Perspectives on CFD V&V in Nuclear Regulatory Applications

Transcription

1 Perspectives on CFD V&V in Nuclear Regulatory Applications Christopher Boyd Senior Technical Advisor for Computational Fluid Dynamics US Nuclear Regulatory Commission 1

2 Foreword Not a discussion of the regulations and methods most commonly used for licensing and decision making. Discussion focused on CFD applications and the associated challenges. Perspective comes from experience within the NRC Office of Nuclear Regulatory Research. 2

3 US Nuclear Regulatory Commission (NRC) Regulates civilian use of radioactive material to protect public health and safety. research and development focused on regulatory activities independent With growing frequency, the NRC is utilizing or reviewing CFD predictions. confidence needs to be established (i.e. V&V) Is the regulator ready? Is the industry ready? 3

4 Regulatory Perspective Industry (the licensee) is responsible for adequate safety analysis. NRC specifies requirements, not methods NRC (the regulator) tries to that the ensure analysis can be trusted. Verification and Validation Independent confirmation Risk informed Risk comes from probability and consequences New modeling approaches require significant effort to build trust. Experience helps! 4

5 Observations Convincing V&V is time consuming and expensive. Shortcuts are common What is good enough? Industry tends to submit methods with a history of regulatory acceptance. Reduces regulatory uncertainty CFD heavily used by industry; less common for licensing Bigger computers have resulted in more complex problems but not necessarily more convincing V&V. practitioners have often focused on qualitative simulation of the next more difficult problem class, rather than achieving quantitative accuracy on the previous problem class. P. Roach, Fundamentals of Verification and Validation, 2009 Continuous Advancement of Simulation Largeness (CASL ) 5

6 Nuclear Safety Analysis Challenges (Range of Scales) facility scale simulations of plumes and other phenomena ~ km containment scale simulations 10 s m component scale simulations ~ m fuel pin scale simulations ~ mm 6

7 Wide Range of Phenomena supersonic flows buoyancy driven flows D. Lucas, D. Bestion, NURETH Multiphase flows 7

8 Reviewing the State of the Art How much mixing do I need? in this field, there is at least as much artistry as science. P. Roach, Fundamentals of Computational Fluid Dynamics,

9 Best Practice Guidance (room for interpretation) NEA/CSNI/R(2007)5 BPG document Use a finer and more regular grid in critical regions. If wall values are the target values tetrahedral meshes should either be avoided or used with great care. For RANS or T RANS select an appropriate statistical model for turbulence. 10 analysts 10 results user effect alive and well 9

10 2014 CFD Benchmark Exercise (OECD/NEA) simple geometry known behavior ideal boundary and initial conditions adequate resources history of benchmark studies in similar facilities PANDA, MISTRA, TOSQAN, He Contours significant related publications Hypothetical reactor accidents are not so easy. Andreani, M, Synthesis of the OECD/NEA PSI CFD Benchmark Exercise, Proceedings of OECD/NEA & IAEA Workshop, CFD4NRS 5, Zurich,

11 ESBWR Modeling of Core Bypass (Two Approaches) Approach 1 46M tet. cells non conformal detailed simplified complete blockage Core bypass Region mesh SLCS injection nozzles control rod model Independent Approach 2 45M hex. cells conformal simplified 11

12 Uncertainty in large nuclear safety Problems Limited CFD grade data Limited information on Boundary and Initial conditions Generated from 1D system codes CFD grade experiments are rare Thermal boundary conditions can be challenging Uncertainty distribution Geometric simplifications are common impact on uncertainty? Grid Convergence Index (GCI) #$&%!! Difficult in full scale, complex reactor geometries Unstructured Mesh, wall treatments, post processing techniques 12

13 2016 GEMIX Benchmark Exercise (OECD/NEA) Blind Benchmark Simple geometry Small scale Significant increase in CPU time for simple uncertainty estimate. GCI significant effort even with pure hexagonal mesh even if particular results are inconsistent, the method adds value and should be attempted OECD/NEA PSI CFD Benchmark Exercise, USNRC Office of Nuclear Regulatory Research OECD/NEA & IAEA Workshop, ASME V&V CFD4NRS 6, 2016 Boston,

14 State of the Art (summary) CFD predictions of complex reactor issues typically require creativity (compromises) in order to produce practical solutions. significant deficiencies in test data (limited scale and data) compromised model and V&V process There seems to be an art to V&V as well as CFD Adherence to V&V standards and best practice guides is lacking and needs to be encouraged. In the mean time, we must determine if the models that are completed can be trusted for the intended purpose. 14

15 USNRC Interests with CFD Preparation for anticipated reviews of Licensee Submittals licensing basis calculations supporting calculations (more common) Safety Research Effective Regulation confirmatory studies, independent analyses, reduced uncertainty Future licensing of specific models for specific applications CASL (VERA tools), other? Future Reactors designed using CFD from the start 15

16 Preparation for Reviews (Focused on Mission) Currently using commercial CFD tools Developing Methods and Experience Best Practice Guidelines, V&V methods Benchmark studies, CFD grade data Cooperative research, both national and international Regulatory guides (Generic >> application specific) Goal is to improve regulatory effectiveness and build confidence in regulatory decisions. 16

17 Safety Research Generic Safety Issues (GSIs) Best Practice Guidelines (BPGs) Support for safety evaluations Support for system code models Experimental programs scaling support validation CFD grade data bases Mach number downstream of a jet exit 17

18 Generic Safety Issue 193 GSI 193: BWR ECCS Suction Concerns PPOOLEX Facility Courtesy of Lappeenranta University of Technology FLUENT CFD Predictions VOF methods 18

19 Safety Research GSI 193 example BWR Mk I studies 0.1m pipe Puma Tests full scale studies 19

20 Safety Research Support for System Codes induced SG tube failure example PWR Severe Accident Flow CFD CFD used to extend test data (not a substitute for test data). prototypical conditions additional variables design variations System Code System Code Support model 3D phenomena in 1D system codes adjusted using CFD predictions and test data reference: NUREG

21 Safety Research BWR Lower Plenum SLCS mixing BWR Test Facility CFD Methods used to track Boric Acid Solution injected during ATWS scenario 21 USNRC Office of Nuclear Regulatory Research

22 Effective Regulation Confirmatory Studies Independent Analysis Answering questions general fluid flow issues confirmation for system codes filling gaps in test data Standard Review Plans Regulatory Guides Inadvertent Actuation of Isolation Condenser 22

23 Licensee Submittals Examples responses to regulators request for additional information (RAI) 3D calculations providing background or support for system codes licensing basis calculation (less common) CFD analysis for Fuel Storage and Transportation Casks is a good example of successful licensing basis calculations. 23

24 CFD for Licensing Basis Dry Cask Applications T(K) 10+ years of experience Multiple Submissions / Reviews Independent Analysis by NRC staff Multiple Test Programs Best Practice Guidance NUREG 2152 B.P.G. Comparison with Data ¼ cask T contours 24

25 CFD for Licensing Basis (new approach) US APWR Advanced Accumulator multi phase modeling (cavitation) CFD validated at ½ scale scaling effects a concern Reference: Nuclear Engineering and Design 249 (2012) years for full review 25

26 US APWR Licensing Nuclear Engineering and Design 249 (2012)

27 CFD for Licensing Basis General Reactor Issues Reviews balance several factors margin? uncertainty? risk? Are best practices applied (or demonstration of a better approach) Are V&V methods appropriate for the application, risk, and safety margin? One of a kind or first of a kind calculations Significant work required to build confidence documentation, V&V issues Industry tends to stick with accepted methods 27

28 Barriers for CFD in Regulatory Applications Regulatory Uncertainty What is good enough? How long will process take? Cost ($$) (licensee pays) regulatory reviews are expensive V&V is expensive Taking the path of least resistance submit methods with a history of regulatory acceptance Modeling limitations limited time/computer resources adequacy of models limited test data 28

29 NRC Roles and Regulatory Support Regulatory Guides application specific Requests for Additional Information (RAIs) detailed questions help licensee understand regulatory needs Detailed Safety Evaluation Reports documented concerns Standard review plans staff guidance Interaction with public (or other agencies) public meetings, publications, technical reports, cooperative research DOE cooperation Industry cooperation ASME (V&V standards, NQA 1) OECD/NEA activities / International agreements Conference publications 29

30 Industry Roles in Developing Regulatory Acceptance Topical Reports Demonstration of model submitted for review and acceptance Support for Best Practice Guidelines and V&V development >> acceptance >> application CFD grade data bases training Interaction with regulator public meetings, test programs, etc. Regulatory Submissions with documented V&V Code developers and vendors including V&V tools 30

31 Training is important Modern codes are easier to use Easier to get the wrong answer Colorful Fluid Dynamics should be validated with respect to purpose Many training programs focus on how to use a code but not how to solve and validate a problem. Analogy: MS Word tutorials don t prepare you to write a technical report. V&V methods should be included in CFD training programs Occasional CFD use is not recommended experience is needed Benchmark studies are an excellent learning opportunity. especially blind benchmarks 31

32 Summary CFD is playing a larger role in nuclear reactor regulation. safety research / supporting calculations licensing basis calculations growing in frequency DOE/CASL is helping push the state of the art The wide spread adoption of best practice guidelines and verification and validation methods will build confidence. expensive but necessary progress is slow but interest is growing user effects remain The ASME V&V symposium is a good example of progress being made. 32