1. 1 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Modelling of H2 Dispersion and Combustion Phenomena Using CFD Codes H. Paillère, E. Studer, A. Beccantini, S. Kudriakov, F. Dabbene and C. Perret* CEA Saclay – *CEA Grenoble

2. 2 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Outline Introduction Description of phenomena related to H2 risk issues CFD modelling at CEA Dispersion / Distribution model validation Combustion model validation Mitigation model validation Outline of a validation matrix ? Necessity for new experiments in support of code validation On-going activities and conclusions

3. 3 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Introduction (1/2) CEA has more than 10 years experience in the field of hydrogen safety, mainly in the field of nuclear safety (e.g. TMI accident in USA, 1979, 400kg H2 burn inside containment) Since 2001, CEA is actively engaged in the development of new energy technologies including hydrogen & fuel cell systems (over 250 engineers and technicians) Safety has been recognized as an important issue to ensure the success of these technologies CEA is a member of the HYSAFE Network of Excellence

4. 4 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Introduction (2/2) In many industrial fields, safety demonstrations have relied mainly on expertise, supported by experiments. Increasingly, numerical tools, in particular, Computational Fluid Dynamics (CFD) codes, are being used for safety assessment, e.g. to predict explosive cloud formations or to assess consequences of explosions or flames Requirements on use of CFD results for safety assessment: –Code validation on representative experimental data (need to identify relevant phenomena & associated test data) –Expertise of code users –Application / knowledge of Best Practice Guidelines An international consensus on a CFD code validation matrix would be an additional factor to support CFD for H2 safety

5. 5 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Description of phenomena related to H2 risk issues Review of accident scenarios and identification of main phenomena (on- going PIRT exercise for example, within HYSAFE)  List of phenomena (also identification of state of the art in modelling, existence of experimental data) Accidents involving H2 usually involve: –Release (gaseous or liquid) –Dispersion into environment, confined, semi-confined or open atmospheres –Action of passive or active mitigation systems to reduce the risk –In case of ignition, and depending on geometry & other parameters Diffusion flames Jet fires Slow deflagrations Flame acceleration DDT Very wide range of flow regimes involving chemical & heat transfer processes, from nearly incompressible buoyant flow to fully compressible reactive flow Challenge for physical models & numerical algorithms Validation of models requires a very extensive effort

6. 6 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Mainly relies on CAST3M code, which is an in-house research code, used for many different applications Not a “black box” code  knowledge of models & methods… and their weaknesses Development of numerical algorithms “best suited” to the physics –Pressure based methods for dispersion calculations –Density based methods (shock capturing) for explosion modelling Trying to apply BPG (grid sensitivity, parametric studies, etc) … but very often beyond code & current computers’ capabilities Participation to benchmarks (e.g. HYSAFE) Other codes are being/will be assessed & used by CEA (e.g. FLUENT as well as two-phase flow codes) CFD modelling at CEA

7. 7 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Dispersion / Distribution model validation Main physical models: –Laminar diffusion –Turbulence –Buoyancy forces Test cases: –Russian-2 experiment (HYSAFE benchmark) –AECL Large Scale Gas Mixing Facility Helium tests –CEA MISTRA Helium tests MH1 and MH2 All confined atmosphere cases – in 3D geometries with no internal obstacles

8. 8 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Russian-2 experiment (HYSAFE bench., see 120004) Experimental data not of very high quality (no repeatability, no information on wall temperatures) Grid and model (mixing length) sensitivity studies were performed Under-prediction of diffusion in the lower part of the facility (as with most codes) Effect of buoyancy forces due to temperature effects? Grid convergence study

9. 9 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 AECL LSGMF He tests Study of dynamics of buoyant jet Side opening in facility so constant pressure was maintained Comparison of different turbulent models was made, RNG k-  vs. standard k  Better accuracy was found using RNG k-  Similar conclusion found using CFX (GRS)

10. 10 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 MISTRA He tests MH1 and MH2 Release of He in purely confined geometry (small pressurisation) to study physics of stratification, and mixing by diffusion Very well instrumented facility (gas sampling, thermocouples, LDV)  detailed field measurements, including temperatures which might play a role in additional buoyancy effects Good agreement between CFD calculations & experimental data (velocity & concentration profiles)

11. 11 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Combustion model validation Focus on confined/semi-confined H2, large scale experiments, many of which performed in framework of nuclear safety studies: –HDR E12.3.2 Test, slow deflagration –Battelle BMC Ex29 Test, slow deflagration –RUT STH 06, fast deflagration –RUT STM4 Test, detonation Model implemented in CAST3M code: Compressible Euler equations with CREBCOM model. Diffusion and turbulence effects taken into account through experimental-based correlations (flame speed) – predictability of such model? Validation efforts aimed at showing ability to calculate dynamic loads (conservative values) rather than detailed physics

12. 12 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 HDR E12.3.2 test: slow deflagration Slow deflagration through interconnected volumes Ability to model flame acceleration and pressure effects by choosing appropriately model constants Grid sensitivity results show limitations of such simplified combustion model

13. 13 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 BMC Ex29 test: slow deflagration Slow deflagration through interconnected volumes (similar to HDR test) Ability to model flame acceleration and pressure effects by choosing appropriately model constants Grid sensitivity results show limitations of such simplified combustion model

14. 14 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 RUT STH06 test: fast deflagration Flame acceleration modelled using CREBCOM model Presence of shock waves (precursor and reflected shock waves) Good agreement with experimental data

15. 15 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 RUT STM4 Test: detonation Flame acceleration & transition not modelled Simulate detonation entering canyon & reflection Results not dependent on CREBCOM model parameters. Code gives similar results with global Arrhenius model Ability to capture pressure peaks if grid sufficiently fine & second-order schemes used

16. 16 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Mitigation model validation Mitigation effects can also be evaluated using CFD For example: use of catalytic recombiners as a risk-reducing measure in confined environments or to decrease H2 content of release (for ex. Boil-off system for LH2 engines) CEA has performed experiments to qualify recombiner systems & developed model incorporated in CAST3M code Model development for other types of mitigation devices & appropriate experimental validation to be carried out in HYSAFE

17. 17 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 Outline of a validation matrix? Current identified gaps: –Separate Effect Tests (eg. Pure diffusion) –Low momentum release in confined atmospheres –Release in obstacle-laden environments –Release in partially / open atmospheres –Combustion in presence of gradients –Diffusion flames & jet fires Need for new experiments

18. 18 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 New tests at CEA: Diffusion column experiment Some doubts as to the mixing processes in Russian-2 test: –Other than molecular diffusion? GADIFAN Separate Effect Test Study of diffusion mixing. No injection, light gas separated from air by diaphragm that can be opened without disturbing the flow He or H2 can be compared Column can be tilted at various angles (to introduce buoyancy)

19. 19 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 New tests at CEA: detailed mixing experiment GAMELAN test for study of He injection & mixing Use of non-intrusive measurement techniques (LIF) for gas concentration measurements Also LDV measurements of velocity

20. 20 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 New tests at CEA: compartmented tests Objective: study mixing processes in a confined, compartmented geometry (public multi-storey car park) Additional physical processes: –Jet impinging –Additional turbulence created by flow around obstacles Use of large scale (7m high, 100m3) facility MISTRA facility (developed for other applications)

21. 21 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 New tests at CEA & INERIS: garage experiments Garage tests “à la Swain”, performed in the framework of HYSAFE’s insHYde project, for typical European garage layouts Detailed measurements for CFD code validation (various concentration measurement techniques will be tested) He release (CEA) & H2 release (INERIS) experiments Study of geometric configurations, effect of ventilation

22. 22 Paper 100071 – Paillère et al. International Conference on Hydrogen Safety, ICHS, Pisa, September 8-10, 2005 On-going activities and conclusions Presentation of validation efforts performed at CEA for H2 risk assessment More work on model development & validation needed Gaps in experimental data identified: –Availability of existing (published) data –need for better experiments (quality of experimental data) –Need for new types of tests, including “Separate Effect Tests” & “Coupled Effect Tests” On-going experimental programme at CEA in support of code modelling & validation Main objective: stimulate discussions on an “internationally”- agreed validation matrix for CFD codes to be used for H2 safety assessment –Within HYSAFE Network of Excellence –Among OECD / IEA Task 19 (H2 Safety) group of experts