Hydrogen vented explosion : experiments, engineering methods and CFD

Transcription

1 ICHS5 September, Brussels, Belgium Hydrogen vented explosion : experiments, engineering methods and CFD S. Jallais Thanks to M. Kuznetsov, S. Kudriakov, E. Studer, V. Molkov, C. Proust, J. Daubech, P. Hooker, E. Vyazmina 1 Air Liquide, Claude-Delorme Research Center Les Loges-en-Josas, France 1

2 Content I. Context II. Phenomenology III. Available experiments IV. General Trends V. Analytical models VI. CFD VII. Conclusions and perspectives 2

3 I. Context 3

4 What is a Vented Explosion? Explosion venting is commonly used to prevent or minimize damage to an enclosure during an accidental explosion Kindling chain : Leak H 2 build up (homogeneous, layer, stratification, ) ignition Vented explosion with internal & external effects Use of hydrogen in «confined zones» Rooms / Garage FC Cabinets Containers Explosion Venting by : Dedicated explosion vents Low resistance elements (doors, windows) Ventilation openings 4

5 What is a Vented Explosion? Parameters: Enclosure Geometry (volume, shape, obstacles, absorbing material, ignition position ) Mixture properties (concentration, distribution, initial turbulence) Vent properties (size, shape, location ) Outdoor conditions (external obstacles) Questions for the safety engineers : What should be the size of a vent in order not to exceed P max? What are the external effects if the structure resists? 5

6 II. Phenomenology 6

7 Phenomenogy Explosion in a closed volume P P AICC 29.6 % H 2 in air P(AICC) = 7.38 bar 7

8 Phenomenology Vented explosion Back Wall ignition 8

9 P 1 or P_ext 9

10 Phenomenology P 2 or P_vib due to the flame-acoustic oscillations 10

11 Example of experimental pressure signal Two peaks : P 1 (or P_ext external explosion ) P 2 (P_vib flame accoustics oscillations ) High-frequency oscillations are present (especially for P 2 ) Filtering is necessary 11

12 II. Available Large Experiments 12

13 Pasman TNO (1974) Cylinder - V = 0.95 m 3 A v = 0.2 m 2 Pstat = bar A v = 0.3 m 2 - Pstat = bar 29.6 % H 2 Central ignition Initial turbulence?? Only one peak?? Data Filtering?? 13

14 AECL Experiments (Kumar et al. 1989, 2006, 2009) Spherical + duct (6.85 m 3 ) (1989) End, central and near vent ignition 10 to 20 % H to m 2 Unusual geometry Filtering?? Rectangular (120 m 3 ) Quiescent (2006) 8 to 12% H m 2 End, central and near vent ignition Low concentration (buoyancy effects??) Filtering?? Turbulent (2009) Central ignition 14

15 Bauwens et al. FM Global FM Global 4.6 * 4.6* 3 m 64 m 3 Vents : 5.4 & 2.7 m 2 3 ignition positions (BW, CI, FW) 12 to 20% H 2 24 trials (4 with obstacles) 13 CI / 9 BW / 2 FI 5 trials with vent panel P stat from 0.4 to 4 kpa Initial turbulence (10 tests - C 3 H 8 ) U from 0 to 0.4 m/sec Acoustic Dampening Wall Material (C 3 H 8 ) 1 ; 5 & 20 % of internal surface area Filtering 80 Hz low pass Effect of the external structure?? 15

16 Ineris experiments (1/2) Small cylinder (1 m 3 ) Diameter 0.94 m Lenght 1.4 m Av = 0.13 m 2 10 to 27 % H 2 Filtering?? Large cylinder (10.5 m 3 ) Diameter 1.6 m - Lenght 5.5 m Av = 2 m 2 14 to 23 % H 2 Filtering?? 5.5 m 16

17 Ineris experiments (2/2) H * L * W = 2 * 2 *1 m 4 m 3 19 trials 17 BW / 1 CI / 1 FI 9 with obstacles Two configurations Av = 0.49 & 0.25 m 2 10 to 24.8 % H 2 Flame visualization H 2 /Air mixture seeded with NH 4 Cl Filtered by an 100 Hz low-pass filter 17

18 On going experiments : Hyindoor EU Project KIT Setup 1 * 1* 1 m 3 Av = 0.2 to 1 m 2 Obstacles Homogeneous Layers Stratification BOS 10% H m 2 HSL Setup 5 * 2.5 * 2.5 m 3 Av = 0.8 to 3.2 m 2 Homogeneous and real stratification 18

19 III. General Trends 19

20 Effects of H2% Generally, P 1 and P 2 increase with H 2 % Below 10%, P 1 and P 2 are relatively low Ex 1 : 4 m 3 10%H2 A v = 0.49 m 2 Pmax = 4 mbar Ex 2 : 64 m 3 12% H2 A v = 5,4 m 2 Pmax = 9 mbar From 12 to 20%, strong non linear behavior Above 20%, warning : DDT becomes highly possible. Especially in large volume or with obstacles 20

21 Effect of vent area Smaller vent size increases peak pressure for all configurations and ignition locations!!!!! S_vent S_vent dp time 21

22 Ignition location No one ignition location is the most severe for all cases 22

23 Effect of obstacles inside the enclosure Obstacles increase the first peak and decrease the second peak!!!! dp time 23

24 Effect of initial turbulence Initial turbulence increases the overpressure Kumar (2009) 120 m 3 Bauwens & Dorofeev () C3H8 - φ = 1 24

25 Effect of stratifications and layers Stratification No experiments For engineering S Lmax and average expansion ratio σ(average) Layers Few experiments with partial volume vented deflagration m = H / h(layer) Also Pappas (1994) And Buckland (1980) 25

26 Effect of Acoustic Dampening Wall Materials Tamanini (1992) 25-mm thick layer of Kaowool ceramic fiber blanket a) 9.5 % CH 4 +air 1.35 m m 2 Bauwens and Dorofeev () 100 mm thick layer of ROCKWOOL insulation 1, 5 and 20% of the internal chamber surface area b) % C 3 H 8

27 III. Engineering models 27

28 Engineering models (1/2) Cubbage & Simmonds (1955) Yao (1974) Bradley and Mitcheson, 1978 Tamanini, 2001 NFPA 68. Very poor predictive capabilities Bauwens et. al., 2010, 2012 model Calculations of the 2 peaks 3 ignition positions (FI, CI, BWI) Obstacles considered 2 tuned parameters Molkov et al model Only one peak No specified ignition location (BWI) No obstacles 28

29 Bauwens et al Pressure inside External explosion pressure Flame speed Flame surface = f(ignition, obstacles) First pressure peak: Second pressure peak: Flame speed affected by acoustics Laminar flame speed Flame-wrinkling factor S u # 0,9. S L.Le -1 # 2 S L 29 This is a two-parameter model! Minor improvement in 2012

30 CFD Benchmark HYSAFE (2010) Case : Pasman 1 m 3 / 29,6% H 2 experiments Participants : UU (Fluent LES), KIT (COM3D), JRC (REACFLOW), KI (b0b) Rather good agreement 30

31 FM Global CFD work : Openfoam & in-house combustion model Back-wall ignition Front ignition 18% H 2 18% H 2 2,7 m 2 5,4 m 2 Good agreement for P 1 Flame-acoustics interactions (P2) are not taken into account in the model! Bauwens et. al. «Vented explosion overpressures form combustion of hydrogen and hydrocarbon mixtures» IJHE 36,

32 Hyindoor Benchmark Objectives : fixing %H 2 and (A^0,5)/L p constant, V influence Participants : UU, CEA, KIT, NCSRD, others?? Case V (m3) a (cm) A (m2) L(m) (A^0,5)/L H2(%) FMGlobal ,4 4,6 0,51 18,3 PRISME 0, , ,15 0,50 18 KIT ,25 1 0,

33 Conclusions and perspectives Few valuable experimental data well filter or with information on filtering Need of experimental data on the influence of : Initial turbulence (paper 237) Layers / Stratification Vent covers Obstacles Distribution of vents Promising FM engineering model : To be improved to take into account : Layers / stratification Obstacles effect could be improved 33 CFD modeling Join the Hyindoor benchmark!! Challenging P 2 modeling!! Structural behavior : Engineering models : impulse and coupling with classical structure model CFD including dynamic structure response

34 ICHS5 September, Brussels, Belgium Hydrogen vented explosion : experiments, engineering methods and CFD S. Jallais Thanks for your attention Simon.jallais@airliquide.com 34