Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Flow, Turbulence and Combustion 70: 325–348, 2003.

325
© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

An Attempt to Realize Experimental Isotropic


Turbulence at Low Reynolds Number

MADJID BIROUK1 , BRAHIM SARH2 and ISKENDER GÖKALP2


1 Department of Mechanical and Industrial Engineering, University of Manitoba, Winnipeg,
Manitoba R3T 5V6, Canada
2 Laboratoire de Combustion et Systèmes Réactifs, Centre National de la Recherche Scientifique,
45071 Orléans Cedex 2, France

Received 6 September 2001; accepted in revised form 14 September 2003


Abstract. This paper presents an attempt to realize experimental isotropic turbulence at low
Reynolds number. For this aim an experimental apparatus, a turbulence chamber “Box”, was de-
signed and built to generate a turbulent flow field in the center of the chamber. The turbulent airflow
field was generated by eight electrical fans placed symmetrically at the eight internal corners of the
externally cubic chamber. The turbulence intensity was controlled by the fans speed. Laser Doppler
velocimeter (LDV) in single and two-point velocity measurements was used to fully characterize
the turbulent field inside the chamber. The main results indicate that the turbulence is homogeneous
and isotropic with a quasi-zero mean velocity within a spherical region of 20 mm radius from the
center of the chamber. The measured turbulent integral length scale was found to be constant and
independent of the turbulence intensity (or fans speed). Furthermore, a noticeable spectral inertial
subrange as prescribed by the Kolmogorov theory has not been observed at the range of Reynolds
number explored here, where Reλ < 100. But rather a scaling region characterized by an exponent
that is lower than the Kolmogorov value, −5/3, has been identified. Moreover, the value of this
exponent showed no defined trend, while the width of the inertial scaling region expands as the
microscale Reynolds number increases.

Key words: frequency spectrum, isotropic turbulence, LDV, Reynolds number, turbulence chamber.

1. Introduction
The simplest form of turbulent flow that can be envisioned is isotropic turbulence,
which is by definition statistically homogeneous and invariant in all three direc-
tions. The main approach has been to use a grid in a wind tunnel, which generates
nearly isotropic turbulence with an intensity that is determined by the free stream
velocity and an integral length scale determined by the spacing of the grid. We
cannot review here the vast literature on grid-generated turbulence, as it is not
the scope of the present article and, we refer the interested reader to the relevant
literature [1–17] and references cited therein.
Despite its success in certain applications, “conventional” (or static) grid-
generated turbulence has several limitations that are difficult to overcome. Firstly, it
is difficult to achieve very high Reynolds number, as the free stream velocity cannot
326 M. BIROUK ET AL.

be increased indefinitely. Secondly, achieving a high Reynolds number requires us-


ing a very large average velocity, which complicates measurements that are usually
made in the laboratory frame. Recently, in an attempt to overcome these limitations
Makita and his collaborators [16, 17] developed the so-called “active” (or dynamic)
grid-generated turbulence that is capable of producing higher Reynolds number
than that achieved usually by ordinary static grid. This technique has recently been
improved and utilized by Mydlarski and Warhaft [13, 14], where a Taylor Reynolds
number of the order of 700 was achieved [14].
Another method for generating isotropic turbulence is by using a closed vessel
stirred with fans. This method was first developed by Semenov [18], who designed
and built the first experiment in which turbulence was generated by means of fans
placed at 90◦ intervals around the periphery of a nearly spherical chamber with an
internal equivalent diameter of 260 mm. Later, similar set-ups based on Semenov’s
concept were developed by different authors [19–25]. Although these authors have
demonstrated that the turbulence field inside the chambers was homogeneous and
isotropic, with small values of the mean velocity components as compared to their
associated fluctuating values, they were only able to do so for two orthogonal
directions because their apparatuses were not symmetric along the third axis; as
only four fans, placed at 90◦ intervals around the chamber internal periphery were
used by all these authors.
In the present experimental investigation, we attempted to improve the isotropy
of turbulence generated in fan-stirred vessel. The geometric symmetry of the flow
generated by the fans inside the chamber is assured in all three directions, which is
achieved by placing eight fans in the eight corners of a cubic chamber, as described
below in Section 2. This article attempts, therefore, to contribute to the study of
the behavior of experimental “turbulence box” by providing more details on the
turbulence structure inside the vessel (or chamber) as characterized by single and
two-point LDV measurements. The next section provides a detailed description of
the experimental set-up and data analysis used in this experiment. The experimental
results and their discussion are then presented in Section 3, followed by concluding
remarks in Section 4.

2. Experimental Set-up and Data Post-Processing


2.1. E XPERIMENTAL SET- UP
This experiment was conducted in the Combustion Laboratory and Reactive
Systems (LCSR, CNRS) in Orléans, France. The turbulence chamber described
hereafter was developed to study the effects of turbulence on droplet gasification,
either by vaporization or by combustion [26, 27]. The turbulence chamber which
has a cubic shape with external dimensions of 400 × 400 × 400 mm3 is made of
10 mm thick aluminum. Internally, the corners of the cube are filled, as shown in
Figure 1, to approach a spherical volume. The chamber has been made optically
accessible via four clear acrylic resin circular windows mounted on the four lateral
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 327

Figure 1. Schematic diagram of the experimental set-up.

walls of the cubic chamber. Two windows have a diameter of 180 and 6 mm thick
and, the remaining two windows have a diameter of 90 and 10 mm thick. The
turbulence field inside the chamber is generated by eight electrical fans with five
blades each, fixed at the chamber corners. The internal and external diameters of
each fan are 60 and 110 mm, respectively. The fan speed, N, ranges between 650
and 2700 rpm. The fans are positioned inside the chamber so that the diameter of
the nearly spherical volume created in the center is around 240 mm. The chamber
is movable along two axes; X and Y , but immobile in Z direction. The maximum
displacement in each direction is 100 mm.
The turbulence chamber is also designed to perform two-point velocity mea-
surements. The two-point spatial displacement system allows the second photo-
multiplier (PM2) to move and focus on the second movable probe volume (see
Figure 1). This probe volume is formed using a 50 mm diameter spherical mirror
with a radius curvature of 300 mm, and a focal length of 150 mm. The spherical
mirror is mounted on a two-dimensional rotating system and placed on the same
axis opposite to the laser beam transmitter. The mirror and its supporting frame
(e.g., two-dimensional rotating system) are mounted on a linear positioning sup-
porting system in order to displace the second probe volume along the Z axis (e.g.,
direction of the incident laser beam). Both photo-multipliers have a focal length of
300 mm and are arranged in the forward scattering mode with an angle of 14◦ off
axis.
An argon-ion laser is used to provide the light source and the beam emerging
from the laser is split into two green-color beams with a wavelength of 514.5 nm.
One beam passes across the Bragg cell to add frequency shift for allowing the flow
328 M. BIROUK ET AL.

Table I. Dimensions of the laser probe


volumes.

r (mm) d (mm) l (mm) δ (µm)

0 0.1800 2.6860 3.8670


29.44 0.1878 3.1189 4.2796

direction and low velocities to be measured. For velocity measurements in one


point in space, the two beams, which are separated by a distance of 40 mm, pass
across a lens with a focal length of 300 mm to form a fixed probe volume. This
latter is characterized by the values reported in the first row in Table I. For two-
point velocity measurements, the initial and final dimensions of the movable probe
volume are reported in the first and second rows of Table I, respectively; where
d, l and δ are the probe volume diameter, length and fringe spacing, respectively.
These values are presented just to show the initial and final dimensions of the mov-
able probe volume as it changes with r; where r is the spatial separation distance
between the two laser probe volumes. The airflow inside the turbulence chamber is
seeded by silicon oil drops of a diameter of approximately 1 µm.

2.2. DATA POST- PROCESSING


LDV signal analysis, in this experiment, has received careful treatment/attention
in order to obtain a good signal-to-noise ratio. Prior to acquiring any experimental
data the conditions of the experiment, such as, shift frequency, flow seeding, and
the operating laser power have been subjected to a preliminary testing in order
to obtain an LDV signal with an acceptable signal to noise ratio. A digital signal
processor (TSI Model IFA 750) based on autocorrelation technique was used to
extract the velocity information. This technique used by this processor is well
explained in [29].
The acquired LDV signal which consists of the arrival time, the transient time
(or the burst length time of the signal) and the corresponding instantaneous velocity
of seeding particles crossing the probe volume is stored on the hard disk of a PC
in order to extract the information characterizing the turbulent airflow field. In the
present experiments, the acquisition data rates achieved were approximately 2.5
and 1.5 KHz for the random and coincidence modes, which were adopted for the
single and two-point velocity measurements, respectively.
The flow mean velocity components and their mean fluctuating values were
calculated for each spatial location/coordinates (x, y) along X and Y axes, within
the spherical volume of 20 mm radius from the center of the chamber. The two-
orthogonal instantaneous velocity components measured along X and Y directions
are expressed as: U = U  + u and V = V  + v  , respectively. A number of
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 329

80 blocks of 1024 data points each have been used to calculate the mean values
of the orthogonal velocity components, U  and V , and their mean fluctuating
components, u and v. These values were, respectively, obtained using the following
expressions:
N N
i=1 Ui ti Vi ti
U  = N and V  = i=1 N
,
i=1 ti i=1 ti


  N
 (Ui − U )2 ti
u = u =  i=1 N
2 and
i=1 ti

  N
 (Vi − V )2 ti
v = v =  i=1 N
2 ,
i=1 ti

where Ui , Vi , u , v  and ti are the flow instantaneous orthogonal velocity compo-


nents, their instantaneous fluctuating components, and transit time of tracer/particle
i, respectively.
Skewness (S) and flatness (F) factors were calculated using the following
expressions:
u3 u4
S= 3/2
andF = 2
.
u2 u2
The Eulerian autocorrelation coefficient for both orthogonal velocity fluctuating
components were calculated as follows:
u (t0 )u (t0 + t) v  (t0 )v  (t0 + t)
R11 (t) =  and R22 (t) =  .
u2 (t0 )u2 (t0 + t) v (t0 )v (t0 + t)
2 2

The calculation of the autocorrelation was performed using the code developed by
Doudou [30] with a class number of 2500 and block number of 80, that is 80000
data points. This calculation code is based on a non-uniform frequency sampling
method, which uses discrete time intervals. For more information on this method
the interested reader is encouraged to consult [30–34]. The autocorrelation coeffi-
cient was then calculated for a maximum time of 350 ms. Figure 7a, displayed in
the next section, shows a typical Eulerian autocorrelation coefficient.
The frequency spectrum of the two orthogonal flow velocity components u
and v  have been calculated, using a maximum block number of 80 that is 80000
data points, by splitting the long signal into smaller segments which are processed
through a standard FFT routine and ensemble averaged.
Two-point velocity measurement consists to create two distinct laser probe
volumes in space separated by an adjustable distance r. The simultaneous mea-
surement of one velocity component at two different laser probe volumes in space
330 M. BIROUK ET AL.

(j )
allows obtaining directly the lateral space correlation coefficient Rii . To obtain si-
multaneous measurements in the two probe volumes a small temporal coincidence
window is required. The lateral spatial correlation coefficients are obtained for the
two orthogonal fluctuating velocities u and v  along the Z axis (which is denoted
by superscript (3) in the two formula below; while subscripts (11) and (22) refer to
X axis and Y axis, respectively) as follows:

u (z0 )u (z0 + r) v  (z0 )v  (z0 + r)


(3)
R11 (t) =  and (3)
R22 (t) =  .
u2 (z0 )u2 (z0 + r) 2 2
v (z0 )v (z0 + r)

It is well known that the laser Doppler velocimetry signal occurs at random and
unpredictable times. For two-point velocity measurements each of the two acqui-
sition channels should present a signal within a narrow window of time, which
is the coincidence time, to be validated. In the present experiment the selected
coincidence time was 1 µs allowing a validated signal with a data rate of 1.5 kHz.

3. Results and Discussion


Preliminary velocity measurements of the flow field generated inside the turbulence
chamber were performed along the two orthogonal X axis and Y axis to check the
homogeneity and isotropy as well as the flow two orthogonal mean velocity com-
ponents. These measurements showed that the geometrical set up of the turbulence
chamber is capable of producing quasi-zero mean velocities, U  and V , only
within a spherical volume of 40 mm of diameter in the center of the chamber.
Outside of this region the two values of the mean velocity components become
gradually important compared with their mean fluctuating velocity components u
and v when moving away from the center of the chamber toward the fans. There-
fore, only results which characterize the turbulent flow field in the spherical volume
of 40 mm diameter in the center of the chamber are presented hereafter. Note that
in the present experiment we did not attempt to acquire the velocity measurement
outside of the 40 mm spherical region. Each of the two orthogonal instantaneous
velocity components U and V were measured in 17 locations/coordinates along
X and Y axes. These locations (or coordinates) are: (x, y) = (0, 0); (±5, 0);
(±10, 0); (±15, 0); (±20, 0); (0, ±5); (0, ±10); (0, ±15); (0, ±20); where all the
coordinates are in mm.
The single-point velocity measurements along the X axis and Y axis allowed the
determination of the probability density function (PDF) distributions. Figures 2 and
3 show the normalized PDF evolution for various values of the turbulence kinetic
energy q, where q = 3u2 /2 for isotropic turbulence. These figures are presented
for three typical flow locations, (x, y) = (0, 0), (x, y) = (0, +20) and (x, y) =
(+20, 0), to show that the PDF distribution behaves similarly in the center point of
the chamber as well as away from the center. In addition, Figures 2 and 3 show that
the PDF distribution is identical for each turbulence kinetic energy (or rotational
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 331

Figure 2. Typical PDF distribution for the u component in two measurements coordinates
(a) (x, y) = (0, 0) and (b) (x, y) = (0, 20) (×: q = 0.10 m2 /s2 ; : q = 0.21 m2 /s2 ; +
q = 0.36 m2 /s2 ; : q = 0.58 m2 /s2 ; ◦: q = 0.85 m2 /s2 ; : q = 1.45 m2 /s2 ).
332 M. BIROUK ET AL.

Figure 3. Typical PDF distribution for the v component in two measurements coordinates
(a) (x, y) = (0, 0) and (b) (x, y) = (20, 0) (×: q = 0.10 m2 /s2 ; : q = 0.21 m2 /s2 ; +:
q = 0.36 m2 /s2 ; : q = 0.58 m2 /s2 ; ◦: q = 0.85 m2 /s2 ; : q = 1.45 m2 /s2 ).
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 333

fan speed, N) in the center of the turbulence chamber where (x, y) = (0, 0) as well
as away from the center of the chamber at (x, y) = (20, 0) and (x, y) = (0, 20).
The PDF distributions in the whole central spherical volume of 40 mm diameter
can be found in [28]. These figures indicate clearly that the PDF distribution is
Gaussian within the whole measurement spherical volume of 40 mm diameter.
These results have been verified by calculating the skewness and flatness factors,
which have been found to be near the theoretical values for an isotropic turbulence,
i.e. 0 and 3, respectively (see Figure 4). Note that Figure 4 displays only typical
results along X axis. Similar results for both velocity components along Y axis can
be found in [28].
A summary of the flow field velocity information in the middle of the chamber,
(x, y) = (0, 0), is presented in Figure 5. This figure shows a linear increase of
both mean fluctuating velocity components, u and v, and the turbulence inten-
sity q 0.5 with the fan speed N as: u (mm/s) = v (mm/s) = 0.35N (rpm) and
q 0.5 (mm/s) = 0.426N (rpm). The mean velocities U  and V  are quasi-zero,
and are independent of the fans speed. The ratio u/v presented in Figure 5 lies
between approximately 0.95 and 1.05, which is about 5% of anisotropy of the tur-
bulent flow field. Therefore, these experimental data indicate clearly the isotropic
character of the turbulent flow field generated in the center of the turbulence cham-
ber ((x, y) = (0, 0)). In addition, the homogeneity and isotropy of the turbulent
flow field have been also verified within the spherical volume of 20 mm radius
from the center of the chamber. Figure 6 displays the ratio of the two orthogonal
mean fluctuating velocity components, u/v, along X and Y axes for various tur-
bulence intensities. This figure shows clearly that the turbulence is homogeneous
and isotropic within the entire spherical region of 20 mm radius from the center of
the chamber. However, the anisotropy of the turbulent flow tends to increase when
moving away from the center.
The integral temporal scale for u and v velocity components have been calcu-
lated, by integrating the Eulerian autocorrelation coefficient using the following
relation:
∞
= Rii (t) dt,
0

where i = 1 for the u component and i = 2 for the v component. Figure 7a


presents a typical example of the Eulerian autocorrelation coefficient for a given
fan speed. Figure 7b displays the integral time scale for both orthogonal velocity
components and shows a decreasing exponential trend with increasing turbulence
kinetic energy. This result is in accordance with the evolution of the low frequency
range of the energy spectra (energetic eddies) where the value of the spectrum near
zero frequency decreases as the flow turbulence intensity increases. The value of
this scale obtained using this approach was found to be smaller but very close to
that obtained by integrating the Eulerian autocorrelation coefficient. This slight
334 M. BIROUK ET AL.

Figure 4. Skewness and flatness factors for (a) u and (b) v fluctuating velocity components
along X axis.

difference is mainly due to the lack of sufficient spectra data points near zero
frequency. The average value of the turbulence integral time scale extracted from
the Eulerian autocorrelation coefficient was found to follow a power law that is
approximated as  ≈ 12/q 1/2 , where  and q are expressed√in ms and m2 /s2 ,
respectively. However, when plotting  versus u, where u = 2q/3, a modified
form of this expression was found as  ≈ 10/u.
Turbulence integral length scale for the two orthogonal fluctuating velocity
components u and v in the Z axis direction, which are respectively L(3) (3)
11 and L22 ,
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 335

Figure 5. Variation of the flow mean velocity components, mean fluctuating values, turbu-
lence intensity, and u/v ratio versus fan speed.

Figure 6. Evolution of u/v along X and Y axes for various turbulence intensity levels (×:
q = 0.10 m2 /s2 ; : q = 0.21 m2 /s2 ; +: q = 0.36 m2 /s2 ; : q = 0.58 m2 /s2 ; ◦:
q = 0.85 m2 /s2 ; : q = 1.45 m2 /s2 ).
336 M. BIROUK ET AL.

Figure 7. (a) Typical Eulerian autocorrelation coefficient for a given fan speed; (b) turbulence
integral time scale versus turbulent kinetic energy.

have been obtained by integrating the surface under the lateral spatial correlation
coefficient curve (see Figure 8a) using the following relation:
∞
L(3)
ii = Rii(3)(r) dr,
0

where i = 1 for the u component and i = 2 for the v component. The values of L(3) 11
and L(3)
22 , are plotted in Figure 8b. As can be seen from this figure, the values of the
turbulence integral length scales L(3) (3)
11 and L22 are constant and independent of the
turbulence kinetic energy (or fans rotational speed). Furthermore, the average value
of L(3) (3)
11 is approximately equal to that of L22 which is around 8.6 mm. Since this
scale is of g-type we can write Lg = 8.6 mm, and we will refer to it in the following
sections as L for simplicity reasons. This result confirms the dependence of the
turbulent integral length scale on the geometry of the system generating turbulence
and not on the turbulence intensity level (or the fans rotational speed).
Fansler and Groff [20] calculated the integral length scale using the following
theoretical relation L = u ×  and found a value of 25 mm for their turbulence
chamber. While, Semenov [18] used also the same expression and found that the
turbulence integral length scale increases with fan speed. Recently, the Leeds group
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 337

Figure 8. (a) Spatial correlation coefficient for the u component versus the separation distance
between the two measuring volumes for various q (×: q = 0.10 m2 /s2 ; : q = 0.21 m2 /s2 ; +:
q = 0.36 m2 /s2 ; : q = 0.58 m2 /s2 ; ◦: q = 0.85 m2 /s2 ; : q = 1.45 m2 /s2 ) and (b) integral
(3) (3)
length scales (L11 and L22 ) versus turbulent kinetic energy, q.

[25] measured the integral length scale and found a value of 20 mm, which is
reported to be independent of the fan’s speed. The present measurements shed
some more light on the relationship between the integral length scale and other
turbulence parameters and confirm that the turbulent integral scale in the case of
fans generating turbulence is not a function of the turbulence intensity or flow
velocity. In a way it is similar to the case of grid-generating turbulence where the
integral turbulent scale depends on the grid dimensions.
In the present investigation we used the measured turbulence integral length
and temporal scales to verify the theoretical expression L ≈ u × . Figure 9 shows
the variation of the ratio of the measured integral length scale over the measured
integral time scale versus the turbulence kinetic energy. The evolution of this ratio

is found to follow a power law that is expressed as L/ = 0.70 q with , L
and q are expressed in s, m and m2 /s2 , respectively. This relationship suggests

that L ≈ 0.70 ×  × q. If we use the mean fluctuating value u instead of the
turbulence kinetic energy q, this empirical expression becomes L/ = 0.80 × u,
which suggests that L ≈ 0.80 ×  × u, with the proportionality coefficient equal
to 0.80 which is around the theoretical value 1. It is worth mentioning here that
338 M. BIROUK ET AL.

Figure 9. Ratio of turbulence integral length scale over integral time scale, L/, versus
turbulent kinetic energy, q.

in order to confirm and rationalize this empirical relationship other experiments


allowing the variation of the turbulence integral length scale as well as producing
a fairly wider range of turbulence intensities are required.
The ratio of turbulence integral scale to Kolmogorov scale, L/η, is plotted in
Figure 10. Kolmogorov scale is estimated as [8] η = (ν 3 /ε)1/4 , where ν is the
air kinematic viscosity taken at room temperature and ε is the turbulence energy
dissipation rate estimated using the following expresion [8]: ε = A × q 3/2 L. In
this relation, the proportionality coefficient A is assumed to be 1 and q and L
are the measured turbulence kinetic energy and turbulence integral length scale,
respectively. Figure 10 shows a linear relationship between L/η and ReL . How-
ever, this relation differs slightly from the theoretical expression reported in [9]:
L/η = (ReL )3/4, by the fact that the proportionality coefficient is around 1.25 in
the experimental relation instead of 1 in the theoretical one [9]. Figure 10 displays
also the order of magnitude of Reλ estimated using the following relation [8]:
Reλ = (15ReL /A)1/2 , where A is assumed to be 1. The values of Reλ with other
turbulence characteristics are summarized in Table II.
The frequency spectra for the two orthogonal velocity components u and v
have been calculated for all the aforementioned flow field measurement locations.
In this article we plotted the normalized spectra versus the normalized frequency.
Figures 11a and 12a show the u spectrum non-dimensionalized by the turbulence
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 339

Figure 10. Ratio of turbulence integral length scale over Kolmogorov scale, L/η and Reλ
versus ReL .

Table II. Reλ and L/η for various turbulence kinetic energy
levels.

q (m2 /s2 ) 0.10 0.21 0.36 0.58 0.85 1.45

Reλ 45 56 64 74 80 92
L/η 47 67 89 106 127 157

integral length scale for three different turbulence conditions at two typical flow
measurement locations. These two figures show that the spectra collapse well ex-
cept for the highest frequency region where the scaling is no longer valid. These
results show similar trend to that reported by Gamard and George [35]. Figures 11b
and 12b show the u spectrum non-dimensionalized by the Kolmogorov length scale
for the same turbulence conditions and flow measurement locations displayed in
Figures 11a and 12a. These figures (11b and 12b) show that when scaling by
the Kolmogorov scale the spectra collapse reasonably well except for the lowest
frequency region. Although there is a shortage of data, which forms the plateau at
the lowest frequency region of the spectra, the trend shown in Figures 11b and
12b is reasonably in good agreement with those reported in [13, 35–37]. Fig-
ures 11 and 12 show also that an intermediate scaling frequency region may be
340 M. BIROUK ET AL.

Figure 11. Non-dimensionalized frequency spectrum for the u component, for three typical
turbulence levels, at flow location (x, y) = (0, 0).
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 341

Figure 12. Non-dimensionalized frequency spectrum for the u component, for three typical
turbulence levels, at flow location (x, y) = (−20, 0).
342 M. BIROUK ET AL.

Figure 13. Compensated spectra by the Kolmogorov scaling-law exponent for (a) u compo-
nent and (b) v component, for the flow location (x, y) = (0, 0).
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 343

observed. This scaling region appears to extend with Reynolds number. Recall that
the Kolmogorov theory predicts the existence of such a region, inertial subrange,
at infinite Reynolds number. It has been confirmed experimentally and numeri-
cally that the existence and expansion of the inertial subrange (or Kolmogorov
scaling-law E(n) ∝ n−5/3 ) depend on the microscale Reynolds number Reλ (where
Reλ = uλ/ν), see for example [8, 35–43], and references cited therein. To verify
the existence of an inertial subrange as prescribed by Kolmogorov theory we plot-
ted in Figure 13 typical compensated frequency spectra n5/3 Eu (n) and n5/3Ev (n)
versus the frequency, n, for the flow measurement location (x, y) = (0, 0). In
this figure, the inertial subrange region, if there is any, should appear as a hor-
izontal flat region. Figures 13a and 13b confirm that no remarkable flat region
may be observed; however, a scaling region characterized by a positive slop can
be identified. We plotted in Figures 14a and 14b the same frequency spectra, for
u and v velocity components, but compensated by a modified form of the Kol-
mogorov scaling-law exponent as proposed by Mydlarski and Warhaft [13], e.g.,
n5/3−p Eu (n) and n5/3−p Ev (n). The value of the exponent p that would make the
scaling region horizontally flat was found by trail and error method. These figures
show clearly an inertial scaling region that expands as Reλ increases. However,
the trend or transition of the modified exponent −5/3 + p seems to be not well
defined at the range of Reynolds number explored in this experiment. This is
mainly due to the width of the scaling region which is not sufficiently large to
yield an accurate value of p particularly at the lowest Reynolds number Reλ ∼ 45.
We anticipate that the value of p would decrease with Reλ for relatively higher
Reynolds number. Mydlarski and Warhaft [13] also found similar behavior of this
exponent at Reλ < 100; however, they found it to decrease with Reynolds number
for Reλ > 100. Our average values of p, which is denoted as n in [13] and µ in
[35], were found to be comparable to those reported in [35], but slightly smaller
than those reported in [13]. Similar results to those presented in Figures 13 and 14
have been obtained for the other flow measurement locations within the spherical
volume of 40 mm of diameter in the center of the chamber, and Figure 15 displays
a typical example.
Although a Reλ ∼ 700 was reached in [14], the inertial region scaling-law ex-
ponent was found to be below the Kolmogorov value, which is 5/3. The authors of
this article predicted that the 5/3 scaling region will not occur until Reλ ∼ 10,000.
This prediction seems to be plausible as Sreenivasan and Dhruva [38] who studied
atmospheric isotropic turbulence with Reλ ranging between 10,000 and 20,000
and found that in fact spectral scaling region exists and expands over three spec-
tral decades. While Saddoughi and Veeravalli [36], who studied local isotropy in
turbulent boundary layers, reported that a microscale Reynolds number of more
than 1,000 is needed to achieve a convincing locally isotropic inertial subrange.
An improved version of the present apparatus is under development to produce
higher Reynolds number than those achieved in the present experiment, which will
allow us to further investigate the dependence of p on Reλ . Clearly the spectra
344 M. BIROUK ET AL.

Figure 14. Compensated spectra by the modified form of the Kolmogorov scaling-law
exponent for (a) u component and (b) v component, for the flow location (x, y) = (0, 0).
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 345

Figure 15. Compensated spectra by (a) the Kolmogorov scaling-law exponent and
(b) the modified form of the Kolmogorov scaling-law exponent for u component, at
(x, y) = (0, −20).
346 M. BIROUK ET AL.

plots displayed in Figures 14 and 15b show consistency with recent findings of
Mydlarski and Warhaft [13] that the power-law exponent in the inertial subrange,
for low Reynolds number range, is not the Kolmogorov theoretical value −5/3, but
rather a modified exponent form that is a function of Reλ . In addition, the results
of the present investigation agree with the conclusions made in [13, 35–38] that
the Kolmogorov theory is indeed an accurate description of high Reynolds number
turbulence.

4. Concluding Remarks
The major concluding remarks of this study are as follows. The turbulence chamber
developed in this investigation proves to be capable of generating an isotropic and
spatially homogeneous turbulence with quasi-zero mean velocity within a spherical
region of 20 mm radius from the center of the chamber even though the Reynolds
number achieved in this experiment is considered to be small, Reλ < 100. The
turbulence integral length scale has been found to be constant and independent of
the turbulence intensity level. The present experiment suggests that different values
of this scale may be achieved by changing the dimensions of the fan.
An inertial subrange, as prescribed by the Kolmogorov similarity theory, has not
been observed at the Reynolds number range explored in this experiment, Reλ <
100. But rather a noticeable scaling region, characterized by an exponent lower
than the Kolmogorov theoretical value, 1.67, has been shown. In addition, the value
of the modified scaling-law exponent −5/3 + p has shown no defined trend for
Reλ < 100; however, the width of the corresponding scaling region expands as
Reλ increases. This suggests that a convincing inertial subrange would appear only
if the microscale Reynolds number is sufficiently high. An improved version of this
turbulence chamber, which will allow generating of higher Reynolds number than
that obtained in this experiment, is currently under development at the University
of Manitoba. The upgraded turbulence chamber will allow investigating further the
dependence of the Kolmogorov inertial scaling law on Reynolds number.
The velocity component w (along the beam direction, Z axis) was not mea-
sured in the present work because the chamber was not designed to move along Z
direction. However, the symmetric position of the eight fans inside the turbulence
chamber and the equality of the two other orthogonal fluctuating components u and
v ensures that w would have similar value to u and v. This will be verified once the
new version of this turbulence chamber is operational.

Acknowledgments
The authors wish to thank Dr. Michel Trinité (Director of Research at CNRS-
CORIA, Rouen, France) for his contribution to two-point LDV measurements.
The first author would like also to thank Ram S. Azad, Professor Emeritus of
Mechanical Engineering, University of Manitoba, Canada, for helpful discussions.
AN EXPERIMENTAL ISOTROPIC TURBULENCE AT LOW REYNOLDS NUMBER 347

References
1. Taylor, G.I., Statistical theory of turbulence. Proceedings of the Royal Society A 151 (1935)
421–478.
2. Bachelor, G.K. and Townsend, A.A., Decay of vorticity in isotropic turbulence. Proceedings of
the Royal Society A 190 (1947) 534–550.
3. Batchelor, G.K. and Townsend, A.A., Decay of isotropic turbulence in the initial period.
Proceedings of the Royal Society A 193 (1948) 539–558.
4. Corrsin, S., Turbulence: Experimental methods. Encyclopedia of Physics 8 (1963) 568–590.
5. Comte-Bellot, G. and Corrsin, S., The use of a contraction to improve the isotropy of grid
generated turbulence. Journal of Fluid Mechanics 25 (1966) 657–682.
6. Comte-Bellot, G. and Corrsin, S., Simple Eulerian time correlation of full-and narrow-band
velocity signals in grid generated, isotropic turbulence. Journal of Fluid Mechanics 48 (1971)
273–337.
7. Kistler, A.L. and Vrebalovich, T., Grid turbulence at large Reynolds numbers. Journal of Fluid
Mechanics 6 (1966) 37–47.
8. Hinze, J.O., Turbulence, second edition. McGraw-Hill, New York (1975).
9. Tennekes, H. and Lumely, J.L., A First Course in Turbulence. MIT Press, Cambridge, MA
(1973).
10. Comte-Bellot, G., Fluid Mechanics Notes, Ecole Central de Lyon, France (1983).
11. Roach, The generation of nearly isotropic turbulence by means of grids. International Journal
of Heat Fluid Flow 8 (1986) 82–92.
12. Mohamed, M.S. and LaRue, J.C., The decay power law in grid-generated turbulence. Journal
of Fluid Mechanics 219 (1990) 195–214.
13. Mydlarski, L. and Warhaft, Z., On the onset of high-Reynolds-number grid-generated wind
tunnel turbulence. Journal of Fluid Mechanics 320 (1996) 331–368.
14. Mydlarski, L. and Warhaft, Z., Passive scalar statistics in high-Peclet-number grid turbulence.
Journal of Fluid Mechanics 358 (1998) 135–175.
15. Makita, H. and Sassa, K., Active Turbulence Generation in a Laboratory Wind Tunnel.
Advances in Turbulence, Springer-Verlag, Berlin (1991).
16. Makita, H., Realization of a large-scale turbulence field in a small wind tunnel. Fluid Dynamics
Research 8 (1991) 53–64.
17. Wang, H. and George, K.W., The integral scale in homogeneous isotropic turbulence. Journal
of Fluid Mechanics 459 (2002) 429–443.
18. Semenov, E.S., Measurements of turbulence characteristics in a closed volume with artificial
turbulence. Combustion, Explosion, and Shock Waves 1 (1965) 57–62.
19. Ohta, Y., Shimoyama, K. and Ohigashi, S., Vaporization and combustion of single liquid fuel
droplets in a turbulent environment. Bulletin of the JSME 18 (1975) 47–56.
20. Fansler, T.D. and Groff, E.G., Turbulence characteristics of a fan-stirred combustion vessel.
Combustion and Flame 80 (1990) 350–354.
21. Kwon, S., Wu, M.S., Driscoll, J.F. and Faeth, G.M., Flame surface properties of premixed
flames in isotropic turbulence: measurements and numerical simulation. Combustion and
Flame 88 (1992) 221–238.
22. Andrews, G.E., Bradley, D. and Lwakabamba, S.B., Turbulence and turbulent flame propaga-
tion – A critical appraisal. Combustion and Flame 24 (1975) 285–304.
23. Abdel-Gayed, R.G. and Bradley, D., Dependence of turbulent burning velocity on turbu-
lent Reynolds number and ratio of laminar burning velocity to RMS turbulent velocity. The
Proceedings of the Combustion Institute 16 (1976) 1725–1735.
24. Atzler, F. and Lawes, M., Burning velocities in droplet suspensions. ILASS-Europe,
Manchester, July 6–8, 1998.
348 M. BIROUK ET AL.

25. Gillespie, L., Lawes, M., Sheppard, C.G.W. and Woolley, R., Aspects of laminar and turbulent
burning velocity relevant to SI engines. SAE Technical Paper 2000-01-0192 (2000).
26. Birouk, M. and Gökalp, I., A new correlation for turbulent mass transfer from liquid droplets.
International Journal of Heat and Mass Transfer 45 (2002) 37–45.
27. Birouk, M., Chauveau, C. and Gökalp, I., Turbulence effects on the combustion of single
hydrocarbon droplets. The Proceedings of the Combustion Institute 28 (2000) 1015–1021.
28. Birouk, M., Turbulence effects on the vaporization and burning of n-alkane hydrocarbon
droplets. Ph.D. Thesis, University of Orléans, France (1996).
29. Menon, R., Jenson, L. and Buddhavarapu, J., Comparaison of signal extraction techniques in
LDV signal processing. Laser Velocimetry Advances and Applications. SPIE 2052 (1993) 35–
42.
30. Doudou, A., Ph.D. Thesis, University of Rouen, France (1990).
31. Jones, R.H., Aliasing with unequally spaced observations. Journal of Applied Metrology 11
(1972) 245–254.
32. Mayo, W.T., Spectrum measurements with laser velocimeters. In: Proceedings of Dynamic
Flow Conference, Baltimore, MD (1978).
33. Mayo, W.T., Shay, M.T. and Ritter, S., The development of new digital data processing tech-
niques for turbulence measurements with a laser velocimeter. Final Report AEDC-TR-74-53
(1974).
34. Gaster, M. and Roberts, J.B., Spectrum analysis of randomly sampled signals. J. Inst. Math.
Applies 15 (1975) 195–216.
35. Gamard, S. and George, K.W., Reynolds number dependence of energy spectra in the overlap
region of isotropic turbulence. Flow, Turbulence and Combustion 63 (1999) 443–477.
36. Saddoughi, S.G. and Veeravalli, S.V., Local isotropy in turbulent boundary layers at high
Reynolds number. Journal of Fluid Mechanics 268 (1994) 333–372.
37. Saddoughi, S.G., Local isotropy in complex turbulent boundary layers at high Reynolds
number. Journal of Fluid Mechanics 348 (1997) 201–245.
38. Sreenivasan, K.R. and Dhruva, B., Is there scaling in high-Reynolds-number turbulence.
Progress of Theoretical Physics, Suppl. 130 (1998) 103–120.
39. Batchelor, G.K., The Theory of Homogeneous Turbulence. Cambridge University Press,
Cambridge, (1956).
40. Betchov, R., An inequality concerning the production of vorticity in isotropic turbulence.
Journal of Fluid Mechanics 1 (1957) 497–506.
41. Dumas, R., Contribution a l’étude des spectres de turbulence. Ph.D. Thesis, Université d’Aix-
Marseille, France (1962).
42. Szablewski, W., One-dimensional spectra of the turbulence energy of homogeneous isotropic
turbulence. Zeitschrift für angewandte Mathematik und Mechanik (ZAMM) 66(12) (1985) 585–
594.
43. Ruetsh, G.R. and Maxy, M.R., The evolution of small-scale features in homogeneous isotropic
turbulence. Physics of Fluids 4 (1992) 2747–2760.

You might also like