Flame Surface Fractal Characteristics in Premixed Turbulent

Combustion at High Turbulence Intensities

Esen Cintosun

University of Toronto, Toronto, Ontario M3H 5T6, Canada

Gregory J. Smallwood

National Research Council Canada, Ottawa, Ontario K1A 0R6, Canada

and
mer L. Glder

University of Toronto, Toronto, Ontario M3H 5T6, Canada

DOI: 10.2514/1.29533
Because of the geometrical nature of the wrinkling of the ame surfaces in premixed turbulent combustion, the
fractal approach is considered to be one of the most appropriate in power-law subgrid-scale models that are widely
usedinlarge eddy simulations. However, the available experimental data inliterature onfractal characteristics of the
premixed turbulent ame surfaces shows signicant discrepancies among data sets, and the several different fractal
algorithms used for evaluation may not produce consistent results. In this work, we analyzed premixed turbulent
ame front images froman extensive experimental study using three different fractal algorithms and determined the
fractal parameters: namely, fractal dimension D, inner cutoff "
i
, and outer cutoff "
o
. These images were obtained
using laser-induced uorescence of OH and Mie scattering on two Bunsen-type burners with diameters of 11.2 and
22.4 mm, withames of propaneair with equivalence ratios of 0.8 and 1. Nondimensional turbulence intensity u
0
=S
L
(ratio of uctuating velocity to laminar ame velocity) was from 0.9 to 15, and the Reynolds number, based on the
integral length scale, was from 40 to 467. The ame front surface areas were calculated for each set of results and
compared with experimental turbulent burning velocities. The three fractal algorithms used for analysis in this
work gave signicantly different results for the same ame image sets. Further, nondimensional turbulent burning
rates computed from the fractal parameters did not agree with the experimental turbulent burning data, except for
those fractal results obtained by the box-counting method and only for u
0
=S
L
6. Implications of these results for
amelet models are discussed. In addition, the perimeter-ratio approach recently adopted to estimate the wrinkled
ame surface area is discussed, and it is argued that the results obtained by the perimeter-ratio approach might not
be unique.
I. Introduction
O
NE of the subgrid-scale models widely used in large eddy
simulation of premixed turbulent ames is the power-law
approach. These are based on the amelet models in which the
reaction is assumed to take place in thin layers, separating fresh gases
from burned products, being wrinkled by the turbulence [1]. In these
models, the turbulent ame speed can be represented as the product
of the laminar ame speed S
L
, corrected for the effects of stretch
(strain and curvature) and the ame-wrinkling . In several studies,
it is proposed to model using the fractal theory (see, for example
[14]). This approach requires submodels for the fractal dimension
and the inner and outer cutoffs, although the outer cutoff is dependent
on the subgrid scale. Because of the geometrical nature of the
wrinkling of the ame surfaces, the fractal approach is considered to
be one of the most appropriate in power-law models. The ame-
wrinkling can be expressed as
=(A
T
=A
o
) =a("
0
="
i
)
D2
(1)
where "
0
and "
i
are outer and inner cutoffs, respectively, and Dis the
fractal dimension. The prefactor a is a variable with a laminar limit of
1 and an intense turbulence limit of 2, but may not be bounded
between 1 and 2 [5]. In most modeling exercises a is taken as a
constant of order 1. A
T
=A
o
is the ratio of the wrinkled ame surface
area to the ow cross section.
It is argued in [6] that in the nondimensional turbulence intensity
u
/
=S
L
(range from 1 to 15) the experimentally determined mean
fractal dimension is about 2.2 and does not showany dependence on
turbulence intensity. This value of the fractal dimension is much
lower than the values found by the previous studies that showed that
the fractal dimension asymptotically reaches to 2.352.37 when the
nondimensional turbulence intensity exceeds 3. It is pointed out that
the probable reason for this discrepancy is the image analysis method
used in the previous studies [6]. The insensitivity of the fractal
dimension of ame front surfaces to turbulence intensity has been
conrmed by independent experiments [7,8]. However, the
conclusions reached in [6] (specically, the insensitivity of the
fractal dimension to turbulence intensity and the numerical value of
the fractal dimension of the premixed turbulent ame front surfaces)
have been questioned in [8,9]. To resolve the controversy regarding
the magnitude of fractal dimension and its dependence, if any, on the
turbulence intensity of medium to high values, further work is
required. In addition, the question of whether the increase in ame
surface area by turbulence is sufcient to explain the observed
increases in turbulent burning rates at high enough turbulence
intensities should be answered, and the conditions at which a
premixed ame front is no longer a passive surface should be
identied.
As discussed in [6], fractal analysis techniques based on different
methods may yield different results. In this work, we analyzed
previously obtained ame front images from an extensive
Presented as Paper 1349 at the 45th AIAA Aerospace Sciences Meeting
and Exhibit, Reno, NV, 811 January 2007; received 29 December 2006;
revision received 5 July 2007; accepted for publication 20 July 2007.
Copyright 2007 by . L. Glder. Published by the American Institute of
Aeronautics and Astronautics, Inc., withpermission. Copies of this paper may
be made for personal or internal use, on condition that the copier pay the
$10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood
Drive, Danvers, MA 01923; include the code 0001-1452/07 $10.00 in
correspondence with the CCC.

Graduate Student, Institute for Aerospace Studies, 4925 Dufferin Street.

Competency Leader, Combustion, Institute for Chemical Process and

Environmental Technology, Building M-9, 1200 Montreal Road.

Professor, Institute for Aerospace Studies, 4925 Dufferin Street, Member

AIAA.
AIAA JOURNAL
Vol. 45, No. 11, November 2007
2785
experimental study using three different fractal algorithms and
determined the fractal parameters: namely, fractal dimension D,
inner cutoff "
i
, and outer cutoff "
0
. The ame front surface areas were
calculated for each set of results and compared with experimental
turbulent burning velocities. Additional analysis of the images
consisted of determining the perimeter ratios (as implemented in
[9,10]) and of comparing these ratios to turbulent burning velocities.
II. Experimental Data
The experimental setup used to obtain the data analyzed in this
study is described in detail in [6]. A brief outline will be given here.
The turbulent premixed conical ames were produced by two
axisymmetric Bunsen-type burners with inner nozzle diameters of
11.2 and 22.4 mm. Premixed turbulent propaneair ames with
equivalence ratios of 0.8 and 1.0 were stabilized by using an
annular propane pilot for low-turbulence ames and a hydrogen pilot
for high-turbulence ames. Perforated plates positioned three nozzle
diameters upstream of the burner rim controlled the turbulence
levels. The turbulence parameters were measured by LDV under
reacting conditions in which the ow is seeded by ne silicone oil
droplets. The length scales and turbulence intensities u
/
reported in
Table 1 were measured on the burner centerline at the nozzle exit.
The instantaneous ame fronts were visualized both by laser-
induced uorescence (LIF) of OH and by Mie scattering. A tunable
excimer laser (Lambda Physik EMG 150T MSC) was used for both
techniques. The dimensions of the laser sheet at the burner centerline
were about 17 cm by 100 m (full width at half-maximum) in the
vertical and horizontal planes, respectively. The sheet thickness was
less than 150 mover the full ame width. The optical detector was
a large pixel format CCD detector (1242 1152 pixels) giving a
ame image spatial resolution of 150 m. At each condition, a
minimum of 100 ame images were captured.
The ame front contours were obtained from the LIF of OH and
Mie scattering images by the methods described in [6]. The ame
front contours then were analyzed using the three different fractal
algorithms described in the next section.
III. Fractal Analysis
In this work, we used three different implementations of the fractal
analysis. These are the techniques that have been used in the majority
of the applications of fractals to physical problems. The three
methods are the caliper technique, Minkowski circle technique, and
the box-counting technique. These are discussed in detail in [11],
therefore only brief descriptions will be given here.
In the caliper technique, the length of a curve on a plane is
estimated by counting how many ruler lengths it takes to cover the
whole length of the curve fromone end of the curve to the other. This
process is repeated for a set of ruler lengths. Aloglog plot of number
of rulers (i.e., the length of the curve) versus ruler lengths yields
information to determine the fractal dimension and inner and outer
cutoffs. Such plots are known as Richardson plots [11]. Figure 1
shows a typical Richardson plot for one of the ame images analyzed
in this study.
Fractal analysis using the Minkowski circle method involves
moving a circle with a specied diameter continuously along the
contour line while keeping the center of the circle on the contour line.
The area covered by the circle is determined and plotted as a function
of the circle diameter, for several realizations with different circle
diameters, on a loglog scale to obtain the Richardson plot.
Fractal analysis using the box-counting method consists of
covering the fractal image with squares (boxes) and then counting
how many boxes are needed to cover the fractal curve completely.
This exercise is repeated with different sizes of squares. Similar to
previous two methods, a loglog plot of the number of boxes versus
the box size gives the desired Richardson plot from which one can
Table 1 Summary of experimental conditions and derived fractal parameters for OH results
Flame set
a
, mm u
/
=S
L
Re

Caliper method "

i
="
o
=D
b
Box-counting method "
i
="
o
=D
b
Minkowski method "
i
="
o
=D
b
K1 1.0 1.77 6.47 308 0.66/11.39/2.24 0.38/38.80/2.32 2.82/19.97/2.39
K2 1.0 1.77 6.47 308 0.62/10.91/2.23 0.39/39.54/2.30 2.65/19.68/2.39
K3 0.8 1.91 9.1 326 0.45/9.97/2.19 0.39/35.82/2.30 3.56/21.04/2.42
K4 1.0 1.65 10.44 463 0.59/11.71/2.24 0.37/35.98/2.31 1.89/18.70/2.37
K5 0.8 1.66 15. 467 0.59/11.16/2.25 0.37/36.51/2.32 2.12/19.01/2.40
K6 0.8 1.77 7.73 257 0.55/12.67/2.18 0.37/39.28/2.31 3.30/20.62/2.38
K7 1.0 1.53 5.3 218 0.60/11.44/2.22 0.38/38.04/2.31 2.68/20.31/2.38
K9 1.0 1.75 8.58 404 0.52/10.44/2.22 0.37/33.42/2.34 2.31/20.87/2.38
K10 0.8 1.66 12.9 402 0.47/9.38/2.20 0.37/35.24/2.33 2.71/20.42/2.40
K11 1.0 1.56 7.09 297 0.51/11.03/2.20 0.39/34.89/2.33 2.63/20.77/2.39
K12 1.0 1.57 7.86 332 0.62/11.32/2.24 0.37/38.49/2.31 2.84/21.15/2.38
K13 0.8 1.68 11.27 355 0.61/10.03/2.25 0.36/35.83/2.31 2.29/21.27/2.38
B1 1.0 1.51 0.84 34 1.07/16.98/2.20 0.45/43.74/2.27 3.87/28.16/2.33
B2 0.8 1.8 1.5 50.6 0.58/23.49/2.12 0.45/60.07/2.23 7.56/48.89/2.36
B3 0.8 2.48 2.0 96.1 1.20/19.42/2.24 0.43/56.32/2.26 5.01/40.91/2.37
B4 0.8 1.96 6.16 227 0.62/7.77/2.21 0.40/24.77/2.32 2.55/16.34/2.38
B5 1.0 2.06 4.31 239 0.60/7.61/2.24 0.39/24.09/2.31 2.21/15.62/2.37
B6 1.0 1.76 3.27 155 0.55/8.96/2.19 0.41/26.07/2.31 2.75/17.88/2.38
B7 1.0 1.76 3.27 155 0.61/9.35/2.18 0.40/25.69/2.31 3.09/18.17/2.38
B8 0.8 4.55 145 0.49/9.29/2.16 0.41/26.44/2.31 3.16/19.13/2.37
a
Flame set B refers to a burner diameter of 22.4 mm, whereas K refers to a burner diameter of 11.2 mm. Only representative ames from set B are listed.
b
Values for "
i
and "
o
are in millimeters.
0.1 1 10 100
0.5
0.6
0.7
0.8
0.9
1
1.1
F
l
a
m
e

c
o
n
t
o
u
r

l
e
n
g
t
h

(
n
o
r
m
a
l
i
z
e
d
)
Caliper length,
inner cutoff,
i
outer cutoff,
o
slope = 1-D
Fig. 1 Example of a Richardson plot used to evaluate the fractal
parameters.
2786 CINTOSUN, SMALLWOOD, AND GLDER
evaluate the fractal parameters. The fractal dimension obtained by
this method is also known as the box dimension.
The caliper technique we adopted is the algorithm used in [7,8].
The Minkowski circle technique is the implementation that was used
in [12]. The box-counting technique is the implementation used in
[13] and was based on the description given in [14].
IV. Results and Discussion
One of the important aspects of planar measurements with laser
sheets is that the contour information is projected on a 2D plane,
although the ame surface is essentially three-dimensional. The laser
sheet is not always perpendicular to the ame sheet, and so the front
contours are projections on 2D planes. These contours represent a
self-afne transformation of the real 3D contours because certain
portions of a ame front contour are scaled by afnities, which are
linear transformations with different contraction ratios in different
directions. These projected contours can be classied as self-afne
fractals. The statistical properties of a self-afne contour or prole
y(x) are invariant under the scaling transformation x x, y
H
y
[15]. The Hurst exponent H can be related to a fractal dimension of
the contour through D
2
=2 H [16], whereas D=D
2
1 [11].
Thus, the 2D measurements and estimation of fractal dimension
reects the three-dimensional characteristics of the ame surfaces if
we assume that, to a rst approximation, the probability density
function of the ame crossing angle is single modal and peaks at a
value corresponding to the Hurst exponent H. More generally, if
each contour H is constant, then the contour is unifractal, where
0 <H <1. If different H are needed for different segments, then the
contour is multifractal [17].
Figure 2 shows the comparison of the fractal dimension D
evaluated for several ames images obtained by both LIF of OH and
Mie scattering, using the box-counting method. The differences in
fractal dimensions of OHand Mie images are within the uncertainties
of D.
Fractal dimensions evaluated by using the caliper, box-counting,
and Minkowski circle methods are shown in Figs. 35, respectively.
In all three gures, the fractal dimension does not show any
signicant dependence on nondimensional turbulence intensity,
especially for medium to high intensities. Another surprising
observation is that the mean values of D show a strong dependence
on the algorithmused in extracting the fractal dimension. The caliper
method yields a D value close to 2.2 (Fig. 3), in agreement with the
previous observations [6,7]. The fractal dimension determined by the
box-counting method shows a mild dependence on u
/
=S
L
for
u
/
=S
L
6 (Fig. 4). At higher intensities, Dsettles to a constant value
of about 2.32. The fractal dimensions determined by the Minkowski
circle method are shown in Fig. 5. Similar to box-counting results,
there seems to be a very mild dependence on u
/
=S
L
for u
/
=S
L
6,
but at higher intensities, fractal dimension stays constant at about 2.4.
The inner and outer cutoffs also showed variations among the
estimations from three different algorithms (Table 1).
In Figs. 68, we compare the experimentally determined
nondimensional turbulent burning velocities S
T
=S
L
with those
estimated from the measured fractal parameters, evaluated by
different algorithms, using the following closure:
S
T
=S
L
=(A
T
=A
o
) =a("
0
="
i
)
D2
(2)
where a was taken as unity. Figure 6 compares the S
T
=S
L
determined
from the experimentally measured values using the approach in
[6,10] to the nondimensional turbulent burning velocity estimated
from Eq. (2) using the fractal parameters evaluated by the caliper
technique. Similar comparisons are shown in Figs. 7 and 8 for S
T
=S
L
evaluated by the box-counting and Minkowski circle methods,
respectively. There is a clear and signicant discrepancy between
measured values of S
T
=S
L
and those determined by Eq. (2). The only
agreement seems to be for the box-counting method case (Fig. 7) for
u
/
=S
L
6.
0 1 2 3 4 5 6 7
2.00
2.05
2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
Mie scattering
PLIF of OH
F
r
a
c
t
a
l

d
i
m
e
n
s
i
o
n
,

D
Nondimensional turbulent intensity, u'/S
L
Box counting method
Fig. 2 Comparison of the fractal dimensions using the box-counting
method from Mie scattering images and OHimages fromidentical ame
conditions.
0 2 4 6 8 10 12 14 16 18
2.00
2.05
2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
F
r
a
c
t
a
l

d
i
m
e
n
s
i
o
n
,

D
Nondimensional turbulence intensity, u'/S
L
Caliper method
Fig. 3 Variation of the fractal dimension evaluated using the caliper
method with nondimensional turbulence intensity.
0 2 4 6 8 10 12 14 16 18
2.00
2.05
2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
F
r
a
c
t
a
l

d
i
m
e
n
s
i
o
n
,

D
Nondimensional turbulence intensity, u'/S
L
Box counting method
Fig. 4 Variation of the fractal dimension evaluated using the box-
counting method with nondimensional turbulence intensity.
CINTOSUN, SMALLWOOD, AND GLDER 2787
Several turbulent ame propagation formulations, including those
based on fractal concepts, are founded on the fundamental
assumption that S
T
=S
L
is proportional to the ratio of the wrinkled
ame surface area A
T
to the owcross-sectional area A
o
. If the fractal
geometry approach [i.e., Eq. (1)] is yielding a true measure of the
wrinkled surface area of the ame front, then Eq. (2) may not be a
reasonable assumption for the turbulent premixed ames in regimes
in which amelet models are assumed to be valid. This experimental
evidence can be interpreted to mean that the ame surface area
increase is not the dominant mechanism in increasing the turbulent
burning velocity under the conditions corresponding to the thin-
reaction-zones regime. When the nondimensional turbulence
intensity u
/
=S
L
exceeds about 67, the ame surface area increase
estimated by the fractal analysis does not explain the observed
increases in the turbulent burning velocity. Of course, this is based on
the assumption that premixed turbulent ame surfaces possess fractal
characteristics. In viewof the fact that the ratio of outer to inner cutoff
is about 1015, fractal assumption may not be entirely justied.
A relatively new concept to get an estimate of the ame surface
area is the ame perimeter ratio P
R
. This approach is used in [9,10].
The actual length of the ame contour perimeter is divided by the
ow characteristic length to estimate the surface area ratio, that is,
S
T
=S
L
=(A
T
=A
o
) =bP
R
(3)
where b is a constant of unity [9]. The perimeter ratio is calculated as
P
R
=P=L
C
(4)
where P is the perimeter length of the main ame contour, and L
C
is
the ow characteristic length taken as the burner diameter. This
approach was also used in this work to estimate the ame surface area
change with turbulence intensity. S
T
=S
L
, estimated from Eq. (3), are
compared with those obtained experimentally in Fig. 9. The
perimeter-ratio approach overestimates the turbulent burning
velocity by more than a factor of 2. Experimental u
/
=S
L
values
evaluated at mean progress variable (c) =0:05 show a relatively
better agreement with the perimeter-ratio estimates (Fig. 9).
However, the perimeter-ratio approach is based on a central
assumption that may not always be valid.
The perimeter of an irregular (non-Euclidean) planar object
depends on the yardstick length used to measure it. This is the
original question that led Richardson to his studies of the irregular
objects: Howlong is the coastline of Great Britain? [18]. As can be
seen in any Richardson plot (for example, Fig. 1), the length of the
perimeter increases as the yardstick used to measure it decreases. In
case of a ame front contour, the perimeter ratio measured will
depend on the yardstick (caliper length, box length, or circle
0 2 4 6 8 10 12 14 16 18
2.00
2.05
2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
F
r
a
c
t
a
l

d
i
m
e
n
s
i
o
n
,

D
Nondimensional turbulence intensity, u'/S
L
Minkowski (circle) method
Fig. 5 Variation of the fractal dimension evaluated using the
Minkowski circle method with nondimensional turbulence intensity.
0 2 4 6 8 10 12 14 16
0
2
4
6
8
10
12
S
T
/S
L
caliper method
S
T
/S
L
at <c>=0.5
N
o
n
d
i
m
e
n
s
i
o
n
a
l

t
u
r
b
u
l
e
n
t

b
u
r
n
i
n
g

v
e
l
o
c
i
t
y
,

S
T
/
S
L
Nondimensional turbulence intensity, u'/S
L
Fig. 6 The nondimensional turbulent burning velocity obtained from
fractal parameters using the caliper method is compared with
experimental S
T
=S
L
data.
0 2 4 6 8 10 12 14 16
0
2
4
6
8
10
12
S
T
/S
L
box counting method
S
T
/S
L
at <c>=0.5
N
o
n
d
i
m
e
n
s
i
o
n
a
l

t
u
r
b
u
l
e
n
t

b
u
r
n
i
n
g

v
e
l
o
c
i
t
y
,

S
T
/
S
L
Nondimensional turbulence intensity, u'/S
L
Fig. 7 The nondimensional turbulent burning velocity obtained from
fractal parameters using the box-counting method is compared with
experimental S
T
=S
L
data.
0 2 4 6 8 10 12 14 16
0
2
4
6
8
10
12
S
T
/S
L
Minkowski (circle) method
S
T
/S
L
at <c>=0.5
N
o
n
d
i
m
e
n
s
i
o
n
a
l

t
u
r
b
u
l
e
n
t

b
u
r
n
i
n
g

v
e
l
o
c
i
t
y
,

S
T
/
S
L
Nondimensional turbulence intensity, u'/S
L
Fig. 8 The nondimensional turbulent burning velocity obtained from
fractal parameters using the Minkowski circle method is compared with
experimental S
T
=S
L
data.
2788 CINTOSUN, SMALLWOOD, AND GLDER
diameter) size and the spatial resolution of the contour image. For
example, the same ame image captured by two CCD cameras with
different pixel resolutions may give two different perimeter ratios.
Therefore, the perimeter ratio determined as described in [9,10] may
not be unique and can yield incorrect ame surface area information.
V. Conclusions
We analyzed premixed turbulent ame front images from an
extensive experimental study using three different fractal algorithms
and determined the fractal parameters: namely, fractal dimension D,
inner cutoff "
i
, and outer cutoff "
0
. The three fractal methods
included the caliper technique, the box-counting method, and a third
algorithm based on the Minkowski circle method. These images
were obtained using laser-induced uorescence of OH and Mie
scattering on two Bunsen-type burners with diameters of 11.2 and
22.4 mm, from ames of propaneair with equivalence ratios of 0.8
and 1. Nondimensional turbulence intensity u
/
=S
L
was from 0.9 to
15, and the Reynolds number, based on the integral length scale, was
from 40 to 467. The main conclusions of this study can be
summarized as follows:
1) The three fractal algorithms used for analysis in this work gave
signicantly different results for the same ame image set. The
sensitivity of the fractal dimension to turbulence intensity was very
weak for u
/
=S
L
6, if any, and showed a very mild dependence for
u
/
=S
L
6.
2) Nondimensional turbulent burning rates computed from the
fractal parameters using the available closure did not agree with the
experimental turbulent burning data. Agreement was only seen for
those burning rates fromfractal results obtained by the box-counting
method for u
/
=S
L
6.
3) If the premixed turbulent ame surfaces have fractal
characteristics and Eq. (1) yields a correct estimation of the ame
surface area, then the current experimental results indicate that the
ame surface area increase may not be the dominant mechanism in
increasing the turbulent burning velocity under the conditions
corresponding to the thin-reaction-zones regime. When the
nondimensional turbulence intensity u
/
=S
L
exceeds about 67, the
ame surface area increase estimated by the fractal analysis does not
explain the observed increases in the turbulent burning velocity.
4) The results obtained by the perimeter-ratio approach, a method
recently adopted to estimate the wrinkled ame surface area, might
not be unique and could be misleading.
Acknowledgments
The work reported in this paper was supported by a Collaborative
Research Opportunities (CRO) grant from the Natural Sciences and
Engineering Research Council (NSERC) of Canada.
References
[1] Fureby, C., AFractal Flame-Wrinkling Large Eddy Simulation Model
for Premixed Turbulent Combustion, Proceedings of the Combustion
Institute, Vol. 30, Pt. 1, 2005, pp. 593601.
doi:10.1016/j.proci.2004.08.068
[2] Grinstein, F. F., and Fureby, C., LESStudies of the Flowin a Swirl Gas
Combustor, Proceedings of the Combustion Institute, Vol. 30, Pt. 2,
2005, pp. 17911798.
doi:10.1016/j.proci.2004.08.082
[3] Charlette, F., Meneveau, C., and Veynante, D., A Power-Law Flame
Wrinkling Model for LES of Premixed Turbulent Combustion, Part 1:
Non-Dynamic Formulation and Initial Tests, Combustion and Flame,
Vol. 131, Nos. 12, 2002, pp. 159180.
doi:10.1016/S0010-2180(02)00400-5
[4] Charlette, F., Meneveau, C., and Veynante, D., A Power-Law Flame
Wrinkling Model for LES of Premixed Turbulent Combustion, Part 1:
Dynamic Formulation, Combustion and Flame, Vol. 131, Nos. 12,
2002, pp. 181197.
doi:10.1016/S0010-2180(02)00401-7
[5] Gouldin, F. C., An Application of Fractals to Modeling Premixed
Turbulent Flames, Combustion and Flame, Vol. 68, No. 3, 1987,
pp. 249266.
doi:10.1016/0010-2180(87)90003-4
[6] Glder, . L., Smallwood, G. J., Wong, R., Snelling, D. R., Smith, R.,
Deschamps, B. M., and Sautet, J. C., Flame Front Surface
Characteristics in Turbulent Premixed Propane/Air Combustion,
Combustion and Flame, Vol. 120, No. 4, 2000, pp. 407416.
doi:10.1016/S0010-2180(99)00099-1
[7] Yang, S. I., and Shy, S. S., Measurement of Fractal Properties of
Premixed Turbulent Flames and Their Relation to Turbulent Burning
Velocities, Chinese Journal of Mechanics, Series A, Vol. 17, No. 2,
2001, pp. 93101.
[8] Chen, Y.-C., and Mansour, M. S., Geometric Interpretation of Fractal
Parameters Measured in Turbulent Premixed Bunsen Flames,
Experimental and Thermal Fluid Science, Vol. 27, No. 4, 2003,
pp. 409416.
[9] Lee, T. W., and Lee, S. J., Direct Comparison of Turbulent Burning
Velocity and Flame Surface Properties in Turbulent Premixed Flames,
Combustion and Flame, Vol. 132, No. 3, 2003, pp. 492502.
doi:10.1016/S0010-2180(02)00495-9
[10] Filatyev, S. A., Driscoll, J. F., Carter, C. D., and Donbar, J. M.,
Measured Properties of Turbulent Premixed Flames for Model
Assessment, Including Burning Velocities, Stretch Rates, and Surface
Densities, Combustion and Flame, Vol. 141, Nos. 12, 2005, pp. 1
21.
doi:10.1016/j.combustame.2004.07.010
[11] Mandelbrot, B. B., The Fractal Geometry of Nature, W. H. Freeman,
New York, 1982, Chap. 5, pp. 28, 33.
[12] Foucher, F., and Mounaim-Rousselle, C., Fractal Approach to the
Evaluation of Burning Rates in the Vicinity of the Piston in a Spark-
Ignition Engine, Combustion and Flame, Vol. 143, No. 3, 2005,
pp. 323332.
doi:10.1016/j.combustame.2005.06.007
[13] Cintosun, E., Analysis of Premixed Turbulent Flame Front Images
Obtained by Mie Scattering and Laser-Induced Fluorescence of OH,
M.A.Sc. Thesis, Inst. for Aerospace Studies, Univ. of Toronto, Toronto,
Ontario, Canada, Jan. 2006.
[14] Flemisch, B., Fractal Dimension, Matlab Database [online
database], http://matlabdb.mathematik.uni-stuttgart.de/index.jsp [re-
trieved Aug. 2007].
[15] Zaiser, M., Grasset, F. M., Koutsos, V., and Aifantis, E. C., Self-Afne
Surface Morphology of Plastically Deformed Metals, Physical Review
Letters, Vol. 93, No. 19, 2004, p. 195507.
doi:10.1103/PhysRevLett.93.195507
[16] Mandelbrot, B. B., Self-Afne Fractals and Fractal Dimension,
Physica Scripta, Vol. 32, No. 4, 1985, pp. 257260.
doi:10.1088/0031-8949/32/4/001
[17] Mandelbrot, B. B., and Frame, M., Fractals, Encyclopedia of Physical
Science and Technology, 3rd ed., edited by R. E. Meyers, Vol. 6,
Academic Press, New York, 2001, pp. 185207.
[18] Mandelbrot, B. B., How Long is the Coast of Britain? Statistical Self-
Similarity and Fractional Dimension, Science, Vol. 156, No. 3775,
1967, pp. 636638.
doi:10.1126/science.156.3775.636
C. Kaplan
Associate Editor
0 2 4 6 8 10 12 14 16
0
2
4
6
8
10
12
14
16
18
20
22
24
Perimeter ratio
S
T
/S
L
at <c>=0.05
S
T
/S
L
at <c>=0.5
P
e
r
i
m
e
t
e
r

r
a
t
i
o

a
n
d

S
T
/
S
L
Nondimensional turbulence intensity, u'/S
L
Fig. 9 The nondimensional turbulent burning velocity obtained from
the perimeter ratio is compared with experimental S
T
=S
L
data.
CINTOSUN, SMALLWOOD, AND GLDER 2789