0% found this document useful (0 votes)
7 views

Notes Lecture 13

Uploaded by

Bajaj
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
7 views

Notes Lecture 13

Uploaded by

Bajaj
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 31

Summer 2013

Lecture 13
Flame instabilities

Moshe Matalon

Darrieus-Landau Instability
Darrieus (1938) Landau (1944)

They treated the flame as a surface of density discontinuity, with a constant


flame speed and temperature, similar to the leading order hydrodynamic theory,
and examine the stability of a planar flame front.
burned
Euler equations ⇢b
Dv ⇢u n
∇ ⋅ v = 0, ρ = −∇p + ρg
Dt
unburned Vf
Rankine-Hugoniot conditions
[ρ (v ⋅ n − V f )] = 0 [ v ⋅ n] = (σ − 1) S L
[ v × n] = 0 [ v × n] = 0
x = f ( y, z , t )
[ p + ρv ⋅ n(v ⋅ n − V f )] = 0 [ p ] = −(σ − 1)ρu S L2
€ €
Flame speed relation


S f ≡ v ⋅ n −V f = SL
unb.
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 1
Summer 2013

Plane flame (in a frame attached to the front)


consider first the case with no gravity

SL
!# S (x < 0)
L
v=i"
x =0 $# SL
σ (x > 0)

SL !# 0 (x < 0)
p − p0 = " 2
unburned #$ −(σ −1)ρu SL (x > 0)

We will retain the notation that the top expression in the bracket is for x < 0 and
the bottom expression for x > 0

Moshe Matalon

Introduce small disturbances (denoted by prime) and restrict attention to 2D


u = u(x) + u!(x, y, t) v = v!(x, y, t) p = p(x) + p!(x, y, t) f = f (y, t)

x
u!x + v!y = 0
ρut! + ρ u SL u!x + ρu!u!x + ρ v!u!y = − p!x y

ρ v!t + ρ u SL v!x + ρuv!v!x + ρ v!v!y = − p!y


unburned

u!x + v!y = 0 u!x + v!y = 0


ρut! + ρ u SL u!x = − p!x ) p!xx + p!yy = 0
ρ v!t + ρ u SL v!x = − p!y ρut! + ρ u SL u!x = − p!x

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 2
Summer 2013

burned
(1, − f y , − fz ) ft ρb
n= Vf =
1+ f y2 + fz2 1+ f y2 + fz2 ρu n

n ~ (1, − f y ) t ~ ( f y , 1) unburned Vf
V f ~ ft
x = f ( y, z , t )
across x = 0
[ u" − v"f y ] = 0 [u!] = 0
[v!+(u + u!) f y ] = 0 [v!] = −(σ −1)SL f y
[ p"] = 0 [ p"] = 0
€ at x = 0-

u + u! − v!f y − ft = SL ft = u!(0− )
€ €
Moshe Matalon

u!x + v!y = 0
p!xx + p!yy = 0
ρut! + ρ u SL u!x = − p!x
Solve in the regions x < 0 with ⇢ = ⇢u and x > 0 with ⇢ = ⇢b

Jump conditions across x = 0

[u!] = 0 and
[v!] = −(σ −1)SL f y ft = u!(0− )
[ p"] = 0

These conditions are to be satisfied across the flame front, i.e. across x = f(x,t), but
since f << 1 they can be related to values across x = 0 using a Taylor expansion.
Since
€ the basic state consists of piecewise constants functions, u’(x =f) ~ u’(0),
v’(x =f) ~ v’(0) and p’(x =f) ~ p’(0) .

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 3
Summer 2013

burned
Normal mode analysis 2⇡/k x

iky + ω t A
v' = V ( x) e y
z
iky + ω t unburned
p' = P( x) e
iky + ω t k wavenumber
f ' = Ae
= 2⇡/k wavelength

This is the method of normal modes, where it is assumed that an arbitrary


disturbance introduced as an I.C., can be expanded in a Fourier integral (or
series) over k. Being a linear problem it suffices to examine the response to
one Fourier component.

In three-dimensions, disturbances are of the form f 0 ⇠ ei k·z+!t , where k =


(k1 , k2 ) is the wave-vector and z = (y, z). The 2D results (in this case) can be
easily extended to 3D if
q
k = |k| = k12 + k22

Moshe Matalon

We are interested in the behavior of the solution as t ! 1.

! = !R + i!I

f = Aeiky+!t = Aei(ky+ !I )t e!R t

If the perturbation goes to zero as t ! 1, which is assured when !R < 0 for all
values of k, the basic state (i.e., the planar flame) is stable.
For instability it is sufficient that !R > 0 for one mode (one value of k)

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 4
Summer 2013

When substituting into the system of PDEs, the problem reduces to a set of
ODEs for the determination of V (x), P (x), A with k real , and ! in general a
complex number.

Pxx − k 2 P = 0
ρωU + ρ u S LU x = − Px
U x + ikV = 0

[U ] = 0 $
!
[V ] = −ik (σ − 1) S L A # at x = 0
[ P] = 0 !
"

U (0 − ) = ω A

Moshe Matalon

The associated boundary conditions are homogeneous. It is, therefore a homo-


geneous boundary value problem with homogeneous BCs, for which the trivial
solution U = V = P = A = 0 is a solution. The question is whether there are
other nontrivial solutions. We are therefore faced with an eigenvalue problem
where ! is the eigenvalue.

In the present problem the only parameters are , SL , in addition to k and !.


Solution to the eigenvalue problem exists only if

F( , SL ; k, !) = 0

! = !( , SL ; k) dispersion relation

Dimensional considerations show that

! = !DL( )SL k

The dimensionless factor !DL( ) and, in particular its sign, remain to be deter-
mined.
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 5
Summer 2013

It is instructive to examine first, the case for which 1⌧1


Let 1⌘✏
~ ~ ~ ~
V = εV , U = εU , P = εP , ω = εω

Pxx − k 2 P = 0 Pxx − k 2 P = 0
ρωU + ρ u SLU x = −Px ρ S U = − P
u L x

x
U x + ikV = 0 U x + ikV = 0

[U] = 0 "  =0 "


$ [U]
[V ] = −ik(σ −1)SL A # at x = 0 $
[V ] = −ikSL A # at x = 0
[P] = 0 $ $
%  =0
[ P] %

U (0 − ) = ω A  − ) = ω A
U(0
Moshe Matalon

"
P $ −C1ekx
Pxx − k 2 P = 0 =#
2
ρu SL $ −C e−kx
ρ S U = − P
u L x

x % 2

U x + ikV = 0 " kx
U $ C1e
=#
SL $ C e−kx
 =0
[U] " % 2
$ "
[V ] = −ikSL A # at x = 0 V $ iC1e
kx
$ =#
 =0
[ P] SL $ −iC e−kx
% % 2

1
 − ) = ω A
U(0 C1 + C2 = kA, C1 = C2 =
2
kA,
1
ω = kSL
2

1
ω = (σ − 1) S L k
2

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 6
Summer 2013

u0 (0+ ) = u0 (o ) = ft the flame front is convected by the flow

v 0 (0+ ) v 0 (o ) = ( 1)SL fy the disturbed flame is equivalent to a flat vortex


sheet at x = 0 of strength ⇠ ( 1)fy that
always increases in time (because !R > 0)

The concentrated vorticity, through the Biot-Savart law induces an axial velocity u
proportional to ft that convects segments of the flame intruding towards the burned gas
further upstream and those intruding towards the unburned gas further downstream.
x
burned

y
unburned

Moshe Matalon

burned

unburned

A different explanation that appears in the literature, uses a mass conservation


argument (based on a quasi-steady picture). A streamtube of initial area A must be
larger at the flame than further downstream when the sheet is convex towards the
unburned gas (f < 0) before expanding back to size A. Upon crossing the flame, the
streamtube first converges and then expands to A far away at infinity. An area
increase, just ahead of the flame, causes a velocity decrease there. The flame (which is
moving at its own speed) will encounter a lower speed and will have to move faster,
thus enhancing the initial wrinkle.
The “quasi-steady” assumption in this argument is not totally satisfactory, because the
perturbed flow field is strongly unsteady!

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 7
Summer 2013

Remove the limitation on , and consider 1 = O(1)

' − C (1 + ωˆ )e kx
P $ 1
= ˆ
&
ρu S L2 $C2 ( ω − 1)e − kx & 1 + ωˆ (ωˆ / σ ) − 1 0 #& C1 #
% σ $ !$ !
$1 − (σ − 1) / ωˆ 1 ωˆ / σ !$ C2 ! = 0
U '$ C1e kx
=&
S L $C e − kx + C e − kωˆ x /σ $ 1 −1 − 1 !"$% C3 !"
% 2 3 %
V $
' iC1e kx
=& − kx − iC ωˆ e − kωˆ x /σ
SL
$%− iC2 e 3
(σ + 1)ωˆ 2 + 2σωˆ − σ (σ − 1) = 0
σ
where ωˆ = ω / kS L −2σ ± 4σ 2 + 4σ (σ 2 −1)
ω̂ =
2(σ +1)

For > 1, one of the two roots is always positive


1
ω = σ + 1 &% σ3 +σ 2 −σ −σ#
"S L k
$ !
Plane flames are unconditionally unstable
Darrieus-Landau (DL) or hydrodynamic instability.
Moshe Matalon

Short waves k 1 grow faster then long waves k ⌧ 1 !

Thermal expansion is responsible for the instability


For weak thermal expansion ! ⇠ 12 ( 1)SL k
1⌧1

The hydrodynamic instability appears to contradict the fact that planar flames
have been observed in the laboratory. Note that the result is not valid for short
enough waves, . There may be stabilizing influences of di↵usion at the shortwave
range, i.e. when ⇠ lf as we shall discuss later. nevertheless, the instability,
which is a result of thermal expansion is always present, and is the dominant
phenomenon in large scale flames, where di↵usion play a limited role.

The DL instability has many ramifications in combustion.

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 8
Summer 2013

 Formation of cusps and crests

Since the flame propagates normal to itself at a constant speed, successive


locations of the front can be constructed geometrically using Huygen’s
principle. The displacement over a time interval t is constructed from a
family of spheres of radius SLt centered on the flame surface. The envelope
draws the new location of the flame front.
The convex/concave sections of the flame behave differently during the
propagation. The convex sections (towards the unburned gas) of the flame
expand so that a possible break in the front heals itself. The concave
sections contract forming cusps.
0 t = 0.0

0.024

0.048
-0.2
f(x,t) 0.072

0.096
-0.4
0.120

0.144
-0.6

0 0.2 0.4 0.6 0.8


x
Moshe Matalon

 Bunsen burner flames


Bunsen flame experiments show that as the flow rate through the
burner is reduced the opening angle of the wedge flame widens and the
flame height becomes shorter. One might expect that when the
incoming velocity U approaches SL the flame will eventually become
flat. Instead the flame adopts a different shape with two sharp crests
instead of one. The hydrodynamic instability does not permit flames
that are too flat.

Uberoi et al., POF 1958


Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 9
Summer 2013

 Flames in tubes
Flames in tubes are generally curved (convex towards the unburned
gas)

Uberoi, POF 1959

Moshe Matalon

burned
Gravity effects x
x =0
y
& SL iky + ω t z
v=i% + V ( x) e g>0
σS iky + ω t
$ L f = Ae
unburned
& − ρ u gx iky + ω t
p − p0 = % 2 + P( x) e
$− (σ − 1)ρ u S L − ρ b gx

The only difference from the previous discussion is in the jump


condition for the pressure.
[ p] x= f = −(σ −1)ρ u SL2
# ∂p &
[ p] x= f = [ p] x=0 + % ( ⋅ f +
$ ∂x ' x=0
−(σ −1)ρ u SL2 +[P] x=0 eiky+ω t + (ρu− ρb )g f = −(σ −1)ρ u SL2
σ −1
[P] x=0 = − ρu g A
σ

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 10
Summer 2013

1
ω=
σ +1 { }
(σ 3 + σ 2 − σ )SL2 k 2 − (σ 2 −1)gk − σSL k

Fr = kSL2/g Froude number


1 hp 3 i
 !⇠ + 2 SL k + . . . for k 1
+1
Influence of gravity dies out as k ! 1, but the short waves remain unstable
r
( 1)g 1/2 SL
!⇠ k k + ... for k ⌧ 1
 +1 +1
Long waves are stabilized by gravity for downward propagation (g > 0), but
remain unstable for upward propagation (g < 0).

Moshe Matalon

1. How does the result of Darrieus and Landau reconcile with


the fact that planar flames have been observed in the
laboratory ?
2. What is the influence of diffusion, which have been
neglected in this analysis ?
3. What are the nonlinear consequences of the hydrodynamic
instability ?

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 11
Summer 2013

Thermodiffusive Instabilities
!
Hydrodynamic instability
1
ω = σ + 1 &% σ 3 + σ 2 − σ − σ #"S L k
$ !
! ! 0 as !1
=1
k

Thermodi↵usive instabilities can be studied by filtering out the hydrodynamic


disturbances. This is most easily done by setting ⇢ = const.

reactive-diffusive system lf
Y
DT T
ρc p − λ∇ 2T = QBρY exp(− E/RT )
Dt
DY
ρ − ( ρ D)∇ 2Y = −BρY exp(− E/RT )
Dt
0 x
planar flame
Moshe Matalon

Introduce dimensionless variables using Dth /SL for length, Dth /SL2 for time,
QYu /cp for temperature, and normalize the mass fraction w.r.t. Yu .
DT
r2 T = !
Dt
DY 1
Le r2 Y = !
Dt

We will consider 1 Le 1 small, i.e. near equi-di↵usion flames, then the only
parameter is the reduced Lewis number ` = (Le 1) where is the Zel’dovich
number

The Planar Flame.


⎧1− e x (x < 0)
Y= ⎨
lf ⎩ 0 (x > 0)
Y T ⎧T + e x (x < 0) ⎧ xe x (x < 0)
T= ⎨ u H = Ta − (1− Le−1 )⎨
⎩ Ta (x > 0) ⎩ 0 (x > 0)

0
x
lR
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 12
Summer 2013

iky + ω t
T = T (s ) + φ ( x) e
iky + ω t
h = h (s ) + ψ ( x) e

f = A eiky+!t
Linearized equations.
at x = 0
φ xx − φ x − (ω + k 2 )φ = 0 ( x < 0) [φ ] = A, [ψ ] = −A
φ = 0 ( x > 0) [φ ′] = A − 2 ψ +
1

(ψ xx − ψ x − (ω + k 2 )ψ = −(φ xx − k 2φ )  [φ ′] + [ψ ′] = −A

This is an eigenvalue problem for !; solutions exist only if the following solv-
ability condition is satisfied F(!, k; `) = 0, or ! = !(`, k). The growth rate !
is in general complex, ! = !R + i!I
Stability is assured when !R < 0 for all k; the solution is unstable if there exists
a mode for which !R > 0.

Moshe Matalon

Dispersion relation

(
4 λ2 (1− λ ) +  (1− λ ) 2 − 4k 2 = 0 )
b 0 ! 3 + b1 ! 2 + b 2 ! + b3 = 0
λ = 1 + 4(ω + k 2 )
bi = bi (`, k)
Sivashinsky, CST 1987

b0 = 64 b1 = 192 k 2 + 32 + 8 −  2
b2 = 2(2 + 8k 2 + )(1 + 12k 2 ) b3 = k 2 (2 + 8k 2 + ) 2

Using the Routh-Hurwitz criterion

Conditions (2 + 8k 2 + ) > 0
for stability:
128 (2k ) 4 + (−3 2 + 16 + 128 )(2k ) 2 + (32 + 8 −  2 ) > 0

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 13
Summer 2013

Conditions 2 + 8k 2 + ` > 0
for stability: 128(2k)4 + ( 3`2 + 16` + 128)(2k)2 + (32 + 8` `2 ) > 0

!I 0
= 6=
0 !I

Stable
Unstable Unstable
Steady cellular Pulsating and/or
flames travelling waves

-2
32/3
4(1 + 3 ) 

Moshe Matalon

burned
x
x =0
f = A ei(ky+!I )t e!R t = A ei(ky+!I )t y

marginal states correspond to !R = 0 iky + ω t


f = Ae
reaction sheet unburned

!I = 0 ) f = A eiky cellular flames

k=0 ) f = A ei!I t planar pulsating flames


!I 6= 0 ) k 6= 0 ) f = Ae ik(y+ct)
c = !I /k
traveling waves along the flame surface

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 14
Summer 2013

Cellular flames
ω
 < −2

k*
k1 k
 > −2

The critical Lewis number E(Tb − Tu )


(Le −1) =−2
for cellular flames: RTb2

Cellular instability is predicted for > 2⇡/k ⇤ . The first unstable wavenumber
will be k1 where k1 = 2n⇡/d where d is the flame dimension (i.e. the size of
the burner or the tube) and n an integer. Experiments, however, show that the
cells have an intrinsic size that is independent of the dimension of the flame.

Moshe Matalon

Hydrocarbon-air flames
The range of Lewis numbers,
Dth (N 2 ) = 0.19 cm 2 /s from rich to lean mixtures
DC 3 H 8 −N 2 = 0.11 cm 2 /s
Le ~ { 0.86, 1.73 }
DO2 −N 2 = 0.22 cm 2 /s
 ~ { − 3, 11 }
Hydrogen-air
Lean H2 –air mixture
€ Dth (N 2 ) = 0.19 cm 2 /s
Le ~ 0.31
DH 2 −N2 = 0.61 cm 2 /s
 ~ −10

The critical Lewis number for the onset of cellular flames seems to agree, in
general, with the experimental observations. For hydrocarbon-air flames cellu-
lar flames are observed on the rich side of stoichiometry (except possibly for
methane), while for hydrogen-air flames on the lean side).

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 15
Summer 2013

Observations de l'instabilité de Darrieus-Landau


Photos J.Quinard 1990

Saturation non linéaire, taux de croissance faible Saturation non linéaire, taux de croissance forte

Mécanisme de base de l'instabilité connu depuis presque 50 ans


Validation expérimentale des taux de croissance fait seulement en 1998-2000
– La flamme est instable, donc
Moshela flamme plane n'existe pas dans la nature !
Matalon
–1) Difficulté expérimentale pour préparer une flamme initialement plane
–2) Difficulté expérimentale pour contrôler la perturbation initiale

Goutelas 2004 30

T
Y

T
Le < 1
Y Thermal diffusivity < Mass diffusivity

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 16
Summer 2013

Planar Pulsating flames


The critical Lewis number E(Tb − Tu )
(Le − 1) ≈ 10.93
for pulsating flames: RTb2

The critical Lewis number for the onset of oscillations is much larger than one
implying that oscillations are likely to be observed in lean mixtures of heavy fuels,
or rich mixtures of light fuels. The critical value, however, is quite large and is not
likely to be reached in common combustion mixtures.

In the presence of heat loss the critical Lewis number is reduced significantly and
the instability may be accessible. Indeed it has been observed in porous plug
burners when the flame was sufficiently close to the burner.

Moshe Matalon

Combined effects of hydrodynamic and


diffusive -thermal Instabilities

The early work of Markstein in the 1950’s was phenomenological in nature,


extending the Darrieus-Landau results by assuming a dependence of the
flame speed on curvature.

The more rigorous results are based on the hydrodynamic asymptotic theory,
which includes the influences of diffusion resulting in the internal flame
structure assumed thin but of finite thickness, and extends the DL growth
rate p
3 + 2
!= SL k
+1
| {z }
!DL
to include higher order terms in k.

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 17
Summer 2013

Planar Flames
Pelce & Clavin, (JFM, 1982); Matalon & Matkoswky (JFM, 1982); Frankel & Sivashinsky (CST, 1982)

ω = ω DL SL k − l f [ B1 + β ( Leeff − 1)B2 + Pr B3 ] SL k 2

DL instability Heat conduction Viscous effects


stabilizing stabilizing
Species diffusion stabilizing
in lean CnHm /air mixtures rich
H2 /air mixtures

The coefficients ωDL , B1, B2, B3 > 0 depend only on thermal expansion σ.
Recall Leeff is a weighted average of the fuel and oxidizer Lewis numbers
with a heavier weight on the deficient component.

Moshe Matalon

*
Leeff < Leeff
DL

*
Leeff > Leeff

kc

k
*
Leeff < Leeff The short wavelength disturbances are also unstable
Hydrodynamic instability is enhanced by diffusion effects
Diffusive-thermal instability
*
Leeff > Leeff The short wavelength disturbances (λ > λc= 2π/kc) are
stabilized by diffusion Hydrodynamic instability
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 18
Summer 2013

Comparison of linear growth with DNS for H2 /air flames; Altanzis et al., JFM
(2012); ETH LAV.
0.4 Le=0.4 simulations
Matalon et al. JFM03, Le=1
Le=1 simulations
Darrieus - Landau
0.3 Matalon et al. JFM03, Le=0.4

2
(a) 0.2
1.75
1.5 0.1
1.25
1 0

0.75
-0.1
0.5 0 0.1 0.2 0.3 0.4 0.5
ω

0.25
0
Leeff=0.4 simulations
-0.25 Matalon et al. JFM03, Leeff=1
-0.5 Leeff=1 simulations
-0.75 Darrieus - Landau
Matalon et al. JFM03, Leeff=0.4
-1
0 0.5 1 1.5 2
Moshe Matalon k

ω = ω DL SL k − l f [ B1 + β ( Leeff − 1)B2 + Pr B3 ] SL k 2

ω DL
k=
l f [ B1 + β ( Leeff − 1)B2 + Pr B3 ]

Unstable
Steady cellular
flames Stable

-2 
neutral stability curves
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 19
Summer 2013

Numerical Results
Jackson & Kapila 1984, CST, 41, 191

Neutral stability curves in the limit of large activation energy, but with σ = O(1). The asymptotic
expression for k << 1 agrees well with these results.

Moshe Matalon

Effect of strain with σ = 5 Effect of gravity with σ = 5

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 20
Summer 2013

Spherical Flames (R >> lf)


Bechtold & Matalon, (C&F, 1987); Addabbo et al. (ProCI, 2002)

r = R(t) [1 + A(t) Sn(θ,φ)]


R

growth-rate: 1 dA R # l &
= %ω − f Ω (
A dt R $ R '

1 dA R & l )
= 'ω DL − f "# B1 + β (Leeff −1)B 2 + Pr B 3 $%*
A dt R ( R +

hydrodynamic
diffusion

Moshe Matalon

R(t)
*
Leeff > Le eff The amplitude A first decay, but after the flame reaches a certain critical
size Rc it starts growing in time. The instability is hydrodynamic in nature.
*
Leeff < Leeff The amplitude grows in time immediately upon its incept. The flame is
diffusively unstable
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 21
Summer 2013

Growth rate computed for a lean propane/air fame σ = 5.9, β(Leeff -1) = 4.93

Moshe Matalon

Marginal Stability Curves

Lean propane/air: σ = 5.9, Leeff =1.49

Λmin= 2πR/n ~ const.


wavenumber n

R
n~

Growing modes:
Λmin < Λ < Λmax

n = const.

Λmax = 2πR/n ~ R
Pecr
Peclet number Pe = R/lf
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 22
Summer 2013

Iso-octane (ϕ = 1.4 – 1.6)


Methane (ϕ = 1.0)

Bradley et al. (C&F, 2000)

Moshe Matalon

Nonlinear Effects
The linear theory predicts whether disturbances of given wavelength grow or
decay in time. The consequences of an instability needs to be determined from
a nonlinear analysis.

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 23
Summer 2013

Hydrodynamically unstable flame


L = 0.075 corresponding to Le > Le⇤ , for =4
critical wavelength c = 1.7

L=1 L=2 L=4

steady propagation:
Moshe Matalon

Propagation speed of the hydrodynamically unstable flame

supercritical
1.25
σ = 5.5
U / S speed U/SL

subcritical
1.20
σ=5

1.15
propagation
L

σ=4
~

1.10

σ=3
1.05

1.00
40 60 80 100 120 140 160
γ
Moshe Matalon =( 1)L/L

Moshe Matalon
University of Illinois at Urbana-Champaign 24
Summer 2013

Steadily-propagating wrinkled flames in quiescent mixtures


x = U t + f (y)
= 80 = 160

U = 1.08 U = 1.15

Moshe Matalon

Diffusion Flame Instabilities


First observation of a thermodiffusive instability in diffusion flames
J.E. Garside and B. Jackson 1951
Polyhedral diffusion flame

Formation of a polyhedral flame, previously observed by


Smithells & Ingle (1892) and by Smith & Pickering (1929)
only in premixed combustion.

The photograph is a polyhedral diffusion flame produced


when a mixture of H2 diluted in CO2 is burned at
atmospheric pressure upon a cylindrical tube of 1.1 cm in
diameter. The surface of the flame was formed of five
triangular cells in the shape of a polyhedron. By adjusting Garside & Jackson, Nature 1951
the flow rate, the flame could be made to rotate about a Polyhedral premixed flame
central vertical axis.

Similar polyhedral flames were observed when N2 was used


as the diluent gas instead of CO2.

Moshe Matalon
S. Sohrab

Moshe Matalon
University of Illinois at Urbana-Champaign 25
Summer 2013

1 cm

N2 diluted H2/O2 diffusion flame at the base of a splitter plate burner


Dongworth & Melvin, 1976

Under normal conditions the base of the flame close to the burner partition, was straight.
At sufficiently high flow velocity and for when the hydrogen concentration was
substantially reduced the base became cellular. Same behavior was observed when the
nitrogen in the fuel stream was replaced with Ar but not with He.

Moshe Matalon

Stability analysis
F (!, k; D, LeF , LeO , , ) = 0

Damköhler number D
Dth Dth
Lewis numbers LeF = DF , LeO = DO
YF /YO
mixture strength = ⌫F WF /⌫O WO

thermal expansion = ⇢u /⇢f

Instabilities in di↵usion flames have been typically observed at high flow rates, or
near-extinction conditions. They are predominately di↵usive-thermal in nature
with hydrodynamics playing a secondary role.

The Burke-Schumann solution of complete combustion (corresponding to the


limit D ! 1) is unconditionally stable. Instabilities result only when there is
sufficient leakage of reactants through the reaction sheet with a flame temper-
ature significantly below Ta .

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 26
Summer 2013

Tf
Ta
D*

Flame temperature
ble stable
unsta BS limit

Dext Damköhler number D

 Instabilities develop when the temperature is lowered significantly below the adiabatic
temperature and there are significant reactant leakage, i.e., for Dext< D < D* namely at
high flow rates or near-extinction conditions.
 D may be controlled by varying the flow rate U, or the burning temperature Ta through
dilution (nature or amount) or the mixture strength ϕ, or by varying the stretch rate K
(tubular flames).
 Instabilities are predominantly thermo-diffusive in nature with hydrodynamic effects
playing a secondary role; the mode of the instability depends on the Lewis numbers LeF ,
LeO and on the initial mixture-strength ϕ.
Moshe Matalon

LeF
high-frequency
instability
Planar
pulsations

1
Stationary oscillatory
cells high-frequency cells
instability

LeO
1
Cellular flames - more likely to occur when ϕ < 1 (lean conditions) - cell size λ = 0.5 - 2 cm
Pulsating flames - more likely to occur when ϕ > 1 (rich conditions) frequency ωI = 1 - 6 Hz

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 27
Summer 2013

1 cm

Dongworth & Melvin, 1976

Nitrogen-diluted H2-O2 Diffusion flame - splitter-plate burner

Under normal conditions - base remains straight; at high flow rates and at high dilution rates
base is cellular. Same behavior when diluted in Ar but not in He.

High flow rate D < D*


in N2 and Ar LF = 0.33 – 0.35
in He LF = 1.02

Moshe Matalon

Pulsations – Jet Flames


Füri, Papas & Monkewitz, 2000

Fuel Oxidizer
Stream Stream
LeF LeO Pulsations

4% C3H8 65% O2
96% N2 35% N2
1.90 1.09 0.33 No
6% C3H8 40% O2
94% N2 60% N2
1.86 1.05 0.76 No
8% C3H8 30% O2
92% N2 70% N2
1.83 1.03 1.32 Yes
12% C3H8 19% O2
88% N2 81% N2
1.78 1.00 3.04 Yes
24% C3H8 19.5% O 2
76% N2 80.5% N2
1.63 0.93 5.56 Yes
42% C3H8 15.5% O 2
58% N2 84.5% N2
1.45 0.85 11.24 Yes
80% C3H8 14.5% O 2
20% N2 85.5% N2
1.20 0.73 19.23 Yes
100% C3H8 14% O2
0% N2 86% N2
1.10 0.69 23.26 Yes
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 28
Summer 2013

Cellular Diffusion Flame


H2 & O2 diluted in CO2 (LeF = 0.35, LeO = 0.86, ϕ = 0.24)

=1
9.1 mm
planar flames

extinction

Robert & Monkewitz

Transition from planar through cellular


flames all the way to extinction as the
mixture strength is reduced.

Moshe Matalon

Planar pulsating diffusion flame

Flame intensity

Flame position

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 29
Summer 2013

CO2 diluted H2/air tubular diffusion flame

Increasing stretch rate from 207 s-1 to 386 s-1

Shopoff, Wang & Pitz 2011


Moshe Matalon

Effect of thermal expansion on cellular flames


= ⇢u /⇢f

LeF =LeO = 0.75 , ϕ = 0.5

D fixed
σ = 6.1
growth rate !

4.8
4
3.6

wavenumber k
Thermal expansion further destabilizes the cellular flame
extends the range of conditions for cellular flames (increases D*)
Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 30
Summer 2013

Effect of thermal expansion on pulsating flames


= ⇢u /⇢f

LeF = 1.5 , LeO = 1.1 , ϕ = 2

D fixed
σ = 2.3

growth rate ! 2.9

3.5

wave number k
Thermal expansion has a stabilizing influence on planar oscillations;
reduces the range of conditions for oscillations (decreases D*)

Moshe Matalon

Moshe Matalon
University of Illinois at Urbana-Champaign 31

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy