J . F h i d Mech. (1988), vol. 197. p p .

139-155 139
Printed in Great Britain

Turbulence reduction by screens

By JOHAN GROTH AND A R N E V. JOHANSSON
Department of Mechanics, The Royal Institute of Technology, S-100 44 Stockholm, Sweden

(Received 13 April 1987 and in revised form 6 May 1988)

Turbulence suppression by use of screens was studied in a small wind tunnel

especially designed and built for the purpose. Wide ranges of mesh sizes and wirc-
diameter Reynolds numbers were covered in the present investigation, enabling the
study of sub- and super-critical screens under the same, well-controlled, flow
conditions. For the latter type small-scale fluctuations, produced by the screen itself,
interact with the incoming turbulence. In the immediate vicinity of the screen the
turbulence was found to be highly anisotropic and the intensities were higher than
on the upstream side. Downstream of a short initial decay region, where the
intensities decrease rapidly, the return to isotropy was found to be much slower than
for the unmanipulated turbulence. The latter was generated by a square rod grid,
and was shown to become practically isotropic beyond a distance of roughly 20 mesh
widths from the grid. The role of the turbulence scales for the overall reduction
effectiveness, and for the optimization of screen combinations for application in low-
turbulence wind tunnels was studied.

1. Introduction
The effects on turbulence of screens or wire gauzes is a classical topic for both
theoretical and experimental turbulence investigations, as may be seen from the
review material in Corrsin (1963) and Laws & Livesey (1978). The latter also deals
with various other aspects of flow through screens, such as effects on mean velocity
distributions and mean flow direction. In the present study we will focus on the
turbulence reduction processes in the case of a uniform, perpendicularly oncoming
stream with moderate turbulence intensities, which is a problem of considerable
practical importance. For instance, the use of screens to improve flow quality in wind
tunnels was first suggested by Prandtl (1932). Limitations on power input and
cooling devices are factors that contribute to the importance of optimizing the use
of screens for turbulence reduction at a given pressure drop.
The problem of turbulence suppression by screens also contains many interesting
basic features of how turbulent eddies of disparate scales interact, etc. On the
upstream side there is a pressure redistribution, and a stream contraction into the
centre-plane of the screen, whereas on the downstream side thin shear layers are
present and vortices are shed if the Reynolds number (Re,) based on wire diameter
is large enough ( 240). The processes may be expected to be quite different above
and below this value of the Reynolds number, which hence is one of the primary
parameters to characterize the effects of a screen. Another parameter that influences
the pressure drop across the screen is the solidity, u,defined as projected solid area
per unit total area. For values of CT higher than about 0.5 flow instabilities due to jet
coalescence may occur (see e.g. Baines & Peterson 1951). Screens for turbulence
140 J . Groth and A . V .Johansson
suppression, hence, always have solidities below this value. The change in flow
direction can be described by a refraction coefficient, a (see e.g. Laws &, Livesey
1978), defined as the ratio between the flow angles on the downstream and upstream
sides, respectively.
A complete theoretical treatment of the flow through a screen is, as yet, not a
tractable problem, but various theoretical modelling efforts have been made. Taylor
& Batchelor (1949) studied the effect of screens both on steady mean flow
disturbances and on turbulence by a linear, inviscid analysis, neglecting the
generation of turbulence by the screen itself (see also Batchelor 1976). Assuming
isotropy for the incoming turbulence, they could determine reduction factors for the
streamwise and lateral intensities, which depend solely on a: and the static pressure
drop coefficient for the screen.
Dryden & Schubauer (1947) put forward a simple empirical formula for the
reduction factors. The bulk of their experiments were carried out with the screens
placed in the settling chamber of a wind tunnel, while the actual measurements were
made in the test section. They noted that for a given total pressure drop it is
advantageous to use several screens with relatively low pressure drop coefficients.
Schubauer, Spangenberg & Klebanoff (1950) studied damping characteristics of
screens both in the super- and the sub-critical ranges. They also investigated the
dependence of the critical Reynolds number on solidity in the range 0.19-0.57.
Several experimental investigations of turbulence reduction by screens and other
flow manipulators have been carried out by Nagib and coworkers. Tan-Atichat,
Nagib & Loehrke (1982) studied the effects of screens and perforated plates on the
streamwise turbulence intensity in two small wind tunnels (diameters 7 and 15 cm),
for a variety of incoming flow conditions. They found that for supercritical screens
the mesh size should be chosen such that the lengthscales of the turbulence generated
by the screen are smaller than those of the incoming turbulence, but not much
smaller. The motivation given for this was that under these conditions the spectral
transfer of energy to dissipative scales would be maximized.
Turbulence was, in the present study, generated by means of a square rod grid, and
wide ranges of the screen mesh size and Reynolds number, Re,, were covered.
Subcritical screens (Re, < 40) were shown to give a large reduction, but the
maximum turbulence suppression for a given total pressure drop was found for
cascade type of combinations in which the screen furthest downstream has a low
supercritical Reynolds number, whereas the upstream ones are chosen to be
relatively coarse. However, the mesh width should be significantly smaller than the
typical scale of the incoming turbulence. Very few data exist in the literature
regarding the behaviour of the lateral turbulence intensities behind screens. I n the
present study the manner in which the flow behind screens, as well as grids, returns
to istropy was studied.

2. Experimental apparatus and procedure

The present flow facility was designed specifically for the study of turbulence
damping by screens and permits different screens and screen combinations to be
readily interchanged. It is built as an open-circuit blowing type wind tunnel with a
2 m long (0.45x 0.45 m2)test section. Mufflers were placed on both sides of the fan to
obtain a low noise level. The fan motor was equipped with guide vanes to reduce swirl
and a conical afterbody long enough to prevent separation. Two honeycombs and
 Turbulence reduction by screens 141

Wire
Meshes diameter
No. (in.-l) (mm) Solidity Re, K
1 2.1 2.50 0.37 830 0.68
2 4.9 1.00 0.35 330 0.66
3 7.9 0.50 0.29 170 0.49
4 12.1 0.50 0.42 170 0.99
5 19.0 0.24 0.33 80 0.79
6 34.0 0.19 0.44 65 1.64
7 147.0 0.04 0.41 13 2.90
Grid 1.0 6.00 0.42 2000
TABLE1. Basic data for the grid and the screens. Reynolds numbers and resistance coefficient (K)
values are given at a mean velocity of 5 m/s.

four screens eliminate to a high degree the remnants of swirl and other flow
disturbances introduced by the fan.
Turbulence was generated by means of a biplane grid, made of wooden rods of
square cross-section (6 x 6 mm'). The solidity was 0.42 and the mesh width 25 mm.
For the bulk of the experiments presented here it was placed 20 mesh widths
upstream of the position where the screens to be studied were inserted. This ensures
that the screens will be far enough downstream to avoid the initial region of strong
anisotropy, and that the turbulence has become laterally homogeneous (see e.g.
Hinze 1959). Combinations with up to five screens were used here, but the set-up
allows for practically any number of screens. The separations could be varied from
0.05 m t o 0.30 m by means of inserts.
The flow behind the grid had a mean velocity ( U ) of about 5 m / s , and was
reasonably uniform in the streamwise direction and in cross-stream planes. The
difference between the maximum and minimum velocity values within a streamtube
0.1 x 0.1 m2 along the centreline of the test section was less than 4% of the mean
velocity. Without grid or screens the turbulence level was approximately 0.4% a t the
beginning of the test section. The lateral dimension (0.45 m) was sufficiently large to
ensure negligible influence from the wall boundary layers, and to provide a good
homogeneity in the cross-stream direction of the grid turbulence. A zero pressure
gradient, and hence constant free-stream velocity, was ensured by a slightly
increasing cross-sectional area in the streamwise direction. This was achieved by
giving one of the walls an opening angle of 0.7".
The seven different types of screens used cover a wide range of mesh sizes and wire
diameters (see table 1). The solidity was in all cases below 0.44 to avoid jet
coalescence. The coarsest screen had a mesh size of half of that of the grid or 70 times
that of the finest screen. The latter is subcritical a t the velocities used here, its
Reynolds number (Re,) based on the wire diameter being only about 10. For
Reynolds numbers below roughly 40 no turbulence will be produced by vortex
shedding. Schubauer et al. (1950) found a critical Reynolds number of about 50 for
a solidity of 0.4. The present findings (figure 1) indicate that the transition from
subcritical to supercritical behaviour is smooth in terms of the pressure drop.
The static pressure drop across a screen is, in wind-tunnel applications, the penalty
to be weighed against the positive effect of turbulence reduction. A natural aim
would hence be t o maximize the damping for a given pressure drop. The dependence
142 J . G o t h and A . V . Johansson

A
A

A
.

O J . . , , .... I , . . . ....
I . . , , ,..I
1oo 10' 102 103
Re,
FIGITRE 1 . The static pressure drop functionfas a function of the wire diameter Reynolds number
Data for all the seven screens a t ten different velocities are included Each screen is represented by
a separate symbol.

of the static pressure drop coefficient, K , on the Reynolds number and the solidity,
cr, can be expressed as (see Pinker & Herbert 1967 or Laws & Livcsey 1978)

K %
!!!E
!L = f(&,) 1- (1 - cry
0.5~17~ (1-u)2 '
where p is the fluid density and f(RPd) is an empirical function The function f(RPd)
can be seen in figure 1 to reach a constant value of about 0.45 for high Re,-values,
but increases quite dramatically with decreasing Reynolds numbers below 100.
Hence, subcritical screens give a high static pressure drop coefficient. For a given
screen the K-value is about three times larger at Me, = 10 as compared t o that at
Re, = 100. Taylor & Davies (1944) made pressure drop measurements in the
Reynolds number range 250-600. They found a trend that well agrees with that in
figure 1 and their asymptotic pressure drop coefficient corresponds to f = 0.45 for
r = 0.35. For higher solidities, however, their asymptotic values are higher (up to
0.6). Pinker & Herbert (1967) present pressure drop data for Reynolds numbers
above 300 with u in the range 0.36-0.68. Their f-values decrease slightly with
Reynolds number and level off at about 0.50, which is somewhat higher than the
value found here. It is noteworthy that the data in figure 1 cover Reynolds numbers
all the way between 2 and 850.
The turbulence measurements were carried out with standard hot-wire single and
X-probes, as well as a hot-film X-probe. These could be traversed in the streamwise
direction (along the z-axis) and vertically (along the y-axis). The anemometer signals
were sampled and stored with a MINC P D P 11/23 on which also all data evaluation
was made.
In order to correctly evaluate the degree of anisotropy behind grids and screens it
is essential, when using hot wires or hot films, to be aware of the influence of
tangential cooling. This can be taken into account in two different ways, viz. either
by calibrating the probes against flow velocity and angle or by using the cornpletc
cosine law including the tangential cooling coefficient k . For an X-probe in the (x,y)-
plane with wires in 45' to the mean flow direction the effective cooling velocities can
 Turbulence reduction by screens 143
where u, and w denote the velocity fluctuations in the streamwise (x)and lateral
(y, z ) directions. The measured u and v components are calculated as Ueff,--+Ueff,,
respectively, divided by .\/2 and a factor (1+ k2)a, where the latter arises from the
calibration procedure. This yields the following lowest-order relations between
measured and true values of the turbulence intensities :

The same correction as for v,, is also valid for the mean value of uv.The k-value for
hot wires was det,ermined for various length to diameter ratios by Champagne,
Sleicher & Wehrmann (1967). For l / d = 200 (as used here) the k-value was found to
be 0.2, which means that the measured vrm,-value is 8% too low if the tangential
cooling is neglected. A cylindrical hot-film X-probe was found to give practically
identical results as the hot wires when its tangential cooling sensitivity was
accounted for. The hot-film (DISA 55R61) k-factor deduced from comparisons
between the measured vrm,-values from the hot-film and hot-wire probes was 0.32.
This agrees well with that (0.35)determined by Alfredsson & Johansson (unpublished
results). The correction for the hot-film probe, hence, corresponds to 23% of the
measured vrms.

3. Results
The turbulence intensities (normalized by U ) produced by the grid were
approximately 5.5% both in the streamwise and the lateral directions a t x = 0, i.e.
a t 20 mesh widths ( M )downstream the grid. This is the position where screens are
inserted and if only one is used, x is hence measured from that screen. It has been
reported in several earlier investigations that turbulence behind a grid retains a small
degree of anisotropy over very large downstream distances. For instance, Comte-
Bellot & Corrsiri (1966)found a practically constant value of 2irm,/urmsslightly above
0.9, over as long a distance as 400 mesh widths, whereas values around 0.8, were
reported by Grant & h'isbet (1957) and Kistler & Vrebalovich (1966).The persistence
of anisotropy in experimental situations may be influenced by large-scale non-
isotropic turbulence on the upstream side of the grid. This factor, however, is not
likely to have influenced the above results. The rate of return to isotropy can also
depend slightly on a macroscale (i.e. integral scale) Iteyriolds number.
The present results indicate that the turbulence behind the grid indicate returns
to a practically isotropic state within about 20 mesh widths (figure 2 ) . The effects of
tangential cooling were here accounted for in evaluating v,,, (see 92). For a k-value
of 0.2 one would get an 8 % correction for the v,,,-data of Comte-Bellot & Corrsin
(1966), and thereby a v,,,/u,,,-ratio very close to unity. Thus, tangential cooling of
the hot wires may, at least partly; explain the lack of return to isotropy behind grids
reported in most earlier studies.? It is interesting to find that Batchelor & Townsend
(1947) concluded in an indirect manner, from u-component measurements of the
Taylor microscale, that isotropy behind a grid is reached after roughly 20 mesh
widths. I n figure 2 vrms/urm, is greater than 1 close to the grid whereafter it decreases
towards unity. A rapid decrease occurs in the region where the cross-stream
t Kistler & Vrebalovich used a n angle calibration of their X-wire. This involved fairly large
uncertainties for t,he determination of vrmswith the analogue terhnique used. They estimated an
error limit of 10 O/o in the wire calibrations, but this could probably be an over-optimistic value.
144 J . Groth and A . V . Johansson

0 20 40 60 80
x'jM
FIGURE
2 . The anisotropy measure vrms/urmsas a function of distance downstream of the grid
(Re, = 2000).

0.3
A

0.03 I I I I I I I I
10 1 10
(x' - xb)/M
FIGURE 3. Decay of the grid turbulence. u, (U/U,,,)~ and A, (U/v,,J2 as a function of ( d - x h ) /
M , where 2' is the distance measured from the grid, zb is the virtual origin and M is the mesh width.
The solid line represents a power-law dependence with exponent 1.0.

inhomogeneities are substantial. Beyond ten mesh widths the variation of the ratio
between the intensities is very slow, and the value is typically only a few per cent
below unity far downstream.
The grid-turbulence intensity urmSwas found to conform well t o the standard type
of power-law x-dependence (see Batchelor & Townsend 1948), i.e.

where x' is the distance downstream of the grid and xh is a virtual origin. Figure 3
shows the decay of turbulence intensities in the region x' > 20 M , i.e. where
 Turbulence reduction by screens 145

u,,,
U

OJ
0 20 40 60 80 I0
xlM
FIGURE
4. Typical example of the effect of a supercritical screen on the streamwise turbulence
intensity (mesh width M = 0.52 em, u = 0.35 and Re, % 330).

approximate isotropy prevails. The best fit to the experimental data in this region
was found for C = 25.2, a virtual origin of 6.0M and an exponent n of 1.0.This is also
the exponent that can be derived from simple dimensional arguments (see Tennekes
& Lumley 1983). The approximate isotropy for x’ > 20M implies that u,, and v,,,
obey the same decay law.
Hanarp (1981) also found a decay-law exponent, in the same region as above, of
1.0.Several other investigators (e.g. Baines & Petersen 1951 ; Comte-Bellot & Corrsin
1966) have found larger values of n, which can be explained by the fact that data
closer to the grid have been included in the determination of the power law constants.
Inclusion of data from the strongly anisotropic region leads to a larger exponent due
to the more rapid decay in this region. When all data beyond 0.8M were included we
obtained a decay law with n = 1.32, in close agreement with the results of Wahrhaft
& Lumley (1978) who obtained n = 1.34 (with zero virtual origin), and with the data
of Comte-Bellot & Corrsin.
The effect of a supercritical screen on the streamwise turbulence intensity
(urms/tJ)is illustrated in figure 4.The measurements on the upstream side were here
done with the probe penetrating the screen. The turbulence intensity decreases as the
flow approaches the plane of the screen due to the contraction effect and a pressure
redistribution on the upstream side. Near the screen on the downstream side the
turbulence intensity is higher than it would be without a screen. This is due to the
vortex shedding from the wires, and the wire-wake shear layers, that give rise to high
intensity small-scale turbulence, which however, decays rapidly. When only one
screen is used the distance required for the intensity to fall below the incoming level
was found t o be about 15 mesh widths for all the supercritical screens. This distance
becomes longer if the incoming turbulence level and macroscale are reduced, which
is the case in screen combinations.
Figure 5 ( a ) shows the downstream variation of the streamwise turbulence for all
the screens used. All but the subcritical screen (no. 7 ) initially have a higher intensity
than that given by the grid alone, although in this figure only data lower than the
grid turbulence a t x = 0 are included to facilitate the comparison between the
different curves. The initial decay occurs on a lengthscale determined by the mesh
width, which can be seen from the collapse of the different curves when x is
normalized by M (figure 5 b ) . The mesh width seems to be a relevant lengthscale, a t
146 J . Groth and A . V . Johansson

xlM

FIGURE
5 . u,,,/U for one subcritical and six supercritical screens (numbers refer t o table l ) , ( a ) as
funct,ion of z in cm and ( b ) as function of x / M .

least for x < 50M. For large x-values the rate of decay for all the screens is smaller
than that for the grid (see figure 5 a ) .
Measurements of the autocorrelation RUU(q)wherc 7 is the spatial separation,
computed from the temporal separation by use of Taylor's hypothesis substantiate,
in an illustrative way, the predominance of small scales just downstream of a
supercritical screen (figure 6 a ) . This is manifested by the fact that both the Taylor
microscale, A , and the macroscale, A , a t x = 2 cm arc much smaller (for screen no. 6
in this case) than for the grid alone. The microscale here is two mesh widths, and the
macroscale is about five times larger. The ratio between the two scales is about the
same for the grid alone, but here the macroscale is roughly equal to the mesh size of
 Turbulence reduction by screens 147

1.0 - (4 h(mm) A(mm)

I:
-I' ~

4.9 23.1
-_- 1.2 8.2
...._ 3.1 21.8

---___-----------
\

----------=
0
I
I I I I I I I
0 20 40 60
11 (mm)

0 20 40 60 80
71 (-1
6. Autocorrelation functions R,, for: -,
FIGCRF: the grid, ---, a supercritical screen no. 6 and
---. a subcritical screen no. 7 at ( a ) x = 2 ern and ( b ) z = 42 cm.
the grid. The macroscale was determined from autocorrelations, whereas the
microscale was computed from r.m.s.-values of u and its time derivative as

For the subcritical screen no. 7 these scales are almost as large as for the grid alone
(at x = 2 cm). Although no vortex shedding occurs for the subcritical screen, fluid
elements become highly stretched during the passage of the screen. The distorted
turbulence exiting the screen will undergo a period of relaxation and energy transfer
from the lateral components to the streamwise one. This could be seen from the
148 J . Groth and A . 8.Johansson

0 50 100 150 200

x (cm)

7. The anisotropy measure V ~ , ~ / U , , , , ~for ---, a supercritical no. 6 and - - - , a subcritical

FIGURE
screen no. I as a function of x.

downstream development of the anisotropy measure wrms/urms(see figure 7). The

decreasing trend of vrms/urmssignifies an energy flux from the lateral components to
the u-component, provided that the dissipation rate is about equal for all three
components. One can readily find from the Reynolds-stress transport equations that,
if diffusion is negligible and the dissipation tensor is isotropic, the intercomponent
energy transfer to uimsfrom the lateral components is proportional to the streamwise
derivative of uims- vim,.
For the subcritical screen the strong energy flux from the lateral components to
the streamwise component also leads to the effect that the average size of the most
energetic (large-scale)turbulent eddies tends to decrease with increasing x-values for
small streamwise distances. This is in contrast with isotropic decaying turbulence,
and may be illustrated by the fact that the macroscale decreases by about 50 % from
x = 2 cm to x = 7 cm downstream of the subcritical screen, whereafter it increases
monotonically and slowly.
Downstream of the initial rapid decay period, one may note that the energetic
eddies are of considerably smaller scale for both sub- and super-critical screens as
compared to those of the grid alone. This can be illustrated by the autocorrelations
a t x = 42 cm (see figure 6 b ) .
The lack of turbulence generation a t the subcritical screen ensures a slow
monotonic downstream development of the Taylor microscale (figure 8). Cor-
respondingly, the fact that the turbulence behind the supercritical screen is
dominated by the small scales is evidenced by a quite small microscale close t o the
screen, whereas the rapid decay of this part of the turbulence exhibits itself in a rapid
increase of the Taylor microscale. The variation of the measured microscale
downstream of the grid shows a fair agreement with that estimated theoretically
under the assumption of isotropy.
I n order to quantify the scales of the turbulence generated by supercritical screens
power spectra were computed for various positions behind the different screens, as
well as for the grid alone. The spectrum for a supercritical screen (no. 6) from the end
of the initial decay region (x = 2 cm or 2 7 M ) clearly yields the dominance of the
 Turbulence reduction by screens 149

0 60

FIGURE 8. Downstream development of the Taylor microscale for 0, the grid, A, a supercritical
no. 6 and 0 , a subcritical screen no. 7 . The solid line represents the microscale for the grid
calculated as h2 = 20v ( x ' - x i ) / U , which can be derived- from (3.1) with n = 1 (under -the
assumption of isotropy).

kE

FIGURE 9. Wavenumber power spectra a t z = 2 cm, for -, the grid, ---, a supercritical no. 6,
---,a subcritical screen no. 7 and ~ . -, .a combination of two screens no. 6 with 10 cm separation.
The spectra are shown in absolute scale, i.e. the area under each curve is equal to u : ~ ~ .

generated small-scale turbulence, and a strong suppression of the large scales (figure
9). I n absolute terms the energy is reduced by a factor of three for wavenumbers
corresponding to the incoming macroscale, whereas above 0.8 mm-l the energy is
increased by a maximum one order of magnitude. A sharp peak is seen a t
wavenumbers around 1.7 mm-l, or equivalently, a t wavelengths of about 5 mesh
widths or 20 wire diameters. However, neither of these numbers was found to be
universal, i.e. the same for different screens. Instead, the appropriate scaling involves
both quantities mentioned. The width of this power spectral peak was found to
increase with increasing wire-diameter Reynolds number, as should be expected. The
150 J . Groth and A . V . Johansson
position of the high wavenumber peak was found to be independent of the incoming
turbulence. This could be concluded from spectral measurements behind two screens
in which case the incoming turbulence to the downstream screen is significantly
different in scale from that to a single screen. For comparison the spectrum a t x =
2 cm for the subcritical screen is also included in figure 9, and shows only a moderate
change in the wavenumber energy distribution as compared with the grid alone. This
is in accordance with the micro- and macro-scale data (figure 6).
To study the development of the dominant high wavenumber peak, spectra were
measured in the immediate vicinity of a fairly coarse screen through which the hot
wire could be protruded. A comparison between a position one mesh width upstream
and one in the plane of the screen showed that the maximum in the incoming power
spectrum was shifted towards lower wavenumbers, roughly by a factor of 0.5,
reflecting the stretching of fluid elements. At one mesh width downstream, the broad
peak corresponding to the maximum a t x = 0 could still be clearly seen, but now
accompanied by a high wavenumber peak from the screen-generated turbulence.
From x = 2,V and throughout the initial decay region the spectrum is dominated by
the screen-generated turbulence, although the peak is slowly moved towards lower
wavenumbers due to the influence of dissipation.
Both sub- and super-critical screens reduce the streamwise turbulence intensity
more effectively than the lateral ones, in agreement with the analysis of Taylor &
Batchelor (1949).The ability to reduce the streamwise intensity increases, for a given
solidity, with decreasing mesh size (see figure 5 a ) , but the same trend is weaker for
vrms, which results in a strong anisotropy downstream of fine mesh screens. For the
subcritical screen vrms/urmshad a maximum value of about 2.1 (figure 7 ) , which
implies, since the lateral components are of nearly equal intensity, that 90 YOof the
energy was contained in the lateral components.
For both types of screens the return to isotropy was found to be very slow, and
take place on a lengthscale that is substantially larger than that for the
unmanipulated grid turbulence. The latter returned to approximate isotropy within
20 mesh widths (i.e. 50 em), whereas the flow behind the screens did not quite reach
isotropy within the 2 m long test section. This may seem surprising. but can be
understood from the following reasoning. The turbulent kinetic energy, q. decays on
a lengthscale which can be estimated as Ldecay = q / ( -dq/dx). From the Reynolds-
stress transport equations (see e.g. Launder, Reece & Rodi 1975) one can show that
the return to isotropy occurs on a lengthscale L = qU/E, where E is the rate of
dissipation. This return is caused by the pressure-strain correlation terms (see Rotta
1951). Under the assumption of Taylor’s hypothesis the lengthscales L and Ldecayare
identical. L,,,,, was computed from the variation of the turbulent energy, and it was
found to be typically a factor of five larger (in the region 0 . 1 4 . 5 m ) for the
supercritical screen (no. 6) than that for the grid. This together with the initial high
degree of anisotropy behind the screens explains the large distance required for the
return to isotropy.
The larger JJdecay for the flow behind screens reflects a smaller rate of dissipation
than for the grid alone. For isotropic turbulence E can be expressed as

(3.3)

This relation indicates the qualitative dependence of E on turbulence intensity and

microscale even for anisotropic cases. The smaller dissipation downstream of screens,
 Turbulence reduction by Screens 151

12
-
10 -
- A=5
I

I I I I I I I I I I I I I I
0 50 100 150
(a) x
10.The development of u,,,/U downstream : -,
FIGURE one screen no. 4 with Re, % 170 and
_ _ _ , a combination of two such screens with various separations, A (in cm).

as compared with that for the grid, is coupled to a strong reduction in intensities
while the microscale is only moderately affected (see figures 5a and 8).

3.1. Screen combinations

When screens are used in combination their separation should a t least correspond to
the length of the initial decay region. Figure 10 shows the turbulence intensity
downstream of a pair of supercritical screens (no. 4) for various separations between
25 and 150 mesh widths. The former roughly corresponds to the length of the initial
decay region (see figure 5b). The total reduction depends only slightly on the
separation within this range. It may here be worth mentioning that Dryden &
Schubauer (1947) found no effect of varying the spacing (from 5 to 70 cm) between
the screens in a wind-tunnel stagnation chamber when the measurements were
carried out in the test section.
One may, rather arbitrarily, construct a reduction factor as the ratio between the
incoming intensity and that far, say 1 m, downstream of a screen. This would give
a reduction factor of 2.9 for a single screen (no. 4), whereas a combination of two gives
a total reduction factor of 4.8. The downstream screen, which encounters a more
' fine-grained ' turbulence with lower intensities, hence gives a smaller turbulence
reduction.
For a given allowable pressure drop the amount of turbulence reduction obtained
depends strongly on mesh sizes, and on how different types of screens are combined,
as is illustrated in figures 1 1 (a,b ) for two-screen combinations. Firstly, it is seen that
a combination of two different supercritical screens, the coarsest being the upstream
one, gives a substantially greater reduction than a subcritical screen alone (figure
l l a ) , although the K-values for the combination and the subcritical screen are
approximately equal. Hence, although subcritical screens give high reduction
factors, their penalty in terms of pressure drop makes them undesirable for practical
purposes. For wind-tunnel applications they also have the disadvantage of being
prone to clogging by small dust particles.
A cascade combination (see figure l i b ) gives a slightly larger reduction, for
moderate x-values. than if the coarser screen were to be replaced by one more of the
152 J . Groth and A . V .Johansson

FIGURE 1 1 . u,,,/U downstream of two-screen combinations, all with 10 cm separation. ( a )

Comparison between c],a cascade combination of supercritical screens (no. 4+no. 6) and A,a
single subcritical screen no. 7. K = 2.7 for both cases. ( b ) Comparison between - - - , the no. 4+no.
6 and ---, a reversed cascade combination no. 6 + n o . 4.Included are also --no. 4+no. 4 and
-, no. 6 + n o . 6.

finer type, although the latter combination has a 25% higher K-value. A reversed
cascade combination of the same screens gives higher turbulence intensities over a
long region in the streamwise direction, which makes them unsuitable for use in
wind-tunnels. However, far downstream the intensity level becomes, somewhat
surprisingly, equal to that of the regular cascase combination. As should be expected,
two coarse screens give less reduction than either type of cascade, for all x-values.
The streamwise turbulence intensity was measured behind various two-screen
cascade combinations with a fine-mesh supercritical screen (no. 6) as the downstream
one. A constant separation of 10.5 cm was used, which corresponds to more than 20
 l’urbulence reduction by Screens 153
mesh widths for all cases. This allows for the initial period of rapid dccay to take place
upstream of the second screen. The damping varied fairly little between different
combinations although Re, of the upstream screen varied between 65 and 330, This
is interesting since it allows a given turbulence reduction t o be achieved for quite
different pressure drops. The typical lengthscales generated by each screen in a
combination should be substantially smaller than the incoming macroscale, but aside
from this the damping ability of the combination is relatively insensitive to the
specific lengthscale ratio and total pressure drop.
Dryden & Schubauer (1947)gives an empirical formula for the turbulence damping
by screen combinations. The damping is calculated as the product of factors
(1+ K i ) 0 , 5( i = 1, N ) , where K i is the respective K-value and N is the number of
screens. This would imply an approximately 25 % lower turbulence intensity
behind the no. 6+no. 6 combination as compared to the no. 2+no. 6 case, while
measurements showed almost equal levels. The latter combination has a pressure
drop coefficient which is only two thirds of that for the former. One should here
remember that the dissipation obscures the generality of conclusions about reduction
factors, and makes comparisons with formulae, such as the one just described, rather
doubtful.
Various three-screen combinations were tested and the lowest turbulence level a t
x = 1 m was 0.78% (to be compared with the incoming level of 5.5%), which was
obtained for a typical cascade combination (no. 3+no. 4+no. 6). This was further
complemented with one each of the two finder screens resulting in a total K-value of
5.8, and a total intensity reduction factor of about 12 (defined as above). The formula
given by Dryden & Schubauer (1947) would give a value of 6.4 for this five-screen
combination.

4. Summary and conclusions

Turbulence generated by a (biplane square rod) grid was found t o undergo a
transition to approximate isotropy from the initially strongly anisotropic state
within a region roughly 20 mesh widths long. The persistent anisotropic character far
downstream of the grid as found by some earlier investigators, can partly be
explained by influence of tangential cooling of the hot wires used. This part of the
heat loss from the wires acts to reduce the apparent intensity of the lateral
component (see (2.3)).I n the approximately isotropic region (x’> 20M) the kinetic
energy, and uims,were found to obey a ‘linear decay law’, i.e. with an exponent of
1.0. This exponent increased if parts of, or the whole, anisotropic region were
included. This explains the discrepancies reported in the literature regarding the
decay law exponent.
The pressure drop coefficient for screens was found to conform well to the general
form (equation (2.1))given by Pinker & Herbert (1967) where the dependence on
solidity is given and the wire diameter Reynolds number enters in a multiplicative
function f.The asymptotic value off for high Reynolds numbers, Re, > 300, was here
found to be approximately 0.45, which is somewhat lower than that found by Pinker
& Herbert.
The turbulence damping ability of a screen increases with decreasing mesh size for
a given solidity. Subcritical screens thus give a large turbulence reduction, but their
‘penalty ’ in terms of a high pressure drop makes them inferior for practical use as
compared with cascade combinations of supercritical screens. Also, a t the velocities
154 J . Groth and A . V .Johansson
typically used in the stagnation chamber in the middle to upper velocity range in a
low-speed tunnel ( - 5 m/s) very small wire-diameters are needed for screens to be
subcritical (0.040mm was used in the present case), which makes them susceptible
to dirt contamination (and expensive).
The flow in the region immediately downstream of a supercritical screen has a
distinctly different character from that far downstream. Both regions were found to
exhibit several interesting features that differ from those of the unmanipulated flow.
In the immediate vicinity of the screen the turbulence intensities are substantially
higher than on the upstream side. The energy spectrum is dominated by the screen-
generated turbulence, and the dominating wavelengths are a few mesh widths
independent of the scale of the incoming turbulence. Within the first 15-25 mesh
widths, ‘the initial decay region ’, the intensities decrease rapidly, and for a single
screen fall below the incoming levels. The dominating spectral peak was found to
persist throughout this region, but is slowly shifted towards lower wavenumbers due
to the influence of dissipation, The increase in energy for small scales is contrasted
by a strong suppression of the low wavenumber regime. Within the initial decay
region the turbulence was found to be strongly anisotropic for both sub- and super-
critical screens, causing an energy transfer from the lateral to the streamwise
component. This was most clearly seen for subcritical screens, for which the
maximum anisotropy corresponded to a situation with 90% of the energy in the
lateral components. Both types of screens exhibit a much slower return to isotropy
than the grid. This was shown to be consistent with a larger lengthscale of decay
beyond the initial decay region. Hence, the anisotropic character behind screens
persists into the ‘far-field region ’, which also is characterized by a smaller rate of
decay than for the grid alone.
When screens are used in combination their separation should be chosen larger
than the length of the initial decay region. As long as this criterion is satisfied the
separation only slightly affects the damping. For a given total pressure drop the
maximum turbulence reduction was found for cascade type of screen combinations
where the last screen (in the streamwise direction) was chosen with a low, albeit
supercritical, Reynolds number. This ensures small scales produced by the last screen
and thereby a short initial decay region. To minimize the pressure drop the other
screens should be chosen with increasing mesh size in the upstream direction. The
pressure drop decreases with increasing Reynolds number (for a given solidity) for
Re, < 300, but levels off above this value, which hence may serve as a guide for
choosing the coarsest screen.
A combination of a relatively coarse screen followed by a finer one gave a t least as
effective a turbulence reduction as a pair of the finer type of screen, although the
latter had a considerably higher pressure drop coefficient. This invalidates any
quantitative description of damping solely based on screen properties such as the
pressure drop coefficient. Experiments with two-screen combinations also showed
that the damping ability depends only slightly on the typical lengthscale (e.g. the
macroscale) of the incoming turbulence, at least as long as this scale was substantially
larger than the mesh size.
As an illustration of the reduction achievable with cascade screen combinations
realistic for application in low-speed wind tunnels, it may be mentioned that a set of
five screens, with Re, from 170 down to 65, was found to give a total reduction,
1 m downstream of the last screen, of a factor 12 compared with the incoming
intensity.
 Turbulence reduction by screens 155
We wish to thank Professor Henrik Alfredsson for actively taking part in the
construction of the wind tunnel, and for many helpful comments, and Magnus
Hallback for assisting with some of the measurements. Support from the Swedish
Board for Technical Development is gratefully acknowledged.

REFERENCES
BAINES,W.D. & PETERSEN, E. G. 1951 An investigation of flow through screens. Trans. A S M E
73, 467480.
BATCHELOR, G. K. 1976 The Theory of Homogeneous Turbulence, chapter 4. Cambridge University
Press.
BATCHELOR, G. K. & TOWNSEND, A. A. 1947 Decay of vorticity in isotropic turbulence. Proc. R.
Soc. Lond. A 190, 534-550.
BATCHELOR, G. K. & TOWNSEND, A. A. 1948 Decay of isotropic turbulence in the initial period.
Proc. R . Soe. Lond. A 193, 539-558.
CHAMPAGNE, F. H., SLEICHER, C.A. & WEHRMANN, 0. H. 1967 Turbulence measurements with
inclined hot-wires. J . Fluid Mech. 28, 153-182.
COMTE-BELLOT, G. & CORESIN,S. 1966 The use of a contraction to improve the isotropy of grid-
generated turbulence. J . Fluid Mech. 25, 657-682.
CORRSIN,S. 1963 Turbulence: Experimental Methods. Handbuch der Physik, vol. 8, part 2, pp.
524-590. Springer.
DRYDEN, H. L. & SCHUBAUER, G. G. 1947 The use of damping screens for the reduction of wind-
tunnel turbulence. J . Aero. Sci. 14, 221-228.
GRANT,H . L. & KISBET.I. C. T. 1957 The inhomogeneity of grid turbulence. J . Fluid Mech. 2,
263-272.
HANARP, L. 1981 An experimental study of grid-generated freestream turbulence and its influence
on velocity and temperature boundary layers on circular cylinders. PhD thesis, Dept of
Applied Thermo and Fluid Dynamics, Chalmers University of Technology, Goteborg,
Sweden.
HINZE,J. 0. 1959 Turbulence, chapters 2.4, 3 and 4. McGraw-Hill.
KISTLER,A. L. & VREBALOVICH, T. 1966 Grid turbulence at large Reynolds numbers. J . Fluid
Mech. 26, 3 7 4 7 .
LAUNDER, B. E., REECE, G. J. & RODI,W. 1975 Progress in the development of a Reynolds-stress
turbulence closure. J . Fluid Meeh. 68, 537-566.
LAWS,E. M. & LIVESEY,J. L. 1978 Flow through screens. Ann. Rev. Fluid Mech. 10, 247-266.
PINKER, R. A. & HERBERT, M. V. 1967 Pressure loss associated with compressible flow through
square-mesh wire gauzes. J . Mech. Engng Sci. 9, 11-23.
PRANDTL, L. 1932 Herstellung einwandfreier Lufstrome (Windkanale). Handbuch der Exp-
erimentalphysik, Leipzig, Germany, vol 4, part 2, p. 73. N A C A Tech. Mem. 726.
ROTTA,J. C. 1951 Statistische Theori nichthomogener Turbulenz. 2 . Phys. 129, 547-572.
SCHUBACER, G. G., SPANGENBERG, W. G. & KLEBANOFF, P. S. 1950 Aerodynamic characteristics
of damping screens. N A C A Tech. Note 2001.
TAN-ATICHAT, J., XAGIB,H. M. & LOEHRKE,R. I. 1982 Interaction of free-stream turbulence
with screens and grids; a balance between turbulence scales. J . Fluid Mech. 114, 501-528.
TAYLOR, G. I. & BATCHELOR, G. K. 1949 The effect of wire gauze on small disturbances in a
uniform tream. &. J . Mech. Appl. Maths 2, 1-29.
TAYLOR, G. I. & DAVIES,R. M. 1944 The aerodynamics of porous sheets. Aero. Res. Counc. R &
M 2231.
TENNEKES, H. & LUMLEY, J. L. 1983 A First Course in Turbulence, chapter 3.2. The MIT
Press.
WAHRHAFT, Z. & LUMLEY,J. L. 1978 An experimental study of the decay of temperature
fluctuations in grid-generated turbulence. J . Fluid Mrch. 88, 659-684.