Combustion instabilities
T. Poinsot
Institut de Mcanique des Fluides de
Toulouse
CNRS, Universit de Toulouse, France
Copyright Dr T. Poinsot 2013
INSTABILITIES ARE EVERYWHERE IN MECHANICS
IN FLUID MECHANICS TOO... AND IN REACTING FLOWS
2
Reacting flows are dominated by
mechanisms which exhibit instabilities:
- hydrodynamics --> Kelvin Helmholtz in shear layers.
(C. M. Ho and
P. Huerre JFM 1984, 16.)
- instabilities at interfaces between liquid fuel and air
- kinetics: the existence of chemical reactions leads to
additional instabilities of flame fronts --> thermodiffusive
instability, formation of cells
- structures: vibration of the combustor walls can couple with
the flames
Ch. 9, 10
3
A classification of instabilities in combustion (Barrere, Wiliams,
Putnam) introduced in the 60s is based on the size of the
components involved in the instability:
- Intrinsic: modes which are due only to local, short distance
interactions between kinetics and flow. Ex: formation of cells
(small size, high frequencies)
- Chamber: modes due to the whole flame where vortices, created
by the instabilities of the shear layer, couple with the flame front
(mid size, mid frequencies)
- System: modes where the whole combustor is involved: long
wavelengths, low frequency
EXAMPLE OF SYSTEM INSTABILITY: THE POGO EFFECT
Unsteady heat release
Oscillation of
reactant flow rate
Unsteady pressure
Vibrating rocket structure
Saturn V fist stage
Soviet Union's N1-L3 rocket
INTRINSIC INSTABILITY
Growth of a spherical laminar
flame ignited by a spark.
PROPANE / AIR:
Perfectly smooth laminar
spherical front
Spark
Time
Fuel / air laminar mixture
Law et al 30th Symp. (Int.) on Comb.
6
HYDROGEN / AIR:
Formation of cells on the
flame front
Spark
Fuel / air laminar mixture
Time
Law et al 30th Symp. (Int.) on Comb.
7
WHEN DO WE OBSERVE THERMODIFFUSIVE INSTABILITIES ?
- Controlled by the Lewis number of the deficient reactant. For
example for a lean H2/air flame, Lewis of the deficient fuel= Lewis
(H2)=0.3
- If this Lewis number is less than a critical value given by
asymptotic analysis (typically 0.5), cellular instabilities can appear
- Whether the growth rate of these instabilities is sufficiently large
to play a role in real (turbulent) flames depends on the flame itself.
Most turbulent combustion models do not incorporate effects of
Lewis numbers but they should.
This also happens in turbulent flames
Boughanem and Trouv
10
CHAMBER INSTABILITY: acoustics + cells
11
CHAMBER INSTABILITY: puffing
Gravity + combustion
12
CYCLE TO CYCLE INSTABILITIES
- In piston engines, one cycle every 100 or 1000 cycles can
fail or burn too slowly or too fast (knock, rumble)
13
14
A specific class of combustion instabilities:
thermoacoustics
In combustion chambers, acoustics and combustion
can couple, leading to unpleasant consequences....
15
There are two main classes of instabilities in
flows (Monkewitz, Huerre):
U<0
CONVECTIVE (AMPLIFIER)
ABSOLUTE (RESONATORS)
Absolute instabilities usually are linked to information propagating upstream
16
FLOW
Convective instability (amplifier)
FLOW
FEEDBACK LOOP
Absolute instability (resonator)
Ch. 8 Sec. 8.5.2
17
VORTEX
FORMATION
INLET FLOW
RATE CHANGE
ACOUSTIC
FEEDBACK
18
VORTEX
COMBUSTION
PRESSURE
WAVE
In combustion chambers, we find:
Acoustics because the flames are confined
Recirculation zones
==> information can propagate upstream by
convection or by acoustic waves
FLOW
SOUND
ABSOLUTE INSTABILITIES ARE EXPECTED !
19
Concentrate on confined flames in combustion chambers
surrounded by walls:
Thermoacoustics (coupled instabilities between acoustics
and combustion) become possible -> these absolute
instabilities are the worst instabilities in combustors
20
Not a new problem: Thermoacoustics is known since Lord
Rayleigh Nature 1878.
Unsteady heat release
Creation of vortices
OR
Oscillation of flow
rates
Unsteady pressure
Acoustic waves
21
Why acoustics and flame can couple:
1/ Flames make noise...
Free flames:
- Make noise
- Are not influenced
by noise
22
2/ In a confined domain (as in most
combustors), noise travels and comes back
The acoustic waves produced by combustion can reflect on
walls, inlets, outlets and come back to the flame zone.
Turbine
23
From
compressor
Combustor
Visualization of perturbations (mean is substracted)
Pressure
24
Entropy
Vorticity
3/ Flames are sensitive to noise
When acoustic waves come back to the flame, they can create
new perturbations, closing the instability loop
Unsteady heat release
Creation of vortices
OR
Oscillation of flow
rates
Unsteady pressure
Acoustic waves
25
A specific feature of combustion chambers:
SWIRL.
Swirl is a rotation of the flow along its axis
26
Swirl: why ?
Flame stabilization !: flames do not propagate at high
speeds. Typically CxHy+Air flames move at sL=50 cm/s.
CxHy/O2 flames at a few m/s.
As soon as the flow speed is larger than a few m/s,
flames can be stabilized only by a recirculation zone
containing burnt bases to ignite the incoming gases.
27
How do you create a recirculation zone ?
Solution 1: dump geometry. Backward facing step
AIR
Propane
28
Temp
Reaction
Velocity
Mixing
Yf Yo
29
Another solution to create recirculation is SWIRL:
Swirl creates a low velocity region in the jet axis.
Strong swirl allows to create recirculation zones
These zones do not touch walls.
(a)
Low swirl
30
(b)
Medium swirl
(c)
High swirl
31
How is swirl introduced ?
32
33
34
Have you seen swirled jets before ? Yes, in the sky !
35
So SWIRL is good for combustion chambers !
But a swirled flow exhibits instabilities:
- the same as in jets: Kelvin Helmholtz
+
- new modes due to rotation (and since we have
introduced negative velocities, absolute modes !)
Rotation axis
36
37
Something that turns can lead to precession:
For a swirled flow: the rotation axis itself turns.
Not necessarily in the same direction as the
mean flow
38
One example of swirled flow instabilities:
the Crow mode.
Page 39
40
Swirled flows have another specific
instability: precessing vortex cores
41
PVC visualized in LES of a swirled reacting flow:
White surface:
flame surface
Red surface: low
pressure surface
The links between thermoacoustics and swirling
flow instabilities such as PVC remain unclear today.
We will discuss them in more details later.
42
HOW CAN WE STUDY
THERMOACOUSTICS ?
Why acoustics and combustion interact
Effects of combustion instabilities in gas turbines
Examples of studies of combustion
Instabilities in gas turbine
configurations
Ch. 8
43
Interaction acoustics / combustion
Not a classical topic: USUALLY, pressure waves are not important
in subsonic flames. They are a by-product of combustion and
produce noise (combustion noise can be important !). But there is
no need to account for them in computing the flames themselves.
In thermoacoustics, when acoustics DO modify flames, we need to
develop a theory able to compute reacting flows and acoustics.
Textbooks:
Crighton, Dowling, Ffowcs Williams,Heckl and Leppington Modern methods in
analytical acoustics 1992, Springer -> acoustics
Poinsot and Veynante Theoretical and numerical combustion 3rd ed, 2012
download at www.cerfacs.fr/Elearning -> combustion theory and simulations
Williams Combustion theory 1985. -> theory
44
CLASSICAL METHODS TO STUDY
RESONATORS:
Two approaches:
- Approach 1: take the conservation equations, linearize them,
look for eigenmodes. Objective: find frequency and growth
rate as well as mode structure (p(x))
- Approach 2: define a proper energy of the system: assume it
changes harmonically, find the frequency and growth rate.
Objective (less ambitious): find instability criteria
Before looking at flames: let us do a small exercice on the
pendulum
45
The linear oscillator with no forcing:
x
v=dx/dt
Spring force: F = -kx
Mass m
Approach 1: just solve it !
With:
46
The linear oscillator with no forcing:
energy definition
x
v=dx/dt
Multiply by the velocity v=dx/dt
and integrate:
E is the total energy of the system
Here E is constant.
47
The linear oscillator with forcing:
F
x
v=dx/dt
If we do not know F, cant say much
==> Approach 1 is limited
BUT Approach 2 can be used:
Multiply by v and integrate:
This is the total energy of the system
48
The total energy of the system
will grow if:
v
F
This is an instability criterion: if the force F and the
velocity v are such that the integral of their product is
positive (ie if F and v are in phase), instability will
grow because the total system energy grows.
If you have a swing at home, you knew this already ?
49
The linear oscillator with forcing: comparing the
two methods in a case where F = a v
Approach 1:
Instability if Re() < 0 if if a > 0
Approach 2: instability if criterion is positive:
50
Linear vs non linear. Exp. growth vs limit cycles
The previous equations are linear.
They can tell us whether:
- The system is stable: Re() >0
- The system is neutral: Re() =0
- The system is unstable: Re() <0.
If the system is unstable, the instability will grow. Where it
will stop cannot be predicted with this linear approach.
After the instability starts, different outcomes can be
obtained at longer times:
- The system can reach a limit cycle
- The combustor can explode
- The operator may stop combustion because of vibrations
- The system may quench on its own
51
LIMIT CYCLES
Overshoot zone
Non-linear zone:
limit-cycle
Pressure oscillations
Linear zone:
exponential growth
20
40
Linear phase
Exponential growth
of the amplitude
Acoustic solvers
52
60
80
Time
100
120
140
Non linear phase
Limit cycle
?
LES and expts
DESTRUCTION
Limit cycles are only ONE form of
final results in thermoacoustics:
If the loss term is not sufficient, the amplitude of oscillations
might grow until something really unpleasant occurs:
- the combustor can explode because the structure
does not resist
53
In industrial gas turbines, the system switches off when
pressure oscillation or structure vibration levels are too high
- the flame can also quench, unable to resist to oscillations.
This can happen without making any noise.
How do we know ? Thanks to active control (see review on active
control of combustion: McManus, Poinsot and Candel, PECS, 1993, 19).
54
ACTIVE CONTROL OF COMBUSTION INSTABILITIES:
-
VORTEX FORMATION
AND COMBUSTION
PRESSURE
WAVE
INLET FLOW
RATE CHANGE
ACOUSTIC
FEEDBACK
Unsteady sensor
(microphone, optical)
Loudspeaker
EXTERNAL
ACTUATION
SYSTEM
55
What is active control of combustion instabilities ?
Example for a laminar Bunsen flame
56
When the gain is small, the instability is reduced
When the gain is larger, the instability is killed
Noise level
without control
57
58
CONTROL
Active control research for thermoacoustics has been very
strong in the 90s. Patents were taken in many labs (EM2C) and
companies (GE). Industrial demonstrations on real engines
worked:
Industrial gas turbines: active control was used in Siemens
gas turbines (Seume, J., Vortmeyer, N., Krause, W., Hermann, J.,
Hantschk, C., Zangl, P., Gleis, S., and Vortmeyer, D. Application of active
combustion instability control to a heavy duty gas turbine. ASME Journal of
Engineering for Gas Turbines and Power 120 (1998), 721726.)
For aeroengines, certification issues killed the idea
But for scientific investigations, active control remains a great
tool because it allows to let the instability start on request
59
Initiating an instability on demand with active control:
60
Other example: a turbulent burner. This combustor is
unstable but it can be stabilized using active control:
Microphone
Loud speaker
DELAY LINE
+AMPLIFIER
61
Point A: with and without control
62
Poinsot et al J. Phys. 1992
Active control allows not only to
control unstable modes but also
to extend the domain of
operation of the burner (in a fuel/
air flow rate diagram):
- with control, point A is unstable
(limit cycle) without control and
becomes stable with control
- with control, point B does not
burn without active control. With
control it burns and is stable
63
TOTAL HEAT RELEASE
For point B: the flame starts oscillating and ... Quenches !
STABLE UNSTABLE
FLAME
FLAME
QUENCHING
FLAME
QUENCHED
FLAME
THIS IS AN EXAMPLE WHERE FLAME
QUENCHING
IS
DUE TO
CONTROL ON
CONTROL
OFFFLAME INSTABILITIES
64
Noise
In this case, instability leads to quenching
Even though noone hears this...
Heat release
65
