Validation of CFD Calculations Against Impinging Jet Experiments

1
Validation of CFD Calculations Against Impinging
Jet Experiments

• Prankul Middha and Olav R. Hansen, GexCon, Norway
  
• Joachim Grune, ProScience, Karlsruhe, Germany
• Alexei Kotchourko, FZK, Karlsruhe, Germany
• September 11, 2007

2
Motivation

• CFD calculations increasingly used for
  quantitative risk assessments
• Validation of tool primary requirement
• Important to focus on realistic scenarios while
  carrying out validation of CFD tool
• Need to reproduce the complex physics of the
  accident scenario
• Validation of tools for combined release and
  ignition scenarios
• Recent experiments performed at FZK present an
  opportunity to perform real validation against
  a complex experiment
• Possibility to develop risk assessment methods
  for hydrogen applications
• (Caution Not large scale)

3
Experimental Details (1)

• Release of hydrogen in a workshop setting
  followed by ignition
• Nine different release scenarios
• Total hydrogen inventory fixed (10 g)

4
Experimental Details (2)

• Two different geometrical configurations
• Released H2 ignited using at two different
  ignition positions (0.8 and 1.2 m above the
  release nozzle)

Plate Geometry
Hood Geometry
5
CFD Tool FLACS (1)

• Solution of 3D compressible Navier-Stokes
  equations using a finite volume method over a
  cartesian grid
• Implicit method (SIMPLE algorithm) for pressure
  correction
• 2nd order scheme in space and 1st order scheme in
  time (2nd order available)
• Standard k-e model with several important
  modifications
• Model for generation of turbulence behind
  sub-grid objects
• Turbulent wall functions for adding production
  terms to the relevant CV across the boundary
  layer
• Model for build-up of proper turbulence behind
  objects of a particular size (about 1 CV) for
  which discretization produces too little
  turbulence
• A distributed porosity concept which enables
  the detailed representation of complex geometries
  using a Cartesian grid
• Large objects and walls represented on-grid, and
  smaller objects represented sub-grid
• Necessary as small details of obstacles can
  have a significant impact on flame acceleration,
  and hence explosion pressures

6
CFD Tool FLACS (2)

• Combustion Model
• Flame in an explosion assumed to be a collection
  of flamelets
• 1-step reaction kinetics, with the laminar
  burning velocity being a measure of the
  reactivity of a given mixture
• A beta flame model normally used that gives the
  flame a constant flame thickness (equal to 3-5
  grid cells)
• Burning velocity model
• A model that describes the laminar burning
  velocity as a function of gas mixture,
  concentration, temperature, etc. Le effects
  accounted for H2.
• A model describing quasi-laminar combustion
  (increase in burning rate due to flame wrinkling,
  etc.)
• A model that describes ST as a function of
  turbulence parameters (intensity and length
  scale) and laminar burning velocity (based on
  Bray et al.)

7
Purpose of Simulations

• Simulations performed prior to experiments with
  the primary purpose of aiding the design of
  experiments, if possible
• Identify scenarios for ignition (cloud size
  reactivity)
• Optimal ignition position and time
• Expected overpressures
• gt Avoid un-interesting tests, optimise use of
  resources
• Secondary purposes
• Evaluate prediction capability (topic of current
  presentation)
• Demonstrate efficiency of calculations
• Development of risk assessment methods
• Presented at LPS, Houston
• Connection with HyQRA (HySafe) and IEA Task 19

8
Representation of geometry and grid

• Grid used
• 5 cm standard grid (2.5cm for explosion)
• Stretch outside interesting region
• Refine towards leak (21mm and 4mm leaks)

9
Dispersion Simulations Plate geometry
Small flammable volume with plate only
Small nozzle (4mm) gt no flammable cloud
10
Dispersion Simulations Plate geometry
11
Dispersion Simulations Hood geometry
Flammable cloud inside confinement for low
momentum
Small nozzle (4mm) gt no flammable cloud
12
Dispersion Simulations Hood geometry
13
Dispersion Results Comparison with Experiments
Concentration dependence on distance from nozzle
100mm nozzle
21mm nozzle
Plate Geometry
14
Dispersion Results Comparison with Experiments
Lateral distribution of concentration
100mm nozzle (0.7 g/s)
21mm nozzle (3.0 g/s)
Plate Geometry
15
Dispersion Results Comparison with Experiments
Photograph of plume vs. Predicted shape
Plate Geometry, 21mm nozzle (3.0 g/s)
16
Dispersion Results Comparison with Experiments
Concentration dependence on distance from nozzle
Hood Geometry, 21mm nozzle
17
Dispersion Results Comparison with Experiments
Concentration dependence on distance from nozzle
Hood Geometry, 100mm nozzle
18
Dispersion Results Comparison with Experiments
Photograph of plume vs. Predicted shape
Hood Geometry, 21mm nozzle (3.0 g/s)
19
Explosion Simulations (Pre-calculations)
Worst-case explosion overpressures (quiescent)
Plate geometry
Hood geometry

• Ignition of non-homogeneous clouds

20
Possible to scale overpressures with cloud size ?

• Aim Development of QRA methodology
• Concept of equivalent stoichiometric cloud size
• Obtained using reactivity- and expansion-based
  weighting
• Expected to give similar explosion loads as the
  real cloud

Cloud Size
Overpressures
21
Explosion Results Comparison with Experiments
Ignition 1.2m from release nozzle (Calculations
performed subsequent to experiments to match
ignition position)
Experiments
Simulations

• Possible different time of ignition for 100mm
  hood leads to higher simulated pressure

22
Explosion Results Comparison with Experiments
Ignition 0.8m from release nozzle (Calculations
performed subsequent to experiments to match
ignition position)
Experiments
Simulations

• Local pressure transient around ignition
  influences simulated pressures near ignition
  location

23
Conclusions

• Leak scenarios well predicted in general
• Less interesting scenarios simplified somewhat
  with respect to grid definition to save time,
  which led to some underprediction
• Predicted pressure levels with FLACS similar to
  those observed in experiments
• Possible to scale predicted overpressures with
  equivalent gas cloud size
• Work important to build confidence in CFD tools
  for QRA calculations

24
Acknowledgements

• FZK and coauthors for interesting experiments and
  access to experimental data
• Look forward to larger scale controlled studies
  in similar setups
• European Union for support through the NoE HySafe
• Norwegian Research Council for support for
  hydrogen modelling activities