Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Axisymmetric description of the scale-by-scale
scalar transport
Luminita Danaila
Context:
ANR ‘ANISO’: F. Godeferd, C. Cambon, J.B. Flor
ANR ‘Micromixing’: B. Renou, J.F. Krawczynski,G. Boutin, F. Thiesset CORIA, Saint-Etienne-du-Rouvray, FRANCE
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
OUTLINE
II. Analytical development
III. Validation with experimental data
I. Context, previous work and motivation
V. Conclusions
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Kolmogorov : Local isotropy
Universality
I. Context and motivation
FLOW
Re
)()( xurxuu LLL
Real space:
x
rx
22 )()( xurxuu LLL
A simple analytical plateform: relation between the second-and the third-order moments at a scale r
energy flux at a scale r
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Scale-by-scale energy budget: Kolmogorov (1941)
3 4δu ε r
5
λR
Non-universality for moderate Reynolds numbers
I. Context and motivation
3 24< δu > = ε r 6ν δu
5 r
Antonia & Burattini, 2006
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
• For different REAL flows (moderate Reynolds, locally isotropic or anisotropic …)
• Necessity to account for explicitly the non-negligible correlation between large-and small scales
I. Context and motivation
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
2
3 2 2
2
1 4 4 4+ δu ε 2ν + δu δu
3 r r 3 r r r t
Isotropy (local) and integration with respect to r
r
244
0
3+ s δu ds
r t
3 24< δu > = ε r 6ν δu
5 r
Method: Navier-Stokes in 2 space points:
Increments:
I. Context and motivation
)()( xurxuu LLL
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
fuI
< ε > r 24 6ν
δu5 ε r r
3δu
ε r
*r
Finite Reynolds numbers- flows:
Grid turbulence, round jet,
channel flow (axis, near wall) ...
Conclusion:
Energy transferred at a scale r= turbulent diffusion + molecular effects+
large-scale effects: shear, decay, mean temperature gradient …
I. Context and motivation
Kolmogorov, 1941Saffman 1968, Danaila et al. 1999, Lindborg 1999
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
fuI
< ε > r 24 6ν
δu5 ε r r
3δu
ε r
Real- Finite Reynolds numbers- flows: Slightly heated grid turbulence, grid turbulence with a Mean scalar gradient ..
Same conclusion : Energy transferred at a scale r … large-scale effects
I. Context and motivation
Similar questions hold for scalars and turbulent kinetic energy
r
I
r
u
dr
dk fsL
2
22
3
4
r
I
r
quq
dr
d fqL
2
22
3
4 R.A. Antonia et al. 1997Danaila et al., 2004Burattini et al., 2005
Yaglom, 1949Danaila et al. 1999
Kolmogorov, 1941
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
II. Analytical development
r
I
r
quq
dr
d fqL
2
22
3
4
Shortcome: local isotropy was supposed …
Question: which is the anisotropic/axisymmetric form of
Scalar equation simpler development
Note: The axisymmetric form of Kolmogorov equation
Chandrasekar 1950, Lindborg 1996, Antonia et al. 2000, Ould-Rouis 2001 ….
Problem: a large number of scalars which are difficult/impossible to be determined from experiments
?
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
II. Analytical development
),(2 rq
)()(2)(2 2 rqur
rux
uuru
x
uU i
ii
i
From:
rxxrq
r
2)(2 2
2
2
x
rx
n
)cos(
r
r
…. All the terms in Eq. (1) depend on 2 variables
Several ways:
isotropy dependence on r
integration over a sphere of radius r dependence on r
(1)
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
II. Analytical development
nrNrrMrqu aa ),(),()(2
Chandrasekar, 1950
Development similar to Shivamoggi and Antonia, (Fluid Dyn. Res., 2000)
22 1**),(),( rrMrqu a
),(**),(),(2 rNrrMrqu aaII
Measurable:
u andIIu
aa Nrr
Mr
r
21
3
),(21
22 2
2
2
2
2
2
2
rIqrr
Injection: decay, production..
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
II. Analytical development
rr
r
dsPDs
sr
dsqs
sr
dsqs
sr
rG
0 2
22
30
2
2
22
2
22
3
0
2
2
22
3
1)1(2
12
12
3
2),(
C. Cambon, L. Danaila, F. Godeferd, Y. Gagne and J. Scott, in preparation
With: Vrrr
Na21
and
Vrr
Na
2
* 1
*aa NMG
Eq. (2)
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
III. Validation with experimental data: EXPERIMENTAL SET-UP*
Volume PaSR: V = 11116 cm3
Injection velocity: UJ = 4.5 - 47 m/s Return flow= porous top/bottom plates Residence time: tR = 8 -46 ms
Reynolds number 60 Rl 1000 (center)
*Prof. P.E. Dimotakis of Caltech was responsible for the conceptual and detailed design of the PaSR and contributed to the initial experiments.
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
A forced box turbulence
x
y z
Injection zone = impinging jets
« Mixing » zone = stagnation zone
Return flow (top/bottom porous)
III. Validation with experimental data
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
III. Validation with experimental data
I
II
Energy Isotropy? iiuuq2
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
3
PIV
( u) 4
r ε 5
JETS
The sign changes at
Large scales (inhomogeneity)
III. Validation with experimental data
Third order ‘classical’ Structure functions : Selection of one particular direction
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
III. Validation with experimental data
),(2 zxq ),(2 rq
H
H
V
Vr
r
x
x
r
),( r
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
III. Validation with experimental data
04/ 2/
3
2),(rG
Eq. (3)
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
III. Validation with experimental data
]2/,0[
d
drGrR
0
0
32
),()(
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
IV. Conclusions
Theory:
-Scale-by-scale energy budget equation for kinetic energy (scalar) in axisymmetric
turbulence (axis of a round jet, axis of two opposed jets ..)
-Measured quantities: u and v (components perpendicular and parallel to )
-all the terms in Eq. (2) can be determined experimentally
The flow:
• Pairs of impinging jets; Return flow by top/bottom porous locally axisymmetric flow
Results :
- better agreement with asymptotic predictions (high Reynolds, anisotropic flows) along the direction normal to the axisymmetry axis (homogeneous plane) than along the direction parallel to
n
n
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Large pannel of structures
Large-scale instabilities (jets flutter)
Mechanisms controlling the mixing?
H/D=3 H/D=5
Instantaneous fields of the mixing fraction
IV. Description of the scalar mixing: fluctuating field
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Re 104D (mm), 2H (mm), H/D
10, 60, 3
10, 60, 3
6, 60, 510, 60,
310, 60, 3 6, 60, 5 6, 60, 5 6, 60, 5
Qv (m3/h) 60 86 60 129 155.2 100 129 155.2
Vinj (m/s) 6.63 9.50 18.42 14.26 17.15 30.70 39.60 47.65
TR (ms) 43.56 30.39 43.56 20.26 17.15 26.14 20.26 17.15
P (bar) 1.40
(m²/s) x10-5 1.089
6089 8728 10149 13092 15751 16915 21821 26252
26252
6089
8 injection
conditions
DVinj Re
III. Validation with experimental data
2 Geometries:
H/D=3
H/D=5
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
The other tests• 2-rd order SF with the Kolmogorov constant
• Normalized dissipation which L? Attention to initial conditions versus universality .. However, a reliable test
for • The most reliable test is the 1—point energy budget
equation, when the pressure-related terms might be neglected (point II).
III. DESCRIPTION of the FLOW: fluctuating field; PIV for determining small scale properties
3/23/22 rCu K
3'u
LC
5.0C
NDISSIPATIO
PRODUCTIONDIFFUSIONVISCOUS
DIFFUSIONPRESSUREDIFFUSIONTURBULENTDECAY
R
Uw
y
Vv
R
Uu
y
U
R
Vuv
y
q
R
q
RR
q
pvy
puRRR
vqy
uqRRRy
qV
R
qU
²²²1
111²
2
1²
2
11²²
22
2
22
150R
/r
3/2r
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
Velocity field: 1) Particle velocimetry– Resolution and noise limitations– PIV resolution linked to size of interrogation/correlation window, e.g., 16, 32,
and 64 pix2, and processing algorithm choices• Does not resolve small scales: the smallest 100% =1.7 mm• Problem to estimate energy dissipation directly
– Towards adaptive/optimal vector processing/filtering
2) LDV in (1 point and) 2 points – Simultaneous measurements of One velocity component in two points of the
space: spatial resolution 200 * 50 microns; sampling frequency= 20 kHZ
Scalar field: PLIF on acetone Small-scale limitations set by spatial resolution (pixel/laser-sheet size)
The smallest resolved scale 100% =0.7 mm
Signal-to-noise ratio per pixel– Adaptive/optimal image processing/filtering
III. Validation with experimental data
Torino, October 27, 2009
CNRS – UNIVERSITE et INSA de Rouen
JETS
III. Validation with experimental data
Third order ‘classical’ Structure functions