11

PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW

THROUGH SQUARE-MESH WIRE GAUZES
By R.A. Pinker* and M. V. Herbert*
The body of previous literature relating to the pressure loss characteristics of woven wire
gauzes fails to provide adequate information on the separate effects of Reynolds number
and Mach number in the range of usual interest for duct flow experiments. For this
reason, tests have been carried out on a selection of eight single gauzes placed normal to
the axis of a duct, of sizes (wire diameter and mesh) such as might be commonly used to
give values of porosity between 0.3 and 0.7. Pressure loss measurements were carefully
made, and covered independently duct Mach numbers up to choking and a 50-fold range
of Reynolds number. The test data are presented so as to provide a single relation between
the incompressible loss factor and a form of Reynolds number, plus a compressibility
correction which depends on gauze porosity.

INTRODUCTION pressure loss hitherto has been the lack of sufficient

MUCH has already been written concerning the behaviour reliable and relevant data, especially for compressible
of wire gauzes or screens placed in an air stream. Rami- conditions.
fications of the subject embrace everything from, on Accordingly, the present experimental investigation
the one hand, abstruse mathematical theory applied to the was undertaken to improve this situation, for the case of a
effect of gauzes on turbulent flow properties to, on the single gauze of woven wire construction, with square mesh,
other, quite crude empirical correlations of data on their placed normal to the flpw, and completely spanning the
pressure losses. Amongst this literature are included duct. Closely spaced multiple gauzes, or other forms of
several analytical treatments deriving the optimum porous material, are specifically excluded from the con-
resistance of a gauze for improving the uniformity of in- clusions. Sizes of gauze (wire diameter and mesh), and
compressible flow-both steady and unsteady. However, duct conditions, have been selected to be typical of normal
it is not always possible to employ a single 'optimum' type, usage. The data obtained should enable a sufficient pre-
and gauzes with a wide range of porosity find applications diction of pressure loss to be made for any gauze in the
in experimental work associated with flow in ducts, such range of porositfr 0.3 to 0.7.
as figures largely in the field of aircraft engine research.
Furthermore, the flow is frequently-indeed, most Notation
usually-travelling at velocities far higher than justify A Cross-sectional area of duct.
the assumption of incompressible flow. d Wire diameter, inches.
For such reasons, the attention of several authors has
already been afforded to the problem of describing the f Volumetric porosity, as used in reference (6).
f A function.
resistance imposed by any gauze, in such a way as to pel'rnit G Geometrical parameter.
ready prediction of its pressure loss at any particular flow 1 Hydraulic mean diameter, as used in reference
conditions. Unfortunately, however, previous papers are
not adequate for successful prediction in the circumstances (6).
M Mach number.
of most frequent interest in this field, namely moderately m Number of strands per inch.
high Reynolds numbers and Mach numbers; their separate N nmd/2.
effects have been recognized, but never properly isolated. P Pressure.
A serious handicap to a complete specification of gauze AP Pressure loss.
The M S . of this paper was first received at the Institution on 25th Q Mass flow.
May 1966 and in its revised form, as accepted by the Council for Re Reynolds number.
publication, on 16th September 1966. 23 T Temperature.
* Ministry of Aviation, National Gas Turbine Establishment, Pye-
stock, Farnborough, Hants. U Mean axial velocity in duct.
JOURXAL MECHANICAL ENGINEERING SCIENCE Vo19 No I I967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

12 R. A. PINKER AND M. V. HERBERT

U Axial component of velocity variation (sub- Collar’s analysis has come in for serious criticism (2) (3).
script 1 refers to far upstream condition and I n reference (2), Batchelor says that Collar unjustifiably
3 to far downstream condition). neglects induced transverse velocity components at the
P Flow factor = U/v (ft-‘). gauze, which are of equal order of magnitude to the small
CI Orthogonal porosity. variations in axial velocity being considered (although
a’ Sinusoidal porosity. transverse velocities far upstream and far downstream are
P Ratio of flow angles defined with reference to of course zero). Later, Taylor and Batchelor (3) point out
equation (4). that the analysis published by Collar in reference (I)
6 Fraction of dynamic head defined with reference depends on the assumption that the change in momentum
to equation (11). of the stream outside the filament is zero, which is not
A ’ Pressure loss coefficient. true. This assumption leads to an incompatibility in static
P Viscosity. pressure across the stream at the gauze, in that the pressure
V Kinematic viscosity. is changing longitudinally along the filament but not along
P Density. the remainder of the stream, which has one constant value
of static pressure everywhere upstream of the gauze and
Subscripts another constant value everywhere downstream. (The
opt Optimum. latter feature of Collar’s theoretical model must be a
0 Incompressible. cause of concern to any reader, without ever having en-
S Static. countered the detailed reasoning in reference (3).)
t Total head. However, both references (2) and (3) agree that equation
1,2,3,4 Forms of G and Re defined in the section on (1) is in fact valid for the special and hypothetical case of
‘Pressure drop data and correlations’. a gauze in steady flow which produces no transverse
velocity components. Batchelor (2) offers another analysis
PREVIOUS WORK which allows for the existence of these transverse veloci-
Theoretical performance of a gauze ties, and emerges with equation (1) again, but he admits
in reference (3) to having erroneously neglected the change
A quite simple analysis by Collar (I)*, for incompressible in transverse velocity through the gauze. Other results
steady flow in a duct, produces the relation given by Batchelor in reference (2) for unsteady flow are
-=- 2-h
ug of interest, in that a value of h = 2-5 supposedly removes
u1 2 S h - -
* * (1) completely any longitudinal fluctuations present in the
where u1 and us are small variations from the mean axial approach flow and gives some reduction of transverse
velocity far upstream and far downstream respectively of a fluctuations, which value of h is very close to that in
gauze placed across the duct, these variations being equation (3); but these results must presumably suffer
considered to exist in a filament of the flow, the remainder from the same defect as his steady-state analysis.
of which is at effectively constant velocity throughout; A rigorous treatment of the steady-state problem is
and where A is the gauze pressure loss coefficient, defined carried out by Taylor and Batchelor (3), who derive
as _ - 1+/3-px
243
AP u1- 1+/3+h
. . . * (4)
A=- . . . .
(%)
For incompressible flow, the changes of total head and
where /3 is the ratio of downstream to upstream angles by
which the flow local to the gauze deviates from being
normal to its plane. Clearly equation (4) is zero when
static pressure (AP, and AP,) are of course equal, so that h = 1+1//3, and for the special case of ,b = 1 (when
it is unnecessary at this stage to distinguish between A, and transverse velocities are unchanged through the gauze)
A,. The dynamic head in the denominator is that of the we come back to equations (1) and (3), as in reference (2).
approach flow. It is immediately clear from equation (1) Another special case cited is that of flow everywhere axial
that u3/ul = 0 when A = 2; that is to say, a gauze for downstream of the gauze, for which /3 = 0 and u3/u1 =
which A = 2 should remove any longitudinal variations in 1/(1+A). It is claimed, however, that neither condition is
steady flow. Thus for incompressible steady flow fulfilled in practice, and that in fact 0 < /3 < 1. Data are
published in an Appendix to reference (3) by Dryden and
A,, = 2 . .* (3). . Schubauer, which corroborate this condition for P, and
Some very limited experimental evidence is included fit the empirical relation
by Collar, which gives some support to equation (3), the
data relating to an air speed of 28 ft/st. If h > 2, then p=- 1.1 (when X > 0.7) . . (5)
velocity variations change sign through the gauze. d(l+A)
* References are given in Appendix 3. Hence
t Other low speed data also tend to support a value of A,,,around 2
(0.C. MacphaiI ‘Experiments on rurning vanes a t an expansion’,
R and M 1876, 1939).
J O U R N A L MECHANICAL ENGINEERING SCIENCE V o l 9 No I 1967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 13

Equation (6) leads to valent to a contraction in which the flow is isentropic

A,, = 2.765 . . . * (7) (total pressure constant), followed by a sudden enlarge-
which supplants equation (3) as the theoretical solution ment in which total pressure falls and static pressure rises,
for steady-state incompressible flow. the initial and final flow areas being equal. Defining a
Despite the long time which has elapsed since the screen 'porosity' as
publication of reference (3) (1949), it is still common to effective open area
-a=
hear the value A,, = 2 quoted amongst those seeking to total area
use a gauze for improving flow uniformity. To be fair (so that porosity = 1 -blockage), it can readily be shown
there appears to be no certain evidence that A N 2-8 is that for incompressible flow
preferable to 2.0 in practice, although Collar's original
data (Fig. 4 of reference (I)) do in fact better support the
higher value.
Of much greater practical importance, probably, is the or
effect of compressibility which is introduced in most
experimental applications. As has long been realized,
increase of approach Mach number leads eventually to the
onset of choking in a gauze (4), and the pressure loss coeffi-
It is not explicitly stated that A, varies with Re, but this
may intuitively be expected.
cient rises to infinity. For a gauze of A = 2.8 when M = 0
A second approach, for porous material of any sort, is
(A, = 2.8), it is physically impossible to operate above
to regard all losses as being due to surface friction in the
M 2: 0.3. The corollary seems to be that A,, must fall as interstices, it being assumed that there is no separation
M is increased. from and no wakes behind the material of a screen. This
treatment is used by Grootenhuis (6) both for single
Pressure drop data and correlations gauzes 'and for close-packed assemblies, but it is clearly
It is necessary to start by distinguishing clearly between less reasonable for a single gauze. His analysis produces
two major effects on the pressure loss coefficient of any the relation
particular geometry of gauze-namely, the effect of
Reynolds number (normally correlated under incompres- A, = 16(1 -f 1f(Re) . . (9)
f"
~

sible flow conditions), and that of Mach number at a given

Reynolds number. The hope may be entertained that the where f is a 'volumetric porosity', and the form of f(Re)
variation with Mach number can be expressed in a form can only be determined from experiment.
which is effectively independent of the particular value of The third model, proposed by Wieghardt (7), applies
Reynolds number, provided that value is constant. Thus only to single gauzes constructed of round wire, and
the two effects may be separable for the purposes of corre- treats the gauze as analogous to a number of isolated
lation and prediction. cylinders placed across the flow. He then introduces the
Some previous authors have created confusion by pre- well-known empirical relation between drag coefficient of
senting data which purport to show the effect of Reynolds a transverse cylinder and Reynolds number, which is
number, but include significant variation of Mach num- adjusted so as to use the maximum velocity between the
ber (e.g. reference (5)), or vice versa (e.g. reference (4): wires of the gauze rather than the approach velocity. This
Adler quotes no levels of Re, but as all his data were ob- leads to
tained at a total pressure of 1 atm, Re must necessarily 1--a
have varied). In none of the known literature is it possible A, = ( ~ f(Re)
2 . * (10) .
to find data for compressible conditions in which the two The group (1--a)/a" as a correlating parameter was earlier
effects can be properly isolated (Appendix 1). One can only suggested by Collar (I), who put the empirical value 0.9
infer that, in the range of flow conditions of present in place of f(Re) in equation (10).
interest, the level of Re is sufficiently high for this para- Another analysis by Taylor and Davies (8) for any
meter to play a comparatively small part, while the screen gives
dominant effect is that of Mach number. 1--6
Considerable attention has, on the other hand, been A, = --
-a2
1 . . . .
paid to the incompressible case, and the way in which
geometrical properties can be correlated so as to produce where 6 is the fraction of dynamic head at the vena
a unique relation between pressure loss coefficient and contraeta which is recovered as static pressure down-
some form of Reynolds number. This situation requires stream. This becomes the same as equation (8) if
to be examined closely, since geometry is of course the -6 = 2c4l-a) . . .
(12) -
third effect of major importance in specifying the pressure or the same as equation (10) at constant Re if
loss of gauzes. Three different theoretical models have
been used.
s = a(1-a) . . . * (13)
In the first, any form of screen is treated as being equi- Since 0 < a < 1, equations (12) and (13) imply 0 < -6 < 3
VoI 9 No I I967
JOURNAL MECHANICAL ENGINEERING SCIENCE
Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015
14 R. A. PINKER AND M. V. HERBERT

or 0 < 6 < respectively, either of which are at least therefore, it must become a matter of seeking which pair
not impossible. It seems more likely, however, that 6 of parameters offers the best correlation.
should not depend heavily on a. Another special case is The literature contains three examples of 'universal'
6 = 0, which leads to correlations, that is to say single curves which supposedly
1-2 embrace all previous incompressible data. Chronologic-
A, = - * *
ally, the first is due to Annand (g), who analysed a lot of
data and obtained by indirect means a single curve in
This relation has been used by Annand (g), who seems to terms of G, and Re,; on first sight this appears the most
regard it more as empirically convenient than as having useful of the three, but as origin31 data are not compared
much theoretical justification, and who presents correla- with the curve, the degree of scatter cannot properly be
tions of data in the form judged. The second comes from Grootenhuis (6) and is in
terms of G, and Re4; many data for single gauzes are
(15) shown, but the scatter about the mean curve reaches as
much as a factor of 2, which situation he attributes to
It will be seen from the foregoing that there is general
experimental error by previous authors; when applied to
agreement in favour of the expression
multiple gauzes and close-spaced packs, the scatter about
A0
= f(Re) . . . . the same curve is of similar magnitude. A third curve has
been derived from the work of Cornell (5), who presents a
where G is some geometrical parameter, definable in plot of G, versus Re, (also without data) showing a spread
terms of a form of porosity. But this is only a small step of lines with variation of a t ; Bushel1 (10) has suggested
forward, since there is a wide choice both as to G and the that these may be approximated by a single line when
form of Re, so that a rather bewildering number of per- plotted in terms of G, and Re,.
mutations are available for attempting a correlation of One forms the impression that many authors have been
incompressible data. As regards Re, there is a choice of content to achieve quite a crude measure of correlation
two obvious velocities, either the approach velocity ( U ) amongst the mass of previously published data; the
or that through the interstices (U/a), and of numerous resultant scatter may or may not then be due to experi-
dimensions, such as wire diameter (d), minimum passage mental error-there can be no means of telling. But this
width [(l/m)-d], or any species of hydraulic mean dia- amount of progress is not entirely satisfactory; for practical
meter. Listing some of the possibilities, and writing purposes it is not good enough to know the pressure loss
- u of a gauze within, for example, a factor of 2. Our present
V=- concern is limited to the comparatively narrow range of
V
Re normally encountered in duct flow experiments, and
we have for this purpose fresh data have been obtained quite
accurately. These will be compared with the three
Re, = Pd 'universal' curves in a later section of this paper; no
Pd attempt is made at correlation with individual old data.
Re, = -
a TEST ARRANGEMENTS
(l--a)2 P 1 Each gauze in t&n was fixed in a 33-in diameter duct,
G3 = -
a2
Re3 = ;(;-d) machined from solid aluminium alloy, in the manner
shown in Fig. 1, so as to present a smooth surface to the
airflow with no blockage of the gauze passages near the
wall of the duct. A standard orifice plate was used to
where, according to reference (a), measure mass flow; this was placed 56 in upstream of the
rrmd (l-md), gauze holder (16 pipe diameters), and was preceded by
f = I-- and I = 228 in of straight pipe, which included a transition from
4 rrm
It is customary to take for a the area of gauze obtained by 4-in to 33-in diameter with a taper half-angle of 0-4".
orthogonal projection, giving what is called here the For a distance of 20 in immediately on either side of the
orthogonal porosity (see a later section of the paper), so gauze holder the pipe was precision bored.
that The rig was arranged in such a way that the air supply
could either be connected directly to the gauze, exhausting
cc = (1-md)2 . . . . (17) through a throttle to atmosphere, or be used to drive an
So far as the utility of equations (8), (S), (10) and (15) ejector by means of which suction was applied to the gauze.
is concerned, the actual aerodynamic situation is too A range of total pressure from 4 to 100 lb/in2 abs. was
complicated for one to expect that any simple theoretical thus available at the test section.
analysis will be sufficient, and there is little reason to
prefer any one of the models mentioned above to any t The combination of GI and Re2 is that proposed by Wieghardt (7)
from his isolated cylinder theory, and his own attempts at correla-
other: indeed, none carry any conviction. In practice, tion conjirm its inadequacy.
JOURNAL MECHANICAL ENGINEERING SCIENCE
Vo19 No I I967
Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015
 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 15

I 1 I I I I :D PRESSURE LOSS
IDUE TO GAUZE

PRESSURE LOSS
DUE TO FRICTION

Fig. 1. Experimental arrangement

Measurement of the gauze pressure loss followed the The tests were conducted in two groups:
method of Morgan (11), which requires a series of wall (1) P constant. M was varied from 0.1 until choking
static tappings upstream and downstream of the gauze. occurred for all eight gauzes (Table 1).
Some distance on either side of the gauze, the pressure (2) M constant (= 0.1). Pwas varied throughout the
distribution obtained will have a slight and continuous available range for three selected gauzes; these were of
slope, due to wall friction. The static pressure loss of the higher porosity in order to reduce the effect of com-
gauze (APJ is then taken as the difference in level between pressibility.
the upstream and downstream pressure distributions
extrapolated linearly to the plane of the gauze (Fig. 1). RESULTS AND DISCUSSION
The total pressure loss (0,) can be obtained by calcula-. Pressure distribution
tion, knowing the mass flow passing.
As already explained, the main interest in this work lay Let us look first at some typical examples of the actual
with fairly high duct Mach numbers, and for convenience wall static pressure distribution, shown in Fig. 2. The
a lower limit of M = 0.1 was arbitrarily chosen. At this general shape is similar in each case and consists of
condition the plant allowed a 25-fold variation of the flow (1) a slight and steady fall well upstream of the gauze,
factor 7 (Reynolds number per foot). This factor was due to wall friction;
employed as the other main variable, because of uncer- (2) a more rapid fall just in front of the gauze, caused
tainty as to the correct characteristic dimension required by gradual acceleration of the flow;
to make it non-dimensional. (3) a large and abrupt fall approximately in the plane
of the gauze;
Table 1. Details of gauzes tested (4) a fairly sharp rise immediately behind the gauze,
corresponding to diffusion of the flow; this pressure
d, m, a, a’, recovery continues to form
wire diameter, strands orthogonal sinusoidal (5) a slight but distinct ‘hump’ about 0.5 in from the
in per inch porosity porosity plane of the gauze;
0.020* 10 0.640 0.741 (6) finally, a gentle fall which is at first rather
0.039* 7 0.529 0.642 steeper than that due to the wall friction far downstream
0.022 14 0.479 0.592
0.048 7 0.441 0.552 of the gauze.
0-022* 16 0.420 0.529
0.0148t 24 0.416 0.524 It is clearly reasonable, as suggested by Morgan (II), to
0.022 18 0.365 0.463 take the effective pressure loss introduced by the presence
0.022 20 0.314 0.394 of the gauze as the difference in level between the lines
for upstream and downstream wall friction as indicated
* Used in tests with varying V. in Fig. 1.
t ‘Odd man out’ in Figs 3,4, 6 and 9.
A generally rather similar picture to Fig. 2 was ob-
Note: The wire diametersof the gauzes were measured as accurately served by Simmons and Cowdrey (12)~who measured in
as possible, allowing for their deformed condition; values given
are not nominal sizes. detaiI the axial variation of both total and static pressure
JOURNAL MECHANICAL ENGINEERING SCIENCE Vol9 No I 1967
2
Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015
16 R. A. PINKER AND M. V. HERBERT

M= 12
).294
DUCT DIAMETER
3.5 in
-x-x-x-x-x-x-x. 11
x. 16/0.022 GAUZE

10

i.
.PLANE OF GAUZE

7
1s

0
F L O W DIRECTION
3

12 REFERENCE TAPPINGS
IN FIG.?
1

I l l 1 I I I I I

I I 1 I
I
DISTANCE FROM GAUZE-In 0
0.1 0.2 0.3
M
0.4 0.5 6

Fig. 2. Pressure distribution Fig. 3. Static pressure loss

through one of the passages in a square grid of rods, remaining one (24/0.0148),the mean curve for which
0.25-in diameter at 0-9-in pitch and presumably unwoven. appears as a dashed line in Figs 3 and 4, agreed with the
In their case the influence of wall friction was absent,.and others as regards pressure loss at low Mach number, but
they could deduce that, with an approach velocity of not as regards the value of choking Mach number (iM*)t.
30 ft/s, the mixing process in the wake region behind the All the data contributing to Figs 3 and 4 were obtained
rods was complete 9 in downstream of the grid; this with 7 N 4 . 3 ~ lo6.
distance corresponds to 10 mesh widths or 36 rod dia- By extrapolating these curves of A, and A, to M = 0 (a
meters. The static pressure recovery exhibited a very fairly easy matter in view of their comparative flatness in
steep initial rise, followed by one more gradual which that region) values of A, can be obtained. These are plotted
terminated at the end of the wake region, where AP, and against porosity ( a ) in Fig. 5, and will only apply to the
A P, became equal. one particular value of 7. Choking Mach numbers can
The cause of our slight downstream hump-item ( 5 ) be taken with confidence from the curves in Figs 3 and 4,
above-is not very clear; no counterpart was found by and Fig. 6 shows M* plotted against a. That relation
Simmons and Cowdrey. For the gauze used in Fig. 2, should be independent of 7.
10 mesh widths = 0.625 in, and 36 wire diameters = 0.79 It is clear from Fig. 6 that the orthogonal porosity (a),
in, so that we may suppose this hump to be just within defined in equation (1 7), considerably underestimates the
the wake region. There seems some likelihood that it is effective open area of a gauze. This discrepancy becomes
associated only with wall pressure and not with static even greater when it is realized that any choked flow
pressure in the stream. passage must have a discharge coefficient below unity
(possibly around 0.95 for a gauze, although this is no
Effect of Mach number at constant flow factor
t In seeking an explanation for this difference, the possibility of a
The variation of A, and A, with M is shown in Figs 3 and 4. signifcant ‘sag’ or bowing was examined. But in order to produce a
Of the eight gauzes tested, data points are included for change in M* suficient to bring it into line with the other results, this
gauze would have to distort more than 10 times as much as the
seven which formed an entirely consistent family. The calculated effect of aerodynamic loading.
JOURNAL MECHANICAL ENGINEERING SCIENCE Vol9 No I 1967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 17

more than a guess). While the orthogonal porosity is gauze, there does not seem to be any great merit in the
hallowed by tradition, and may be retained as a convenient ‘volumetric’ porosity used by Grootenhuis (6).
descriptive label, some interest attaches to seeking a form Two approaches present themselves, based on different
of porosity with more physical significance. For a single representations of the bending in a square-weave wire
mesh. The first is illustrated in Fig. 7, and assumes that a
wire is locally straight between its points of tangency with
wires crossing it; considering the minimum flow area to be
in the plane X-X, this is bounded approximately by four
intersecting ellipses. Use of a porosity evaluated on this
basis results in some improvement, but the open area is

v ‘4.3 x 106 ft-‘

____-

I
TWO POINTS -

I
M I N I M U M FLOW AREA I N PLANE X-X

SHADED AREA REPRESENTS APPROXIMATE UNDERESTIMATE

0.4 0.6 0.8 I N POROSITY BY ORTHOGONAL PROJECTION
a
Fig. 5. Incompressible pressure loss Fig. 7. Porosity of gauze-I (elliptical representation)
JOURNAL MECHANICAL ENGINEERING SCIENCE Vol9 No I 1967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

18 R. A. PINKER AND M. V. HERBERT

still underestimated. An alternative approach assumes the

centre-line of each wire to be formed into a sine wave, as
illustrated in Fig. 8; this time the bounding curves in the
plane of minimum flow area do not have any overlap, so
that this area is further increased. What we have called
the sinusoidal porosity is then given by

. . * (18)
where N = r m d / 2 .
As Fig. 9 shows, use of equation (18) leads to an open

Fig. 10. Compressibility ratio for static pressure loss

M I N I M U M F L O W A R E A IN PLANE X - X

SHADED A R E A REPRESENTS UNDERESTIMATE IN

POROSITY BY ORTHOGONAL PROJECTION

Fig. 8. Porosity of gauze-11 (sinusoidal representation)

Fig. 9. Choking Mach number-11 (sinusoidal porosity) Fig. 11. Compressibility ratio for total pressure loss
JOURNAL MECHANICAL ENGINEERING SCIENCE V o l 9 No I 1967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 19
-
v=4.3x106ft-’
6 6

4 4
10 A0

2 2

0
0 2 4 6 8 ‘0 2 4 6 8 10 12 14
1 --a - 1-a2
G , = z 2- (t2

2 2

0
0 2 4 6 8 ‘0 2 4 6 8 10 12 14
-(q-a)* c-, 16( 1 - f )
3 a2 f*

Fig. 12. Correlating functions of geometry

area greater than one-dimensional theory, to an extent It may be expected that Figs 10 and 11 are applicable at
that would imply a discharge coefficient between 0.9 and any constant value of r,
and such is assumed in their use
1.0. Hence where a ‘true’ porosity is required, the sinu- later.
soidal relation should be used rather than equation (17).
A process of interpolation, applied to Figs- 3, 4 and 6, Geometric parameters
has been used to derive families of curves for &/Ao and Mention was made in a previous section of the confusing
h/’O for even Of Over the range 0*3to Oe7, and situation which exists with regard to various forms ofthe
these are presented in Figs 10 and 11. The variation of
geometric parameter used in correlating incompressible
static pressure loss coefficient is quite well described by
flow data. The values of A, from Fig. are plotted in
the empirical relation
Fig. 12 against the four parameters G,, G2,G3 and G,, as
. . defined earlier. In each case the curve is made to pass
* (19) through A. = 0 at porosity = 1.

v- f t-’

Fig. 13. Results with varyingflow factor at M = 0.1

JOURNAL MECHANICAL ENGINEERING SCIENCE Vol9 No 1 1967
Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015
20 R. A. PINKER AND M.V. HERBERT

17
v -ft-’
Fig. 14. Correlation results for incompressibleflow

The general form of equation (16) implies that a plot of shown in Fig. 14 in the form .h,/G,, which parameter can
)loagainst G should be a straight line through the origin. be seen to produce quite a good correlation. A significantly
It will be seen from Fig. 12 that either G , or G , nearly worse correlation was achieved by use of the function
fulfils this condition, but that G, and Ga do not. At this ho/G,; this is not included, but a measure of the difference
stage, therefore, the field can be reduced to only two can be gained by comparing Figs 15 and 17.
runners .
Comparison with published correlations
Effect of flow factor at low Mach number Three supposedly ‘universal’ curves are to be found in the
As already explained, 0.1 was the lowest convenient literature, as discussed in an earlier section. Figs 15, 16
value at which to hold Mach number constant, while and 17 demonstrate the relative merits of these so far as
varying 7 over a 25-fold range. Results for the three the present work is concerned. That due to Annand (9)
gauzes tested in this way appear in Fig. 13. These data (Fig. 15) is clearly much superior, and indeed shows very
for A, were then corrected to M = 0 by means of the good agreement with our data. Grootenhuis’s curve (6)
curves for &/Ao given in Fig. 10; the amount of this (Fig. 16) is a poor fit, while his parametric groups do not
correction from M = 0.1 is very small, since only higher give a satisfactory correlation (see also Fig. 12 and the
porosity gauzes were used. The resultant values of A, are section on ‘Geometric parameters’). The curve derived by

1‘6

1.4

1.2

0.4

0’ 2

Re, =r d

Fig. 15. Comparison with Armand

JOURNAL MECHANICAL ENGINEERING SCIENCE Vol9 No I 1967
Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015
 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 21

05 Y

-
h0
64
0.4. \

0.3

0.2
- \ X I

Y
1

\
1

-
%

--f 0.
I.
* a .
4
i t *

.
0
1
.
4.

0 -
0
A4
.
.**
4

0.1-

Bushel1 (10)from the work of Cornell (5) (Fig. 17) lies used in correlations; but one can state a definite pre-
reasonably well among the data, but, as mentioned above, ference for the geometric parameter G, as that which is
the use of G I gives a notably worse correlation than G,. most effective.
Two aspects of these comparisons may be noted. (2) Both Annand (9) and Cornell (5) suggest that a
curve of ho/G, after falling steadily with increase of Re
(1) Because our data fall outside the range of up to some value around lo3, then rises again between
Reynolds number where the curves become steep, the lo3 and lo4. There would seem to be an element of
particular form of G ( G l , G,, etc.) is more important wishful thinking in this suggestion, based on an oft-
than the particular form of Re (Rel, Re,, etc.). It is sought analogy with the drag coefficient of an isolated
difficult to discriminate between the latter forms, since cylinder (see discussion in references (5) (7) and (9)).
the term Fis common, and any characterisuc dimension But the evidence in favour of such a rise in Xo/G has
only varies by a factor of about 2 between the three always been very slender; after removing some data
gauzes tested, which is small compared with the total contributed by Cornell (5) in which the effect of Mach
variation of Re. In the region of present interest, there- number rather than Re is clearly dominant (as men-
fore, it probably matters little which form of Re is tioned in Appendix I), one is left with only the three

Fig. 17. Comparison with Cornell1Bushell

JOURNAL MECHANICAL ENGINEERING SCIENCE

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

22 R. A. PINKER AND M. V. HERBERT

data points for a grid of rods from reference (12)quoted The following specific observations from this work may
by Annand (9). These are shown in Fig. 15, the values be noted:
being taken from the original source*. While there can (1) The conventional porosity of a gauze, as obtained by
be no doubt that the measurements of Simmons and orthogonal projection, considerably underestimates
Cowdrey (12)were carefully made, they were dealing the minimum flow area. More reasonable agreement
with velocities of only 10 to 30 ft/s where pressure losses with experiment is afforded by a form of porosity
become very small. The difference in level of their based on the assumption that the centre-line of each
three points is in fact no more than the scatter amongst wire is bent into a sine wave.
our own data, whose mean trend is distinctive and nearly (2) Several forms of geometrical parameter are used in
flat. It seems, therefore, that the suggestion of a rise in the literature dealing with the correlation of pressure
A,/G in this range of Re can be discounted. loss under incompressible flow conditions. The form
used by Annand (9) (called here G2)is significantly
Prediction of pressure loss more effective than any other.
It is apparent from the preceding section that, so far as (3) It is still not clear which of several forms of Reynolds
incompressible flow conditions are concerned, an almost number is the most satisfactory partner in this cor-
horizontal extension of the correlation curve published by relation.
Annand (9) covers very satisfactorily the region of current (4) Previous work has established that the incompressible
interest. The parametric groups are easily evaluated for pressure loss coefficient falls with increase of Reynolds
single gauzes, unlike those of Grootenhuis (6) for ex- number up to some value around lo3; there has been
ample. In general, therefore, it seems that one should some published speculation as to whether the curve
employ Fig. 15 to obtain A,. Where, however, one is con- rises again between lo3 and lo4. In fact it becomes
cerned with gauzes having dimensions similar to those almost flat.
used in the present tests, as in practice will probably most (5) Of the ‘universal’ curves for incompressible pressure
often be the case, then it may be more convenient to dis- loss given by other authors, that of Annand (9) agrees
pense altogether with the rather uncertain forms of Rey- very well with our data, while that of Grootenhuis (6)
nolds number, and work directly from the correlation in does not. In addition the latter’s geometric parameter
terms of 7 given in Fig. 14. is unsatisfactory (see item (2) above).
For compressible flow at any constant v,
the effect of
Mach number can then be obtained from the curves of ACKNOWLEDGEMENTS
&/Ao and AJA, in Figs 10 and 11. The whole procedure is Crown copyright, reproduced with the permission of the
set out in Appendix 2 for easy reference. Controller, H.M. Stationery Office.
CONCLUSION APPENDIX 1
Test data have been obtained which cover
SOME NOTES O N P R E V I O U S P A P E R S R E L A T I N G T O THE
(1) duct Mach numbers from 0.1 to choking, at EFFECT OF A P P R O A C H MACH NUMBER
constant flow factor (= Reynolds number per foot), The only comprehensive data on this effect known to have been
for eight gauzes; published previously are those of Adler (4), and since 1946 that
(2) a 25-fold range of flow factor, at 0.1 Mach num- paper has always been the standard work of reference. Unfor-
tunately Adler’s tests suffered from two defects. First, as men-
ber, for three gauzes. tioned elsewhere in this paper, he varied Reynolds number
These data have been analysed so as to produce simultaneously with Mach number, so that one must not follow
Cornell’s suggestion (5) of taking ho from an incompressiblecorre-
(1) a correlation curve from which the incompres- lation and &/Ao from Adler’s curves. Secondly, he did not deter-
sible pressure loss of a gauze may be found within the mine the choking Mach number (M*) clearly, and his interpolated
present test range; plots show a discontinuity at choking which does not in fact exist.
(2) comparison with three published ‘universal’ Cornell (5) seems to have obtained an effect of Mach number
without realizing it. The set of data for a gauze of solidity 0.665
correlations of other incompressible data; (a = 0-335)given in his Table 2 show a clear upward trend of At
(3) the relation between gauze porosity and choking with M-the highest value of M tested was 0.1834, but a gauze of
duct Mach number; this porosity should choke at M N 0.23 according to our Fig. 6.
(4) curves which give the variation with Mach Despite this trend with M, all the data for that gauze appear in
Cornell’s Fig. 3 for ‘low velocity, high Reynolds number flow’.
number of static and total pressure loss coefficient for From this figure he deduces that, at some Rez > lo3, the curve of
any constant flow factor. Ao/G1rises again; but if the data for this particular gauze are re-
* In evaluating ho/G,for use in Fig. 15, we havafollowed Annand moved, there is no clear evidence for anything but the curve
in taking the porosity as u = (1-md)2, where d = 0.25 in and remaining level. This confusion by Cornell has been noted by
l l m = 0.9 in. But the rods composing this grid were apparently Morgan (11), but is of course preserved in the shape of incompres-
not woven together, so that, while having the same area projected sible correlation curve proposed by Bushell (10)(see Fig. 17 of the
orthogonally as a woven system like a gauze, they would be ex- present paper), since this is derived directly from Cornell’s own
pected to have a greater effective porosity, more nearly that of
two separate lines of rods each having u = ( 1 -md). Despite this, paper.
the agreement with gauze data in Fig. 15 is very good, whereas A few more data at high Mach number are provided by Morgan
there is no agreement when the higher porosity is used. (II), who claims to recognize the distinct effects of M and Re, but
J O U R N A L MECHANICAL E N G I N E E R I N G SCIENCE Vol9 No I I967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015

 PRESSURE LOSS ASSOCIATED WITH COMPRESSIBLE FLOW THROUGH SQUARE-MESH WIRE GAUZES 23

who gives no values of Re for his tests. It seems likely, in poht of pressure given on p. 112 of Graphs for use in calculations of com-
fact, that he has done just the same as Adler and varied both to- pressible aimw (Oxford University Press, 1954) since
gether, since the description of his rig mentions air being sucked V = Pt x (Re per ft per unit pressure)
through the gauze from a bellmouth, which suggests constant total
pressure and temperature, and hence varying Re. Morgan makes a (4) Look up Ao/Gain Fig. 14 at particular V.
number of other unfortunate mistakes in his discussion: (5) Hence obtain A,.
(6) Look up &/Ao or &/Ao in Figs 10 and 11 at appropriate M
(1) He says that some data published by Simmons and Cow- and a.
drey (12) (and quoted by Annand (9) as suggesting a rise of (7) Hence obtain A, or At.
Ao/Gz at high Re) are, like those data of Cornell’s mentioned
above, showing what is really an effect of Mach number. This APPENDIX 3
cannot be so, as the three points in question were obtained at
velocities not greater than 30 ftjs (see the section on ‘Comparison REFERENCES
with published correlations’). (I) COLLAR,
A. R. ‘The effect of a gauze on the velocity dis-
(2) He compares his values of As with curves from Adler (4, tribution in a uniform duct’, A.R.C. R 13M 1867,
said to be for A, but actually those for At. 1939 (February).
(3) He states that ‘Adler’s experimental work was confined (2) BATCHELOR, G. K. ‘On the concept and properties of the
to low solidity ratios, and the higher ranges of his curves are idealized hydrodynamic resistance’, Australian Council
based upon calculation’. That is untrue, as Adler in fact tested for Aeronautics, Report No. 13, 1945 (January).
a gauze of solidity 0.742, whereas his interpolated plots only go (3) TAYLOR, G. I. and BATCHELOR, G. K. ‘The effect of wire
up to solidity 0.65. gauze on small disturbances in a uniform stream’,
The last two of these mistakes have previously been noted by Quart. Jl Mech. appl. Math. 1949 2 (Pt 1).
Bushell (10). (4) ADLER,A. A. ‘Variation with Mach number of static and
total pressures through various screens’, N.A.C.A.
Wartime Report No. L5F28, 1946 (February).
APPENDIX 2 (5) CORNELL, W. G. ‘Losses in flow normal to plane screens’,
Trans. Am. SOC.mech. Engrs 1958 80, 791.
H O W T O F I N D THE P R E S S U R E LOSS C O E F F I C I E N T (6) GROOTENHUIS, P. ‘A correlation of the resistance to air flow
O F A GAUZE of wire gauzes’, Proc. Instn mech. Engrs 1954 168 (No. 34),
(I) Knowing wire diameter in inches (d) and strands per inch (m), 837.
work out orthogonal porosity (7) WIEGHARDT, K. E. G. ‘On the resistance of screens’,
Aero. Quart. 1953 4 (February).
OL = (l-md)’
(8) TAYLOR, G. I. and DAVIES,R. M. ‘The aerodynamics of
1 porous sheets’, A.R.C. R & M 2237, 1944 (April).
(2) Hence obtain Gz = 2-1
(9) ANNAND,W. J. D. ‘The resistance to airflow of wire gauzes’,
(3) For required duct flow conditions work out V, given that J l R . aeronaut. SOC.1953 57 (March).
(10) BUSHELL, K. W. Unpublished work at N.G.T.E., 1963.
= = 9 (units = ft-1) (11) MORGAN, P. G. ‘High speed flow through wire gauzes’,
v 4 J l R . aeronaut. SOC.1959 63 (August).
where A is duct area. Given the duct Mach number ( M ) and total (12) SIMMONS, L. F. G. and COWDREY, C. F. ‘Measurements of
pressure and temperature (Pt and Tt), Y can conveniently be ob- the aerodynamic forces acting on porous screens’,
tained from the curves of Reynolds number per foot per unit A.R.C. R & M 2276, 1945 (August) (published 1949).

IOURKAL MECHANICAL ENGINEERING SCIENCE Vol9 No I 1967

Downloaded from jms.sagepub.com at HOWARD UNIV UNDERGRAD LIBRARY on March 7, 2015