This action might not be possible to undo. Are you sure you want to continue?

# Experimental investigation of the grid-generated turbulence features

**in a free surface ﬂow
**

Fre´de´ric Murzyn

a,

*

, Michel Be´lorgey

b

a

School of Civil Engineering and The Environment, Faraday building, Highﬁeld, University of Southampton, SO17 1BJ Southampton, UK

b

Morphodynamique Continentale et Coˆ tie` re, UMR 6143, CNRS, Groupe Me´ canique des Fluides, 24 rue des tilleuls, 14000 Caen, France

Received 13 September 2004; received in revised form 2 December 2004; accepted 17 December 2004

Abstract

The present study aims to investigate the features of a grid-generated turbulence occurring in a current ﬂow with a free surface

ﬂow. The interest is focused on the length and time scales of the turbulence. These are the macro, the micro and the Kolmogorov

scales. To analyze the ﬂow, a 2D LDV system has been used to measure U, W , u

0

and w

0

. This non-intrusive and optical technique is

really accurate (in terms of space and time resolution). Furthermore, it does not disturb the ﬂow and provides a high data rate. Both

horizontal and vertical velocities are recorded at the same time according to a coincidence window (s

cw

). Bias measurements are

avoided by using a ﬁltering technique during data processing. The improved homogeneity and isotropy of the turbulence down-

stream of the grid allows the use of the Taylor hypothesis. Thus, all length and time scales of the ﬂow can be estimated. Results

are discussed as well as the inﬂuence of the upcoming mean velocity ðUÞ on the turbulence properties.

Ó 2005 Elsevier Inc. All rights reserved.

Keywords: LDV; Grid-generated turbulence; Isotropy; Turbulence length/time scales; Current; Taylor hypothesis

1. Introduction

In terms of coastal engineering, all studies are inter-

esting because they can bring solutions to diﬀerent envi-

ronmental problems such as pollutant dispersion, air/

sea exchanges, beach planning and so on. Lots of prob-

lems particularly occur closed to the coasts. Building a

breakwater with respect to the shoreline properties

[29] or a bridge between a continent and an island,

designing an artiﬁcial reef for surfers and suitable for

ﬁshes [17], devising a pipeline for gas transportation,

understanding the problem of pollutant dispersion in

an estuary (to avoid its contamination by chemical

materials) are examples of research topics where coastal

engineers have strong interests. Before surﬁng the waves

over this artiﬁcial reef or protecting a bay against wave

action and storms, researchers have to deﬁne the best

shape of these structures taking into account the in-situ

conditions (water depth, wave height/period, current

speed. . .), their behavior in extreme conditions (storms,

risks of damages). . . Laboratory experiments, numerical

modelling and/or in situ measurements are three diﬀer-

ent ways to proceed to get as many information as pos-

sible (eﬃciency of the structure, costs, key parameters

inﬂuencing the ﬂow around these structures. . .). Among

that, sediment transport is one of the most important

points they have to take care as it is responsible for ero-

sion or accretion. For instance, scour erosion around a

cylinder can lead to the destabilization of the supports,

then to instabilities of the pipe leading to its destruc-

tion. As a consequence, the spreading of the gas or

0894-1777/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.expthermﬂusci.2005.02.002

*

Corresponding author. Tel.: +44 23 8059 4656; fax: +44 23 8067

7519.

E-mail addresses: fpjf@soton.ac.uk (F. Murzyn), belorgey@meca.

unicaen.fr (M. Be´lorgey).

www.elsevier.com/locate/etfs

Experimental Thermal and Fluid Science 29 (2005) 925–935

the petrol into the oceans then means pollution of the

environment.

The turbulence plays an important role in these sedi-

ment transport problems [4,23,24]. Everywhere in the

ﬂow (near the bottom, inside the water column or at

the air/water interface), it has a strong inﬂuence. In

the lower part of the ﬂow, burst and sweep phenomena

are responsible for highly turbulent events and particle

entrainment [15]. Above this area, turbulent structures

aﬀect particle path and modify their settling velocity

[7,13]. In terms of environmental consequences, this

has a real impact as it has been explained by Jing and

Ridd [18] for the particular case of the Cleveland bay

in Australia for coral reefs.

In the last two decades, many studies have been

undertaken to understand the interaction between tur-

bulence and sediment transport [10,12,23–25]. Results

mainly show that the length and time scales of the turbu-

lence are one of the most important parameters [4].

In the present paper, we aim to investigate this point

by describing the behavior of these characteristics scales.

Experiments have been carried out in a ﬂume (current

ﬂow), downstream of a square mesh grid. This grid-gen-

erated turbulence is homogeneous and isotropic which

makes the study easier. Indeed, the turbulence proper-

ties may be estimated by using Taylor hypothesis. On

the one hand, the role of the current velocity is analyzed.

On the other hand, these results bring information on

how turbulence develops behind an obstacle in rivers

ﬂows.

2. Experimental set-up, data acquisition and analysis

2.1. Experimental set up

The experiments were conducted in the wave/current

ﬂume of the ‘‘Morphodynamique Continentale et

Coˆ tie`re’’ laboratory at the University of Caen/Basse

Normandie. It is 16.5 m in length, 0.5 m in width and

0.7 m in height (L

c

= 16.5 m, l

c

= 0.5 m and

h

c

= 0.7 m). Side walls are made of glass. Thus, optical

measurement systems such as LDV can be used for ﬂow

visualizations and velocity measurements. The bottom

Nomenclature

d water depth

d

b

bar width of the grid

h

c

channel height

I

x,z

longitudinal and vertical turbulence intensi-

ties

L

c

channel length

l

c

channel width

L

f

macro length scale of turbulence

M grid mesh size

t time

T duration of acquisition

TT transit time

T

t

macro time scale of turbulence

Re

M

, Re

k

g

Reynolds numbers

~x, ~y, ~z spatial directions

U, W instantaneous horizontal and vertical veloci-

ties

U (or U

mean

), W (or W

mean

) mean horizontal and ver-

tical velocities

U

max

maximum horizontal velocity

U

max,p

maximum horizontal velocity (vertical pro-

ﬁle)

u

0

, w

0

horizontal and vertical turbulent ﬂuctuations

of the velocities

Symbols

e dissipation rate

j turbulent kinetic energy

g

K

, s

K

, v

K

Kolmogorov scales

k

f

micro length scale of turbulence

r Solidity coeﬃcient

s

t

micro time scale of turbulence

s

cw

coincidence window

t kinematic viscosity

Acronym

LDV Laser Doppler Velocimetry

Fig. 1. The wave/current ﬂume at the University of Caen/Basse

Normandie.

926 F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935

of the channel is made of PVC. Fig. 1 shows this appa-

ratus. Underneath the channel, a PVC ‘‘high-pressure’’

tube (0.315 m of diameter) ensures the recirculation of

the water (constant volume).

The current is generated by a pump. The ﬂow rate is

accurately deﬁned by controlling the rotation frequency

of the pump. Experiments can then be made at diﬀerent

times in the same conditions. The mean horizontal

velocity ðUÞ can be up to 2 m/s depending on water

depth (d). For the present study, the ﬂow depth (d) is

constant and equal to 0.35 m (d = 0.35 m).

Turbulence can be generated by diﬀerent means. For

instance, a cylinder may be installed across the ﬂow or

the bottom roughness can be increased. Here, the turbu-

lence is generated by a square mesh grid placed in a

cross section of the ﬂow. According to Comte-Bellot

and Corrsin [9], the mesh size (M) and the grid bar

thickness (d

b

) are chosen to get a solidity coeﬃcient

(1) of 0.33. Fig. 2 shows the grid. This grid is made of

15 meshes in width and 18 in height. The rigidity of

the grid is ensured by a frame made of duralumin.

r ¼

d

b

M

2 À

d

b

M

_ _

¼ 0:33 ð1Þ

In our experimental set up (recirculating channel),

a honeycomb (0.2 m in thickness, square cells of

0.004 m · 0.004 m) is located upstream of the grid to re-

duce the level of the turbulence intensity before the grid

(less than 4%). It breaks the possible vortex generated at

the entrance of the ﬂume [21].

A calibration of the ﬂume has been made before these

experiments to investigate the ﬂow properties with and

without the grid. In both case, closed to the side walls

and above the bottom, the boundary layer thickness

has been estimated to less than 0.08 m. Further details

on this calibration are presented by Murzyn [21]. From

that, the best emplacement for the grid has been deﬁned

2.3 m downstream of the honeycomb.

Based on U and M, the Reynolds number character-

izing the ﬂow is (2):

Re

M

¼

UM

t

ð2Þ

Three diﬀerent mean horizontal velocities have been

studied (U ¼ 0:225 m/s, 0.25 m/s and 0.325 m/s) corre-

sponding to Re

M

of 7425, 8250 and 10,725.

~x is to the main ﬂow direction (x = 0 at the grid posi-

tion) and~z is the vertical direction (z = 0 on the bottom).

U and W are respectively the horizontal and the vertical

velocity components. U is positive in the ﬂow direction

and W is positive upward.

2.2. Data acquisition

Data are acquired through the BSAFlow software of

DANTEC. Several parameters are recorded such as the

transit time (TT), the arrival time (AT) and the instanta-

neous velocities (U and W).

For turbulent ﬂows, data acquisition must be made

very carefully. The data collection duration (T) and

the number of acquired samples (N) are important. To

avoid any bias on velocity measurements, acquisition

must be done according to two criteria:

1. T must be long enough compared to the time scales of

the ﬂow (this point will be discussed farther);

2. N must be large enough to escape any problems dur-

ing data processing. According to Belmabrouck [2]

and Belmabrouck and Michard [3], N = 2000 seems

to be a minimum.

Data rate must be high enough as well in order to

follow the ﬂuctuations of the ﬂow. Particle seeding

may improve it. Finally, we decided to choose T = 50 s

and N = 5000 (when one of these conditions is reached,

data collection stops). The data rate was always higher

than 50 Hz.

U and W are simultaneously recorded with respect to

a coincidence window (s

cw

). s

cw

depends on the mean

TT of a particle (moving at U) across the volume of

measurement. It can reach 10 times this mean TT [14].

s

cw

has been ﬁxed at 0.8 ms.

Fig. 3 presents a partial record of U and W as a func-

tion of time (Re

M

= 8250, x/M = 20). More than 1000

samples are shown here.

2.3. Data analysis

2.3.1. Determination of mean and turbulent velocities

For a current ﬂow, U is constant over T and is then

deﬁned from an averaged of the whole set of data (3)

whereas velocity ﬂuctuations are calculated following

(4):

U ¼

1

N

N

i¼1

U

i

ð3Þ

Fig. 2. The grid (M = 0.033 m; d

b

= 0.006 m).

F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935 927

u

0

¼

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

1

N

N

i¼1

ðU

i

ÀUÞ

2

_ _

¸

¸

¸

_

ð4Þ

During the measurements, few particles which do not

follow the ﬂow might be recorded. Nevertheless, they do

not modify U [5,6,11] if they do not represent more than

5% of N. No ﬁltering technique is then needed to get U

and W .

This is diﬀerent for u

0

and w

0

[11]. At this stage, ‘‘exo-

tic’’ data must be removed to avoid any bias in u

0

and

w

0

. We have decided that all samples separated from

the mean value for more than 3 standard deviations

are deleted. Less than 1% of the data are removed. This

has no eﬀect on data interpretation as expected by Srik-

antaiah and Coleman [27].

2.3.2. Taylor hypothesis

Theoretically, the turbulence developing downstream

of a grid is homogeneous and isotropic [9]. Time and

space properties of the ﬂow are related to each others

according to U (5). This is the Taylor hypothesis, also

called ‘‘frozen turbulence hypothesis’’.

o

ot

¼ ÀU

o

ox

ð5Þ

Some others conditions must be satisﬁed for a grid-

generated turbulence. Particularly, the level of turbu-

lence needs to be lower than 20% and U must always

be the same along ~x. The most important consequence

is that the dissipation rate (e) can be deduced from the

spatial evolution of the turbulent kinetic energy (j).

2.3.3. Length and time scales of turbulence

There are diﬀerent scales of turbulence. The macro

scales correspond to the biggest energetic structures,

the micro scales are deﬁned as the smallest scales which

do not dissipate energy. Lastly, the Kolmogorov scales

are the smallest dissipative scales of the ﬂow. Here, we

are interested in all of them. The following relationships

show how they can be deduced from u

0

, e and m:

L

f

¼

u

03

e

ðaÞ T

t

¼

L

f

u

0

ðbÞ ð6Þ

k

f

¼

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

30t

u

2

e

¸

ðaÞ s

t

¼

1

15

k

f

u

0

ðbÞ ð7Þ

g

K

¼

t

3

e

_ _

1=4

ðaÞ s

K

¼

t

e

_ _

1=2

ðbÞ

v

K

¼ ðteÞ

1=4

ðcÞ ð8Þ

Based on the estimation of the j and e, this ‘‘energetic

technique’’ is well-adapted to many ﬂows and has

been successfully applied by Comte-Bellot and Corrsin

[9], Wei and Willmarth [30] or Benedict and Gould

[5,6].

3. Experimental results

A detailed preliminary analysis of the non-disturbed

ﬂow is presented by Murzyn [21]. It particularly shows

that the mean turbulence level is below 4% downstream

of the honeycomb.

The ﬁrst part of the results concerns evolutions of U,

W , u

0

and w

0

downstream of the grid. We only show re-

sults for Re

M

= 8250 but same trends are obtained for

Re

M

= 7425 and 10,725. Longitudinal and transversal

proﬁles are performed at the centerline of the channel

(z/d = 0.5).

3.1. Mean and turbulent velocity proﬁles

Fig. 4 presents both evolutions of U and W as a func-

tion of the distance to the grid x/M.

These plots show that U and W do not strongly

change when x/M increases. When the ﬂow passes

through the grid, it is accelerated. Nevertheless, it rap-

idly recovers a constant level. Its direction is not aﬀected

by the grid. W =U

max

does not exceed 5%. This is low but

three times greater than it was for the non-disturbed

ﬂow. For x/M > 15, U and W remain almost constants.

The vertical proﬁles of U and W are represented on

Fig. 5 for x/M = 5, 15 and 30. The boundary condition

(z/M = 0) impose that U = W = 0. U

max,p

is the maxi-

mum horizontal velocity obtained for the corresponding

proﬁle. Thus, U=U

max;p

and W =U

max;p

are smaller or

equal to 1.

Our results indicate diﬀerent behaviors depending on

x/M. Closed to the grid (x/M = 5), U can be negative for

smallest z/M. In this region, a recirculating area takes

place due to the frame of the grid. For x/M = 5, proﬁles

0 5 10 15 20

-0.04

-0.02

0.00

0.02

0.04

0.06

0 5 10 15 20

0.20

0.22

0.24

0.26

0.28

0.30

U

t

t

W

Fig. 3. Time series of velocity components (U and W, Re

M

= 8250).

928 F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935

are not ﬂat and uniform. As x/M increases, this non-

homogeneity tends to disappear. From x/M = 15, all

proﬁles become ﬂat revealing a homogeneous ﬂow.

Fig. 6 describe the evolution of U and W along the

transverse axis ð~yÞ at x/M = 20, 40 and 60. U

axis

is the

mean velocity at on the centreline of the channel.

A symmetric behavior is displayed around y/M = 0

independent of x/M depicting a homogeneous ﬂow on

the transversal axis.

At this stage, we can conclude that the mean ﬂow

downstream of the grid is homogeneous for 15 < x/

M < 60, À3 < y/M < 3 and 3 < z/M < 8. This deﬁnes

the new region of the future investigations. The aim of

the following analysis is the characterization of the

homogeneity of the turbulent ﬁeld downstream of the

grid.

Fig. 7 deal with the evolutions of horizontal (I

x

) and

vertical (I

z

) turbulent intensities. These ﬁgures reveal

that both I

x

and I

z

rapidly decrease in the closest region

downstream of the grid (0 < x/M < 15). After that, it

nearly remains constant. For x/M > 15, turbulence

intensities are 25% higher than those measured for the

non-disturbed ﬂow [21].

Results also show that I

x,z

are lower than 20%. This is

in good agreement with Cadiergue [7] and Michelet et al.

[20]. These authors found I

x

= 3% at x/M = 30.

0 10 20 30 40 50 60 70

0.7

0.8

0.9

1.0

1.1

x/M

U

mean

/U

max

0 10 20 30 40 50 60 70

0.00

0.02

0.04

0.06

x/M

W

mean

/U

max

Fig. 4. Longitudinal proﬁles of U and W downstream of the grid (Re

M

= 8250).

0

2

4

6

8

10

-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2

x/M=5

x/M=30 x/M=50

z/M

U

mean

/U

max.p

0

2

4

6

8

10

-0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12

x/M=5 x/M=15

x/M=30 x/M=50

z/M

W

mean

/U

max. p

x/M=15

Fig. 5. Vertical proﬁles of U and W at diﬀerent positions x/M (Re

M

= 8250).

-4 -2 0 2 4

0.0

0.2

0.4

0.6

0.8

1.0

1.2

x/M=20

x/M=40

x/M=60

y/M

U

mean

/U

axis

-4 -2 0 2 4

0.00

0.02

0.04

0.06

0.08

0.10

x/M=20

x/M=40

x/M=60

y/M

W

mean

/U

axis

Fig. 6. Transversal proﬁles of U and W at diﬀerent positions x/M (Re

M

= 8250).

F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935 929

Cenedese et al. [8] gave 0.07 < I

x

< 20. For x/M > 100,

the viscous dissipation becomes more important.

Fig. 8 describe the evolution of I

x

and I

z

from the

bottom (z/M = 0) to the free surface for three sections

downstream of the grid. On the bottom, I

x

= I

z

= 0

(boundary condition).

The bottom has a strong inﬂuence on the shape of I

x

and I

z

. For z/M < 3, they reach 20% and 18% respec-

tively. This is caused by the frame of the grid which cre-

ates a strong turbulent area near z/M = 0. The free

surface has also an inﬂuence on I

x

and I

z

. For z/M > 9

and x/M = 15, I

x

and I

z

take values of 13% and 10%

meaning that the inﬂuence of the air/water interface is

less important than the bottom one on the turbulent

intensities.

Lastly, Fig. 9 represents the behaviors of I

x

and I

y

along ~y for diﬀerent positions downstream of the grid.

These results indicate that turbulence intensities exhi-

bit some diﬀerences around the centerline. For y/M < 0,

lower values are found. Nevertheless, the variations are

quite small round y/M = 0. A default on the parallelism

of the walls is one of the possible reasons to explain that

point. One other reason could be the persistence of a

‘‘small mean vortex’’ in the channel.

0 10 20 30 40 50 60 70

0.0

0.1

0.2

0.3

0.4

x/M

I

x

=u'/U

moy

0 10 20 30 40 50 60 70

0.0

0.1

0.2

0.3

0.4

I

z

=w'/U

moy

x/M

Fig. 7. Turbulence intensities I

x

and I

z

downstream of the grid (Re

M

= 8250).

0

2

4

6

8

10

0.0

x/M=15

x/M=30

x/M=50

z/M

I

x

0

2

4

6

8

10

0.0 0.1 0.2 0.3

x/M=15

x/M=30

x/M=50

z/M

I

z

0.1 0.2 0.3

Fig. 8. Vertical proﬁles of I

x

and I

z

at diﬀerent positions x/M (Re

M

= 8250).

-4 -2 0 2 4

0.00

0.02

0.04

0.06

0.08

0.10

x/M=20

x/M=40

x/M=60

y/M

I

x

-4 -2 0 2 4

0.00

0.02

0.04

0.06

0.08

0.10

x/M=20

x/M=40

x/M=60

I

z

y/M

Fig. 9. Transversal proﬁles of I

x

and I

z

at diﬀerent positions x/M (Re

M

= 8250).

930 F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935

These last results (Figs. 7–9) justify that the turbulent

ﬂow is homogeneous in the domain deﬁned by 15 < x/

M < 60, À3 < y/M < 3 and 3 < z/M < 8.

Fig. 10 helps us to characterize the isotropy. It shows

the evolution of u

0

/w

0

along the horizontal axis on the

centreline of the channel.

Although the data scattering (lower on the left part),

this plot is interesting. Indeed, it shows that the mean

value of u

0

/w

0

is 1.1. This is in good agreement with

the results of Gibson and Dakos [16] obtained in a wind

tunnel. These authors also argue that a small anisotropy

(10%) is often found for a grid-generated turbulence.

Farther downstream, u

0

/w

0

takes higher values meaning

the beginning of region with a small anisotropy (x/

M > 50). Using a contraction, Comte-Bellot and Corrsin

[9] got a mean value of 1 (perfect isotropy). Unfortu-

nately, this arrangement is not possible for our experi-

ments due to the presence of the free surface.

As a conclusion, an isotropic region is deﬁned start-

ing at x/M = 15 for À3 < y/M < 3 and 3 < z/M < 8. In

this area, the Taylor hypothesis can be applied. From

now, we only consider this domain.

3.2. Turbulent kinetic energy and dissipation

The Taylor hypothesis states that turbulence scales

can be deduced from the measurements of the turbulent

kinetic energy j ¼

1

2

u

2

þv

2

þw

2

_ _ _ _

and its dissipation

e ¼

oj

ot

_ _

. This so-called ‘‘energetic method’’ has been

successfully used by Tennekes and Lumley [28], Mich-

elet [19], Michelet et al. [20] and Cadiergue [7].

Our LDV system is only 2D. Thus, the transversal

component of the velocity (V) is not available. For

homogeneous and isotropic ﬂow, we suppose that

v

2

¼ w

2

[1]. Fig. 11 presents evolutions of j and e

according to x/M.

On the one hand, these graphs tend to prove that the

decay of the turbulent kinetic energy is mainly concen-

trated closed to the grid (x/M < 15). On the other hand,

the decay of the dissipation is more evident [21]. In the

isotropic region, it follows a power law (e#(x/M)

À2

).

Turbulence length and time scales can now be esti-

mated. First of all, we are interested in the length scales.

Then, we will discuss about the time scales. Finally, the

Kolmogorov scales will be presented. Results are shown

for three diﬀerent mean horizontal velocities (U ¼

0:225 m/s, 0.25 m/s and 0.325 m/s). The corresponding

Reynolds number are Re

M

= 7425, 8250 and 10,725.

3.3. Length scales of turbulence

In Fig. 12 are plotted the evolutions of the macro

length scale of turbulence as a function of x/M. The

length scales are normalized with M.

As seen, these results show that L

f

/M grows regularly

with x/M. At x/M = 15, it is equal to 0.5. Then it in-

creases to more than 2 (at x/M = 40). Farther down-

stream, the data scattering becomes signiﬁcant. On the

one hand, this can be explained as a return of the ﬂow

0 10 20 30 40 50 60 70

0.8

1.0

1.2

1.4

u'/w'

x/M

Mean level

Fig. 10. Isotropy of the ﬂow (Re

M

= 8250).

0 10 20 30 40 50 60 70

10

-6

10

-5

10

-4

10

-3

10

-2

0 10 20 30 40 50 60 70

10

-5

10

-4

10

-3

10

-2

x/M

x/M

Fig. 11. Turbulent kinetic energy and its dissipation downstream of

the grid (Re

M

= 8250).

0 10 20 30 40 50 60 70

0

1

2

3

4

5

6

Re

M

=7425

x/M

L

f

/M

Re

M

=8250

Re

M

=10725

Fig. 12. Macro length scale of turbulence.

F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935 931

to an anisotropic state for the highest x/M as mentioned

by Gibson and Dakos [16]. On the other hand, L

f

is de-

ﬁned as a function of u

03

(5a). Lastly, no inﬂuence of U

is noticed on L

f

. This size structures is imposed by the

mesh size (source of the turbulence).

In terms of comparison, the agreement is clearly good

with previous studies. Bailly and Comte-Bellot [1] gave

the same order of magnitude (L

f

/M = 0.47 at x/

M = 42) whereas Cadiergue [7] found that 1 < L

f

/

M < 6 for 20 < x/M < 80 with an increase of data scat-

tering as well. For Yearling and Gould [31], this ratio

evolves between 0.2 and 0.5 when 2 < x/M < 20 with a

linear growth. For cylindrical bars, Gibson and Dakos

[16] and Sirivat and Warhaft [26] gave L

f

/M = 2.5 and

3.2 at x/M = 100.

The small diﬀerences may come from the experi-

mental set-up (vertical ﬂow for Cadiergue [7], cylindri-

cal bars for Gibson and Dakos [16] or Sirivat and

Warhaft [26]) and from the data processing technique

used to deﬁne u

0

and w

0

(no data ﬁltering for Cadier-

gue [7]).

Fig. 13 concerns the micro length scale of turbulence

(k

f

is called ‘‘Taylor micro scale’’ as well). Results are

still normalized with M.

Re

M

has almost no inﬂuence on k

f

(or slightly). It

grows from 0.15 to 0.5 whatever U is. The range of

k

f

is then situated between 0.005 m and 0.0165 m. Data

scattering is lower than for L

f

/M. Indeed, this scale is

proportional to

ﬃﬃﬃﬃﬃ

u

2

_

(6a) instead of u

03

for L

f

(Fig.

14).

Comparing the present results with others experimen-

tal studies reveals a good correlation again. When

Cadiergue [7] indicates that 0.6 < k

f

/M < 1.3 for

20 < x/M < 80, Michelet et al. [20] have found 0.5 < k

f

/

M < 0.6 for 30 < x/M < 70. For Belmabrouck [2], the ra-

tio k

f

/M evolves between 0.2 and 0.25 (20 < x/M < 50).

Benedict and Gould [6] argue that 0.25 < k

f

/M < 0.4

for 16 < x/M < 44.

This tends to point out that L

f

/k

f

may not inﬂuenced

by U. Fig. 14 conﬁrms this feeling.

3.4. Time scales of turbulence

Time scales of turbulence are also of importance.

They provide time characteristics of the ﬂow which are

useful for experimental studies. For instance, they help

to deﬁne duration of acquisition (macro time scale, T

t

)

or the mesh size Dt for numerical codes (micro time

scale, s

t

). Here we are ﬁrstly interested in T

t

. This scale

is related to the ‘‘extinction time’’ of the turbulence. It

corresponds to the time scale of the largest eddies. These

structures are responsible for the turbulent diﬀusion and

the energy exchange between the mean and the turbulent

ﬂows. It also gives an estimation of their persistence.

Fig. 15 is the evolution of T

t

(macro time scale) as a

function of x/M.

Fig. 15 displays that T

t

regularly increases from 0.7 s

(x/M = 15) to nearly 8 s at x/M = 50. This always stays

lower than the duration of the acquisition (T = 50 s)

and it does not exceed 15% of T. Thus, it conﬁrms our

hypothesis beforehand supposed saying that T )T

t

. A

turbulent eddy (moving at U) will cover less than 2 m.

It does not have enough time to reach the end of the

ﬂume located more than 10 m downstream of the grid.

0 10 20 30 40 50 60 70

0.0

0.2

0.4

0.6

0.8

f

λ /M

x/M

Re

M

=10725

Re

M

=8250

Re

M

=7425

Fig. 13. Micro length scale of turbulence.

0 10 20 30 40 50 60 70

0

2

4

6

8

10

12

Re

M

=7425

Re

M

=8250

Re

M

=10725

x/M

L λ

f

/

f

Fig. 14. Evolution of L

f

/k

f

.

0 10 20 30 40 50 60 70

0

2

4

6

8

10

12

14

Re

M

=7425

Re

M

=8250

Re

M

=10725

x/M

T

t

Fig. 15. Macro time scale of turbulence.

932 F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935

In the same way, Fig. 16 describes the behavior of the

Taylor micro time scale (s

t

) with x/M. Although s

t

in-

creases with x/M (from less than 0.06–0.4 s), a diﬀerence

appears depending on Re

M

. The Reynolds number

aﬀects the growth of s

t

. The larger Re

M

is the lower s

t

is. According to the previous results (Fig. 13), this could

indicate that the velocity scale of the Taylor micro scale

increases with U.

3.5. Kolmogorov scales

Kolmogorov postulates that the key parameters

governing the small-scale motion are the dissipation

rate (e) and the kinematic viscosity (m). From that,

he deﬁnes space (g

K

), time (s

K

) and velocity (v

K

)

scales representative of the smallest dissipative struc-

tures. Below this domain, the action of the viscosity

rules over the rest. Fig. 17 presents the evolutions of

these scales.

For g

K

, there is an inﬂuence of U. The highest U is

the lowest g

K

is. g

K

goes from 0.0002 to 0.00055 m.

The best ﬁt to these data is a power law such as

g

K

#(x/M)

0.48

.

For s

K

, a similar trend is depicted. The smallest dis-

sipative structures persist during 0.05–0.3 s. They are

transported by the mean ﬂow over less than 0.075 m

(2.3 M).

Lastly, v

k

increase with U and decrease with x/M

from 0.005 m/s to 0.002 m/s.

3.6. Reynolds number associated with micro scales

In relation with k

f

, a Reynolds number ðRe

kg

Þ is asso-

ciated (for isotropic turbulence, k

f

¼

ﬃﬃﬃ

2

p

k

g

). It deﬁnes

the separation between the biggest and smallest eddies.

Theoretically, when Re

kg

< 10, the beginning of the ﬁnal

decay zone of turbulence is reached [1].

Our estimations for Re

kg

are plotted on Fig. 18.

The beginning of the ﬁnal decay region of the turbu-

lence is not reached even for x/M = 70. Bailly and

Comte-Bellot [1] expect this region to start for x/

M > 100. This tends to conﬁrm their expectations.

4. Conclusions

In the present paper, we aimed to investigate the grid-

generated turbulence features (mesh size M) in a current

ﬂow with a free surface (constant water depth,

d = 0.35 m). A 2D LDV system was used to measure

the velocity ﬁelds at mid-depth in this homogeneous

and isotropic ﬂow. Through Taylor hypothesis, turbu-

lence length and time scales have been estimated for

three diﬀerent experimental conditions corresponding

0 10 20 30 40 50 60 70

0.0

0.1

0.2

0.3

0.4

0.5

Re

M

=7425

Re

M

=8250

Re

M

=10725

x/M

t

τ

Fig. 16. Micro time scale of turbulence.

0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

0.0000

0.0001

0.0002

0.0003

0.0004

0.0005

0.0006

0.0

0.1

0.2

0.3

0.4

0.000

0.001

0.002

0.003

0.004

0.005

x/M

K

Re

M

=7425 Re

M

=8250 Re

M

=10725

x/M

K

x/M

K

Fig. 17. Kolmogorov length, time and velocity scales.

F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935 933

to Re

M

of 7425, 8250 and 10,725. Our results point out

that:

• A detailed study of the disturbed ﬂow (mean and tur-

bulent ﬁelds) show that a homogeneous and isotropic

region is found starting 15 M downstream of the grid.

In this area (15 < x/M < 60, À3 < y/M < 3, 3 < z/

M < 8), the Taylor hypothesis is valid and turbulence

length and time scale may be then determined

through an ‘‘energetic method’’;

• The macro length scale of turbulence (L

f

) grows with

the distance to the grid x/M from 0.5 (x/M = 15) to

more than 2 (x/M = 40). Results exhibit not only a

strong agreement with several previous studies but

also demonstrate that U does not inﬂuence L

f

. The

size of the biggest eddies are mainly imposed by the

size of the source of turbulence (M in this case);

• The micro length scale of turbulence (k

f

) increases

with x/M as well. Typical values of k

f

are located

between 0.005 m (0.15 M) and 0.0165 m (0.5 M). As

shown for L

f

, U is not a key parameter governing

their behavior and the ratio L

f

/k

f

is then independent

of Re

M

;

• The macro time scale of turbulence (T

t

) can reach up

to 8 s at x/M < 50. So, the biggest eddies are con-

vected over a maximum distance of 2 m which is

not long enough to get to the end of the ﬂume;

• The micro time scale of turbulence (s

t

) develops with

the distance to the grid from 0.15 s to 0.6 s at x/

M = 60. The role of U is stronger: the highest U is

the shorter s

t

is;

• The Kolmogorov scales have been estimated as well

showing diﬀerent behaviors. At a ﬁxed position x/

M, as U increases, g

K

and s

K

decrease whereas v

K

takes greater values.

• The evolution of the Reynolds number of turbulence

ðRe

kg

Þ does not allow to deﬁne the beginning of the

ﬁnal decay region of the turbulence which may start

at x/M > 100.

All these results provide interesting information on

the turbulence features developing behind an obstacle

in a current ﬂow. Previous studies proved that the free

surface have an inﬂuence on the structure of the turbu-

lence closed to an air/water interface, either in current or

in wave/current ﬂows [22]. This present experimental

study was more focused on the role of the current and

aimed to fully describe the features of a ‘‘well-known’’

turbulence which is often used due to its homogeneous

and isotropic properties. This work could be useful

and bring further knowledge to study sediment trans-

port in ﬂood and ebb tides as well as in rivers ﬂows

where current prevails over waves. Nevertheless, one

should be careful to the experimental set-up. The thick-

ness of the grid frame needs to be as small as possible to

avoid any additional interactions with the ﬂow.

Future works could be undertaken to analyze the

interaction between turbulence and sand particles in

terms of bed load or suspension transport. Studies on

scour erosion can be lead as well to identify key param-

eters governing particle motions. Closed to the air/water

interface, the multiphase ﬂow (air/water/particles) would

require particular devices such as optical probes (air

fraction) or ultrasonic sensors (solid fraction).

Acknowledgement

The authors want to thank the Regional Council of

Basse—Normandie and the European Community for

their ﬁnancial support.

References

[1] C. Bailly G. Comte-Bellot, Turbulence, Cours de DEA, Ecole

Centrale de Lyon, 1999.

[2] H. Belmabrouck, Mesure des e´chelles turbulentes de longueur par

ve´locime´trie doppler laser en deux points, Ph.D. Thesis, Ecole

Centrale de Lyon, 1992, 196 p.

[3] H. Belmabrouck, M. Michard, Taylor length scale measurement

by Laser Doppler Velocimetry, Experiments in Fluids 25 (1998)

69–76.

[4] M. Be´lorgey, A. Arsie´, S. Cadiergue, The importance of the

turbulence scales in coastal engineering, in: Proceedings of

Hydralab workshop, Hanovre, 1999.

[5] L.H. Benedict, R.D. Gould, Concerning time and length scale

estimates from burst-mode LDA autocorrelation measurements,

Experiments in Fluids 24 (1998) 246–253.

[6] L.H. Benedict, R.D. Gould, Regarding Taylor microscales and 1-

D energy spectra determined from spatial and autocorrelation

measurements using LDV, Turbulent ﬂows FED-155 (1993) 79–

83.

[7] S. Cadiergue, Analyse des caracte´ristiques de la vitesse de chute de

particules solides en e´coulement turbulent, Ph.D. Thesis, Univer-

sity of Caen/Basse-Normandie, 1998, 158 p.

[8] A. Cenedese, G.P. Romano, F. Di Felice, Experimental testing of

TaylorÕs hypothesis by LDA in highly turbulent ﬂow, Experiments

in Fluids 11 (1991) 351–358.

0 10 20 30 40 50 60 70

40

80

120

160

200

240

Re

M

=7425

Re

M

=8250

x/M

Re

g

Re

M

=10725

Fig. 18. Evolution of Re

kg

.

934 F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935

[9] G. Comte-Bellot, S. Corrsin, The use of a contraction to improve

the isotropy of grid-generated turbulence, Journal of Fluid

Mechanics 25 (1966) 657–682.

[10] D.T. Cox, N. Kobayashi, Identiﬁcation of intense, intermittent

coherent motions under shoaling and breaking waves, Journal of

Geophysical Research 105 (C6) (2000) 14223–14236.

[11] M.E. DeCroix, R.D. Gould, Laser Doppler Velocimetry errors

due to Bragg cell based frequency shifting, in: Proceedings of

FEDSM, ASME Fluids Engineering Division Summer Meeting,

Washington, DC, 1998.

[12] R. Deigaard, J. Fredsoe, I.B. Hedegaard, Suspended sediment in

the surf zone, Journal of Waterway, Port, Coastal and Ocean

Engineering 112 (1) (1986) 115–128.

[13] F. Dietlin, Contribution a` lÕe´tude de lÕinﬂuence de la turbulence

sur la vitesse de se´dimentation de particules marines, Ph.D.

Thesis, University Pierre et Marie Curie, Paris 6, 1982, 294 p.

[14] J.G. Eriksson, R.I. Karlsson, An investigation of the resolution

requirements for two-point correlation measurements using LDV,

Experiments in Fluids 18 (5) (1995) 393–396.

[15] M. Garcia., Y. Nino, F. Lopez, Laboratory Observations of

Particle Entrainment into Suspension by Turbulent Bursting,

Coherent Flow Structures in Open Channels, John Wiley and

Sons Ltd., 1996.

[16] M.M. Gibson, T. Dakos, Production of temperature ﬂuctuations

in grid turbulence: WiskindÕs experiment revisited, Experiments in

Fluids 16 (1993) 146–154.

[17] A. Jensen, Artiﬁcial reefs of Europe: perspective and future, ICES

Journal of Marine Science 59 (2002) S3–S13.

[18] L. Jing, P.V. Ridd, Wave-current bottom shear stresses and

sediment re-suspension in Cleveland Bay, Australia, Coastal

Engineering 29 (1996) 169–186.

[19] S. Michelet, Turbulence et dissipation au sein dÕun re´acteur agite´

par une turbine Rushton, Ph.D. Thesis, Polytechnic National

Institute of Lorraine, 1998, 195 p.

[20] S. Michelet, Y. Antoine, F. Lemoine, M. Mahouast, Mesure

directe du taux de dissipation de lÕe´nergie cine´tique de turbulence

par ve´locime´trie laser bi composante: validation dans une turbu-

lence de grille, C.R. Acade´mie des Sciences Paris, 326, se´rie IIb,

1998, pp. 621–626.

[21] F. Murzyn, Etude de lÕinﬂuence dÕune onde sur les e´chelles de

turbulence: Application a` la houle, Ph.D. Thesis, University of

Caen/Basse-Normandie, 2002, 202 p.

[22] F. Murzyn, M. Be´lorgey, Wave inﬂuence on turbulence length

scales in free surface channel ﬂows, Experimental Thermal and

Fluid Science 25 (2) (2005) 179–187.

[23] B. Mutlu Sumer, R.J.S. Whitehouse, A. Torum, Scour around

coastal structures: a summary of recent research, Coastal Engi-

neering 44 (2001) 153–190.

[24] B. Mutlu Sumer, L.H.C. Chua, N.-S. Cheng, J. Fredsoe, Inﬂuence

of turbulence on bed load sediment transport, Journal of

Hydraulic Engineering 129 (2003) 585–596.

[25] G. Rashidi, S. Banerjee, Turbulence structure in free-surface

channel ﬂows, Physics of Fluids 31 (9) (1988) 2491–2503.

[26] A. Sirivat, Z. Warhaft, The eﬀect of a passive cross-stream

temperature gradient on the evolution of the temperature variance

and heat ﬂux in grid turbulence, Journal of Fluid Mechanics 128

(1983) 323–346.

[27] D.V. Srikantaiah, H.W. Coleman, Turbulence spectra from

individual realization Laser Velocimetry data, Experiments in

Fluids 3 (1985) 35–44.

[28] H. Tennekes, J.L. Lumley, A First Course on Turbulence, 16th

ed., MIT Press, 1997.

[29] F. Thomalla, C.E. Vincent, Beach response to shore-parallel

breakwaters at Sea Palling, Norfolk, UK, Estuarine, Coastal and

Shelf Science 56 (2003) 203–212.

[30] T. Wei, W.W. Willmarth, Examination of v-velocity ﬂuc-

tuations in a turbulent channel ﬂow in the context of sedi-

ment transport, Journal of Fluid Mechanics 223 (1991)

241–252.

[31] P.R. Yearling, R.D. Gould, A comparison of methods used to

estimate turbulence scales Individual Papers in Fluids Engineer-

ing, FED, vol. 150, ASME ,1993, pp. 67–72.

F. Murzyn, M. Be´ lorgey / Experimental Thermal and Fluid Science 29 (2005) 925–935 935