1. [ID 204] Evaluation of ADREA-HF LES against Hydrogen release dispersion
Mr. Nektarios Koutsourakis1,2
Dr. Alexander Venetsanos2
Prof. John G. Bartzis1
4th International Conference on Hydrogen Safety
San Francisco, California, September 12-14, 2011
1 University of West Macedonia Greece
2 NCSR “Demokritos” Greece
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

2. ADREA-HF CFD CODE Constantly developing
Latest additions:
LES (Large Eddy Simulation)
Parallel solver
Arbitrary number of species
Combustion
Back tracking
GUI pre and post processor (called EDes)
Advantages:
Robust, powerful and general
Specialization in environmental applications
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

3. ADREA-HF CFD CODE ADREA specializes in dispersion applications
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

4. ADREA-HF CFD CODE GUI environment for pre and post processing
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

5. LARGE EDDY SIMULATION Between DNS and RANS
Navier-Stokes math. analysis: impossibly difficult
DNS: Fully-resolved NS
Cost  Re2.75 (for wall-bounded flows  Re3.5)
RANS: Time-averaged NS
Fast, widely tested, usually accurate “enough”
LES: Spatially-filtered NS
Cost near wall:  Re2.4 ! (at outer layer  Re0.5)
Converges to DNS. As in DNS, LES needs: a) averaging b) demanding boundary conditions
Suggested were RANS fails: transient, separated flow
DNS: Thousands of years to simulate 1s of flight
RANS: Good for steady state problems, with no big separations
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

6. ENERGY CASCADE LES solves most of the turbulence
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

7. LES EQUATIONS AT ADREA-HF Compressible volume-filtered Navier-Stokes
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

8. LES EQUATIONS AT ADREA-HF Alternative subgrid model: RNG-LES
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

9. LES AT ADREA-HF Numerics
ADREA/SIMPLER algorithm
Solver: Parallel BiCGstab with parallel Schwarz preconditioner. Speedup with 2 CPUs up to 1.7
Numerics:
Deferred correction central scheme for convection
Crank-Nicolson 2d order for time advancement
Linear upwind with VanLeer limiter for concentration
For more:
Venetsanos, A. G., E. Papanikolaou and J. G. Bartzis, 2010: The ADREA-HF CFD code for consequence assessment of hydrogen applications. Int. J. Hydrogen Energy, 35, 3908.
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

10. TEST CASE HySafe INERIS-6C experiment
1g/s H2 for 240s
Jet 20mm
Room 3.78 * 7.2 * 2.88m
Jet exit velocity 38m/s
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

11. TEST CASE SETUP Simulation details
Standard Smagorinsky with Cs=0.1
Total number of cells in the room:
CFLmax=0.3
Minimum mesh resolution: 0.02*0.02*0.053 m
Simulation time in a workstation: 15 days for the release phase (240s); dtav1= s + 2 days for the first 260s of the post-release phase
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

12. TEST CASE SETUP Geometry, sensors, part of the grid
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

13. RESULTS Comparison with experiment
Photo from experiment
ADREA LES
An asymmetry on the ceiling can be noticed (higher C towards the sensors’ side of the room).
The plume at the higher parts of the jet seems to be slightly wider in the simulation.
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

14. RESULTS Video of the Hydrogen propagation
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

15. RESULTS Concentration contours animation
A flow field develops that transfers the hydrogen along the ceiling and then downwards along the side walls, which have higher concentrations than the neighbouring regions of the same height. Thus the plume takes a mushroom shape.
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

16. RESULTS Comparison with sensors
Each plot should be explained, along hydrogen route: Starting from 16 to 1 to 7 to 12.
Why the slope increases along route, why the difference of release and diffusion decreases, etc.
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

17. DISCUSSION Some remarks
In general very good agreement
High turbulence during release phase, only
Release overestimated, especially near source. Very difficult for CFD and for turbulence modelling to capture the physical processes next to the source
Diffusion underestimated (possible experimental asymmetry?) - improvement with better grid
Better agreement is prevented from experimental inaccuracies and numerical discrepancies
I think release is overestimated due to inadequacy of CFD to see what exactly happens at the emission area. That is a very difficult point that needs very advanced models
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

18. FURTHER EXAMINATION Cs value influence (release phase -coarse grid)
Jet axis sensors
Ceiling sensors
Side wall sensors
Sensor 16
Sensor 14
Sensor 13
Sensor 1
Sensor 4
Sensor 6
Sensor 7
Sensor 8
Sensor 9
Sensor 10
Sensor 12
Only 80s-160s measurements taken into account!
Max diffusion close to 0.12
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

19. DISCUSSION Cs affects the results
Cs (SGS modelling in general) important near jet
At the ceiling, no Cs influence (buoyancy dominates)
Non-monotonic behaviour (increasing the Cs gives better results, but up to a limit)
Worst Cs by far: 0.2
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

20. FURTHER EXAMINATION C variances for each Cs
C variance = mean (Ci’2)
VARIANCE IS A MEASURE OF RESOLVED TURBULENCE if total turbulence is higher (while variance being same), it means that the SGS (unresolved) is higher….
To calculate the variance I only used values from 74s till 166s for all cases.
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

21. DISCUSSION A critical balance: resolved vs. SGS turbulence
As we go away from the jet, variance (and resolved turbulence intensity) drops
Two mechanisms compete: - Modelled SGS turbulence - Resolved turbulence (generally low if SGS is high)
With Cs=0.2, very high artificial viscosity causes more “laminar” flow and lower fluctuations
Resolved turbulence is higher with Cs=0.1, but total turbulence seems to be higher with Cs=0.12
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

22. FURTHER EXAMINATION Evaluation of RNG-LES (run not finished yet)
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

23. DISCUSSION RNG LES seems to be an improvement
RNG LES provides in general better results
Improvement towards the right side for sensors both close to the jet (turbulent flow) and far from it (laminar flow). That was not the case by varying Cs
Resolved turbulence seems to decrease close to the jet and increase far from it
RNG LES is believed to better model the subgrid scales for both the turbulent and the laminar flow
Suggestible when the turbulence intensity varies
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

24. CONCLUSIONS I Acceptable results and physical insight
LES results are good, given all numerical and experimental inaccuracies
Overestimation during release phase (close to jet)
Underestimation during diffusion phase (and far from jet)
Attempt to investigate the balance between resolved and SGS turbulence with parametric analysis of Cs
RNG LES seems to behave slightly better in both turbulent and laminar parts of the flow
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

25. CONCLUSIONS II LES is good, but not for everyday use
ADREA-HF has now tested LES in hydrogen release
LES can provide peak concentrations (good for flammable mixtures or for calculation of dosages)
LES can provide concentration statistics
LES many times more expensive than RANS
LES is not RANS++
There is not good LES – there is good CFD
In general, in any good CFD code LES can be added
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE

26. THANK YOU
Acknowledgements: Greek Scholarships Foundation, Municipality of Aghia Varvara
ICHS 2011, #204 LES MODELLING OF HYDROGEN RELEASE AND ACCUMULATION WITHIN A NON-VENTILATED AMBIENT PRESSURE GARAGE USING THE ADREA-HF CFD CODE