistic methods (i.e. Monte Carlo).Deterministic tools nevertheless rely on many approxima-tions, with the neutron transport equation solved explicitlyafter reducing the complexity of the task at hand (typicallyusing space-homogenization, energy-condensation, andangular approximation techniques) [7]. The problem isfirst solved over a small region of the computationaldomain using approximate boundary conditions, and the“fine-grid” solution then computed is used for producingequivalent average properties locally. In a second step, aglobal “coarse-grid” solution is found for the full computa-
tional domain, in which only average local properties areconsidered, i.e., in which the true complexity of the systemis not represented explicitly. Typically, three to four of such“bottom-up” simplifications are used to model a full reactorcore. Although used on a routine basis for reactorcalculations, the approximations used in each of thecomputational steps are almost never corrected by theresults of the calculations performed in the following stepswhen a “better” (i.e. taking a larger computational domaininto account) solution has been computed.
In the probabilistic approach on the other hand, noequation as such is solved. Rather, the probability ofoccurrence of a nuclear reaction/process of a given type ona given nuclide at a given energy for a given incomingparticle (which can still exist after the nuclear interaction)is used to sample neutron life histories throughout thesystem [8]. Using a very large number of such histories,actual neutron transport in the system can be simulatedwithout requiring any simplification, and statisticallymeaningful results can be derived by appropriatelyaveraging neutron tallies. However, due to the size andcomplexity of the systems usually modelled, Monte Carlotechniques are extremely expensive computing techniques,which limited their use for routine applications in the past.
With the advent of cheap computing resources, boththe deterministic approach and the probabilistic approachare now being used on massively parallel clusters tocircumvent the limitations mentioned above. In thedeterministic case, the process of averaging (“bottom-up”) is now being complemented by a de-averaging process(“top-down”) in an iterative manner, so that a bettermodelling of the boundary conditions can be achieved usingthe information available from the coarser mesh. Themodelling of full cores in a single computational step is alsobeing contemplated. In the probabilistic case, the use oflarge clusters allowsmodelling full reactor cores, and effortsare being pursued to include the feedback effects inducedby changes in the composition and/or density of thematerials [9,10]. Due to the complexity and level of detailsin the deterministic approach based on the averaging/de-averaging process, there are situations where thedeterministic route can become quite expensive, beingalmost on par with the probabilistic route for high-fidelitysimulations.
On the thermal-hydraulic side, the strategy is toaverage in time and in space the local conservationequations expressing the conservation of mass, momentumand energy. The double averaging results in a set ofmacroscopic conservation equations that are tractable for alarge system as a nuclear reactor, unfortunately at theexpense of filtering the high-frequency and small-scalephenomena [7]. In addition, the averaging processintroduces new unknown quantities (expressing forinstance the wall transfer and possible interfacial transferbetween the phases) that are usually determined usingempirical or semi-empirical correlations. These correla-tions are heavily dependent on the flow regimes. Such amodelling strategy is often referred to as a system codeapproach. With the advent of cheap computing power,current efforts focus on modelling much finer scale usingCFD tools instead.
4 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
3.2 CORTEX
For the CORTEX project, since a majority of thediagnostic tasks are based on the inversion of the Green’sfunction, the key objectives in the area of advancednumerical simulation and modelling can be summarized asfollows: (a) the development of modelling capabilities forestimating the transfer function, (b) the validation of suchtools against experiments specifically designed for thatpurpose, and (c) the inversion of the reactor transferfunction using machine learning.
Concerning (a), one of the strategic objectives of theproject is to determine the area of applicability of existingtools for noise analysis and to develop new simulationtools that are specifically dedicated to the modelling of theeffect of stationary fluctuations in power reactors with ahigh level of fidelity. The ultimate goal is to developmodelling capabilities allowing the determination, for anyreactor core, of the fluctuations in neutron flux resultingfrom known perturbations applied to the system. Twotracks are followed. Existing low-order computationalcapabilities are consolidated and extended. Simultaneous-ly, advanced methods based on deterministic neutrontransport and on probabilistic (i.e. Monte Carlo) methodsare developed so that the transfer function of a reactorcore can be estimated with a high resolution in space,angle and energy. Since the modelling of the response ofthe system to a perturbation expressed in terms ofmacroscopic cross-sections is equally important as themodelling of the actual perturbation, large efforts arespent on converting actual noise sources into perturba-tions of cross-sections. For that purpose, emphasis is puton developing models for reproducing vibrations of reactorvessel internals due to FSI. Finally, the evaluation of theuncertainties associated to the estimation of the reactortransfer function is given particular attention, togetherwith the sensitivity of the simulations to input parametersand models.
Concerning (b), although the tools allowing estimatingthe reactor transfer function can be verified againstanalytical or semi-analytical solutions for simple systemsand configurations, the validation using reactor experi-ments specifically designed for noise analysis applications isessential. Two types of neutron noise measurements areconsidered: a so-called absorber of variable strength and aso-called vibrating absorber.
Finally, concerning (c), the backtracking of the drivingperturbation (not measurable) from the induced neutronnoise (measurable at some discrete locations throughoutthe core) is performed using machine learning. With thetools referred to above, the induced neutron noise formany possible scenarios of considered perturbations isestimated. The results of such simulations are thenprovided as training data sets to machine learningtechniques. Based on such training sets, the machinelearning algorithms have for primary objective to identifythe scenario existing in a nuclear core from the neutronnoise recorded by the in- and ex-core neutron detectorsand, when relevant, retrieve the actual perturbation (andits location).
3.3 HPMC and McSAFE
The major objectives of the HPMC project were thefollowing:
(a) optimal Monte Carlo-thermal-hydraulics coupling: theobjective was to realise efficient coupling of the MonteCarlo codes SERPENT and MCNP with the thermal-hydraulic subchannel codes SUBCHANFLOW andFLICA4, suitable for full core applications;
(b)
optimal Monte Carlo burn-up integration: the objec-tive was to realise an efficient integration of burnupcalculations in the Monte Carlo codes SERPENT andMCNP, suitable for full core applications;(c)
time-dependence capabilities in Monte Carlo methods:the objective was to develop an efficient algorithm formodelling time-dependence in the Monte Carlo codesSERPENT and MCNP, applicable to safety analysisand full core calculations.Based on the promising results of the HPMC project,the McSAFE project started in September 2017 with thegoal to move the Monte Carlo-based multi-physics codestowards industrial applications, e.g. simulation of deple-tion of commercial LWR cores taking thermal-hydraulicfeedback into account, analysis of transients such as REA.For this purpose, a generic and optimal coupling approachbased on ICoCo and the open-source NURESIM platformis followed for the coupling of the European Monte Carlosolvers such as MONK, SERPENT and TRIPOLI withsubchannel codes, e.g. SUBCHANFLOW and fuelthermo-mechanics solvers, e.g. TRANSURANUS. More-over, dynamic versions of TRIPOLI, SERPENT andMCNP6 coupled with SUBCHANFLOW are developedfor analysing transients. Especially, SERPENT/SUB-CHANFLOW is being coupled with TRANSURANUS forthe depletion analysis of commercial western PWR andVVER cores while considering thermal-hydraulic feed-back. Emphasis is put on the extensive validation of thetools being developed within McSAFE. For the validationof the depletion capabilities, plant data are used, whereasfor the validation of the dynamic capability of the coupledMonte Carlo–thermal-hydraulics codes under develop-ment, experimental data of unique tests e.g. the SPERTREA IV E are used. Finally, high fidelity tools based onMonte Carlo requires a massive use of HPC in order tosolve full cores at the pin level. Methods for optimalparallelization strategy, scalability of Monte Carlo-basedsimulations of depletion problems and time-dependentsimulations, are also scrutinized in the McSAFE project.Since memory requirements for such problems mayrepresent a limiting factor, methods for the optimal useof memory during depletion simulations of large problemsneeds to be further developed.
3.4 NURESAFE
The main objectives of NURESAFE were as follows:
– To enhance the prediction capability of the computationsused for safety demonstration of the current LWRnuclear power plants through the dynamic 3D coupling of
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 5
the codes, simulating the different physics of the probleminto a common multi-physics simulation scheme.
–
To advance the fundamental knowledge in two-phasethermal-hydraulics and develop newmulti-scale thermal-hydraulics models. Emphasis was put on couplinginterface tracking models with phase-averaged models.Moreover, pool and convective boiling were given specialattention, together with the physics of bubbly flow.–
To develop multi-scale and multi-physics simulationcapabilities for LOCA, PTS and BWR thermal-hydrau-lics, thus allowing more accurate and more reliable safetyanalyses. The aim was to develop a European referencetool for higher fidelity simulation of LWR cores for designand safety assessments. The delivery of safety-relevantindustry-like applications was also one of the primaryobjectives of the project, so that the various applicationscould be used by the industry at the completion of theproject.–
To develop generic software tools within the NURESIMsoftware platform and to provide a support to developersfor integration of the codes into this platform.4 Key achievements with respect to advancednumerical simulation and modelling for reactorsafety
4.1 CORTEX
Since the start of the project, the key achievements in thearea of advanced numerical simulation andmodelling alongthe three objectives identified in Section 3.2 can besummarized as follows.
4.1.1 Development of modelling capabilities for estimatingthe transfer function
The work carried out so far is performed along severallines.
In the area of mechanical vibrations, an extensivereview of the past work on vibration of reactor internalswas carried out. The focus was on both obtaining acoverage of all possible sources of neutron noise, aphenomenological description of each corresponding sce-nario, and of the observed neutron noise patterns whenactual plant measurements were available. First simula-tions using thermal-hydraulic perturbations generated by asystem code were later fed into a FEM code modellingmechanical structures.
In parallel to those activities, neutronic capabilities arebeing developed. For coarse mesh approaches, threeparallel tracks are pursued. Nodal codes used for thesimulation of other core transients in the time-domain areused. To use some of these codes, the first step is to generatea set of time-dependent macroscopic cross-sections thatsimulate the movement of the fuel assemblies on a fixedcomputational coarse grid, based on the results of the FSIsimulations. Procedure are being implemented to generatethe whole set of cross sections. In addition to the use of
existing time-dependent tools with a set of time-dependentcross sections, another approach is pursued based on thedevelopment of an ad-hoc software relying on FEM. TheFEM method has a large versatility for solving balanceequation using different spatial meshes and a code is beingdeveloped along those lines. It will offer the possibility inthe future to have a moving mesh following the vibrationcharacteristics determined from the FSI calculations. Themain advantage of the FEM route lies with the fact thatonly static macroscopic cross sections for the initialconfiguration of the core are necessary. Finally, a thirdand complementary approach based on a mesh refinementtechnique in the frequency domain is being developed. Themodelling of vibrating reactor internals requires thedefinition of perturbations on very small spatial domainscompared to the size of the node size used in coarse meshmodelling tools. This makes it necessary to develop meshrefinement techniques around the region where theperturbation exists. This mesh refinement technique iscurrently implemented in a frequency-domain core simu-lator earlier developed. For fine mesh approaches,deterministic methods relying on the method of discreteordinates (Sn) are being developed. Moreover, a neutronnoise solver relying on the method of characteristics isbeing implemented. In probabilistic methods, an equiva-lence procedure between neutron noise problems in thefrequency-domain and static subcritical systems is beingdeveloped. A method using complex statistical weights anda modified collision kernel for the neutron transportequations in the frequency domain have been implementedin a Monte-Carlo code. Likewise, another method usingcomplex-valued weights in the frequency domain has beenimplemented.
As can be seen above, several complementaryapproaches are being developed. They either rely onexisting codes or codes specifically developed for noiseanalysis.Moreover, these codes work either in the time or inthe frequency domain. These tools use either a coarse-meshapproach (possibly with a moving mesh) or a fine-meshapproach regarding the spatial discretization. Finally, bothdeterministic and probabilistic methods are considered.
4.1.2 Validation of the modelling capabilities againstexperiments
Concerning the validation of such tools against experi-ments specifically designed for neutron noise, two researchfacilities are used: the AKR-2 facility at TUD, Dresden,Germany, and the CROCUS facility at EPFL, Lausanne,Switzerland. Pictures of those two facilities are given inFigure 1.
The perturbation was simultaneously recorded by 7 and11 neutron detectors, for the first AKR-2 and CROCUScampaigns, respectively, located throughout the respectivecores, together with the recording of the actual perturba-tion introduced. The data acquisition systems weresuccessfully benchmarked against an industry-grade dataacquisition system from TUV Rheinland ISTec GmbH. Interms of perturbations, AKR-2 has the ability to perturbthe system in two ways: either by rotating a neutron
Fig. 2. Example of the reactor response to a localized absorber of variable strength unrolled as two-dimensional images (courtesy ofUniversity of Lincoln) [11].
Fig. 1. Overview of the CROCUS and AKR-2 facilities.
6 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
absorbing foil (thickness of 0.02 cm� length of 25 cm�width of 2 cm) along a horizontal axis or by moving aneutron absorbing disc (thickness of 1.0mm� diameter of12.7mm) along a horizontal axis. In the former case, the foilrotates at a distance of 2.98 cm from its axis at a frequencyof up to 2.0Hz, whereas in the latter case, the disc is movinghorizontally with a maximum displacement amplitude of20 cm at a frequency up to 2.0Hz. At CROCUS, up to18 fuel rods located at the periphery of the core can bedisplaced laterally with a maximum displacement upto±2.5mm from their equilibrium positions at a frequencyup to 2Hz. The first noise measurements for the three typesof noise sources (rotating absorber and vibrating absorberat AKR-2; vibrating fuel rods at CROCUS) have beenperformed as part of the validation of the data acquisitionsystems.
Since both the perturbations and the correspondinginduced neutron noise are recorded in the experimentsdescribed above, such experiments can be used to validatethe neutronic tools aimed at estimating the Green’sfunction of the reactor and being developed within
CORTEX. Such noise measurements, where both theperturbations and the corresponding neutron noise arerecorded, represent a world premiere.
4.1.3 Inversion of the reactor transfer function usingmachine learning
Preliminary tests were performed using simulated signals,either in the time domain or in the frequency domain.Several scenarios corresponding to different types of noisesources were considered: localized absorbers of variablestrength in the frequency domain, travelling perturbationsalong fuel channels in the frequency domain, fuel assemblyvibrations in the time domain, and inlet coolant perturba-tions in the time domain. First successful machine learningtests on the absorbers of variable strength were based on“unrolling” the three-dimensional induced neutron noiseinto the juxtaposition of two-dimensional images, eachcorresponding to the plane-wise response of the reactor coreto the perturbation [11]. Figure 2 represents such two-dimensional information that was then fed to a Deep CNN
Fig. 3. 3D pin power predicted by SERPENT/SUBCHAN-FLOW for the PWR UOX/MOX core [10].
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 7
to retrieve the actual location of the perturbation. Therecovery of the exact spatial location of the noise source wasthereafter improved by using instead a three-dimensionalCNN, so that the axial coupling information could be fullyexploited in the unfolding [12]. In addition, both theabsorber of variable strength data and the travellingperturbation data were used. The network could bothrecognize the type of perturbation applied and recover theactual location of the perturbation being applied. For thetime-domain data, the different scenarios could besuccessfully identified using a LSTM network.
4.2 HPMC and McSAFE4.2.1 Optimal Monte Carlo-thermal-hydraulics coupling
The HPMC project demonstrated the potentials andcapabilities of Monte Carlo-based multi-physics coupledcodes for improved static core analysis taking localinterdependencies between neutronics and thermal-hydraulics into account. At the completion of the project,two coupled codes, SERPENT/SUBCHANFLOW andMCNP/SUBCHANFLOW, had been developed for staticfull core simulations at the pin level. Those codes weresuccessfully applied to the analysis of a PWR core withUOX andMOX fuel assemblies, while taking local thermal-hydraulic feedback into account and using HPC clusters[9,10]. As an illustrative example, the capability of thecoupled code SERPENT/SUBCHANFLOW to perform apin-level analysis of a full PWR core with local thermal-hydraulic feedback is shown in Figure 3. The problemconsists of 55 777 neutronic nodes (pins and guide tubes),2.2 million fluid cells, as well as 23.4 million solid cells(thermal-hydraulic solver). A total of 4� 106 neutrons percycle and 650 inactive and 2500 active cycles were used inthe SERPENT calculations. The simulation was performedat the KIT IC2 HPC cluster using 2048 cores. A convergedsolution was achieved after 5.8 CPU-year (1.03 days).
4.2.2 Optimum Monte Carlo burn-up integration
Another important outcome was the exploration anddevelopment of various schemes for stable depletioncalculation using Monte Carlo codes such as the SIEmethod [13] for stable steady-state coupled Monte Carlo-thermal-hydraulics calculations.
4.2.3 Time-dependence capabilities in Monte Carlomethods
A highlight of the project was the implementation of atime-dependence option in MCNP5 (dynMCNP) thatrequired source code modifications [14]. This optionincludes the generation and decay of delayed neutronprecursors, possible control rods movement, etc. To reducethe statistical error in the generated reactor power insuccessive time intervals, a method of forced decay ofprecursors in each time interval was implemented.Moreover, variance reduction methods (like the branchlesscollision method) were introduced. Thermal-hydraulicfeedback was also implemented. To let the time-dependentthermal-hydraulic calculations take the heating historyinto account, further extensions of the codes werenecessary.
Finally, various ways for parallel execution of a MonteCarlo calculation using the MPI and OpenMP applicationprogramming interfaces were investigated and their efficien-cy measured in terms of the speedup factor. For applicationon large computer clusters with different computer nodesandmultiple processors per node, the optimumcombinationof MPI and OpenMP was determined. Application ofOpenMP was introduced in the SERPENT2 code. TheMCNP codewasmodified to use all available processor coresfor neutron history simulation [15].
The main achievements close to the midterm of theMcSAFE-project are described hereafter.
Full core multi-physics depletion: Methods for deple-tion of full core using Monte Carlo codes are beingdeveloped. First of all, the efficiency and stability ofMonte Carlo burnup simulations were studied by optimalcombination of free parameters that allow solving full coreproblems [16]. In addition, a collision-based domaindecomposition scheme for SERPENT2 is being developedto solve large-scale high-fidelity problems with largememory demands (e.g. full core pin-by-pin depletion). Forthis purpose, memory-intensive materials are split amongMPI tasks, enabling the memory demand to be dividedamong nodes in a high-performance computer [17].Investigations were also performed to identify thecomputational requirements for depletion calculationstaking thermal-hydraulic feedback into account for 3Dproblems (e.g. 5� 5 fuel assemblies mini-core) [18].Potential bottlenecks and limitations, e.g. huge RAM-requirements which increase linearly with the number offuel assemblies � 40 GB for eight fuel assemblies � couldbe identified. Alternatives were also proposed to overcomethe challenges, such as a collision-based domain decom-position.
Code integration: The European Monte Carlo codesTRIPOLI, SERPENT, and MONK as well as the fuel
Fig. 4. SALOME global view.
8 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
thermo-mechanics code TRANSURANUS were fullyintegrated into the European NURESIM simulationplatform (SUBCHANFLOW � SCF was already part ofthe platform). Each solver owns a specific meshing. Newflexible and object-oriented coupling schemes based on theICoCo-methodology are being developed for each of thecodes integrated into the NURESIM platform. Thefollowing coupled code versions are available: MONK/SCF, SERPENT/SCF, TRIPOLI/SCF.
Dynamical multi-physics calculations: Another impor-tant task in the McSAFE project is to extend general-purpose Monte Carlo codes (SERPENT2, TRIPOLI-4 andMCNP6) to dynamic version that can accurately calculatetransient behaviour in nuclear reactors considering localthermal-hydraulic feedback. New versions of Monte Carlocodes with time-dependent capabilities (called dynam-icMC) are at the end of the development phase for theanalysis of transients. TheseMonte Carlo codes are coupledwith the SCF thermal-hydraulic solver, thus leading to thecoupled codes: dynMCNP/SCF, dynTRIPOLI/SCF, dyn-SERPENT/SCF. The code extensions and modificationsare described in more detail in [14,19,20]. The couplingschemes must be appropriate for massive HPC-simula-tions. The peculiarity of time-dependent Monte Carlo is todescribe the behaviour of delayed neutrons, which have asignificant influence on the statistical uncertainty (stan-dard deviation) of the power prediction. An additionalchallenge is the short lifetime of prompt neutrons (roughly100 ms in an LWR) compared to the large decay time ofprecursors of delayed neutrons for the method develop-ment. To test the dynamic capability of the Monte Carlocodes, different REA scenarios are being developed withinMcSAFE.
4.3 NURESAFE4.3.1 Simulation platform
One of the main outcomes of the NURESIM and NURISPprojects was the release of the NURESIM platform that isheavily used in NURESAFE. The NURESIM platform isbased upon the software simulation platform SALOME.SALOME is an open-source project, (http://salome-platform.org), which implements the interoperabilitybetween a CAD modeller, meshing algorithms, visualisa-tion modules and computing codes and solvers, asrepresented in Figure 4. It mutualises a pool of generictools for pre-processing, post-processing and code coupling.Its supervision module provides functionalities for codeintegration, dynamic loading and execution of componentson remote distributed computing systems, and supervisionof the calculation. Support is provided to developers forintegration of the codes into the SALOME software and forproducing and managing the successive versions of theNURESIM platform on a dedicated repository. Innovativedeterministic and statistical methods and tools forquantification of the uncertainties developed withinNURESAFE give a better knowledge of conservatismsand margins.
The NURESIM platform provides a set of state-of-the-art software devoted to the simulation of normal operationand design basis accidents of LWR (i.e. BWR, PWR, andVVER). The platform includes 14 codes covering differentphysics: neutronics, thermal-hydraulics, fuel thermo-me-chanics at different scales, 2 thermal-hydraulics systemcodes, 2 single-phase CFD codes, 2 two-phase CFD codes, 3sub-channel thermal-hydraulic analysis codes, 2 advancedfuel thermo-mechanics codes, 2 DNS codes, 3 neutron-
Fig. 5. Distribution of power density (MW/m3, left) and coolant temperature (°C, right) at 86 s after the initiation of theMSLB event.
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 9
kinetics codes. All these codes were extensively bench-marked and validated against experiments during thecourse of the NURESAFE project.
SALOME is connected to URANIE, an open-sourceplatform aimed at providing methods and algorithmsabout uncertainty and sensitivity, and verification andvalidation analyses in the same framework (https://sourceforge.net/projects/uranie/). The URANIE and SA-LOME platforms work nicely together. Any calculationscheme developed in SALOME can be used withinURANIE.
Through the link with URANIE, users of theNURESIM platform successfully performed in theNURESAFE project sensitivity analyses andmodel calibra-tion studies.
4.3.2 3D dynamic coupling of codes
Individual models, solvers, codes and coupled applicationswere run and validated through modelling “situationtargets” corresponding to given nuclear reactor situationsand including reference calculations, experiments, andplant data. As safety analysis was the main issue within theproject, all these situation targets consisted in someaccidental scenarios. The challenging “situation targets”were selected according to the required coupling betweentwo different disciplines. Industry-like applications werereleased at the end of the project for the following “situationtargets”:
– Square lattice PWR MSLB; – One selected BWR ATWS; – VVER MSLB.The analysis also included uncertainty quantificationusing the URANIE open-source software.
The BWRATWS analysis framework featured coupledsimulations combining system thermo-hydraulics, 3Dneutronics, thermo-mechanical evaluation of fuel safetyparameters, and uncertainty evaluation. The MSLBtransient analysis provided more accurate assessment of
margins between predicted key parameters and safetycriteria. The outcome of the transient simulation wasevaluated with respect to local re-criticality and maximumreactor power level. As an illustrative example, the resultsof the PWR MSLB are presented hereafter.
A two-step modelling approach was applied. In the firststep, reference results were produced using the platformcodes with higher resolutions of coupling between corenodal and sub-channel scale. In the second step, CFDevaluations were included into the solution. In that way, animprovement in the prediction of the target safetyparameters could be achieved. In order to increase theconfidence of the CFD results, a validation was alsoperformed by comparing the calculation results withexperimental data from the HZDR test facility on coolantmixing ROCOM. The cross-section libraries were createdusing new methods of grid point selection [21]. Variouscombinations of system codes, core thermal-hydrauliccodes and neutronic codes were used. Figure 5 highlightsthe 3D distributions at time t=86 s after the initiation ofthe MSLB.
The obtained results confirmed that the NURESIMplatform is applicable for challenging coupled transients inPWRs. Furthermore, by accomplishing the coupling ofreactor dynamics codes and CFD codes, the superiority ofthe NURESIM platformwas demonstrated. The conductedadvanced calculations demonstrated the excellent statusand the readiness for industrial applications of theNURESIM platform and the integrated codes.
4.3.3 Advanced CFD modelling
Advancement in the fundamental knowledge of CFDmodelling was pursued and new models based on detailedDNS for momentum exchange and boiling heat transfersituations typical of LWR thermal-hydraulics were devel-oped. New benchmark data bases for fundamental andapplied problems were developed. The existing computa-tional multiphase flow strategies were first extended inorder to cope with a wider range of practical applications.
10 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
Novel methods for pool and convective boiling in a channelwere also developed. Advanced strategies for modellingturbulent bubbly flow in a channel and in a rod bundle wereanalysed. Finally, the novel models and simulationtechniques were implemented in codes, validated andapplied in this context. New versions of the CFD platformcodes NEPTUNE_CFD, TransAT and TRIO_U weredelivered to end users, including the most advancednumerical simulation features and the associatedmodellingapproaches for the physics pertinent to both PWRs andBWRs.
Three specific issues were addressed within NURESAFE:
– All-topology flow modelling by coupling interfacetracking models with phase-averaged models.–
DNS and LES of pool and convective boiling [22]. – DNS and LES of bubbly flows [23,24].4.3.4 Multi-scale and multi-physics simulations
In the area of multi-scale and multi-physics simulations ofLOCA, PTS and BWR thermal-hydraulics, multi-scaleandmulti-physics simulation capabilities for more accurateand more reliable safety analyses were developed.
LOCA is usually simulated with industrial versions ofthermal-hydraulic system codes. Although system codesare able to address most safety needs, the status and limitsof the current methods and tools for plant analysis werereviewed during the NURISP project and areas forimprovements were pointed out. Advanced tools andmethods for multi-scale and multi-physics analyses andsimulations of LOCA, including situations with deformedor ballooned rods and possible fuel relocation, weredeveloped. The addition to system thermal-hydrauliccodes of two-phase CFD tools and of advanced fuel modelsallowed revisiting these transients for more accurate andreliable predictions. This required improving and couplingCFD to system codes or improving system codes andsystem codes coupled with fuel thermo-mechanics codes.Furthermore, methods for uncertainty and sensitivityanalysis applied to system codes were improved. In thisframework, a special focus was put on the issue of thequantification of the uncertainties of the closure laws. Thiswork was based on a benchmarking of the possible methodsusing reflooding experimental data (FEBA andPERICLES).
Concerning PTS, better simulation capabilities wereachieved by improving the CFD modelling thanks to theanalysis of new experimental data (including TOPFLOWsteam-water tests and KAERI CCSF test). In addition,sensitivity and uncertainty methods were applied to CFDcodes and state-of-the-art methods on validation, uncer-tainty and sensitivity of CFD applications to reactor issueswere reviewed.
In the field of BWR thermal-hydraulics, progress in thesimulation of two-phase thermal-hydraulics phenomenaspecific to BWR was achieved. This includes dry-outprediction, transient core thermal-hydraulics and steaminjection in pressure suppression pool. CFD codes and sub-channel codes were used, improved and validated duringthe project.
5 Training, education and disseminationactivities
5.1 CORTEX
The dissemination of the project results is carried out alongfive parallel lines of actions: involvement of end users intothe project, organization of workshops, organization ofshort-courses, peer-reviewed publications, and presenta-tions at conferences and meetings.
Concerning the involvement of end users, the projectinvolves, beyond academic partners, research institutes,TSOs, utilities, fuel and reactor manufacturers, as wellas services companies. Those organizations are eitherdirectly contributing to the project as project partners orparticipating to the project via the Advisory End UserGroup, having a consultative role to the consortium.
Three workshops will be organized:
– two workshops on the experiments performed at theresearch reactors and on the validation of the neutronicmodels based on such experiments, where experimentalistsand modellers will present, describe and discuss theirresults;–
one (final) workshop on the demonstration of themethods developed within the project on actual plantdata. During this workshop, the entire consortium will:(a) summarize the findings and the lessons learntthroughout the project, (b) give recommendations ontechniques and instrumentations for core monitoring andsurveillance (in order to improve the reliability andsafety of the nuclear units); and (c) provide an outlook forthe future in this area.Eight short courses were/will be developed:
– two courses on reactor dynamics and neutron noise. Bothcourses were already given and had 47 registeredparticipants in total. The first course covered thefundamentals of reactor kinetics and the theory of smallspace-time dependent fluctuations. The second coursedealt with additional aspects, such as core thermal-hydraulics, its coupling to neutron kinetics and reactorstability, and included hands-on training on the AKR-2reactor at TUD;–
two courses/workshops on signal processingmethods andtheir applications. Both courses/workshops were alreadyarranged and attracted 64 attendees. The first course wasan introduction to basic techniques for signal analysisand their possible applications. The second course dealtwith advanced signal processing methods and statisticalcharacterization of plant measurements, which can beapplied to reactor core monitoring and dynamic sensorsurveillance;–
one hands-on training session on the simulation of reactorneutron noise in power reactors using a time-domainneutron kinetics code. The students will have theopportunity to model different types of disturbances,such as fuel assembly vibrations, inlet disturbances, flowfluctuations, etc. and study their effect on the neutronflux throughout the entire system;–
one course on uncertainty and sensitivity analysis.Emphasis will be put on the application of such methods
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 11
to the estimation of the reactor transfer function and thecorresponding neutron noise;
–
two hands-on training sessions on the two researchfacilities used in the project. The sessions will consist ofthe following exercises: reactor start-up procedures,control rod and critical experiments, and a set of neutronnoise experiments.In the area of publications, after 18 months as a runningproject, the following has been achieved:
– one journal publication (two more under review); – eight conference publications (ten more under review); – seven conference presentations.In addition, most of the deliverables (26 in total � tenwere already delivered) are/will be publicly available.
All the publicly available resources are directlyaccessible on the project website http://cortex-h2020.eu.In addition to the publications and deliverables listedabove, newsletters are distributed once a year. Theconsortium is also heavily using LinkedIn http://linkedin.com/company/cortex-h2020 to inform about theproject. Promotional materials (video, leaflet, poster) arealso available.
5.2 HPMC and McSAFE
The dissemination, education and training activities ofboth projects rely on the following pillars:
– dissemination plan for the identification of end usersand stakeholders (industry, academia, regulators,TSO);–
creation of a public website http://www.mcsafe-h2020.eu;–
organisation of a dedicated training course to be held inApril 2020 where the main tools of McSAFE will bepresented and demos of selected applications will beshown to the community;–
presentation of the main results at internationalconferences, e.g. PHYSOR, M&C, etc., publication ofthe main results in scientific journals, presentationat the NUGENIA Forum, the FISA Conference,etc;–
establishment of a Users’Group consisting of institutionswhich will get access to the use of the codes beingdeveloped and extended within McSAFE, for performingsimulations of own problems. Important feedback fromthe Users’ Group is expected regarding the capabilitiesand user-friendliness of the codes;–
creation of a Technical Advisory Board consisting ofselected experts of the community of stakeholdersand aimed at reviewing the McSAFE developmentsand at providing advice and comments on the maindevelopments;–
delivery of 57 deliverables in total, from which around 30are already finalized. Some of them are publicly availableon the project website;–
education and training of young scientists throughdoctoral programs and through the involvement ofmaster and bachelor students in the project at thedifferent partner institutions.5.3 NURESAFE
In order to foster the dissemination and facilitate the use ofthe platform codes, 15 training sessions of a few days eachwere given to the staff of the NURESAFE partners and toexternal users’ organisations during the course of theproject. The end users of the NURESIM platform and ofthe individual codes could thereafter efficiently use thetools and methods.
Two public NURESAFE general workshops were heldin Budapest on June 16–17, 2014 and in Brussels onNovember 4–5, 2015, respectively, in order to present thenew methods, models and functionalities that weredeveloped. About 50 people attended each of the work-shops.
Many publications were made:
– 12 articles were published in peer-reviewed journals(Annals of Nuclear Energy, International Journal of Heatand Fluid Flow, Multiphase Science and Technology,Nuclear Engineering and Design).–
28 presentations were delivered at internationalconferences (NURETH, ICONE, CFD4NRS, SNA-M&C, …).An active Users’ Group was set up when starting theproject. The objective was to give the opportunity toorganizations which were not members of the NURESAFEconsortium to use and test the new methods and tools.Five universities and companies were members ofthe NURESAFE Users’ Group: 3 non-European and2 European. They provided fruitful feedback on the useof the codes in some challenging situations, especially inthermal-hydraulics.
6 Utilization and cross-fertilization
CORTEX is by essence an international project, since oneof the partners is from USA and another one is from Japan.Moreover, the project gathers academic partners, researchinstitutes, TSOs, utilities, fuel and reactor manufacturers,as well as services companies in order to develop a coremonitoring technique in close dialogue with all relevantstakeholders. This will result in a method directlyapplicable for the industry. Finally, additional interestwas received from the USA for developing a similar methodas the one being developed in CORTEX.
Although neutron noise core monitoring has been usedin a “rudimentary” manner in some plants worldwide, themethodology proposed in CORTEX and relying onmachine learning techniques combined with dedicatedneutron noise simulations has never been attempted.Moreover, the development of neutron noise simulationcapabilities at an industrial level also represents a noveltyin CORTEX. Being able to infer from the detector readingsthe existence, location and features of possible anomalieswould represent a world-premiere.
If successful, the project will also be able to identify theroot-cause of some operational problems during exploita-tion. CORTEX will for instance investigate the increase ofthe neutron noise levels observed in some Pre-KONVOI
12 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
PWRs, events remaining unexplained and which, in somecases, led to reduced power operation or reactor scrams[25–28].
In the area of Monte Carlo simulations, the main toolsbeing developed within HPMC and McSAFE are high-fidelity tools, which can also provide reference solutionsto any low-order solution (e.g. nodal diffusion solvers)used by regulators and the industry in real life situationsand for licensing purposes. Since the tools are able toprovide unique full core solutions at the pin level takinginto account local thermal hydraulic feedback, such toolssubstantially improve the modelling accuracy whenpredicting depletion and simulating static core config-urations. In addition, the dynamic capability added tothe Monte Carlo codes coupled with thermal hydraulicsubchannel codes pave the way for the analysis oftransients (e.g. REA, MSLB) with an unequalledaccuracy as of today. Hence, these tools are very wellsuited for being used by the industry as a complement tolow-order solutions. Finally, for all cases where noexperimental data are available at a fine resolution, thesetools can predict local safety-relevant parameters. Withthe maturity of the being developed Monte Carlosolutions, the project will allow industry-like problemsto be modelled. This will provide a possibility to assessthe adequacy of deterministic based solution methodsthat are routinely used by the industry and that rely onmany approximations and limitations, as highlighted inSection 3.1.
The end users of the NURESIM software platformalso benefit since the end of the project from theimprovements made within the NURESAFE project insimulation capabilities, more precisely when e.g. theyperform industrial studies, safety analyses, optimisationof reactor operation and reactor design. The end users arethe members of the NURESAFE consortium (22organisations) and the members of the NURESAFEUsers’ Group (five organisations). They can be catego-rized into (1) utilities (three utilities operating themajority of the European fleet of nuclear reactors), (2)one reactor and fuel manufacturer and vendor (Frama-tome), (3) three TSOs to safety authorities and (4)universities and research institutes. The standardisedenvironment offered by the platform and the interopera-bility of codes facilitate collaborative work between allpartners. Collaborative work contributes to the increaseof the leadership of European science for nuclear reactorsimulation.
Since the end of the NURESAFE project, further useand development of the software platform are pursuedthanks to:
– a continuous maintenance by CEA of the softwarerepository dedicated to the NURESIM platform;–
further development and maintenance of the general-purpose software SALOME and URANIE (two open-source software supporting the entire platform);–
further development and maintenance of each individualsoftware by code owners.This above resulted in long-term frameworks that havealready been used for many years.
7 Conclusions and future recommendations
Using the NURESIM platform, challenging DNS & LESsimulations were performed within NURESAFE to analysebubbly flow with and without phase change in order tounderstand intricate phenomena that are beyond measure-ments capabilities. New modelling routes were proposedbased on these results and were documented andimplemented in the platform available to all stakeholders.Novel ideas were explored, and some others were furtherrefined, such as combining large-scale and small-scaleprediction techniques. Such techniques should in themedium term replace state-of-the-art methods that arelimited to one flow regime. These novel techniques areapplicable to more complex core-level thermal-hydraulicsituations involving boiling. Solution procedures takingadvantage of the coupling between various codes tacklingdifferent physics and scales were successfully developed.
In the area of Monte Carlo methods, the methods fordepletion and dynamic calculations are close to theirculmination. The developed coupled codes based on theICoCo-methodology are now implemented in the Europe-an simulation platform NURESIM and the testing andvalidation phase will soon start. For this purpose,different benchmark problems of different size are beingdeveloped so that all partners will apply the developedtools for the analysis of those problems. Moreover, thevalidation of the codes under development using plant/experimental data is of paramount importance forMcSAFE. Therefore, plant data of two European reactors(PWR-KONVOI and VVER-1000) are being preparedand documented for the validation of the advanceddepletion capability of the tools. On the other hand,selected SPERT III REA E test data will be used for thevalidation of the dynamic versions of the Monte Carlocodes. Finally, application to LWR and SMR are foreseento demonstrate the extended capabilities of the multi-physics codes. Generally, it can be stated that consider-able efforts are still needed for high-fidelity simulationsbased on Monte Carlo codes in an HPC-environment inorder to perform core analysis with acceptable statisticsfor the key parameters of interest.
Beyond the major developments in computing capabil-ities for normal operation and design basis accidents, themonitoring of reactors and the early detection of anomalieswill become increasingly important, due to the ageing fleetof reactors in Europe. By extending the current simulationplatforms to the modelling of stationary fluctuations andtheir effect, such simulation tools can be used for creatinglarge data sets that can thereafter be used to detect, fromgiven measured reactor parameters, possible anomalies.For such a purpose, machine learning was demonstrated inCORTEX, using simulated test data, to be potentiallycapable of retrieving anomalies. Tests on actual plant dataremain nevertheless to prove the viability of this technique.In addition, although the phenomena considered so far inCORTEX do not require taking the thermal-hydraulicfeedback into account, the estimation of the coupledneutronics/thermal-hydraulics reactor transfer functionmight be necessary for other scenarios.
C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020) 13
In the area of neutron transport, it should also be notedthat the methods being developed would allow modellingfull core in pure transport. The limitations and approx-imations otherwise introduced when pre-generatingassembly-wise macroscopic cross-sections would then beeliminated, thus greatly enhancing the level of faithfulnessof neutron transport simulations for strongly heteroge-neous cores (such as when using new fuel assembly designs,MOX fuel, etc.).
In essence, the different situations needing accuratemodelling require the inclusion of more and more physics.Beyond neutronics, thermal-hydraulics and thermo-me-chanics, other as important physics might need to beincluded: fuel physics, structural mechanics, coolant andradiation chemistry, radionuclide transport, etc. Trulymultiphysics and multi-scale modelling approaches stillneed to be developed at a more mature level for tacklingsuch situations. This includes the development of newmodels, their coupling, as well as the use of the latestadvancements in numerical analysis optimized for HPC. Inthis respect, the development of hybrid methods, such asdeterministic and probabilistic methods in neutrontransport, or DNS, LES, CFD, and macroscopicapproaches in fluid dynamics and heat transfer, shouldbe favoured and optimized. This requires having differentscientific communities collaborating and capitalizing oneach other’s strengths and expertise. With so challengingmodelling targets, the use of machine learning forpredictive modelling should also be considered, wheremachine learning could be used in place of or in addition tomore traditional modelling approaches. The enormousamount of measured data at commercial reactors, researchreactors, and experimental facilities represent a definiteasset, in a machine learning-based modelling strategy thatshould be utilized as much as possible.
A complete list of the papers published within theprojects can be found on the respective project websites:http://cortex-h2020.eu (for CORTEX), http://www.fp7-hpmc.eu (for HPMC), http://www.mcsafe-h2020.eu (McSAFE), http://www.nuresafe.eu (for NURE-SAFE).
Nomenclature
ATWS
Anticipated Transient without Scram BWR Boiling water reactor CAD Computer-aided design CFD Computational fluid dynamics CNN Convolutional neural network CORTEX CORe Monitoring Techniques and EXperi-mental Validation and Demonstration
DNS Direct numerical simulation EPFL Ecole Polytechnique Fédérale de Lausanne FEM Finite Element Method FSI Fluid-structures interaction HPC High Performance Computing HPMC High Performance Monte Carlo Methodsfor Core Analysis
HZDR
Helmholtz-Zentrum Dresden-Rossendorf LES Large eddy simulation LSTM Long short-term memory LWR Light water reactor LOCA Loss-of-coolant accident McSAFE High Performance Monte Carlo Methodsfor SAFEty Analysis
MOX Mixed oxide MPI Message passing interface MSLB Main steam line break NURESAFE NUclear REactor SAFEty SimulationPlatform
NURESIM European Platform for Nuclear ReactorSimulations
NURISP NUclear Reactor Integrated SimulationProject
OpenMP Open multi-processing PTS Pressurized thermal shock PWR Pressurized water reactor RAM Random access memory REA Rod ejection accident SIE Stochastic implicit euler SMR Small modular reactor SPERT Special Power Excursion Reactor TestProgram
TSO Technical support organization TUD Technical University of Dresden UOX Uranium oxide VVER Vodo-Vodyanoi Energetichesky reactorReferences
1. C. Demazière et al., Overview of the CORTEX project, inProceedings of the International Conference on the Physics ofReactors – Reactor Physics paving the way towards moreefficient systems (PHYSOR2018), Cancun, Mexico, April22–26, 2018
2. V. Sanchez et al., High performance Monte Carlo computingprojects: from HPCM to McSAFE, in NUGENIA Forum,Ljubljana, Slovenia, April 14, 2015
3. L. Mercatali et al., McSAFE projects highlights, inInternational Multi Physics Validation Workshop, NorthCarolina State University, USA, June 2017
4. E. Deville, F. Perdu, Documentation of the Interface for CodeCoupling: ICOCO, CEA, Paris, 2012
5. B. Chanaron et al., Advanced multi-physics simulation forreactor safety in the framework of the NURESAFE Project,Ann. Nucl. Energy 84, 166 (2015)
6. B. Chanaron, The European NURESAFE simulation projectfor reactor safety, in Proceedings of the Internationl Confer-encre on the Nuclear Engineering (ICONE22), Prague, CzechRepublic, July 7–11, 2014
7. C. Demazière, Multi-physics modelling of nuclear reactors:current practices in a nutshell, Int. J. Nucl. Energy Sci.Technol. 7, 288 (2013)
8. I. Lux, L. Koblinger,Monte Carlo particle transport methods:neutron and photons calculations (CRC Press, Boca Raton,FL, 1991)
14 C. Demazière et al.: EPJ Nuclear Sci. Technol. 6, 42 (2020)
9. A. Ivanov et al., Internal multi-scale multi-physics coupledsystem for high fidelity simulation of light water reactors,Ann. Nucl. Energy 66, 104 (2014)
10. M. Däubler et al., High-fidelity coupled Monte Carlo neutrontransport and thermal-hydraulic simulations using Serpent2/SUBCHANFLOW � Part I: Implementation and solutionverification, Ann. Nucl. Energy 83, 352 (2015)
11. F. Calivà, et al., A deep learning approach to anomalydetection in nuclear reactors, in Proceedings of the 2018International Joint Conference on the Neural Networks(IJCNN2018), Rio de Janeiro, Brazil, July 8–13, 2018
12. F. De Sousa Ribeiro et al., Towards a deep unified frameworkfor nuclear reactor perturbation analysis, in Proceedings ofthe IEEE Symposium Series on Computational Intelligence(SSCI 2018), Bengaluru, India, November 18-21, 2018
13. J. Dufek, J.E. Hoogenboom, Description of a stable schemefor steady-state coupled Monte Carlo-thermal-hydrauliccalculations, Ann. Nucl. Energy 68, 1 (2014)
14. J.E. Hoogenboom, Demonstration of the time-dependenceaftera control rodmovement,HPCMDeliverableD3.10 (2013)
15. J.E. Hoogenboom et al., Maximum efficiency in massivelyparallel execution of Monte Carlo criticality calculations, inProceedings of the Joint International Conference on theMathematics and Computation (M&C), Supercomputing inNuclear Applications (SNA) and the Monte Carlo (MC)Method (MC2015), Nashville, USA, April 19–23, 2015
16. I. Mickus, J. Dufek, Optimal neutron population growth inaccelerated Monte Carlo criticality calculations, Ann. Nucl.Energy 117, 297 (2018)
17. M. Garcia et al., Development of a spatial domain decomposi-tion scheme forMonteCarloneutron transport, inProceedingsof the International Conference on Nuclear Engineering(ICONE26), London, United Kindgom, July 22–26, 2018
18. D. Ferraro et al., Foreseen capabilities, bottlenecks identifi-cation and potential limitations of Serpent MC transport
code in large-scale full 3D burnup calculations, in Proceed-ings of the International Conference on Nuclear Engineering(ICONE26), London, United Kindgom, July 22–26, 2018
19. M. Faucher et al., New kinetic simulation capabilities forTRIPOLI-4
®
: methods and applications, Ann. Nucl. Energy120, 74 (2018)
20. A. Livensky et al., Modeling the SPERT transients usingSerpent2 with time-dependent capabilities, Ann. Nucl.Energy 125, 80 (2019)
21. S. Sanchez-Cervera et al., Optimization of multidimensionalcross-section tables for few-group core calculations, Ann.Nucl. Energy 73, 387 (2014)
22. S. Lal et al., Direct numerical simulation of bubble dynamicsin subcooled and near-saturated convective nucleate boiling,Int. J. Heat Fluid Flow 51, 16 (2015)
23. Y. Sato et al., Computational fluid dynamics simulationof single bubble dynamics in convective boiling flows,Multiphase Sci. Technol. 25, 287 (2013)
24. H. Li, H. Anglart, CFD model of diabatic annular two-phaseflow using the Eulerian- Lagrangian approach. Ann. Nucl.Energy 77, 415 (2015)
25. Bundesamt für Strahlenschutz, Kurzbeschreibung undBewertung der meldepflichtigen Ereignisse in Kernkraft-werken und Forschungsreaktoren der BundesrepublikDeutschland im Zeitraum Januar 2011, Stand 14.12.2012(2012) (in German)
26. Almaraz Trillo Report, Neutron noise status in Trillo NPP,Technical report CO-12/043, Spain, 2012
27. CSN/PDT/NCTRI/TRI/1503/203, Proposal of technicaldecision on the revision of the Technical OperationalSpecifications relative to the dead band of the filter in TrilloNPP, technical report (in Spanish)
28. M. Seidl et al., Review of the historic neutron noise behaviourin German KWU built PWRs, Prog. Nucl. Energy 85, 668(2015)
Cite this article as: Christophe Demazière, Victor Hugo Sanchez-Espinoza, Bruno Chanaron, Advanced numerical simulationandmodelling for reactor safety� contributions from the CORTEX, HPMC,McSAFE andNURESAFE projects, EPJNuclear Sci.Technol. 6, 42 (2020)