You are on page 1of 24

REFERENCES

Baines, W. D. aud Peterson, E. G., 1951, "An Investigation of Flow Through Screens,"
Manuscript submitted for the SAND--91-142 9C
ASME/AIAA/ASCE/SIAM/AIChE
National Fluid Dynamics Congress DE9 2 00 293 0
June 22-25, 1992
Los Angeles, CA

..
• . ,, ...... . ,,..

REYNOLDS NUMBER DEPENDENCE OF THE DRAG COEFFICIENT


FOR, LAMINAR FLOW THROUGH F!_E-SCALE SCREENS 1

T. J. O'Hern and J. R. Torczynski


Fluid and Thermal Sciences Department
Sandia National Laboratories
Albuquerque, NM 87185-5800

ABSTRACT

The laminar flow downstream of fine-mesh sc:'eens is studied experimentally and numeri-

cally. Two different screen types are examined experimentally, both with open areas greater

than 50°_ and wire dimensions less than 100 /_m. Such screens produce flow disturbances

of much smaller scale than those examined in most previous studies of flow-conditioning

screens and grid-generated turbu._ence. Instead of using standard woven-wire screens, high-

uniformity screens are used which are fabricated b- photoetching holes into 50.8 #m thick

Inconel sheets. The holes thus produced are square with rounded corners, arranged to form

a square array, with a minimum wire thickness (located halfway between wire crossings) of

D = 50.8 #m. A flow facility has been constructed for experiments with these screens. Air at

85 kPa e,nd 295 K is passed through each screen at upstream velocities of 1 to 12 m/s, yield-

ing Reynolds numbers ReD -- pUD/# in the range 2 _< ReD (_ 35. Pressure drops across the

screens are measured at the_c conditions using pressure transducers and manometers. From

these data, the Reynolds number dependence of the drag coefficient CD is determined. Three-

dimensional flow simulations are performed over the Reynolds number range 2.5 < ReD <: 15 ;

using the spectral-element code NEKTON (Nektonics/creare.x). The geometry of the pho-

toetched screens is simulated by a similar geometry with the same open area and minimum

1Supported by the U. S. E'OE under contract DE-AC04-76DP00789. _ ,_,,_ _ _*,_,_

I_iS'_I;,ItBUTION OF THIS DO,..., MEN] IS UNLIMITED

!
¢ W

wire thickness. The drag coefficients are determined from the computed press,:rc differences

across the screens and are in reasonable agreement with the experimental values, although

the agreement degrades slightly with increasing Reynolds number. Comparison with stan-

dard semi-empirical correlations for the drag on screens shows that such correlations are

applicable for the present screens so long as the correct choices for screen open area fraction
0 and minimum wire thickness D are used in the correlation.

From the simulations, the length of the recirculating region downstream of the screen

is found to grow roughly linearly with increasing ReD. Also, the downstream extent of

the recirculating region is different behind different screen locations: it is longest behind

a wire crossing and shortest behind a wire waist. Photorefractive schlieren photographs of

the flow structure immediately downstream of the screen have been recorded. The screen

is electrically heated, and the thermal wake clearly indicates the variation in fluid density

(refractive index) between the heated screen wire wakes and the cold jet flows through the

screen openings. Such schlieren photographs show long, narrow flow structures downstream

of the screen, similar to those seen in the simulations.

DISCLAIMER

This report was prepared as an account of work sponsored by an agency of the Uniled States
Government. Neither the United States Government nor any agency thereof, nor any of their
employees, makes any warranty, express or implied, or assumes any legal liability or responsi- I_"
bility for the accuracy, completeness, or usefulness of any information, apparatus, product, or ._
process disclosed, or represents that its use would not infringe privately owned rights. Refer-
ence herein to any specific commercial product, process, or service by trade name, trademark,
manufacturer, or otherwise dees not necessarily constitute or imply its endorsement, recom-
mendation, or favoring by the United States Government or any agency thereof. The views
and opinions of authors expressed herein do not necessarily state or reflect those of the
United States Government or any agency thereof.

t
J

NOMENCLATURE

Al, A2 screen aspect ratios

CD screen drag coefficient


d woven-wire diameter

D minimum wire thickness (spanwise)

f woven-wire screen drag function

H distance between adjacent wires

Ln nondimensional recirculation length

O open area fraction

/Xp pressure drop across screen

Red woven-wire diameter Reynolds number

ReD wire thickness Reynolds number

U fluid "velocity upstream of screen

W plate thickness (streamwise)

x spati_l coordinates

p fluid density

# fluid viscosity

(') nondimensional ()

()efr effective value of ()

¢ (
f

:1
r • ! 0

INTRODUCTION

The downstream evolution of flow through screens is a fundamental fluid mechanics prob-

lem that is important in a variety of engineering situations. Screens are commonly used for

flow conditioning in wind and water tunnels due to their effect of reducing turbulent length

scales and intensity. As discussed by Roach (1987), several methods for computing the tur-

bulent intensity downstream of a screen require an estimate of the pressure drop across the

screen. The flow through screens is also commonly used to generate quasi-homogeneous grid

turbulence. For these reasons, a large body of published literature (discussed below) is avail-

able on screen flows, and a number of important findings have been made concerning the

behavior of this flow. For example, the flow can be categorized depending on the value of the

screen wire Reynolds number, with the flow laminar for Reynolds number less than 40 and

vortex-shedding occurring for higher Reynolds numbers. In addition, screen nonuniformity

has been shown by a number of investigators to lead to jet coalescence and subsequent for-

mation of larger-scale flow structures (e.g., Bbttcher and Wedemeyer, 1988). However, most

of this work has focused on screens with characteristic wire dimensions several times larger

and with slightly different geometries than those of interest here. For example, Bbttcher and

Wedemeyer (1989) examined flows through screens composed of 0.300 mm diameter wires,

Roach (1987) used wire diameters of 0.914 mm, and Pinker and Herbert (1967) used wire

diameters ranging from 1.22 to 0.376 mm. Notable exceptions are Groth and Johansson

(1988), who examined screens of seven different wire diameters ranging from 2.5 mm down

to 40 #m, and Bernardi et al. (1976) who examined screens with wire diameters as small as

25 #m.

The present study is unique in combining experimental and numerical examination of the

steady laminar flow through highly-uniform fine-mesh photoetched screens. The experimen-

tal and numerical results are also compared with the results of Groth and Johansson (1988)..

The Reynolds numbers of interest are maintained below 40 in order to avoid initiation of

vortex shedding in the wire wakes. The objective of the present experiments and simulations

4
ii
r • ! t

is to determine the dependence of the drag coefficient

Ap
½pu (1)
i on the wire thickness Reynolds number

t pUD
:, Reo = (2)
,' /Z

:1
1
for fixed value= of the screen aspect ratios

H D
J A1 = A2 = (3)

with the quantities p, U, and /Z measured upstream of the screen (see Fig. 1). Roach

(1987) points out that the wire diameter or width D is the appropriate length scale for

consideration of the downstream flow. Therefore, D was chosen as the characteristic length
in the definitions used here.

EXPERIMENTS

The pressure drop is measured for the flow through two different screen geometries (80-

and 100-mesh), with open area fractions O of 0.66 and 0.55, respectively, and minimum wire

thickness D = 50.8/zm. Such screens produce flow disturbances of much smaller scale than

those examined in most previous studies of flow-conditioning screens and grid-generated tur-

bulence reported in the literature. Screen nonuniformity has been shown by a number of

investigators to lead to jet coalescence and subsequent formation of larger-scale flow struc-

tures (e.g., BSttcher and Wedemeyer, 1988). For this reason, high-uniformity photoetched

screens are used in this work instead of standard woven-wire screens. High-uniformity screens

are fabricated by photoetching holes into Inconel (76% Ni, 16% Cr, 8% Fe) sheets of thick-

ness W = 50.8/zm. The holes thus produced


are square with rounded corners, arranged to
¢
form a regular square array (see Fig. 2). The minimum wire thickness D is located halfway "

between wire crossings, and its value is measured for each screen using an optical microscope.

The screen open area is also determined by direct optical microscopic examination. Table 1

lists the pertinent parameters of the two screens examined in this work.
f i

Table 1. Parameters for the two screens of interest.

Mesh (wires/in.) ..... 80 100


Min. Wire Diam. D (/zm) 50.8 50.8
Plate Thickness W (/zm) 50.8 50.8
Wire Separation H (#m) 318 254
Open Area Fraction O _ 0.66 0.54-0.56
Aspect Ratio A1 = H/D 6.2 5.0
Aspect Ratio A2 = D/W
....... 1.0 1.0

An open-loop wind tunnel has been constructed for experiments with these screens.

Screens are mounted to span the 50 by 5 cm rectangular test section, and are held in such

a way that no screen supports protrude into the flow. A steady flow of filtered room air at

85 kPa and 295 K is passed through each screen. Under these conditions, air has a density

of p = 1.01 kg/m 3 and a viscosity of # = 1.82 x 10-5 kg/m.s. Upstream velocities U range

from 1 to 12 m/s, which yield Reynolds numbers ReD = pUD/# in the range 2 _<ReD _ 35.

Static pressure ports are located 2 cm upstream and downstream of each screen. Pres-

sure drops are measured using a MKS Baratron Type 223B differential pressure transducer

(0 to 1 torr), with MKS Model PD_-C-2C power supply and digital readout unit, and an

inclined oil manometer. The pressure transducers used for recording the data have a factory-

supplied accuracy of 0.01% of reading il digit (0.13 Pa). The velocity values are determined

using a calibrated in-line turbine flowmeter (EG&G Flow Technology Model FT-64C1NA-

GEA-2), providing a mean velocity with an uncertainty of approximately -}-0.02 m/s. The

mean velocities are also periodically checked using a hot wire anemometer.

These measured pressure-drop values have not been corrected to account for pressure loss

due to friction in the duct. As shown by Pinker and Herbert (1967), such losses are typically

negligible in comparison to the pressure drops due to the screens.


Fr¢.m the mcasured
.
pressure-drop data, the Reynolds number dependence of the drag coefficient CDis determined ,

(cf. Eq. 1). Figures 3 and 4 show experimentally determined drag coefficients as functions

of the wire thickness Reynolds numbers for the 80-mesh and 100-mesh screens, respectively.

The screen wakes are visualized using a photorefract;_ve schlieren technique (Robey ct

6
al., 1990; O'Hern et al., 1991). In this technique, a crystal of the photorefractive material

barium titanate (BaTiO3) is used as an optical temporal and intensity filter, deflecting the

steady portion of the illuminating beam and transmitting the fluctuating portion containing

phase and intensity perturbations introduced by the flow. Schlieren photographs of the

flow structure immediately downstream of the screen have been recorded. The screen is

electrically heated, and the thermal wake clearly indicates the variation in fluid density

(refractive index) between the heated screen wire wakes and the cold jet flows through the

screen openings (see Fig. 5).


COMPUTATIONS

Three-dimensional flow simulations are performed using the spectral-element code NEK-

TON (Nektonics/creare.x). The following nondimensionalizations (denoted by tildes') are

used in all simulations discussed below: ft = x/H, _ = 1, /) = 0.2, /5 = l, and 0 = 1,

where x represents position. Thus, these simulations correspond to the experimental results

for 100-mesh screens discussed in the previous section (/) = All). The nondimensional

viscosity is scaled by the desired wire thickness Reynolds number ReD according to the

relation
1
[z- 5ReD ' (4)

and the pressure drop is similarly scaled by

ZXp (5)
An- pU2 •

The drag coefficient is thus related to the nondimensional pressure drop by

CD = 2A/5. (6)

Schematic and perspective views of the computational domain and the spectral-element

mesh are shown in Figs. 6 and 7, respectively. The rounded-corner geometry of tl_ pho- :

toetched screens is modeled by a similar geometry with the same open area and minimum

wire thickness. Symmetry is assumed about planes passing parallel to the inflow direction

through each wire midplane and hole midplane but is not assumed along diagonal planes (see

7
Fig. 6). Although this latter symmetry assumption would permit additional computational

savings, it is not made due to the inability of NEKTON to impose a symmetry boundary

condition on a surface that is not normal to one of the coordinate axes. The computational

domain extends over the range 0 <_ _ <_ _,


1 0 ....
< _ < ½, and 0 < _"< 3. The line defined by

= 0 and _ = 0 passes through the center of a hole, and the line defined by _ = _I and_= i

passes through the center of a wire crossing (see Fig. 6). The screen itself is centered at the

streamwise position $ = 1, with leading face at 2 = 0.9 and with trailing face at _, = 1.1.

Thus, the computational domain extends one screen-mesh spacing upstream of the screen

centerplane (_ = 1) and two screen-mesh spacings downstream of the screen centerplane,

Table 2 shows the results of simulations for several different Reynolds numbers in the

range 2.5 < ReD <_ 15 (these results are also included in Fig. 4). Simulations at higher

Reynolds numbers are not performed because the thinness of the boundary layers at the

screen surface requires greater resolution than permitted by the spectral-element mesh shown

in Fig. 7, which in turn significantly increases the required computational effort. The pressure

drop is determined by taking the difference in pressure between the inflow plane (_ = 0) and

the outflow plane (_, = 3). In ali cases, the spatial variation of the pressure within each of

these planes is about 1%. The effect of grid refinement is examined by using successively

higher-order polynomials (5th, 7th, and 9th) to represent the flow within each element. Since

spectral-element methods are expected to exhibit exponential convergence in element order

kNektonics, 1991), an extrapolation is made to infinite order to indicate the accuracy of the

results. This is accomplished by fitting the pressure drop at a given Reynolds number to

an exponential function in element order. The pressure drops determined by extrapolating

this relation to infinite order are typically within 1% of the values determined using 9th-

order elements, which are themselves typically within 1% of the values from the 7th-order

solutions. ¢

Ii _ tl .......
Table 2. Computational results for aspect ratio A1 -- 5.

i
Order 5 Order 7 Order 9 Extrapol.
ReD Ap CD Ap CD Ap CD Ap CD

2.5 4.54 9.09 4.6S 9.36 4.74 9.48 4.79 9.58


5.0 2.98 5.96 3.02 6.03 3.03 6.06 3.04 6.08
10. 2.23 4.45 2.22 4.44 2.22 4.43 2.21 4.42
15. 2.01 4.01 2.00 3.99 1.99 3.97 1.97 3.95

! From the computational flow solutions, the shape and extent of the recirculating region

I immediately downstream of the screen can be examined. The flow in this region has a

I complex vortical structure: streamlines originating near the point (0.5,0.5,0) enter the recir-

culating region, swirl around one or more times in this region, and finally exit this region

near the side walls defined by the planes _ = 0.5 or 9 = 0.5. Figure 8 shows a streamline

originating at the point (0.475, 0.485, 0) that exhibits this phenomenon for the simulation

with ReD -- 2.5.

The maximum extent of the recirculating region can be determined in the following

' manner. The streamline that exits the computational domain at the point (0.5,0.5,3) can be

_: traced backward through the domain. For each of the four Reynolds numbers, this streamline

I is shown in Fig. 9. Note that this streamline does not enter the computational domain at

the point (0.5,0.5,0), as discussed above. The point at which the streamline reattaches to

the line defined by _ = _ = 0.5 indicates the maximum extent of the recirculating region.

A plot of the nondimensional distance Ln between the downstream screen face (4 = 1.1)

and this reattachment point as a function of Reynolds number ReD is given in Fig. 10.

From Figs. 9 and 10, the length of the recirculating region is seen to grow almost linearly

with Reynolds number for ReD > 5 . Recalling that the screen mesh H is five times the

nfinimum wire thickness D, it is seen that the length of the recirculating region is nearly 7D

for ReD --- 15, which is much longer than the recirculating region behind a cylinder of the

same diameter at the same Reynolds number (Van Dyke, 1982).

The shape of the recirculating region is difficult to depict in two dimensions but can be

characterized by its "footprint" on the 9 = 0.5 plane (or, equivalently, on the _ = 0.5 plane).

9
This footprint is given approximately by the curve in this plane for which the streamwise

velocity component vanishes. The footprint of the recirculating region is shown in Fig. 11 for

the simulation with ReD = 2.5. It is seen that the recirculating region is of greatest extent

downstream of a wire crossing (k = 9 = 0.5) but becomes much shorter downstream of a wire

waist (:_ = 0_ 9 = 0.5). The footprints have the same bell-shaped curve for tile other three

simulations but grow increasingly long with increasing Reynolds number. Extrapolation of

these results and the results of Figs. 9 and 10 to larger Reynolds numbers suggests that the

recirculating regions downstream of the wire crossings are related to the long, narrow flow

structures observed in the schlieren photographs (Fig. 5).


DISCUSSION

Figure 4 shows the experimentally and computationally determined values of the screen

drag coefficient CD. Direct comparison of the two data sets can be made only for the 100-mesh

screens over the Reynolds number range 2.5 < ReD < 15. The computational values agree

reasonably well with the experimental values (to within about 20%) at lower values of the

Reynolds number but increasingly overpredict the drag coefficient as the Reynolds number

becomes large. This is probably due to two aspects of the simulation. First, the computa-

tional domain is probably not long enough in the upstream direction. As a result, the flow is

turned more sharply by the screen in the simulations than occurs in the experiments, which

causes larger pressure drops. Second, the model of the screen used in the computations has

sharp corners, at which separation is forced to occur, whereas the real screens have beveled

edges (see Fig. 2). The sharpness of the screen edges is an important parameter in deter-

mining the pressure drop, oince the precise shape of each individual surface will control the

location of separation. As a result, the flow in the experiments probably remains attached

further along the screen surface, and the delayed separation (compared to the simulations)

permits some pressure recovery and reduces the drag slightly.

Groth and Johansson (1988) have proposed a correlation relating the drag coefficient cD

to the wire diameter Reynolds number Red and the open area fraction O of the form

CD = f(Red)[O -_- 1]. (7)

10

!1
This correlation was developed for woven-wire meshes constructed from wires of round cross-

section, with various wire diameters and screen open areas. The values of the function f(Red)

used below are read directly from Fig. 1 of Groth and Johansson (1988), with an approximate
error of +0.01.
, ,

It is of interest to apply this correlation to flow through photoetched screens with the

parameters discussed above. However, due to the obvious differences in wire cross-section and

screen geometry from the Groth-Johansson screens, it is not entirely clear how the Reynolds

number based on minimum wire thickness D, used in the previous sections, compares to the

Reynolds number based on screen wire diameter d used by Groth and Johansson. Three

possibilities for calculation of the Reynolds number are outlined below. First, an effective

open area fraction can be calculated assuming circular wires of diameter d = D as O_ =

(1 - D/H) 2. Second, an effective wire diameter den can be calculated assuming the prescribed

open area O as den = H(1 - v/-O). Third, the actual measured values of D and O for each

screen can be used. Tables 3 and 4 show the Groth-Johansson drag coefficients for 80-mesh

and 100-mesh photoetched screens, determined from their correlation using the three different

methods discussed above. In these tables, the Reynolds number Red is equal to ReD except

when the length scale den is used.

Table 3. Correlation results for aspect ratio A1 = 6.2 (80-mesh screen).

ReD Using Oe_ and D Using O and de_ = 59.7 izm Using O and D

_ _ 4_00_14_,_. _ 00__,, _ _40__8


Red f(Red) Oe_ CD Red f(Red) 0 CD Rea f(Rea) 0 CD

_0_0 _, 0_1_
10.c _00_ 0_1
_ . _, __0.000
11_ 0_0_,
_, _0 _ 0_0_1
_00_ 00_
_
_ _0 1_00_11_1_0,_00_0_0_1_0_0 000_
_, _0 0900_10_0_1_0_00_0
.... _0__0 0_00__
t,.a'

Figures 3 and 4 compare these correlations to the experimental and computational results '

discussed in the previous sections. Figure 3 shows that use of D and Oe_ produces drag

coefficients that are in agreement with the 80-mesh screen experiments, whereas use of den

and O or use of D and O results in a significant overprediction of the drag coefficient. Figure

11
Table 4. Correlation results for aspect ratio A1 = 5 (100-mesh screen).

ReD U_ing Oen and D Using O and deft = 63.9 #m Using O and D
Red f Red') O_fr CD Red f.(Red) 0 CD Rea f!Red) 0 ......
CD
2:5 2.5 4.46 0.64 6.43 3.1 3.67 0.56 8.03 2.5 4.46 0.56 9.76
5.0 5.0 2.71 0.64 3.91 6.2 2.34 0.56 5.12 5.0 2.71 0.56 5.93
10.0 10.0 1.78 0.64 2.57 12.6 1.51 0.56 3.31 10.0 1.78 0.56 3.90
15.0 15.0 1.36 0.64 1.96 18.9 1.16 0.56 2.54 15.0 1.36 0.56 2.98
35.0 35.0 0.90 0.64 1.30 44.1 0.80
....
0.56 1.75 35.0 0.90 0.56 1.97

4 shows that use of D and O produces drag coefficients that are in agreement with the 100-

mesh screen experiments and simulations, whereas use of d_fr and O or use of D and Oefr

results in a significant underprediction of the drag coefficient. The cause of the different

behavior for the two meshes is not known at present. The spread of the values predicted by

the three correlation methods is lower for the 80-mesh than for the 100-mesh screens, one

J
reason being that each wire in a photoetched screen decreases the actual open area from the
t effective value based on wires of diameter d D. For the 100-mesh screens there are 20_

more wires than for the 80-mesh screens, so the deviation from the effective value is larger.

These observations are valid only over the limited Reynolds number range examined here.

f
For example, Roach (1987) states that for a square mesh array of square rods, the same

as the computational geometry considered here and similar to the experimental geometry,

the geometric term in Eq. 7 should be raised to the power 1.09 to achieve the best fit of

available experimental data over the Reynolds number range 102 < ReD < 104. Applying

such a correction to the correlations shown in Figs. 3 and 4 does not lead to a noticeable

improvement in their agreement with the experimental data, indicating that such a power

relationship may not be needed in the low Reynolds number regime considered in the present

study.

¢
,'

12

n lr
CONCLUSIONS

Experiments and three-dimensional computations ha.ve been performed to study the flow

through and downstream of screens. A comparison has been made between experimental and

computational determinations of the drag coefficient for flow through screens at a variety

of Reynolds numbers. Iu the cases examined to date, the experimental and computational

w'aes of the drag coefficient are in reasonably good agreement over Reynoids numbers in

the range 2.5 < ReD < 15. The extent of the recircu]ating region behind the screen wires

is found to grow almost linearly with Reynolds number but to vary significantly in length
behind different spanwise locations along a wire. The maximum extent occurs downstream

of wire crossings and is found to be unusually long when compared to the rech culating region

downstream of a circular cylinder at the same Reynolds number. These long, narrow recir-

culating regions are hypothesized to be related to the long, narrow flow structures observed

i using photorefractive schlieren photography. Comparison t,a empirical correlations reveals


that woven-wire correlations are applicable for the photoetched rounded-corner geometry

with reasonable accuracy so long as the correct values of screen open area fraction 0 and
minimum wire thickness D are used in the correlation.

ACKNOWLEDGMENTS

The authors wish to thank John Lewin and Bob Meyer for their assistance in design

and fabrication of the screen test facility, and to Tom Grasser for his help in performing the

experiments. This work was performed at Sandia National Laboratories, supported by the
U. S. Department of Energy under contract number DE-AC04-76DP00789.

¢
REFERENCES

Baines, W. D. a, Jd Peterson, E. G., 1951, "An Investigation of Flow Through Screens,"

Trans. ASME, Vol. 73, pp. 467-480•

Bernardi, R. T., Linehan, J. H., and Hamilton, L. H., (1976), "Low Reynolds Number Loss

Coefficient for Fine-Mesh Screens," Trans. ASME, J. Fluids Engineering, 762-764•

BSttcher, J. and Wedemeyer, E., 1989, "The Flow Downstream of Screens and its Influence

on the Flow in the Stagnation Region of Cylindrical Bodies," J. Fluid Mech., Vol. 204,

pp. 501-522.

Groth, J. and Johansson, A. V., 1988, "Turbulence Reduction by Screens," d. Fluid Mech.,

Vol. 197, pp. 139-155.

Laws, E. M., and Livesey, J. L., 1978, "Flow Through Screens," Ann. Rev. Fluid Mech.,

Vol. 10, pp. 247-266.

Nektonics, NEKTON User's Guide: Version 2.7fl, 1991, Nektonics, Cambridge MA.

O'Hern, T. J., Robey, H. F., Torczynski, J. R., Neal, D. R., and Shagam, R. N., 1991,

"Measurements of Passive-Scalar Spectra for Grid Turbulence Using a Nonlinear Optical

Technique," Fluid Measurements and Instrumentation Forum, FED-Vol. 108, R. Gore, G.

Jones, H. Hayami, and M. Nishi, eds., American Society of Mechanical Engineers, New York,

pp. 29-33.

Pinker, R. A., and Herbert, M. V., 1967, "Pressure Loss Associated with Compressible Flow

Through Square-Mesh Wire Gauzes," J. Mech. Engr. Sci., Vol. 9, pp. 11-23.

Roach, P. E., (1987), "The Generation of Nearly Isotropic Turbulence by Means of Grids,"

Int. J. Heat and Fluid Flow, Vol. 8, No. 2, pp. 82-92.

Robey, H. F., Albrecht, G., and Moore, T., 1990, "An Optical Technique for the_])irect •

Measurement of the 2-D Spectral Density of a Passive Scalar in a Turbulent Flow," AIAA

Paper 90-1667, American In, titute of Aeronautics and Astronautics, Washington DC.

Van Dyke, M., 1982, An Album of Fluid Motion, Parabolic Press, Stanford CA, p. 28.

lA
._ p, T, p, I.'_,U _ p-,_p
!t .....

screen screen
side view front view

Pig. 1. Schematic of flow through a screen.

Fig. 2. 80-mesh screen geometry photograph.

15
6 .... I .... I .... I .... I .... 1 .... I ....
4>
;. --"-- Measured
_ --A--. Corr (Oeff, D)
t

o ; ....o .... Corr (0. doff)


." _. -...o-.- Corr (0, D)
c,.:) ,-,
t_
,\\
,.._.
•_ _/,.
%',-..:.,
NI--- n. "..%.

0
Q,) '_ . ".',":.::,
(..) =., ...:,
'% "8"-..
(0
o') 2 "_ii _ "'"'.'.".,.
L _"_ li ..,.,,,...... _ .....

a --._._;;.._.;._..
" ........ .....-.:.-.:...
m.-- ......
= :::'" :.;..:.'.:......... .
........ ... ,,._'_...,,._...._;,: :.-.
:;

0 .... t , , , , I .... t .... t I I I ' I , , , , I ....

0 5 10 15 20 25 30 35

Reynolds Number ReD

Fig. 3, A comparison of different methods of determination of the functional dependence

of the screen drag coefficient ep on the wire thickness Reynolds number ReD for the 80-mesh
10 ' ',.,' ' i .... i .... I .... i .... , .... I ....
screen. •
=_ i"_
| Measured
le----
I Simulation
0\_ iA ..... Corr (Oeff. D)
" Io.........
o,rlO,
(,.} lo...... Corr 10, DI
" _\\
-!
.__ 6 ,, _,_-..

_,_

(..) 4 _., ...',,,_., -e


""%,,,, "(_.,_" Ih_, m %¢_
_ii _
.,= ....... ,@.... i_l._ -
_,-.,. "0 .......... ''' '-. III,.'J':.II --.,,.,,- _ .
'let "''' ''"...... " _. = ""=m l _mm_i mi

.Q .._ .., .,, ,,. ,,. ,.., "'''''''*........,,..,o ="''.,..,,,..i

o
0 5 10 15 20 25 30 35

Reynolds Number ReD t /

Fig. 4. A comparison of different methods of determination of the functional dependence

of the screen drag coefficient CDon the wire thickness Reynolds number ReD for the 100-mesh
screen.

16

|
. °

upstream screen downstream


location

Fig. 5. Screen flow photorefractive schlieren photograph.

symmetry assumed
y=O 7 °
x=O x=1/2

Fig. 6. Schematic of screen plane used for computations (tildes are suppressed in the

figure).
17

l1
I

= ,,_
/_ /
/ t
I ]
, IS Z=O

screen
, location
: / Z=l

y=O _z- =3
X=O X= 112

Fig. 7. Perspective view of the computational domain. (tildes are suppressed in the
figure).

18

|
° .....
t •

x=y=O

x+y=l/2 ....

x=y=l/2 ---_
z-O_ z-3-

x=O, y=1/2 ....

x=y - _""_-----

x = 1/2, y = 0
z=O Z-3

x= 1/2

x_o z=O - z=3

y_O .....
z=O z=3

Fig. 8. Several views of the streamline entering the computational domain at the point

(0.475, 0.485, 0) for ReD = 2.5. It enters the domain, swirls around immediately behind the
screen, and exits the domain near the plane _7= 0.5 (tildes are suppressed in the figure).

!9
1
X= 1/2

x=O
z=O z=3

X= 1/2

x=O l__ RED=5


'
z=O z=3

x= 1/2

x=O
z=O z=3

X=x=O
1/2 J-_ ,, I_ -"'_ ReD=
], 15

z=O z_-3 • mmm

Fig. 9. Views of the streamlines that indicate the maximum extent of the recirculating

region (tildes are suppressed in the figure): (a) Re = 2.5; (b) Re = 5; (c) Re = 10;

(d)R_= zs.
2O

,|
m i, •

0.0 ' i i , I , , .i , I i , , i I i , i '


0 5 10 16 20

Re D

Fig. 10. Plot of the nondimensional maximum length of the recirculating region against

Reynolds number.

• Z

_t

z=0.9 z=1.1 z=1.3


Fig. 11. The contour of vanishing streamwise velocity component approximates the

footprint of the recirculating region. The case ReD = 2.5 is shown here (tildes are suppressed

in the figure).
21

You might also like