You are on page 1of 53

Large-Scale Transition to Detonation

Elaine S. Oran
Laboratory for Computational Physics & Fluid Dynamics US Naval Research Laboratory Washington, DC Sixth International Seminar on Fire and Explosion Hazards Leeds, April 2010

In Collaboration with: Vadim Gamezo (NRL) David Kessler (NRL) Alexei Poludnenko (NAS/NRC Postdoc, NRL) Peter Hamlington (NAS/NRC Postdoc) J. Craig Wheeler (University of Texas, Austin) Takanobu Ogawa (Seikei University) R. Karl Zipf, Michael Sapko, Eric Weiss (NIOSH, Pittsburgh Research Laboratory) Various Aspects of the Work Sponsored by: NRL(ONR), AFOSR, NIOSH, NASA, NEDO, DARPA Acknowledge Conversations, Advice, Criticism .... Forman A. Williams (UCSD) K.N.C. Bray (Cambridge University) Derek Bradley (Leeds University) James Driscoll (University of Michigan)

What we all know ...


Laminar Flame
M << 1 Energy Release Expansion Thermal Conduction Molecular Diffusion Radiation ....

Turbulent Deflagration
M<1

Detonation
M>1 Energy Release Compressible Flow Shocks, and Complex Shock Structures ...

( This is a real mess

Products of Reactions xl

Fuel

Wide Range of Conditions: Flamelets Distributed Flames Shocks ...

Products of Reactions

Fuel

xd ~ (1-100) xl

Transitions among these states are not as well understood.

Large-Scale Transition to Detonation Themes How does DDT happen?


How does it start? Are there common precursors?

What are the effects of scale?


Longer time scales? Larger space scales? How does this affect the limits?

When can we expect to be able to predict flame acceleration and DDT?

H2-Air Mixture Ignited in a Channel with Obstacles Movie will show how ...
Starting with a small flame in a channel containing a combustible mixture, a turbulent flame develops and produces shock waves. This leads to the formation of unsteady shock-flame complexes and detonations. Beginning of Movie:

Flame 2 cm obstacles

direction of flow and wave propagation

Explanation of the Movie ... (1)


The initially laminar flame moves slowly into the unreacted material.
Obstacles perturb the flow, which then interacts with and distorts the flame, which becomes turbulent.

Perturbed Flame

Flow perturbations (turbulence) increase the surface area of the flame, enhance the energy-release rate, and thus accelerate the flame and background flow.

The turbulent flame generates compression waves, which eventually coalesce in front of the flame to form a shock. The shock is continuously strengthened by compression waves coming from behind the shock and heating the gas.

Shock-flame interactions are important increase flame area and generate vorticity.

Explanation of the Movie ... (2)


2.255 ms
1 9 ms

2.258 ms

2.260 ms

2.264 ms

52

53

54

53

54

53

54

53

54

55

Onset of a detonation

Propagating detonation wave


2.269 ms

The shock reflects from an obstacle ... creates a hot spot, or ignition center, which can become a spontaneous wave ... A detonation wave results that may or may not survive.

62

63

64

65

Detonation wave
2.301 ms 2.313 ms 2.328 ms

Flame Detonation decays .

A quasi-detonation results for smaller channels.

This phenomenon repeats.


62 62 63 64
2.340 ms

65

66
2.369 ms

Detonation occurs again.


2.375E ms

from the obstacle, and the

66

67

69

70

69

70

71

Detonation Wave Propagation


For large enough channels, the detonation successfully propagates over the obstacle.
Detonation wave
2.869 ms 2.907 ms 2.907 ms

DDT
150 150 155
2.963 ms

155
2.975 ms

160

Detonation wave
2.998 ms

The detonation is partially extinguished, but quickly recovers.

165

170

165

170

170

175

Physical and Computational Models


+ (U) = 0 t (U) + (UU) + P + =0 t E + ((E + P )U) + (U ) + (K T ) = 0 t (Y ) + (Y U) + (DY ) w =0 t 2 (U) U) ( U) I = 3 RT P U 2 P = E= + 2 M ( 1) Q dY w = AY exp dt RT

Solutions of the unsteady Navier-Stokes equations, with models for chemical energy release, diffusive transport, thermal conduction, ... . No turbulence models. Use AMR (Adaptive Mesh Refinement) to obtain resolution as needed, perhaps down to the viscous dissipation scale, DNS (direct numerical simualtion), and principles of MILES (Monotone Implicit Large-Eddy Simulation)

Tn Tn Tn K = 0 D = D0 = 0 Cp 0 K 0 0 Cp = Le = = = Sc = Pr = D D D C D K

Solution Approach
Solve the unsteady, compressible Navier-Stokes equations in one-, two-, and and three-dimensions by three different numerical methods: a lower-order Gudonov method (Gamezo), a high-order FCT method (Ogawa), and, most recently, by a high-order PPM-type method (Poludnenko). Include models for chemical reactions, energy release, thermal conduction, and molecular diffusion, and calibrate them to reproduce basic flame and detonation properties. Resolve the flow down to viscous microscale (sometimes). Use direct numerial simulation (DNS), or AMR (adaptive mesh or block refinement (PARAMESH) algorithms. Usually some part of calcualtion is large-eddy simulation (LES), using basic concepts of monotone-integrated LES (MILES). Simulate specific laboratory experiments, some specifically designed to test the model (e.g., DDT: Thomas et al.; Flame acceleration: Teordorczyk et al.); Natural gas explosions: Kuznetzov et al., Zipf et al.)

Material, Chemistry, and Reaction Wave Parameters Stoichiometric Hydrogen-Air


Quantity T0 P0 0 M A Ea (= Q) q 0 = 0 = D0 Value 293 K 1 atm 8.7345 104 g/cm3 1.17 21 g/mol 6.85 1012 cm3 /g-s 46.37 R T0 43.28 R T0 /M 2.9 105 g/s-cm-K0.7 298 cm/s 7.289 T0 0.1372 0 0.035 cm 1.993 105 cm/s 31.47 P0 16.24 P0 3.457 T0 9.010 T0 9.104 0 1.802 0 0.01927 cm 12 cm Denition Initial temperature Initial pressure Initial density Adiabatic index Molecular weight Pre-exponential factor Activation energy Chemical energy release Transport constants

Input

Output

Sl Tb b xl DCJ PZN D PCJ TZN D TCJ ZN D CJ xd

Laminar ame speed Post-ame temperature Post-ame density Laminar ame thickness CJ detonation velocity Post-shock pressure Pressure at CJ point Post-shock temperature Temperature at CJ point Post-shock density Density at CJ point 1D half-reaction thickness Detonation cell size

What we have seen ..... A turbulent flame develops from a spark


Three stages of development: 1. Flame stretching and folding 2. Flame-front wrinkles from turbulent fluctuations and instabilities 3. Flame surface created by shock-flame interactions.

A shock-flame complex forms Hot spots form (Zeldovich gradients) The systems transitions to detonation

What we have seen ..... A turbulent flame develops from a spark


Three stages of development: 1. Flame stretching and folding 2. Flame-front wrinkles from turbulent fluctuations and instabilities 3. Flame surface created by shock-flame interactions.

A shock-flame complex forms Hot spots form (Zeldovich gradients) The systems transitions to detonation

Shock-Flame Complex
An important and curious transitional (?) state.... Appear as a result of shock-flame interactions:
Evolution of flames in channels with obstacles Shock-flame interactions with boundary-layers Shock-flame interactions with wakes Interactions of Mach stems and flames

Sometimes they are precursors to DDT, sometimes they seem to propagate steadily with no indication of change in state.

Shock-Flame Complex

An Array of Obstacles
1. Initial flame 3. Shock-flame complex forms

2. Flame becomes turbulent

What we have seen ..... A turbulent flame develops from a spark


Three stages of development: 1. Flame stretching and folding 2. Flame-front wrinkles from turbulent fluctuations and instabilities 3. Flame surface created by shock-flame interactions.

A shock-flame complex forms Hot spots form (Zeldovich gradients) The systems transitions to detonation

Hot spots form ... possibly leading to detonation


Temperature Contours
2.255 ms 2.258 ms 2.260 ms 2.264 ms

52

53

54

53

54

53

54

53

54

55

Onset of a detonation
The shock reflects from an obstacle ... creates a hot spot, or ignition center, which can become a spontaneous wave ...

Sometimes the hot spots do not detonate but create a shock and a flame

What we have seen ..... A turbulent flame develops from a spark


Three stages of development: 1. Flame stretching and folding 2. Flame-front wrinkles from turbulent fluctuations and instabilities 3. Flame surface created by shock-flame interactions.

A shock-flame complex forms Hot spots form (Zeldovich gradients) The system transitions to detonation

Consider a Another Calculation Spark Ignition in a Larger Channel


A spark in a H2-Air mixture, formation of weak shocks and a flame that quickly becomes turbulent, transition to a detonation

Note the frequent, continual, repeating shock interactions.

8 cm

An Array of Obstacles
2. The flame becomes turbulent.

1. Initial flame

3. Shock wave formation

4. DDT: shock reflection from obstacles

Focus on Some of What We Have Learned

* *

Dynamic, fast, changing chemically reactive-flow systems are not in equilibrium. Major physical transitions in these systems occur in nonequilibrium, transitional states. Detonations appear from hot spots, which are gradients in reactivity. (Here we saw hot spots appearing from shock reflections.) Shocks and turbulent flames create the environment in the unreacted background gases in which a detonation may occur. Shocks and shock-interations are extremely important in in this flow. The most obvious are shock-flame interactions, which generate vorticity, stretch the flame, and completely change the dynamics of the flow. When to expect agreement with experiments ...?

Focus on Some of What We Have Learned

* *

Dynamic, fast, changing chemically reactive-flow systems are not in equilibrium. Major physical transitions in these systems occur in nonequilibrium, transitional states. Detonations appear from hot spots, which are gradients in reactivity. (Here we saw hot spots appearing from shock reflections.) Shocks and turbulent flames create the environment in the unreacted background gases in which a detonation may occur. Shocks and shock-interations are extremely important in in this flow. The most obvious are shock-flame interactions, which generate vorticity, stretch the flame, and completely change the dynamics of the flow. When to expect agreement with experiments ...?

A range of geometries, initial conditions, and reactive materials has been studied ...
Material: Ethylene and acetylene (low pressure)
Thermonuclear C-O system (white dwarf star) Hydrogen (atmospheric, stoichiometric) Methane (atmospheric, stoichiometric, lean)

Dimension: 1, 2, and 3 dimensions Geometrical Configurations:


Channels and chambers. Vary: height; blockage ratio; obstacle geometry, spacing and symmetry Spherical (white dwarf), open (recent DNS)

Mode of initiation: Shock-flame interaction Smooth or spark

Comparisons to Experiments
Acetylene: Shock-flame interactions (with University of
Wales Aberystwyth). Qualitative and quantitative agreement for timing of DDT with increasing shock strength. Description of physical processes and some controlling mechanisms.

1999

Ethylene: Shock-flame interactions (with University of Wales


Aberystwyth). Qualitative agreement on types of shock-flame interaction with boundary layers. Curious, unexplained similarity (agreement?) between results of 2D and 3D calculations.

2001

Hydrogen: Flame acceleration and transition in channels with


obstacles (with Seikei University, Warsaw Institute of Technology, Shimizu Corp). Comparisons to specially designed experiments on DDT initiation. Some comparisons are excellent; others are ... not so good. Limited 2D and 3D comparisons, some excellent. Sensitivity of some results to small bakground variation.

2006

Methane: Flame acceleration and transition in large channels


with obstacles (experiments by Kuznetsov et al., with NIOSHPittsburgh). Comparisons to experiments; some look excellent. Current work.

2008

Comparisons to Experiments
Acetylene: Shock-flame interactions (with University of
Wales Aberystwyth). Qualitative and quantitative agreement for timing of DDT with increasing shock strength. Description of physical processes and some controlling mechanisms.

1999

Ethylene: Shock-flame interactions (with University of Wales


Aberystwyth). Qualitative agreement on types of shock-flame interaction with boundary layers. Curious, unexplained similarity (agreement?) between results of 2D and 3D calculations.

2001

Hydrogen: Flame acceleration and transition in channels with


obstacles (with Seikei University, Warsaw Institute of Technology, Shimizu Corp). Comparisons to specially designed experiments on DDT initiation. Some comparisons are excellent; others are ... not so good. Limited 2D and 3D comparisons, some excellent. Sensitivity of some results to small bakground variation.

2006

Methane: Flame acceleration and transition in large channels


with obstacles (experiments by Kuznetsov et al., with NIOSHPittsburgh). Comparisons to experiments; some look excellent. Current work.

2008

Possible Explanations for Level of Agreement


Chemical model is good enough -The model carefully controls the heat release in time and space in the flow. This might be enough to impose for now.

The nature of the turbulence -The turbulence is controlled by shock interactions: Shock and flames; shocks and shocks, shock traveling in inhomogeneous material, etc. This definitely requires further investigation.

Turbulence Driven by Shock-Flame Interactions


The initial shocks interact with the flame surface -- a Richtmyer-Meshkov (RM) instablity. This distorts the flame surface, creating more surface area and vorticity. Initially, the disturbance is at a single or a few wavelengths, but this develops quickly to driving on all scales ranging from the system size to the thickness of a laminar flame. Small-scale flame structures burn out quickly, leaving behind a smoother flame surface and vorticity. Subsequent shocks further distort the flame surface and create more vorticity. For spark ignition, steady RM driving appaears more quickly. In the regime of the coupled shock-flame complex, before DDT occurs, there is fairly steady RM driving.

Focus on the Shock-Flame (Richtmyer-Meshkov) Interaction

Stoichiometric methane-air Spark ignition Channel diameter, 52 cm

Some properties of randomly forced, broad-band turbulence, such as RM or shock-generated.

Energy spectrum can have a number of envelopes, including k-5/3 typical of Kolmogorov spectra. Higher moments, such as vorticity or enstropy can behave differently. Intermittancy is suppressed.

Here we can watch the shock interactions as they evolve ...

Turbulence Driven by Other Shock Interactions


Other sources of vorticity and turbulence: Shock-shock interactions Shock interactions with inhomogeneous media Propagating curved shocks Shock interactions with turbulent media These are ubiquitous, not just local in the flow. This is the kind of turbulence that appears in molecular clouds, such as the Orion nebula. The energy spectra appears to be Kolmogorov (-5/3 decay), but it is not driven at the largest scale of the system, but by multiple shock interactions over a broad band of scales.

Experiments at Lake Lynn Experimental Mine

km

Detonation and DDT Experiments at Lake Lynn Experimental Mine

* Cylindrical stainless steel tube

73m long, 1m diameter, pipe 10 mm thick. Fit with pressure transducers & ion probes.

* First series of experiments, Fall 2009:


(Actually, natural gas with < 1% ethane).

Direct initiation of detonation in CH4-air

* Multistage ignition process

Fill 3m x 1m polyethylene bag with CH4-O2; Ignite with blasting cap. Energy input/bag: 38 MJ/m2; 1-2 bags; extra 1/2 bags will be added. Vary percentage of methane from 5% to 17%.

* Approximately 30 experiments performed in 2009. Operation


stopped for winter.

* Modeling effort also underway.

Detonation and DDT Experiments at Lake Lynn Experimental Mine Preliminary Results: Fall 2009
Approximately 30 experiments. Measured detonation velocities for CH4-air agree with CJ predictions.

First results ...


Detonation limits extended for large tubes.
Methane, volume %
5 10 15 20

Dcj, m/s

LLEM Ideal Dcj

Equivalence Ratio

How do we scale this to a 1 km tunnel?

Detonation Cells in Natural Gas and Air


170

Critical Conditions for Onset of Detonation

Critical conditions in rooms and channels with obstacles Mixtures have irregular cell structure Detonation cell size (or mean cell size) is a reasonable parameter. (Dorofeev, 2005)

Simulations of Experiments by Kuznetsov et al.


Stoichiometric CH4 + Air
Evenly spaced obstacles

Spark ignition

S
(cm)

d/2 h

d = 17.4, 52 cm Blockage ratios BR = h/d = 0.3, 0.6 Experiments in a cylindrical tube

Movie shows the evolution of the spark, formation of a flame, transition to a turbulent flame, and hot-spot ignition of a detonation

Initial Results

Reasonably good agreement, but some discrepancies


Average velocity too high after DDT for BR = 0.3 Experiments suggest more frequent transitions from quasi-detonation to choking propagation regime Short-lived detonation for BR = 0.6, but none in experiments

Local detonation ignition events too frequent in simulations


Ignition delay time too short? Physical differences with experiments: energy losses, geometry, volume?
(Kessler, Gamezo, Oran, 2009, 2010)

The same types of physical processes that lead to DDT occur in a wide range of exothermic materials, and over a very large range of scales ... from millimeters to kilometers as long as there is an exothermic material that can support a flame and a detonation.

Thermonuclear Supernovae (Type Ia) Universal Standard Candles


Before ...... NGC4536

After

..... 1981B

Is This a Detonation?

Takes ~ 2s, Releases ~ 10 51 ergs Type Ia 12C + 16O -> higher elements

Background Information
(Based on Theory and Observations)
- White dwarf star exists 108-1010 yrs. Explosion time: ~2 s . - Bright as an entire galaxy .... Releases 1051 ergs (1027 Mton) - All SNIa have similar spectra, light curves, ejecta velocity, density, composition profiles, isotopic abundances ... - Form heavy elements ... 12C + 16O heavier elements. Produce elements from Mg, Si ... to Ni - In the pre-explosion state, Radius is 2x108 cm ( ~ earth), M ~ Msun Density is 109 g/cm3 (center), decreases to ~ 106 g/cm3 - Explode by a thermonuclear process Carbon-oxygen reactions support flames and detonations Reaction temperature 109 - 1010 K Major energy-releasing reaction: 12C+ 12C ( -> He, , Ni, Mg) Reaction proceeds to produce Mg, Si .... Ni (50%) Equation of state: P = ( -1) E, = 1.3 - 1.5

Converged Simulation of SNIa Explosion


0.576 s 1.257 1.393 1.487

1.531 s

1.573

1.613

1.652

1.724 s

1.760

1.883

1.902

Three-Dimensional Simulations of Thermonuclear Supernovae


(Science, 2003; PRL, 2004, ApJ 2005)
1.62 s 1.62 s

DDT path 2
1.77 1.51

Observational Data include: Energy release Time-dependent spectra Results of Simulations: Deflagration stage alone, without DDT, predict observations.

path 1
1.69

DDT

1.87

1.20

1.75

Detonation stage alone (not shown) cannot predict observations. Delayed detonation (DDT) predicts correct energy release (paths 1 and 2). Earlier DDT will produce consistent spectra (path 1).

1.97 s

1.82 s 0.90 0.61 0.30 s

Flame Initiation

Since 2003, all of the much larger-scale, much more resolved computations have NOT taught us significantly new physics about the explosion process. Conclusion: To really know if DDT can occur in this environment, we need to resolve the physics on the microscale!

Flames in Intense Turbulence


Range of resolutions: 64 x 64 x 1028 128 x 128 x 2048 256 x 256 x 4096
2

Kolmogorov

RM

(Combustion & Flame, 2010)

Some Preliminry Results


For the hydrogen-like flame in intense Kolmogorov turbulence:

There is a dynamic feedback between the turbulence and the flame brush.
2

Flamelet structures in the flame brush are very robust. They maintain their laminar-flame-like structure through the intense bending and folding that occurs due to very strong turbulent intensities (Poludnenko & Oran, 2010, CNF). Meanwhile, the heat release in the flame brush tends to suppress vorticity and intermittancy. This helps keep flamelets robust (Hamlington et al, 2010).
UL /SL = 30
Time-averaged flamelet structure in the flame brush

Isocontours of vorticity

Fuel

Product

Large-Scale Transition to Detonation Themes

How does DDT happen?


How does it start? Are there common precursors?

What are the effects of scale?


Longer time scales? Larger space scales? How does this affect the limits?

When can we expect to be able to predict flame acceleration and DDT?

Thank you for your kind attention!

You might also like