School of Aerospace Engineering

Flame Dynamics, Hydrodynamics, and

Acoustics

Tim Lieuwen
Georgia Institute of Technology

Acknowledgements: Air Force Office of Scientific Research (Chiping Li, Program

Monitor) and US Department of Energy (Rich Dennis, Program Monitor)

1
 School of Aerospace Engineering

Unsteady Combustor Physics

• Fluid Mechanics
– Hydrodynamic stability
of inhomogeneous flows

• Acoustics
– Wave propagation in
complex flows

• Combustion
– Response of reaction
fronts to disturbances

2
 School of Aerospace Engineering

Combustion Instability

• One of the largest risks associated

with development of modern
propulsion and energy systems

• Manifest themselves as narrowband 500

oscillations at natural frequencies of 400

Fourier Transform
combustor 300

– Key problem lies in understanding 200

how flames respond to harmonic 100
flow perturbations
0
0 100 200 300 400 500 600 700

Frequency (Hz)

3
 School of Aerospace Engineering

Flame Response to Disturbances

• Class of more general problems dealing

with flame response to flow
disturbances:
– Turbulent combustion: time average
response of burning rate to stochastic
flow fluctuations
– Combustion noise: RMS of burning
rate
– Focus here: ensemble averaged
response of turbulent flow and flame in
harmonically oscillating acoustic field

4
 School of Aerospace Engineering

5
 School of Aerospace Engineering

Hydrodynamic Stability of Reacting

Flows
• Flame response dominated by
large scale coherent structures that
arise from underlying
hydrodynamic instability of flow

• Key mechanisms:
– Kelvin-Helmholtz Instability
– Rayleigh-Taylor Instability

6
 School of Aerospace Engineering

Hydrodynamic Stability of Reacting

Flows
• Key mechanisms:
– Kelvin-Helmholtz
Instability
– Rayleigh-Taylor
Instability

• Dispersive

Bechert , D. ,Pfizenmaier, E., JFM., 1975

7
 School of Aerospace Engineering

Hydrodynamic Stability of Reacting

Flows
• Key mechanisms:
– Kelvin-Helmholtz Instability
– Rayleigh-Taylor Instability

• Dispersive

• Strongly influenced by density

stratification

8
 School of Aerospace Engineering

Example: 2-D Wakes

• Density ratio across
flame is an important
bifurcation parameter
in shear flows

• Most lab scale burners

operate at much higher
density ratios (~6)
than applications
(~1.5-3)

Erickson, Mehta and Soteriou, Aerospace Sciences

Conference, 2006
9
 School of Aerospace Engineering

Results Very Sensitive to Flame -

Shear Layer Offset

10% shift in flame

Absolutely stable position causes
critical flame density
ratio to shift by 70%

Absolutely unstable

Emerson, B., O’Connor, J., Juniper, M. and Lieuwen, T., JFM, 2012.

10
 School of Aerospace Engineering

Hydrodynamic Stability of Reacting

Flows
• Key mechanisms:
– Kelvin-Helmholtz Instability
– Rayleigh-Taylor Instability

• Dispersive

• Strongly influenced by density stratification

• Multiple hydrodynamic modes

– Most naturally unstable/amplified, not necessarily the
11 one subjected to strongest excitation by acoustics
 School of Aerospace Engineering

Transversely Forced Swirling Flame:

Pressure Antinode
-3
10
2.5

Excites strong axisymmetric

2
jet mode
1.5

0.5

0
-4 -2 0 2 4 6

Smith, T.E. et al., In 55th AIAA Aerospace Sciences Meeting, 2017.

12
 School of Aerospace Engineering

Transversely Forced Swirling Flame:

Velocity Antinode
-3
10
2
Excites strong helical
modes
1.5

0.5

0
-4 -2 0 2 4 6

Smith, T.E. et al., In 55th AIAA Aerospace Sciences Meeting, 2017.

13
 School of Aerospace Engineering

14
 School of Aerospace Engineering

Response of Heat Release to Flow

Perturbations
0.5

Q '(ω) / Qo
0.4 G=
u '(ω) / uo
Why does this saturate?
Q’/Qoo

0.3
CH*′ / CH*

Why at this amplitude?

0.2

0.1 What factors affect

slope of this curve (gain
0
0 relationship)
0.2 0.4 ? 0.6 0.8
u′ / u
u’/u o
o

15
 School of Aerospace Engineering

Excited Bluff Body Flames

(Mie Scattering)

Increased
Amplitude of Forcing

16
 School of Aerospace Engineering

Swirling Jet Flames

17
 School of Aerospace Engineering

Excited Swirl Flame - Attached

(OH PLIF)
0° 45° 90° 135°

315° 270° 225° 180°

18
 School of Aerospace Engineering

Quantifying Flame Edge Response

Time
Series

Power
Spectrum
L’(x, f0)

19
 School of Aerospace Engineering

Spatial Behavior of Flame Response

Convective wavelength:
λc= U0/f0at forcing
• Strong response
frequency
- ‒distance a disturbance
Non-monotonic spatial
propagates at mean flow
dependence
speed in one excitation
period

20
 School of Aerospace Engineering

Flame Wrinkling Characteristics

1. Low amplitude flame fluctuation near
attachment point, with subsequent growth
downstream

2. Peak in amplitude of fluctuation, L’=L’peak

3. Decay in amplitude of flame response

farther downstream

4. Approximately linear phase-frequency

dependence
• Can relate slope to axial convection
velocity of flame wrinkle, uc,f

Shanbhogue et al., Proceedings of the Combustion Institute, 2009.

21
 School of Aerospace Engineering

Flamelet Description of Front Dynamics

∂L ∂ G 
:  ∂L+ u ⋅∇ ∂L |
2 F. Williams,
G-equation =
 G s | ∇
G
+  u − v  = S L L1 +   Combustion
∂t ∂t ∂x   ∂x  Theory

22
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Destruction of wrinkles

23
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Destruction of wrinkles

24
 School of Aerospace Engineering

Disturbance Propagation on Premixed

Flame Sheets
• Linearized solution of
G-equation

∂L′ ( x, t ) 1 x ∂un′ x − x′ 1 x
∂x
=
ut ∫ 0 ∂x
( x′, t −
ut
) dx′ + ⋅ un′ ( x =0, t =−
ut
t
ut
)

25
 School of Aerospace Engineering

Wrinkle Convection By Tangential

Flow

u0 uc,f
ut ut
sL

• Small amplitude wrinkles propagate non-dispersively

– Not true for non-premixed flames

• Non-local character of flame response is probably the

26
most dynamically significant feature of forced flames
 School of Aerospace Engineering

Phase Characteristics of Flame Wrinkle

Convection speed of
Flame wrinkle, uc,f

Mean flow velocity, u0

Disturbance Velocity, uc,v

Shin et al., Journal of Power and Propulsion, 2011.

27
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Destruction of wrinkles

28
 School of Aerospace Engineering

Excitation of Flame Wrinkles – Spatially

Varying Disturbance Field
∂L′ ( x, t ) 1 x ∂un′ x − x′ 1 x
∂x
=
ut ∫
0 ∂x
( x′, t −
ut
) dx′ + ⋅ un′ ( x =0, t =−
ut
t
ut
)

• Flame wrinkles generated by velocity

fluctuations normal to it

• Net flame disturbance at location x

is convolution of disturbances at
upstream locations and previous
times

• Convecting vortex is continuously

disturbing flame
– Vortex convecting at speed of uc,v
– Flame wrinkle that is excited
convects at speed of ut

29
 School of Aerospace Engineering

Model Problem: Attached Flame Excited by a

Harmonically Oscillating, Convecting Disturbance
• Model problem: flame excited by convecting velocity field,

un′
= ε n cos(2π f (t − x / uc ,v ))
ut

• Linearized solution:

 
L' − i ⋅ ε n sin θ  i 2π f ( y / ( uc ,v tan θ )− t ) i 2π f ( y / ( ut sin θ )− t ) 
= Real  × e −e 
ut f  2π ut cos θ / u c ,v − 1  ( ) 


30
 School of Aerospace Engineering

Solution Characteristics
• Note interference pattern on
flame wrinkling λint

• Interference length scale:

1
λint ( λ t sin θ ) =
| ut uc ,v − 1|
L’

y / ( λ t sin θ )

31
 School of Aerospace Engineering

Interference Patterns on Flame

Sheets

L’
L’

1
λint ( λ t sin θ ) = Acharya et al., ASME Turbo Expo, 2011
| ut uc ,v − 1|
• High phase acoustic speed disturbances (with long wavelengths)
where uc,v>>ut lead to short length scale disturbances
• Vortical disturbances where ut~uc,v lead to long length scale
32 wrinkling
 School of Aerospace Engineering

Aside: Randomly Oscillating,

Convecting Disturbances
• Space/time coherence of Random
excitation

Lre
disturbances key to

f
L’2
interference patterns Single frequency
excitation

• Example: convecting or

random disturbances to
simulate turbulent flow
disturbances

33
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Destruction of wrinkles

34
 School of Aerospace Engineering

Flame Wrinkle Destruction Processes

: Kinematic Restoration
• Flame propagation normal to
itself smoothes out flame
wrinkles
∂L  ∂L  ∂L 
2

+ u f − v f  = SL 1+  
• Typical manifestation: vortex ∂t  ∂x   ∂x 
rollup of flame

• Process is amplitude dependent

and strongly nonlinear
– Large amplitude and/or short
length scale corrugations
smooth out faster

Sung & Law, Progress in Energy and Combustion Science, 2000.

35
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Destruction of wrinkles
4.Spatially Integrated heat release

36
 School of Aerospace Engineering

Response of Heat Release to Flow

Perturbations
0.5

Q '(ω) / Qo
0.4 G=
u '(ω) / uo
Why does this saturate?
Q’/Qoo

0.3
CH*′ / CH*

Why at this amplitude?

0.2

0.1 What factors affect

slope of this curve (gain
0
0 relationship)
0.2 0.4 ? 0.6 0.8
u′ / u
u’/u o
o

37
 School of Aerospace Engineering

Spatially Integrated Heat Release

38
 School of Aerospace Engineering

Why the 1/St Rolloff?

• Consider spatial integral of traveling wave
disturbance:
  x  u    LF   
LF

∫x = 0 cos ω  t − u  dx = − sin ω  t −   − sin [ωt ]

ω   u  

Traveling Wave Low pass filter due to interference

effects associated with tangential
convection of wrinkles

• Low pass character of heat release response comes

from the wrinkle convection and integration!
39
 School of Aerospace Engineering

Key Research Questions

• Hydrodynamics of swirling
jets; density stratification
effects
• Acoustics – Boundary
conditions in reacting,
flowing systems
• Harmonically excited
turbulent flames

40
 School of Aerospace Engineering

Commercial Aircraft Engine

• Combustion
instabilities emerged
during full engine
testing of GenX engine

41
 School of Aerospace Engineering

42
 School of Aerospace Engineering

Financial Times
23 July 99, Issue 49

Daggers Drawn over Nehuenco

“The Patience of Chile’s Colbun power company has finally run out over the continued non-
performance of the Siemens-built Nehuenco generating plant. Exasperated by repeated break-downs
at the new plant and under pressure from increasingly reluctant insurers – (and with lawsuits looking
likely) – the generator announced that it will not accept the $140m combined-cycle plant - built and
delivered by the Germany equipment manufacturer.

Siemens, together with Italy’s Ansaldo, took the turnkey contract for the 350 MW plant in 1996 and
should have had it in service by May of last year. The startup was delayed till January. Since then
matters have worsened. There have been two major breakdowns and, says Colbun, there have been no
satisfactory explanations.

The trouble could not have come worse for Colbun. The manly hydroelectric generator, which is
controlled by a consortium made up of Belgium’s Tractebel, Spain’s Iberdrola and the local Matte and
Yaconi-Santa Cruz groups, has been crippled by severe drought in Chile, which has slashed its output
and thrown it back – without Nehuenco – onto a prohibitively expensive spot market.”

43
 School of Aerospace Engineering

Wrinkle Convection
Flamelet Step Response problem: ua t<0
u=
Step increase axial velocity from ua to ub, : ub t≥0

u = ua t<0 u = ub t >0 u = ub t >> 0

? .

44
 School of Aerospace Engineering

Wrinkle Convection

ua t<0
u=
ub t≥0

45
 School of Aerospace Engineering

Analysis of Flame Dynamics

1.Wrinkle convection and flame relaxation

processes
2.Excitation of wrinkles
3.Interference processes
4.Destruction of wrinkles

46
 School of Aerospace Engineering

Excitation of Wrinkles on Anchored Flames

∂L′ ( x, t ) 1 x ∂un′ x − x′ 1 x
∂x
=
ut ∫
0 ∂x
′
(x ,t −
ut
′
ut
′
) dx + ⋅ un ( x =0, t =−
t
ut
)

• Linearized solution of G Equation,

assume anchored flame

• Wrinkle convection can be seen un’

from delay term L’

θ
u0 u
t

sL

47
 School of Aerospace Engineering

Excitation of Flame Wrinkles – Spatially

Uniform Disturbance Field
∂L′ ( x, t ) 1 x ∂un′ x − x′ 1 x
∂x
=
ut ∫
0 ∂x
( x′, t −
ut
) dx′ + ⋅ un′ ( x =0, t =−
ut
t
ut
)

• Wave generated at attachment

point (x=0), convects
downstream
 sL ( t ) 
θ ( t ) = cos 
−1

• If excitation velocity is spatially  ( )
u t
uniform, flame response
exclusively controlled by flame
anchoring “boundary condition”
– Kinetic /diffusive/heat loss effects, though not
explicitly shown here, are very important!

48
 School of Aerospace Engineering

Near Field Behavior

Increasing
amplitude, u’ • Flame starts with small
amplitude fluctuations
because of attachment
L’(x=0, t) = 0
Normalized by u’

• Nearfield dynamics are

essentially linear in
amplitude

Shanbhogue et al., Proceedings of the Combustion Institute, 2009.

49
 School of Aerospace Engineering

Kinematic Restoration Effects

• Leads to nonlinear farfield flame
dynamics
• Decay rate is amplitude dependent

x/λ c
Numerical Calculation Experimental Result

50